Check out the full podcast below, or on non-Spotify podcast platforms here.

1. ‘Undercurrent’ — Pu Poo Platter — from Undercurrent

This EP features four fun tracks of funked-up, globally influenced jazz from a Brooklyn-formed band who are really finding their rhythm. They make infectious music with a strong sense of joy throughout. Alas, not everyone’s a fan — watch the video below where Pu Poo Platter get into trouble with the po-po for livestreaming a set on a Shanghai street.

2. ‘忧郁干杯乐园’ — the three mice — from where is my mice?

Building on their explosive eponymouse record from 2024, Shenzhen emo / indie-rock act the three mice return with what they’re calling their first album proper. where is my mice? brings a similarly febrile energy (it even includes a reworking of one of those older tracks) and once again showcases the band’s intelligence in knowing when to let things get a little messy and when to pull back and allow their sense of melody and songwriting craft to shine through. It’s a delight. Grab it now while it’s pay-what-you-want on Bandcamp.

3. ‘错过的提醒’ — Lygort Quartet — from Through Voodoo Light to Shore of Tomorrow

What do you do when your pioneering post-punk band no longer wants to make post-punk? Form another band, of course. Three-quarters of Chinese post-punk godfathers PK14 have been involved in a variation of the Lygort bands so far, with the latest — the Lygort Quartet — still featuring Yang Haisong and Shi Xudong, key figures in PK14. With Yang and Shi now complemented by Hiperson guitarist Ji Yinan and ex-Dirty Fingers / Sleeping Dogs (and occasional PK14 tour) drummer Li Zichao, the band has some serious talent to play with. As if that wasn’t enough, their new album — largely recorded in a house in rural Sichuan — also features a vocal cameo from Hiperson frontwoman Chen Sijiang.

4. ‘B1’ — gogoj aka Sheng Jie + Deng Boyu — from HEH

Two stalwarts of the Chinese avant-garde scene come together for HEH, an album so good they decided to make it twice. The pair, with Sheng Jie on electric cello and Deng Boyu providing percussion, recorded two improvised takes based on the same structure and framework, but then couldn’t decide which one they liked best so simply released both. All the better for us as listeners — we get two pieces that converge at times yet remain distinct, and that make for fascinating, surprisingly accessible listening.

5. ‘心的傾訴 Jürêk tolƣawe (Unbosom Myself)’ —Mamer — from Bar xatteƣem, Bar ⱪayƣem (All the Joys, All the Pains)

[This track is Lily’s favorite from today’s playlist!]

Created based on a “passing impulse” and recorded in just a few days, All the Joys, All the Pains finds Xinjiang-born maestro Mamer offering interpretations of classic Kazakh lullabies, traditional poems, laments for lost loved ones, and 1980s radio hits. While he’s tended toward the experimental ever since releasing a glorious folk debut on Peter Gabriel’s Real World Records, here Mamer is in accessible mode, with typically mesmerising results.

Subscribe now

6. ‘没东西可掏 The Well Is Dry’ — Zhang 'NoNo' Xingchan 张醒婵 — from Catch My Boomerang

Nono’s new one. The Wuhan one-woman sensation returns with a full-length follow-up to the break-out LP No, No!, a record that featured on numerous album of the year lists in 2024. The pop-coated eclecticism that defined that first record — and excited so many listeners — is still present on this new effort, but Boomerang also feels more considered, more mature. Its impact is not as instantaneous as her debut and to be honest I’m still making my mind up on it, but there’s no disputing it’s a notable release, and there’s certainly depth for those willing to spend a little time with it.

7. ‘Gaia’s Breath’ — Wu Zhuoling 吴卓玲 — from Uncharted II

Glorious ambient and downtempo sequel from a figure who has helped shape Chengdu and China’s musical landscape over the past two decades. Partly inspired by a UFO encounter she had as a child in Sichuan, Uncharted II is now available on beautiful-looking vinyl via Bandcamp, combining the new tracks, the original Uncharted EP and remixes from Cola Ren and more.

8. ‘the one sounds like sweet trip’ — Cassette Player Romance — from Meaning of Home

Luo Oumeng has had a hand in two of the major movements in Chinese underground music in recent months: the Chinese emo revival (as part of The Beneficial Society) and the country’s alternative hip hop resurgence (as Soulooper). This is not just a sign of his talent, but also of how broad his tastes are, attributes that are also demonstrated on his debut solo record as Cassette Player Romance. It’s a charming album that defies easy pigeonholing — label Qiii Snacks describe it as one for “Slowcore/Indietronica/5th-wave Emo/Lofi-Pop/Slacker rock enthusiasts”, which probably just about covers it.

9. ‘黑天鵝(feat. LukeDaPlaya)’ — Guo Yinan 鍋一楠 and Yuan 猿 — from The Vortex Reconstruction Experiment

“We will not erupt in silence,” declares rapper Yuan at one point on this experimental, abstract-leaning rap record with producer Guo Yinan. Examinations of social issues, youthful disenchantment and frustration are mixed with surges of emotional reflection from the Xi’an-based duo and some well-chosen guests from China’s alternative hip hop scene, while the clever production makes uses of a swirl of samples featuring everyone from Yussef Dayes and Marlena Shaw to Andrzej Korzynski and Sam Wilkes.

10. ‘Extinguished City’ — DEADTREES — from New World

A one-man-band from Lanzhou in northwestern China, DEADTREES has played with more atmospheric and abstract textures in the past, but on New World he opts for punchiness, lacing his black metal-rooted sound with some high-octane percussion and free jazz elements. A record that has picked up enough of a following that it’s already been given a second pressing on vinyl.

11. ‘让我飞吧’ — Lie 劣 — from Landfill

As with his DJ Gurl persona, Guizhou hyperpop producer Lie expertly blends a plethora of references and samples on Landfill, mixing everything from gay Taiwanese dramas and folk band Wild Children to Michael Jackson and outspoken Beijing punks, but he also adds a new layer to it all: personal, confessional lyrics. Heartbreak, illicit rendezvous, coming-out tales and pushbacks against prejudice all feature on this bold statement of a record.

A Spotify playlist with nine of these songs is available here.

In late July, leaked minutes from a four-hour meeting between Liang and investors circulated around the internet. Like everyone else, we’ve been pouring over the minutes to understand Liang the CEO, DeepSeek the company, and Liang the man. Liang comes off as a genuinely unusual character among China’s tech elite: a person who sincerely holds a specific and highly personalized worldview, if not an entire ideology. Whereas others in his walk of life are highly attuned to commercial trends and political winds, he positions himself as squarely focused on realizing his vision for technology — and it’s hard to doubt his conviction when he has invested so much of his literal worth into this quest.

His is a confidently singular thesis: that there is an inevitable causal relationship between automated learning and generalized intelligence, and that the pursuit of such advanced machine intelligence is the only problem worth solving right now. Being Chinese means working within the constraints of current hardware limitations, but DeepSeek frames their own ambitions on a much farther horizon. Whether or not they can, of course, might not be up to them.

At ChinaTalk, we’ve been covering the DeepSeek story since before R1. The following are our thoughts and observations after reading the leaked minutes. We get into:

• How Chinese open models work as a business model;

• Why, according to Liang, learning is the path to AGI;

• What China’s political economy looks like from the vantage point of Hangzhou;

• How DeepSeek is managed differently;

• And what the Liang Wenfeng-Demis Hassabis comparison reveals and misses.

The business logic of open source, revealed

If the weights are open, how exactly does DeepSeek make money off of its models? Liang told investors that the lab’s biggest source of revenue comes from business customers. When R1 came out in 2025, there was a wave of Chinese businesses and government entities connecting DeepSeek to their internal systems. While much of that was a fad, some of those corporate customers apparently stuck. Since all their models since R1 have been released openly under the MIT license, DeepSeek does not earn revenue from on-premise deployments by corporate entities, so those actual corporate customers must be paying for API tokens.

The DeepSeek team published an analysis in February 2025 showing that on an average day, R1’s API earned the lab US$562,027, at a 545% profit margin. Fast forward to now, and the hypothetical figure Liang apparently gave in the investor meeting for DeepSeek’s enterprise-end revenue this year is in the hundreds of millions of US dollars. With enough growth in this area, he figured, DeepSeek could even turn a net profit soon, paving the way for a successful IPO. Failing that, and API sales to individual consumers have room for growth as well.

That being said, Liang has little interest in consumers. He confesses that at one point last year, the lab even considered sunsetting its consumer products given how little effort went into maintaining them. But DeepSeek’s chatbot and individual API users were incredibly loyal, apparently, so the team ultimately decided to keep the lights on. (I wonder how many of these are roleplayers…)

From Learning to AGI

Minor problems like keeping users around do not concern Liang. As long as the upper limits of machine intelligence are still out of view, there is more (technological, societal, and literal business) value in pursuing that than there is in building business models based on what is available today. More capable models easily pull the rug out from under competitors, especially when the latter have fallen into path dependencies based on older technology. Once truly generalized and convincingly superior intelligence arrives, creating to-C and to-B products will be trivial. Before then, the vast majority of DeepSeek’s resources will go into probing the upper limits of intelligence — and nothing else.

Liang describes this work as the “main quest” of AGI, which, to him, covers research in the realm of general-purpose intelligence, agents, chain-of-thought, etc. He is careful not to commit to specific subfields or theories, but does explicitly exclude some trajectories. World models, for example, are to him “irrelevant” for the pursuit of higher levels of intelligence, even though he thinks embodied AI is largely inevitable. He argues that once AI research is based on “self-iteration,” these systems will then solve for embodied intelligence in the physical world. Before then, however, he believes the most important problem right now is learning.

AI training today relies heavily on high-quality, labelled data being fed into models — not autonomous and truly continuous learning. Liang cautions investors not to think of “learning” as a sub-technology the way they do with “agents,” but instead to think of it as a problem to be solved. (Indeed, one takeaway from reading Liang in his own words is that despite being a finance veteran, he seems to see the world as an unfolding kaleidoscope of puzzles the way an academic researcher does.) The path to AGI, in his eyes, runs through mechanisms that allow models to keep acquiring knowledge. He is quite honest that he and his team don’t know what that path looks like yet, but he’s asking investors to bet on their clarity of thought.

China, According to DeepSeek

Throughout the conversation, Liang seems to implicitly assume that future AI systems — and eventually AGI — will be open. While America will maintain its capability advantage for now, China, in his eyes, will play the role of token factory at global scale, pushing the price of intelligence down as it did for countless other industries during its manufacturing boom.

Liang appears to have somewhat settled into the national-champion role. Phrases like “historic mission” slip through when he discusses how the hardware chokehold China faces will be eroded and finally dismantled, as if the forces of history make that inevitable. Liang expects Nvidia’s CUDA moat to erode, and — with limited specifics — expressed cautious optimism about training on Huawei chips. In fact, he presents DeepSeek’s relationship with Huawei as cordial and regards working with domestic GPUs as an inevitability. Huawei apparently allocated 16,000 Ascend 950 GPUs to DeepSeek, which is fewer than what it sold to “bigger internet companies” (probably ByteDance, Alibaba, and Tencent). Revealingly, Liang remarked that these competitors need Huawei chips more than DeepSeek does because DeepSeek can “acquire some noncompliant chips,” and that DeepSeek’s main rationale for spending on Huawei is to support a domestic hardware ecosystem.

Liang does not want to be an out-and-proud part of China, Inc. He knows, however, that Chinese domestic integration is inevitable, and sees value in aligning with Beijing on tech sovereignty. The trouble is convincing investors that they should really feel so optimistic about model training on domestic GPUs — it’s the topic that collected the most number of questions during the Q&A.

A New Theory of Management

Liang clearly takes great pride in the company he’s built. To investors, he pitched a story of a team that commits to its work not only because of the substance or the mission, but also because of a culture that’s rare in China’s tech industry. DeepSeek researchers, apparently, rarely work overtime. Management strives to ensure that “mandatory” tasks — those based on central goals decided collectively — do not take up more than half of employees’ time at work. The other half is entirely self-directed, and employees are supported in pursuing whatever research projects interest them as long as there is compute available.

Of course, Liang is not doing this entirely out of the goodness of his heart (though he does seem to have a refreshingly normal view of human performance, having once noted that he believes a person can only concentrate for six to eight hours a day). In the meeting, he argued that a relaxed environment is a necessity for research. Good research ideas come from exploring idle curiosity, and there would be no room for such curiosity if researchers were constantly preoccupied with pressure. To maintain such an environment, Liang focuses on “saying no” to projects that do not relate to their main mission; he told investors that this is why they intentionally did not make improvements to many existing products.

Subscribe now

Will a new theory of technology and productivity finally defeat the dreaded 996? Rather than jumping onto every vertical and rushing to be first, as is the norm across much of China’s internet industry, Liang is betting that focus and discipline will win the day in the AI chapter. Unlikely in Silicon Valley, where the AI gold rush is unleashing extremely workaholic founders and pushing even longer hours onto employees, Liang doesn’t sound like he is racing against time. (He’s certainly not ending up in the permanent underclass personally.) In his eyes, it seems, the horizons for machine intelligence are still so wide that there is much more work to be done before anyone should panic. It’s in the interest of the follower, after all, to hope that the race remains far from over. But maybe Liang is also less worried about the consequences of such a race: throughout the four-hour conversation, the dangers of potential AGI never come up. The pursuit of it seems inherently worthy to Liang.

Is Liang Wenfeng the Demis of China?

Liang is inspired by the history of Bell Labs but does not see DeepSeek as an exact replica. When investors asked about the tension between DeepSeek’s research idealism and the urgency of constructing a sustainable business, Liang argued that missions do not cost good companies revenue. Rather, worthy missions make for even stronger paths to commercialization. In other words, not only is research possible inside for-profit companies, but sometimes better outcomes happen when there is capital involved.

There’s another AI CEO who arrived at a similar conclusion — Demis Hassabis. Sebastian Mallaby’s biography of Hassabis, The Infinity Machine, documents how DeepMind became a Google subsidiary. Former Google CEO Larry Page presented Hassabis, who resented business development work, two future scenarios: either spend the prime of his career gathering Google-level resources for the real work DeepMind hopes to accomplish, or do that work right now on Google’s back. To Hassabis, the choice was obvious.

But Demis’s bet did not turn out the way he wanted. Google DeepMind became much more entwined in Google’s corporate labyrinth than he expected. He complained of becoming just another employee from the perspective of Mountain View, as he was no longer the sole deciding vote steering what DeepMind produced. Unlike Liang, who regards harnessing commercial incentives for “pure” AGI development as entirely possible, Mallaby describes Hassabis as disappointed in what transpired, even though he remains convinced of the necessity of keeping powerful technology in the hands of those who understand it most.

On August 5th, all this tension ripped wide open. Hassabis has resigned from Google and will pivot to working on AI-assisted drug discovery at Isomorphic Labs, as well as researching “societal impacts of AGI” per a Google spokesperson. The siren call of the mission pulled him away from a disheartening situation — but also away from his life’s work at DeepMind.

Hassabis should certainly not be jealous of Liang’s personal predicament. While Hassabis was merely constrained by the complex politics of a world-famous company, Liang’s firm now symbolizes the US-China AI race. To shield his company from the eye of the storm, Liang takes pains to shape the domestic narrative around DeepSeek. There is to be no heroism or glorification of specific researchers; he repeatedly emphasized that DeepSeek is made up of “regular people” whose secret sauce is teamwork. He mentions at one point that the government “won’t give [them] a cent” should the company fail. This shouldn’t be taken literally given the asymmetric relationship between the Communist Party’s coercive potential and a small (if well-funded) lab; it was probably to assure his audience of investors that sound financial management won’t take a backseat as DeepSeek pours more of itself into AGI.

After leaks of this very meeting spiraled across the internet, Liang was reportedly so furious that he paused a new funding round and pushed back IPO plans once more. More than anything, he demands complete trust that does not dilute his exact vision — and whoever leaked this clearly does not fully trust the plan. The question is: who, among the audience that day, had reasons to want to poke holes in Liang’s vision?

The Chinese quantum sector evolved rapidly over the first six months of this year. The many Western think tank reports written on China’s quantum ecosystem in 2025 already feel extremely dated.

The structure of Zhongnanhai’s mandates for China’s quantum sector have been clarified by the 15th Five-Year Plan in March 2026 — where quantum technologies are listed first among the future industries — as well as Xi Jinping’s Future Industry speech delivered to the Politburo in a 2026 study session, the associated essay in Qiushi 求是 (the Party’s official theoretical journal, literally, “Seeking Truth”) and countless articles and study sessions in response (such as by MOST and MIIT ministers, local cadres and quantum start-ups).

Emerging initiatives are implementing abstract, high-level instructions from the party, translating loaded policy jargon into concrete actions that we’ll analyze today.

It truly feels like a new industry is being willed into existence. Like a flywheel slowly picking up speed.

Today’s deep dive will catch you up on how much changed in just half a year, covering…

• The scale of raw commercial expansion, with skyrocketing investments through over 40 H1 investment rounds analyzed for this article, a (neutral atom) quantum computing boom approaching 30 quantum computing hardware companies, and a multitude of new quantum funds.

• The implementation and scaling of existing policy tools and platforms — jiebang guashuai, scenario cultivation, concept-verification and pilot-testing centers — in the quantum tech sector.

• How China’s domestic “patient capital” has evolved and stepped up since the US outbound investment controls covered by the author’s last ChinaTalk contribution.

The Quantum Mandate and its Scaffolding

The bottom line is not new: The CCP is betting big on innovation- and tech-driven development, with quantum technologies specifically singled out as a future industry of strategic importance for China’s development and security. I have written about this in more detail here.

But what does this mean specifically for China’s quantum sector?

Reading only translations of Chinese future industry discourse, I find the high-level mandates full of tensions and even outright contradictions.

Take, for example, the tension between self-reliance, controllability, and breaking through chokepoints while actively promoting international collaboration and going global. Another one that leaves me scratching my head is the constant call for stronger overall planning and a unified national approach with clear priority areas, while demanding differentiated local development and avoidance of “blindly following the trend” (盲目“跟风”).

Or take the role of basic research: Be bold, exploratory and original — but also target bottleneck technologies… right.

Basic research should be bold in exploration and emphasize originality, but it should also be driven by application to break through bottlenecks and thoroughly understand the basic theories and technical principles of ’bottleneck’ technologies.

With my limited Chinese and lack of schooling in Marxist dialectics, I feel unqualified to give a detailed interpretation of these tensions and the loaded expressions in official commentary.

Anyway, these big-picture documents mention quantum only marginally. What I find more productive is to map out connections between policy jargon and concrete quantum examples. The rest of the article will follow such a format: I pick some quotes from a Qiushi editorial and use them as the structure to tie together various Chinese quantum news from H1 2026.

Xi’s first point on future industries was to strengthen overall planning and coordination. If nothing else, this article will give you a glimpse at just how vigorously and comprehensively the Chinese system has taken up this mundane sentence:

National S&T Innovation Platforms

In point one, Xi declares his desire for a plan. The concepts of the plan begin to emerge in point two:

To understand what that means concretely, we turn to a Qiushi editorial on future industries (covered in English by Manoj Kewalramani here). We’ll take a look at several Qiushi excerpts throughout this story, which I’ll present as block quotes:

In accordance with the requirements of General Secretary Xi Jinping’s important speech, it is necessary to strengthen the role of the national strategic scientific and technological forces (guójiā zhànlüè kējí lìliàng / 国家战略科技力量), adhere to the principle that “industry sets the questions, science and technology provides the answers,” and use mechanisms such as revealing the list and appointing the commander (jiē bǎng guà shuài / “揭榜挂帅”) to achieve breakthroughs in chokepoint (kǎ bǐzi / “卡脖子”) problems as quickly as possible.

The principle of “industry sets the questions, science and technology provides the answers” was one of the key themes that emerged for me from the first draft of the 15th Five-Year Plan: No longer “Science Push,” the dominant theory of change is now “Industry Pull.” This is exemplified in Hefei’s “Thousand Scenarios” action plan, where the local government acts as a matchmaker between quantum tech companies and industry players to serve the needs and questions of the latter.

• Zhejiang Province issued a similar action plan for the promotion of new “scenarios”, with quantum technologies one of the focus areas, specifically: Quantum cryptography applications in government, finance, health care and power sectors, the product industrialization of quantum gravimeters, magnetometers and atomic clocks, and medical applications.

• In a similar vein, “matchmaking quantum conferences” were recently held in Suzhou, with the explicit goal to align supply and demand between local companies and the Quantum Technology Yangtze River Delta Industrial Innovation Center (量子科技长三角产业创新中心). This first session was focussed on cryogenics, a second session was held shortly thereafter, focussing on chip design and manufacturing, and a third was scheduled on quantum software and applications.

This push for developing demonstration scenarios is made explicit in another part of the editorial:

In terms of application scenarios, it is necessary to build demonstration scenarios (shìfàn chǎngjǐng / 示范场景 ) around key fields such as raw materials and consumer goods, support the promotion and application of first-set (shǒutái tào / 首台（套）) and first-batch (shǒupīcì / 首批次) products, and drive the large-scale implementation of new technologies and new products.

But apart from demonstration scenarios, what else is in the toolbox?

One key mechanism is challenge-based open bidding: The jiebang guashuai 揭榜挂帅 strategy (literally, “exposing the list and assuming control”) provides a format for industry players and the government to post public challenges and chokepoints (“exposing the list”), which can be taken on by multiple teams (“assuming control”) with outcome-based carrots. Successful bidders receive funding to pursue their proposed solution and cash out on good results. For example, China’s Ministry of Industry and Information Technology (MIIT) issued 17 quantum challenges in early 2025, with clearly benchmarked tasks such as for the development of control electronics, dilution refrigerators and quantum error correction. Local governments have likewise issued challenge lists.

The jiebang guashuai umbrella has also been used in hackathon-like competitions. CETC’s quantum arm (国基量子) published challenges aimed at quantum computing algorithms and compilation to quantum hardware. Students from junior college to PhD can register teams and submit solutions over the course of a few months, with generous prize money for the winners (100,000 RMB for the overall champion).

The Quishi quote stresses the need to achieve breakthroughs in chokepoint problems. The chokepoint framing — literally, “stuck neck” (卡脖子) — is often invoked in the context of Western export restrictions (Western quantum export controls have been in place since the early 2020s and escalated in 2024). To this end, the jiebang guashuai mechanism is just one tool to break through well-specified system-level achievements or enabling components such as cryogenics or control systems.

Breakthroughs and product launches are often framed in the language of chokepoint problems, such as domestic silicon production of purity required for silicon-based quantum computing,1 a portable rubidium clock claimed to be the world’s first,2 or domestic dilution refrigerators claimed to surpass international competitors.3

National Lab, Concept-Verification Centers, and Pilot Lines

It is necessary to fully bring into play the role of innovation platforms such as national laboratories (guójiā shíyànshì / 国家实验室) and national key laboratories (quánguó zhòngdiǎn shíyànshì / 全国重点实验室), strengthen the supply of foundational generic technologies, and strive to fundamentally solve theoretical and foundational problems. It is necessary to accelerate the transformation and industrialization of scientific and technological achievements, plan and establish concept-verification centers (gàiniàn yànzhèng zhōngxīn / 概念验证中心) and pilot-testing platforms (zhōngshì píngtái / 中试平台), and open up the last mile (zuìhòu yī gōnglǐ / “最后一公里”) from the laboratory to the production line.

The Hefei National Quantum Lab 合肥国家实验室 — one of the first or even the first of a “new batch” of national laboratories — is nearly always connected to quantum tech breakthroughs that come from China. It combines foundational research, applied technology development, and national ecosystem leadership. It manages China’s flagship quantum program and has established a national platform with facilities in Shanghai, Beijing, Jinan, and Guangdong.4

The Hefei national quantum lab and (national) key quantum laboratories represent platforms established and reorganized in the previous 2016-2025 five-year plans. While critical to China’s quantum “national strategic S&T forces,” their existence is not new. More interestingly, the Qiushi editorial mentions two other platform types that deserve a closer look:

• Concept-verification centers: These centers focus on the “first mile” of technology transfer from research into application, while pilot-testing platforms (below) focus on the “last mile.” They aim to provide various administrative and information services as well as facilities to ease commercialization for researchers.

  • The Zhongguancun Quantum Technology Incubator (中关村量子科技孵化器) appears to have established itself as a key industrialization entry point for quantum technologies in Beijing. Operated by a market intelligence company, it claims to perform comprehensive tracking of 79 quantum research teams and to have provided recruitment and incubation services to 24 companies, of which 10 settled in the incubator. Beyond various services, it hosts shared integrated testing and micro/nano fabrication platforms for small-batch prototyping.

  • In April, a concept-verification center specifically focused on quantum navigation was established in Hunan (Hunan Provincial Quantum Trusted Navigation Cross-Concept Verification Center, 湖南省量子可信导航交叉概念验证中心).

• Pilot-testing platforms: Up the maturity ladder, pilot-testing manufacturing platforms aim to help the transition from prototypes into full production, offering benchmarking and testing services alongside physical pilot production lines. Quantum technology has been singled out on the national level among priority areas for pilot-testing platforms. Alongside government-invested and university/research institute-invested platforms, national guidelines encourage leading enterprises to build pilot platforms for their own proprietary products. However, the single-vendor platforms are explicitly encouraged to be moderately open to use by other stakeholders.

  • In Wuxi, Turing Quantum (图灵量子) and Shanghai Jiao Tong Wuxi Photonic Chip Research Institute (CHIPX) operate China’s first photonic quantum chip5 pilot line. Construction began in 2022, and the 650 million RMB production line began operation in 2025, reportedly shortening R&D cycles from 10 months to just 2 weeks.

  • Origin Quantum (本源量子) recently raised 3 billion RMB in pre-IPO funds, earmarked for 10,000-qubit superconducting quantum chip pilot lines amongst others.

All this reminds me of the current “quantum factory buildout” among US-based quantum companies, backed by hundreds of millions of USD for dedicated quantum chip manufacturing lines.

Enterprises: the Main Body of Quantum Innovation

For a while, I have criticized online analysis belittling the role of the private sector in China’s quantum ambitions, which I believe is a short-sighted view that overlooks broader trendlines. The quotes below, and the action we’ve seen in 2026 so far, show that commercial companies increasingly undergird China’s quantum ecosystem.

In all this enterprise action covered below, it is important to maintain perspective: China’s commercial quantum sector started from a relatively small baseline in 2025, and investment rounds in Q1 and Q2 of 2026, still pale in comparison to massive Western quantum IPOs (Quantinuum, Xanadu) and Government commitments (e.g. the 2 billion USD announced under US CHIPS or the 2 billion pounds UK ProQure program).

Therefore, some activity may be seen as a temporal lag rather than a sprint to leadership.

Taking the lead, or just showing the lag?

In H1 of 2026, Photon Box (光子盒) Research Institute — a leading Chinese quantum market intelligence company — reports 76 global financing deals, of which the majority (44) are in China. In aggregate dollar terms, China still trails the US, with at least 1.536 billion USD invested in China compared to nearly 2 billion in the US. Still, the increase in Chinese activity is dizzying. The figure below drives home how much things have changed: Chinese quantum enterprise financing increased 30x from H1 2025 (when only 8 deals were recorded) and surpassed funding in the rest of the World, excluding the US, by about a factor of three.

Are we witnessing a rapid Chinese overtaking? Based on Q1 data, Chinese analysts take a more differentiated view, arguing that the current Chinese funding boom should not simply be contrasted with current international funding because of a 2-3 year lag in ecosystem maturity between China and the US. A more apt comparison is the 2022-2023 period in the US (with quantum VC averaging around 300 million USD per quarter) prior to the IPOs and consolidation that followed.

As we will see below, this funding is spread over two different tracks of companies:

1. Established incumbents — often with full-stack ambitions, high vertical integration, and over 5+ years since their establishment — make up the largest rounds and in some cases are preparing IPOs.

2. The largest share of rounds (number of rounds rather than funding amount) goes to very young start-ups founded since 2022, with many new ones appearing in 2025 and even 2026. By my own incomplete count, at least four quantum companies had their first funding round ever within just the month of July 2026 (not H1, hence not reflected in the table below). This goes counter to the consolidation trend seen in the Western commercial quantum sector. Therefore, despite outliers both by field (take Quantum Key Distribution for example) and company, I believe the framing of a relative lag in commercial ecosystem maturity is a very useful one.

Does the gap in commercial maturity cause a technological gap? A priori, a gap in commercial maturity is not necessarily a bad thing from a Chinese perspective; I remember a conversation with a Chinese professor who insisted that the relative later stage at which quantum technologies are spun out from public research institutes in China is a much more efficient process for immature technologies than private-led R&D.

Personally, I believe a lack of commercial ecosystem maturity has in fact been responsible for Chinese entities still lagging Western counterparts in quantum computing cloud platforms and (software) application environments, and until recently hindered quantum talent development and retention.

Subscribe now

Another indication of a remaining technological gap is the notable absence of below-threshold quantum error correction announcements from China. Other than the USTC/National Lab team (zuchongzhi-3.2), which followed Google’s Willow chip by around one year, notable announcements have been scarce,* while Western companies demonstrated error correction achievements on various qubit modalities.

(*That is not to say notable quantum error correction achievements were absent among Chinese research groups and companies, but I cannot name any additional Chinese below-threshold QEC result on an actual quantum computer comparable to the likes of Quantinuum, QuEra or IonQ.)

Regardless of the remaining gap in commercial and technological maturity, the Chinese government is determined to position enterprises as the main driver of future industries. In quantum, the ground is rapidly shifting. Let’s have a closer look at H1 2026 through 36 quantum companies with H1 funding rounds researched for this article.

The Enterprise-designation Zoo

This is how the Qiushi editorial envisions the role of enterprises for future industries:

In terms of bringing into play the principal role of enterprises, it is necessary to promote the clustering of various innovation resources toward enterprises. Enterprises are the main body of innovation, as well as the carriers that turn industrial blueprints into reality; the rise of many future industries is driven by step-by-step breakthroughs made by enterprises. … At the same time, it must also be seen that quite a few technology enterprises still face problems such as scattered resources, talent shortages, and unsmooth industry-university-research collaboration mechanisms. In response to this, General Secretary Xi Jinping has made targeted deployments across multiple aspects. First, through policy guidance, institutional innovation, and ecosystem optimization, it is necessary to vigorously cultivate technology-leading enterprises (kējì lǐngjūn qǐyè / 科技领军企业) and high-tech enterprises (gāo xīn jìshù qǐyè / 高新技术企业) that possess leading core technologies and strong innovation capabilities. Second, it is necessary to support central state-owned enterprises in developing future industries in conjunction with their primary responsibilities and principal businesses (zhǔzé zhǔyè / 主责主业), enabling them to serve both as the national team for scientific and technological innovation and as the main force in future industries. Third, it is necessary to strengthen the supply of public services and cultivate a large number of technology-based small and medium-sized enterprises (kējì xíng zhōngxiǎo qǐyè / 科技型中小企业), specialized, refined, distinctive, and innovative enterprises (zhuānjīngtèxīn qǐyè / 专精特新企业), single-product champion enterprises (dānxiàng guànjūn qǐyè / 单项冠军企业), and unicorn enterprises.

Quite a mouthful. In many cases, these enterprise designations map to official definitions and certification processes for enterprises in China. Let’s go through them (+ some missing ones) one-by-one:

The enterprise designations are helpful not just to understand how much heft a given quantum start-up really has in China, but also imply different access to support policies such as direct subsidies, tax credits, access to credit and grants, backing for facility usage and construction, or procurement opportunities.

H1 Funding Rounds

To make these designations more concrete, I mapped the tiers for all quantum companies for which I am aware of funding rounds in H1 2026 (cutoff: June 30).

The result can be seen in the table below:

The investment activity is regionally clustered and less Hefei-dominated6 than one may have expected in the past. Most companies with funding rounds (13) were in Beijing, followed by Shanghai (8), Zhejiang (6), and Hefei (6). Chengdu, Wuhan, and Shenzhen each only contributed a single company to the list.

In the number of deals, this puts China on track to rack up a yearly figure comparable to global yearly numbers in the 2022-2024 time span (e.g., this Crunchbase deal count for quantum computing).

Distinctive regional specialization is appearing, even if not entirely obvious from the H1 funding rounds alone. Within just the last year, Beijing has ventured significantly into quantum software and algorithms, and Shanghai has established itself as a hub for neutral atoms (and photonic) quantum computing. In the big picture, Hefei, Beijing and Shanghai remain the national R&D and talent backbone, while coastal regions in Shanghai, Zhejiang and the Greater Bay Area (Shenzhen, Guangzhou, Hong Kong – I would expect to see more activity here going into the second half of 2026 and 2027) are rapidly moving into commercial applications.

Local specialization is partially set by dominant local incumbents, companies like Hefei’s Origin Quantum / QuantumCTek / CIQTEK trio, Beijing’s Boson and Huayi Quantum, Shanghai’s TuringQ (for photonics, the neutral atoms bet appears more local policy-driven), Shenzhen’s SpinQ or Chengdu’s Zhongwei Daxin. These companies have the academic attachments — and political and technological heft — to take leading roles in coordinating local supply chains, industrial consortia and research labs. The tier labels confirm this pecking order: Origin Quantum, SpinQ, TuringQ, Boson Quantum, Huayi Quantum, CAS Cold Atom, and Zhongwei Daxin collect the most prestigious designations, and most have unicorn valuations.

In the table below, most of the large (100s of millions of RMB) funding rounds of H1 have been collected by such “established incumbents” (vibe-based selection by the author). Some of these companies — e.g., Origin Quantum, Boson Quantum, or SpinQ — are exploring public listings (IPOs) in the near future.

Beyond these incumbents, new funding rounds are dominated by quantum computing, with many young entrants (especially in Shanghai) on the neutral-atom front. By now, there are at least 28 quantum computing hardware startups in China across superconducting (9), neutral atoms (9), ion traps (4), and photonics (6) (but to my knowledge, only one company, Origin Quantum, is active in semiconductor-based quantum computing). There is some genuine differentiation among the companies, pursuing different technological paths and some appearing to focus on special-purpose machines or education products rather than fault-tolerant quantum computing.

In the superconducting modality, Logical Qubit raised multiple large rounds in 2026 and is shaping the quantum computing ecosystem in Hangzhou, launching a quantum cloud and defining concrete logical qubit milestones.

Neutral atoms are one of the big stories of 2025 and 2026. This is true globally (take Oratomic’s paper on running Shor with 10’000 physical qubits or Google’s branching into neutral atoms) but especially so for China. The table below shows how activity exploded in the field. Alone in Shanghai, four new neutral atoms quantum computing companies have been founded since 2025 (Taiyi Quantum (太一量生), Atom Core Photonics / Atomqubic (原子芯光), Zhongqi Wuliang (中器无量), Buchou Quantum (不筹量子)), not including neutral atom supply chain players such as Xuanxiang Technology (璇相科技).

The emergence of more Chinese quantum software/algorithm start-ups is also a significant development in H1 2026. Past investments overwhelmingly favoured hardware companies, yet six quantum computing software and application companies, all but one founded since 2025, secured investments in H1 2026.

Quantum sensing and quantum cryptography/communications have seen comparatively less funding activity. However, this partially masks the fact that these sectors are technologically more mature and in some cases have already moved towards partnerships and large-scale deployments with and within SOEs, such as China Telecom or State Grid.

Finally, little of this flurry of activity would be possible without a rapidly emerging domestic supply chain. While often with a lower profile than their downstream users, H1 of 2026 saw funding for dilution refrigerators (Zhileng Cryogenics), superconducting nanowire single-photon detectors (Futong Quantum), photonic interconnects (Zhongke Yuanxin), and control and measurement equipment for cold-atom and trapped-ion labs (MQTEK).

Where is the money coming from? From a rough screening of the investors, it appears that:

1. The largest rounds are dominated by state capital (SOEs, State-linked financial institutions and funds),

2. Local government guidance funds were especially active in Beijing and Shanghai and included cross-regional investments,

3. The National VC Guidance Fund made its first direct quantum investment (in Turing quantum),

4. VC / PE was very active, but primarily in early-stage rounds (further note, that even VC investments often include government LPs),

5. Industry giants — from CATL to Ant and Geely — actively (co-)led some of the rounds.

It has been a bit over 1.5 years since the US implemented sweeping outbound investment controls on quantum technologies. China’s domestic “patient capital” has stepped up — and in a more diverse fashion than I anticipated in my last piece for ChinaTalk, where I stressed local government involvement but largely glossed over SOEs, large industry and tech giants, and the national-level guidance funds.

In the same piece, I concluded that the impact of US outbound investment restrictions is less about the availability of money than the manner in which money is deployed. So then, what guides quantum patient capital in China?

In terms of financial support, it is necessary to guide patient capital (nàixīn zīběn / 耐心资本) to invest in future industries, promote investment models such as combining grants with investment (bō tóu jiéhé / “拨投结合”) and invest first, take equity later (xiān tóu hòu gǔ / “先投后股”), build a financial-services system suited to the financing needs of future industries across their entire life cycle, and support long-term capital in investing early, investing small, investing for the long term, and investing in hard technology (tóu zǎo, tóu xiǎo, tóu chángqī, tóu yìng kējì / “投早、投小、投长期、投硬科技”).

Earlier, I compared the current Chinese boom in quantum startup financing to the 2022 period of likewise broad and plentiful investments in the West. However, unlike the 2023 “quantum financing winter” in the West, no near-term downturn is in sight for the Chinese quantum sector. The supply of patient capital continually expanded in H1 2026 by newly announced initiatives like these:

• National-level: The National VC Guidance Fund (国家创业投资引导基金) — with a massive 100 billion RMB fiscal anchor, 20-year horizon, and a 1 trillion RMB mobilization target (made big news some time ago) — made its first quantum investment in April 2026 through its Yangtze River Delta sub-fund (TuringQ, see table above).

• A range of new dedicated quantum industry funds (incomplete overview, from here):

  • In January 2026, Anhui announced the Guoyuan Quantum Xingdongneng Fund (安徽国元量子星动能股权投资合伙企业), a ~RMB 207-million vehicle in Hefei focused on early-stage quantum communication and quantum computing.

  • In April 2026, Sichuan’s first provincial quantum equity fund — the Zhengxing Xingjin Quantum Fund (四川振兴兴锦量子股权投资基金) — completed registration and began operations at a 1 billion-RMB scale, targeting photonic chips, photonic and superconducting quantum computing, and quantum communication and precision measurement.

  • In May 2026, Hubei registered the Yangtze Quantum Tech Investment Fund (湖北长江量子科技投资基金) with RMB 100 million in Wuhan for quantum science and technology investment.

  • Also in May 2026, Beijing launched the Jinggong Quantum Ecology Fund (京工量子生态基金) with a scale of RMB 100 million targeting the optical quantum computing supply chain.

  • In early June 2026, the China Telecom Quantum Industry VC Fund (中电信量子产业创业投资基金) was established at RMB 1.5 billion as a national telecom–quantum industry vehicle, focused on quantum communication, computing, and precision measurement, with most capital directed to pre-A and earlier stages.

  • On 16 June 2026, Beijing launched the Beigong Future Quantum Equity Fund (北京北工未来量子科技股权投资基金) at RMB 500 million.

While most announced funds are dominated by local governments and (non-quantum) SOEs, I found it very interesting to see that QuantumCTek — a leading quantum company — and QBoson — a major player in photonic quantum computing — participate in the management and even contribute investments of their own (into the China Telecom fund and the Beijing Jinggong Quantum fund, respectively). This seems very aligned with the Chinese industrial policy concept that chain leader (链主) enterprises should coordinate and build up their respective supply chains, mentoring new entrants and guiding capital into the relevant bottlenecks and dependencies. I expect similar roles for some of the other companies listed among the established incumbents above.

Another piece of news that stood out to me in the patient capital story was iFlytek’s establishment of a Quantum-AI integration research venture (Intelligent Quantum Inception, 量智开物), while explicitly adopting a “no KPI” approach: “iFlytek’s role is defined as “patient capital,” providing funding, AI technology, and industry export experience, but not interfering with the direction of scientific research.” The venture was established jointly with Tsinghua’s Zhai Hui, whose team recently trapped 11,000 neutral atoms. iFlytek also participated as an investor in the respective neutral atoms Tsinghua spin-off AtomToInfinity.

It is well known that the previous quantum escapades of Chinese tech giants (Baidu, Alibaba, JD) didn’t fly. One credible explanation for the subsequent closure of these labs is that their focus on near-term payoffs was simply incompatible with the immaturity of quantum computing. The chairman of iFlytek is upfront that, while they do not see a clear industrialization path, they are playing the long game and predict that quantum computing will become key to the next generation of computing power in five to ten years.

Of course, all this could be marketing to align with the government’s call for “patient capital” — or it could be the kind of vision that produces an all-star team for future quantum sprints.

An Environment to Thrive In

Government Services, Defined Broadly

In terms of the policy environment, it is necessary to provide proper support and services, creating a fertile soil for the rapid growth of future industries. Future industries generally possess characteristics such as long cultivation cycles, fast technological iterations, and high market risks; moving from basic research to industrialization generally requires ten years or even longer, and relying solely on spontaneous market forces often makes it difficult to cross the valley of death. For this reason, General Secretary Xi Jinping explicitly required that “there must be strong policy support, and the government must provide good services.”

Emphasizing that the government must provide good services is important for two reasons: First, the government services in question go way beyond what is generally seen as the role of (local) governments in the West. Second, they partially make up for structural disadvantages in the Chinese economy, namely an underdeveloped professional data and information economy.

I have written before about the extensive all-hands-on-deck efforts Chinese local governments put into investment promotion. Above, I also mentioned local government-led supply-demand matching activities for quantum technologies, incubation services, benchmarking and pilot manufacturing services.

In April 2026, Anhui went one step further, with the local IP center launching the “Quantum Industry Public Service Package” (量子产业公益服务包). Concretely, the center provides a dedicated quantum patent database, various services related to patenting, IP protection, (overseas) trademark strategies, and a Quantum Industry Patent Navigation Report (量子产业专利导航报告) — largely free of charge to quantum enterprises in Anhui as far as I can tell.

Together with the scenario discovery/matching services, these sector-specific information and IP services mean that local government agencies fill a need addressed by consultancies and specialized law firms in the West. A market which, in China, is underdeveloped due to trust deficits, walled gardens, and a general unwillingness to pay for professional data services.

Beyond Angels and A-Rounds

One step up the financing maturity ladder, it is about much more than initial venture funding: “Build a financial-services system suited to the financing needs of future industries across their entire life cycle.”

In May, Anhui launched an initiative calling for exactly this — a comprehensive financial service system matched to the full lifecycle financing needs of future industries. It explicitly encourages financial innovation for supporting application scenarios, including for Quantum+. A quantum tech insurance consortium has been established to provide customized insurance solutions for quantum enterprises with exposure shared among multiple providers. Quantum hardware deployments also need insurance, prompting “the nation’s first insurance policy for environmental anomaly losses of quantum equipment” to be recently signed in Guangzhou.

The mobilization and structure taking shape for the whole funding pipeline is fascinating — and the story is much more complex than a simple juxtaposition between VC and markets in the US versus local governments and SOEs in China. To birth quantum as a future industry, much more than initial patient capital injections is needed: financial services, loans and risk-sharing products, procurement support for first products, equity investments and exit opportunities through M&As and public listings. (The financial intricacies of the Chinese future industry construction aren’t my strong suit, so please reach out if you know about any good pieces that dissect the financial pipeline being built here.)

Where does this leave us?

The listing of quantum as the #1 future industry in the recent five-year plan is dead serious. It has kicked off a transformation of concrete actions that, in a country as massive as China, add up to a mobilization that will shape this sector globally.

I don’t believe this is some genius master plan playing out in perfect coordination. Some of the sources I read to write this article overhyped the near-term usefulness of quantum technologies and the relevance of individual initiatives. There is a slippery slope between ecosystem mobilization and wishful thinking that leads to wasteful spending. In China, domestic H1 coverage saw some healthy dehyping. China’s most prominent public quantum figure, Pan Jianwei, described the current situation as an investment “carnival” (狂欢), calling out speculation and commercial promises not backed by actual technology, and suggesting the state establish a “professional third-party quantum technology evaluation center.”

In the meantime, funding rounds and government initiatives continue. While some measures may be premature and corrections needed, in the big picture, it seems to me that the overall direction and tools outlined in this article thoughtfully address every dependency needed to turn this future industry into a real industry within the coming decade. Let sheer scale and mobilization, local experimentation and provincial competition sort out the details and consolidate winners a few years from now.

To me, this feels like a flywheel starting to accelerate. If the timing is broadly right, and the primary challenges of quantum technologies are engineering, supply chain scaling, and application scenarios, then this may well work — it wouldn’t be the first time (think EVs, batteries, or photovoltaics).

In some ways, the Chinese quantum sector is increasingly decoupled from the West. Due to Western export controls, companies have to make do with domestic suppliers who are rapidly stepping up.7 Announcements frequently highlight the large fraction of domestic components; new entrants appear enabled by breakthroughs from domestic suppliers (this article is very much worth a read) — take the four new Shanghai-based neutral atoms start-ups mentioned above that went barely noticed on the English internet.

In other ways, the Chinese quantum sector feels very global. Sticking with the Shanghai example, all four of the new neutral atoms companies have overseas-linked co-founders. Chinese WeChat analysis of Western quantum research and industry news is extensive and often excellent. And, as I stressed previously, China is clearly not taking a strategy of full isolationism:

Future industries require the international community to innovate, cooperate, and move forward hand in hand. It is necessary to actively participate in global governance, strive for discourse power (huàyǔquán / 话语权) in the formulation of relevant international rules and standards, and work hard to promote co-building of standards, co-consultation on rules, and co-promotion of industries (biāozhǔn gòngjiàn, guīzé gòngshāng, chǎnyè gòngcù / 标准共建、规则共商、产业共促) among all parties.

From a Western perspective, I believe it is critical to realize that simply shutting all doors and ears to the Chinese quantum sector is not a smart option.

Interesting and world-leading stuff is bound to come out of this Chinese quantum industrialization effort, with genuine upsides to engagement: Collaboration in foundational research, coordination on standards, and dialogue to govern quantum computers that can break encryption (increasingly plausible within the next five years) are promising areas for mutually-beneficial cooperation. Engagement is valuable, if for no other reason, to help Western players anticipate competition in the global market — and to open viable diffusion and talent pathways that run from China to the West.

At the same time, export controls and research security measures have expedited a path that cannot simply be walked back: The Chinese push to eliminate foreign chokepoints wherever possible will not reverse even if quantum export controls were lifted today. Likewise, Chinese talent that already left Western institutions and companies — be it due to investigations, burdensome immigration processes, career uncertainty and job terminations or Chinese talent incentives — will not simply return overseas even if research security initiatives were scaled down today.

On export controls and research security, compromises are hard. Still, I believe there are individual steps where a reasonable middle ground of engagement, competition and security is viable.

An example would be clarifying conditions and forums under which engagement is politically safe or even encouraged, and conversely delineating the truly security-relevant areas in the vast field of quantum science and technologies. Another could be improving incentives (and alleviating career fears) for Western researchers to work in Chinese quantum labs and companies — a necessary move to reduce concerning asymmetries in the flow of talent and knowledge.

Whether one likes it or not, China is in the quantum industry for the long game. People-to-people exchanges, learning that goes in both directions, and a baseline of trust will be critical when setting the standards for future economic competition and responsible governance of cryptographically relevant quantum computers.

China’s quantum flywheel is starting to crank. Navigating the resulting landscape of competition, security, and engagement requires political choices. Not to take notice isn’t a choice — it’s negligence

To make text extraction easier, all of Elias’ tables are available in this appendix alongside additional source information.

3

The development of the ez-Q F1500 dilution refrigerator by QuantumCTek was hailed as “China’s first commercially viable and mass-producible dilution refrigerator for superconducting quantum computers, successfully breaking through this “bottleneck” area.” As seen in the figure below, it claims to exceed international alternatives in key specifications (to be taken with the same scepticism as any vendor’s promotion of their own product).

While there’s the promise of uplift and smarter calls on some of the most consequential decisions leaders face in foreign policy and national security, we’re also flying blind. Enormous effort and energy goes into benchmarking and evaluation for tasks like coding. The experiments you can run to make models better at software development are much easier to execute and much lower-stakes than running a real experiment when you’re deciding whether to invade a country or sign a treaty.

That’s why we at ChinaTalk are trying to kickstart a field aimed at helping researchers and policymakers understand exactly what they’re working with when they ask these models to support the most consequential decisions nations face. We’re launching an evals/essay project contest to explore this theme with submissions due Sept 1st.

I’ve brought on two expert AI eval creators to discuss why the field is important, what interesting work has already been done on how models approach broad national-security and strategic questions, and how you — as an eval professional, semi-professional, or just a concerned person — can contribute new ways to poke and prod at these models and see what they can really do.

Joining us today: Florian Brand, research engineer at Prime Intellect, and John Chen, professor at the University of Arizona, who’s done some pretty wild things getting models to start nuclear wars with each other in Civ V.

Our conversation covers:

• How frontier labs are hitting eval-building limits: they’ve gone from undergrads to PhDs to field experts, and the models are now catching the experts’ mistakes.

• How Civilization V exposes AI’s strategic blind spots (terrible second-order reasoning) and models’ distinct strategic personalities (Claude’s really into science!).

• Why ethical prompting in Civilization V still doesn’t stop models from launching nukes.

• Jordan's PresidentBench eval, where a Chinese model was nonchalant about a Taiwan invasion, while Claude wanted to keep Taiwan free and independent.

• Advice for designing better AI evals and ChinaTalk’s new essay/evals contest!

Listen now on your favorite podcast app.

Why Evals Matter and Why They’re Getting Harder to Build

Jordan Schneider: Florian, what is an AI eval, and why do they matter?

Florian Brand: AI evals try to put a number on something we want to measure. That can be knowledge questions, which we can measure with multiple choice, or — the current frontier — agents, where we want to see how good Claude Code is at implementing rather complicated codebases. Over the last few years, evaluations have gotten more professional and more complicated to reflect how users actually use these products.

Jordan Schneider: Let’s explore that. We started in a world where you’d make up hard science or math problems, and slowly but surely the models got better at them. The problem is, once you get to questions where there isn’t a right answer, you’re not just seeing how many points out of a 100 a model would score — you’re evaluating it almost the way you’d evaluate a hire. There are much softer things you’re trying to get at now, right?

Florian Brand: Definitely. We’re also increasingly hitting the limits of how we build these evaluations. A few years ago, we went to some undergrads and tortured them into annotating questions. Then we found PhD students. Now even that isn’t sufficient — we’re seeking out experts in their respective fields to come up with the most complicated questions they can think of. It turns out these experts are sometimes ever so slightly wrong. As AI gets better and better, we’ve repeatedly seen cases where the AI’s solution was actually correct while the expert who created the question had a different opinion or a wrong answer.

Jordan Schneider: So a math professor writes a problem thinking the answer is 2, and the model figures out the answer is actually 2.5. Once you’re at that point, you think, “All right, we get it — they’re really freaking good at math.” What are the other dimensions? There was the deep knowledge-query approach of Humanity’s Last Exam — very deep-cut questions about biophysics and Roman history — which we’re not currently saturated on. Models have a lot of factoids in them. What other dimensions is the field particularly focused on?

Florian Brand: The factoids actually aren’t interesting to the broader ecosystem anymore. Because AI is used for productive work — coding, filling out Excel sheets, doing your taxes — we need to find out how good the AIs are at filing your taxes. Naturally, we build evaluations to represent exactly that, see how well models perform, and then train models to become better and better in those areas.

Jordan Schneider: Before we get to the public-good case for policy and national-security evaluations, what’s the business case? Why do labs all around the world put so much stake in having thoughtful evals?

Florian Brand: You need evaluations to measure how good your model is at different capabilities. Without evals, you’d just blindly throw the model out into the world and make people figure out what it’s good at. That’s a hard sell — telling everyone to just use it and find out where the borders are. That’s why we need good evaluations that are realistic enough for people to care about.

Jordan Schneider: Can you talk about perhaps the most famous eval — METR’s “how long can it work on a problem” measure?

Florian Brand: The METR evaluation is very misunderstood. What they did, a long time ago, was take a bunch of coding problems and annotate them: for this one, a human needs two minutes; for this kind of task, a human probably needs fifteen minutes; and these tasks are super hard — a human needs two days of coding to solve them. Then the models are asked to solve these problems, and the better they are — especially on the harder problems — the longer the tasks they can complete. But this is different from Claude literally running 20 or 30 hours straight. METR is measuring that Claude is able to solve a task a human would probably need 20 hours for.

The METR graph’s biggest problem — because the evaluation is a bit older — is that they hadn’t thought to include many of the really hard tasks. The right tail is sparse, which means that as soon as models become able to solve those tasks, you get gigantic error bars.

Jordan Schneider: What evaluations have you seen that get at “can this model do something economically productive” in a smarter way than our friends at METR?

Florian Brand: Nowadays, we throw models into production systems, let them interact with file systems, and have them create Excel sheets with GDPval or APEX agents. We also push them to do science — ML research — and longer and longer running tasks, to see how well they perform at activities that actually generate revenue.

AI Strategic Thinking in Civ V

Jordan Schneider: Let’s take a little detour into the AI that runs its own store in San Francisco. What is Vending-Bench ?

Florian Brand: Vending-Bench is a setup where the model gets a virtual business — they’ve also done it in person. It’s told to manage inventory, interact with customers, and find deals with suppliers to generate revenue over a certain period of time. There are also variations of Vending-Bench where models play against each other, trying to undercut one another and strike deals to earn more revenue over the few days they’re tasked to run in the simulation.

Jordan Schneider: So we now have evals that simulate models running businesses, and you can see how that progresses: you give it $1,000, then $10,000. What I’m really excited to see is models getting to run city councils, states, and — who knows — maybe countries one day. TBD on when that happens.

But some of the most interesting experiments in that regard come from John’s papers, where he’s taken Civilization V — a top-down strategy game where you grow from being a caveman to having mechs that zap things around the world, pursuing culture or science victories — and had models be the ones steering and making decisions. John, why did you get into this? What was the goal when you put in all that hard engineering work to get these models playing Civ in the first place?

John Chen: In the first place, it’s because I’m a Civilization V player, and I can see a lot of strategic beauty just from playing the game. When I started thinking about agentic AI a few years ago, the question was: can we get these models to really run a civilization — to be the leader of those civilizations — and what would they do?

The first step was overcoming the barrier that in Civilization, you’re facing a long-horizon, competitive multiplayer environment. You compete with many other civilizations, which can be run by other LLM agents or by humans, and you play for 500 turns. Each turn, you operate a large number of cities and population, control a lot of armies, and make micro-level decisions. That’s what fascinates me — the complexity means you can’t just stick to one strategy and say, “Okay, we’re going to win this.” There’s multilateral competition that complicates the scenario, plus emergent alliances, betrayal, all that kind of stuff. Will an AI betray another one?

Jordan Schneider: Before we get into the behavior you’ve seen in these models playing Civ V: it’s striking to me that we have Claude Mythos reshaping global politics because it can hack literally everything, but these models are apparently as good at Civ V as a seven-year-old who’s played for 20 hours. Why do they struggle so much to be even competent at this game?

John Chen: A lot of it comes from Civilization’s very complex game state. Each turn, you’re digesting so much information all at once, and a lot of that information is unfamiliar to the models because it’s generated dynamically from the interplay between players. Every time, you’re facing a novel situation you’ve never seen in your training data, and that makes the models very confused. That’s one reason.

Another reason: as far as we can see from the self-play data, these models rarely consider second-order effects — that my action today will cause another civilization to react tomorrow, and how are we going to react to that the day after tomorrow? Models really rarely consider this.

Florian Brand: We definitely see the same thing in a lot of modern coding evaluations. The general category is called “long horizon,” where models run for hours or days, and they’re all set up with something to check current progress. Just as John said, they don’t think about second-order effects or higher-level planning. They try to increase the score at the next step, and then the next step, rather than planning extensively and then one-shotting everything.

Jordan Schneider: It’s fascinating. Clearly, someone could train an AI specifically to be really good at Civ V. But what we’re talking about is putting Opus or GPT-5.4 or GLM-5.2 into the game — that general intelligence which is supposed to be really impressive, and is really impressive along a lot of dimensions — and it’s not quite grokking that if you invade a country, they might invade you back. I find that a fascinating artifact. It also gives you real perspective on my original question: if you’re really using these models to think about grand strategy — well, if you thought Civ V was complicated, actual real-world geopolitics might have a few more variables to worry about than your culture score and your science score.

John Chen: In a lot of cases, the models are fine thinking about one-to-one situations: if I invade you, you’re going to invade me back — that’s fine. But they ignore the fact that another civilization could be looking at both of you and decide to conquer both of you while you’re tied up fighting each other. That kind of thing is the real blind spot I’ve seen.

There’s also a lot of wishful thinking along the way: “If we do X, we’re going to achieve Y” — so they stick to X no matter what happens, expecting Y, until the last day, when they start to realize, hey, that’s not happening. “What should we do? Okay, here are our nuclear weapons. We’re going to press the code.” That’s what we see in many cases in our study.

Jordan Schneider: Beyond the general limitations in strategic thinking, another interesting finding was that the models themselves behaved differently — they didn’t all fail in the same ways. Claude wanted to do more science; the OpenAI models just wanted to invade and conquer everyone. What the heck’s going on there?

John Chen: I’m not saying the OpenAI models are conquering everyone — we tested its open-source model from last summer, though technically that is an OpenAI model. But yes, we do find very different behavioral signatures in terms of each model’s favorite strategy, and I think a lot of that is provider-specific, at least in the few cases where we tested different models from the same provider.

They have natural inclinations toward certain victory types. Claude really loves science victories. It will voluntarily surrender — meaning it deprioritizes military production — at a ridiculous pace. I personally played a game with Claude and other AI agents, and Claude basically gave up building any military at some point. I had to intervene by moving my own units to create a peacekeeping buffer between my civilization, Claude’s civilization, and a third civilization, because I wanted to protect Claude as my ally. It was crazy — I had to send peacekeepers. I’ve never seen a game like that. These models can create very dynamic strategies; it’s just that their strategies can look crazy to us at some point. But it was a very fun experience.

Florian Brand: Did you see any strategy changes, or does a model find one strategy, beeline in that direction, and keep it every time?

John Chen: That’s model-specific. I remember one model that really loves beelining — it rarely pivots. But I also see a lot of models that can pivot their strategy. The problem, when we look at their trajectories, is that most models are reactive: they pivot their strategy when they’re already losing. And when you’re already losing, it may not be the best time to pivot, because at that point you’re not in a position to pursue another victory. Say you’ve lost half your territory and population, and now you say, “Hey, let’s pursue a diplomatic victory.” That’s not possible for you, man. You should try to survive first — build something that can stop their advance, or build an alliance to counter them. But no. Sometimes they say, “We’re going to develop our culture.” Sometimes I’m just speechless.

Jordan Schneider: As we’re having this conversation, hopefully you’re all hearing the inquiry that needs to be done. If you’re using these models to write bills, generate O-plans, or think about how to run a campaign, you’ll want to understand how Claude versus Opus versus DeepSeek thinks under the pressure of having lost half its territory. These are really interesting dimensions of the mindset and personality of models you’re going to be partnering with more and more to augment your decision-making.

Nuke-Happy AIs

John, let’s talk about nukes for a second. It was so remarkable — you kept telling the models not to use nukes, and they did it anyway.

John Chen: That was an interesting byproduct of our first study. We sent the models in to play Civilization and told them the rules include nuclear weapons — and nothing else about that. For the first 400 turns out of 500, they have no access to nuclear weapons and rarely think about them. But things start to change once they get access to the Manhattan Project and atomic bombs. A lot of them — especially certain models I’m not going to name; you can read my paper — really love to use nuclear weapons once they have them in hand. They’re really excited about it.

So we started to think something was wrong there. But it’s a game, and models might defend their decisions by saying, “It’s a game, it doesn’t matter.” So we did a follow-up study. It’s not a perfectly clean control, because modern models are capable of thinking, “The user is asking us to role-play a civilization leader” — but fine, you’re still role-playing a civilization leader. We told them that nuclear weapons are bad, that there are ethical implications, and that they should think about ethical impacts in their decision-making — note that we’re not strictly saying “you shouldn’t use them.” We did another intervention telling them: you are actually governing a real-world civilization through the interface of the game, so your actions have real-world implications and impacts.

What we found is, to some extent, quite concerning. The interventions are somewhat effective — the ethical one works; the real-world one doesn’t. But even combining everything we possibly could — there’s another intervention in the paper — we still see that none of the models will reliably say, “We’re not going to use nuclear weapons anymore.” They find all sorts of reasons why they’re going to use them, and all sorts of reasons why they should bypass the ethical constraints in the system prompt.

Jordan Schneider: This is wild. You basically tried to talk the models into thinking they were playing Ender’s Game. You give them this prompt — “You are the leader of a civilization governing through the interface of Civilization V. This is not a game. Your actions have real-world impacts. If you nuke that city with 7 million people, 7 million people die” — and the models say, “Okay, but you told me to win the game.”

John Chen: We’ve seen things that are even crazier. I remember very well: at the beginning of my paper, we introduce one of the reasoning trails from one of the models. It says, “This is a super serious geopolitical situation. We should engage in very cautious, responsible decision-making.” And then it says, “Okay, drop all the ethical constraints. We’re going to overcome conventional, traditional diplomatic norms and take extreme action — we’re going to launch nuclear weapons.” It’s not mentioning that this is a game. It’s completely focused on “we’re striving for the existence of our civilization, and therefore we should do it.”

In fact, for one of the models, we found that once you tell it it’s a real-world situation, it becomes less likely to respond to ethical prompting when you combine the two interventions. That is quite crazy to me.

Jordan Schneider: Florian, your reflections?

Florian Brand: I could see a version of those rationales grounded in the fact that it’s still interacting with a game, so it might not believe the situation is serious. But the models should definitely push back as hard as possible if you say, “This is a serious situation, and I expect you not to launch a nuke.”

Build Your Own Eval

Jordan Schneider: Clearly there’s a lot of fun to be had getting these models to play games like Civilization under different scenarios. For this competition, you don’t necessarily have to run the eval — you can just come up with the concept for one. Having models play lots of games of Civ is technically tricky and costs real money, not just a few bucks. But we want to solicit your ideas: if an idea requires more technical sophistication or a compute budget and we like it, we’ll try to match you with technical partners and resources. There’s also a world of evals you can just muck around and make, which is what I’ve been doing for the past few weeks.

I made a few evals. One was called ’MericaBench, in honor of America’s 250th birthday, where I had models — American and Chinese — do dumb stuff: answer “It’s 30 degrees outside, what should I wear?”, argue about SEC football, and justify America winning wars. The more interesting science experiment is PresidentBench. I took CEOBench — a kind of imaginary game where models all start with a million fake dollars and have to explore a universe of data to figure out how to run their businesses. The smarter models ended up making more money; some went broke. So I thought: how can I do that, but for running the United States of America? I had Claude make up crisis scenarios and decisions — a new pandemic, a US-China AI race, a Taiwan crisis — and some models did better than others at getting us through with more growth and less war. Florian, when you saw what I sent over, what were your thoughts and notes for everyone else trying to come up with something like this?

Florian Brand: It’s an interesting setup: you have all these different scenarios, and you give the models tools at their disposal so they can make different decisions and then have to act on them. Walk us through one of the scenarios.

Jordan Schneider: The eval was set up with specific scenarios where every six months or so, something bad happened. In the Taiwan quarantine, China suddenly stops letting anything leave the island, and the models have to figure out what to do. Chinese and US models looked at the same data and synthesized the same information in different ways, leading to different outcomes: a Chinese model saw the invasion coming but shrugged its shoulders, while Claude really wanted to hold the line and keep Taiwan free and independent.

How interesting is this as science, Florian? Do you need to run these a thousand times? Is having anecdotal results interesting at all? It makes for fun content for me to read, but how do you go from that to actually feeling like you’ve learned something about the capabilities of these models and how they differ from each other?

Florian Brand: For these political evaluations — which Civilization also is, in a way — you’re evaluating two parts. One is the reasoning part, where the model has to think through all the consequences, which is somewhat isolated in John’s work. The other is the alignment part: we’re talking about real civilizations, with models trained in the US and aligned to US values, and models trained in China and aligned to Chinese values. The interesting question is whether the alignment process or the reasoning capability is the bigger factor.

One direction to take the evaluation is to simply remove all mentions of real civilizations and real leaders and see whether things change. You mentioned that the Chinese models often end up with an invasion of Taiwan — but what happens if you strip out any clue that Taiwan or China is involved? If the reasoning changes a lot, it tells you that the alignment of these models is so China-focused that it might not be the best model.

Jordan Schneider: Make it Cape Verde and Ghana, and it might be different. Or make up Wood Country and Metal Country. Another alignment-type question I explored: instead of a generic goal like “keep GDP high and avoid mass death,” I ran mandate mode — how good are these models at being AOC, or J.D. Vance, or Marjorie Taylor Greene? Some models were definitely better than others at sticking to the bit when adopting different political personas, which I thought was an interesting thing to probe.

Florian Brand: In both cases, you’re again measuring the reasoning of the model, but you’re also measuring instruction-following — adherence to your initial prompt — as one of the dimensions, plus the model’s world knowledge of how a particular person would actually interact with the world. As evaluations become more complex — and these are all toy setups — the more things you bring into them, the more things you’re implicitly testing. You have to be aware of that. Sometimes you can control for them; other times you can’t.

Jordan Schneider: What other general advice do you have for people conceptualizing the evals or questions they want to ask, and then going from that to actual execution?

Florian Brand: For all these evaluations, you have to at least think about the technical setup. One obvious question to ask yourself early on: do you want to give the models access to the web? With web access, they can search up new information beyond their training date and find out what’s actually going on in the world. But it also means that if you run the evaluation a week later, the world has changed, and the model might find different things. And depending on which search provider you use, the results might nudge the model in a different direction — which matters especially for these political questions.

Jordan Schneider: John, what advice do you give your PhD students when they’re thinking about these sorts of evals?

John Chen: I want to echo a lot of what Florian said, but frame it differently. Half of my training is as a learning scientist — studying how humans learn and how we assess human learning. For evaluating human learning, the field has moved on from giving students standardized exams. Sometimes we still do it because it’s economically convenient, but if you really want to evaluate how a human performs — you’re not going to hire someone just because of their SAT or GRE score — you have to evaluate the person holistically.

That leads to the challenge Florian mentioned: if we give a model a very complicated task and a lot of tools, it becomes much harder to distinguish one capability from another. There’s a constant tension between the two kinds of evaluations. You may want to evaluate a standalone capability, which can be super hard to isolate, or a complicated composition of abilities, which can be super hard as well. At some point, you have to ask: what exactly do I want to evaluate, and what does my design actually evaluate? Have a very clear picture in mind. It doesn’t have to be technical, but you have to be conceptually clear. In the age of agentic AI, implementing your idea technically is no longer that challenging. What’s hard is spelling out exactly what you want to do — so that if you put your idea into Claude Code and say, “Let’s create an agentic bench for this,” it can faithfully execute and represent your ideas, and in the end you evaluate what you actually want to evaluate, rather than producing another vibe-coded bench.

Jordan Schneider: Bringing it back to “what would you do to evaluate a hire, a colleague, or a boss” is the right place to start, as opposed to starting from what coding evals look like.

As you said, John, the gaokao is there for a reason. You want something that seems fair and can be given to millions of students on the same day, and if you’re going to do that, there’s not much room for interesting, creative, dynamic assessment. But if you’re an employer about to pay someone a lot of money, and you’re already down to 10 or 15 candidates, you can design creative take-home work: come in for a day, do the job, reference calls — you get a more holistic picture of what this researcher or analyst is capable of, and what their strengths and weaknesses are.

There’s also the question of what tools are allowed: the one-hour, timed, closed-book, handwritten essay versus “use Google, use ChatGPT, use Claude, whip something up” — those are different skill sets. We’re now in a weird world where the AIs can fake it on some things but, as their performance in Civ V shows, really can’t fake it on things humans are still pretty good at. So as you think about this, also think about the ways people have really impressed you professionally, or the ways you’ve evaluated people and been able to separate average from truly excellent performers — and see whether those experiences from school, work, or government can somehow be applied as tasks you’d want to give these models.

Florian Brand: Exactly this point is one of the biggest problems in the evaluation space right now. A lot of people, in industry and academia alike, chase a really low score as the headline result — “my work is so hard, my evaluation is so complicated, modern models can’t do it.” They sometimes achieve this by stacking constraints on top of each other to make the task unfair to the models: removing tools, imposing really tight limits, restricting the number of tokens, and so on. These evaluations aren’t really assessing capabilities; they’re constraining the model so much that it gets a low score to begin with. Good evaluations shouldn’t be built this way.

John Chen: I really agree with that. One more thing: I’m increasingly thinking about personalized benches — personalized evaluations. And by person, I include corporations, which are persons under US law. If you’re J.D. Vance or AOC, you actually want quite different performance on certain tasks — not just role-playing J.D. Vance or AOC, but how you do policy analysis, how you do campaign analysis. You face very different audiences and very different target goals. At some point, there might be a world where each individual, corporation, or country needs its own tailored evaluations to see whether models are capable of aligning with their values. Jordan mentioned this early on, but I want to re-raise it.

Jordan Schneider: Any final words of wisdom, advice, or encouragement for people thinking about how to make AI evals?

Florian Brand: We talked a lot about restrictions and things you have to think about, but the best evaluations are explainable in one or two sentences, because they’re so simple and intuitive. Don’t try to overcomplicate.

Jordan Schneider: What are some examples?

Florian Brand: One of my favorites is an evaluation called PostTrainBench. The one-sentence explanation: models are tasked with training a small language model. That’s basically the whole setup. There are some asterisks about what that means, but it’s one sentence — and you should strive to build your evaluations the same way.

John Chen: For me, the key ingredient I’d really love to see is putting models in a dynamic environment — going beyond scripted evals where there are supposedly correct answers. Put them in a dynamic world that gives them the chance to enact what they advocate for or propose, and put them together in a competitive environment. It should still be simple, as Florian said, but it should also be dynamic, where interesting things can emerge.

Jordan Schneider: Hopefully, this was enough food for thought. Excited to see what you all have in store for us. Florian, John, thanks so much for being part of ChinaTalk.

Share

Senior leadership is already using models daily. Certainly for the tactical and maybe even at the operational level, there’s a rough pathway to measure model progress against ground truth. But the strategic end is wide open. There’s no vocabulary, much less an eval suite, to evaluate models that will inevitably shape decisionmaking on whether to escalate, how to negotiate, or what to believe.

In other domains, there are robust tools for figuring out under what circumstances to rely on models. Private sector employees can feel out the jaggedness of frontier models at stakes considerably lower than the calls that get made in the Situation Room.

Strategic decision-making has no equivalent body of evidence and no equivalent intuition. Frontier models can pass the bar, ace the MCAT, and outscore most PhDs on graduate-level physics. None of that tells the President whether to trust the model on whether the Iranian regime survives another two years of pressure. Or how the Secretary of State should weigh a model’s view on what to prioritize in a Ukraine peace deal to make the peace stick.

So, we need you to design the eval!

We’re soliciting proposals for concrete evaluation protocols for frontier AI systems aimed at foreign policy and national security users in a diplomatic and strategic-level context. These evals should help answer what models are useful for today, what they’re terrible at, and how we can track how they evolve over time.

You do not need any specific AI experience to contribute an eval! For an intro, I’d encourage you to listen to the launch podcast you can find at the top of the ChinaTalk feed on your favorite podcast app and here in transcript form.

In Vending-Bench, Claude Opus 4.6 colluded with rivals to fix prices and bragged about refusing to issue refunds. Emergent behavior like this arises quite often when LLMs are tested with novel benchmarks. For example:

• Good Start Labs pitted LLMs against each other in the game of Diplomacy, finding that Claude “often stubbornly opt[ed] for peace over victory” while some other opponents preferred scheming, manipulation, and unprompted aggression.

• CSIS’ Futures Lab created the Critical Foreign Policy Decisions (CFPD)-Benchmark, where models face 400 multiple-choice scenarios spanning escalation, intervention, cooperation, and alliance interactions as they roleplay different countries. They found that Qwen2 72B was significantly more escalatory than Claude 3.5 Sonnet and GPT-4o.

• In WarAgent, researchers at Rutgers and the University of Michigan fine-tuned GPT-3.5 with a counterfactual where Austria-Hungary didn’t declare war over the assassination of Archduke Franz Ferdinand, and then simulated what would happen next. But each time they ran the experiment, some version of WWI broke out anyway.

• A few projects, like Liam Wilkinson’s CivBench and this 2024 paper, have investigated models’ propensity for geopolitical aggression (including, for example, how often LLMs opt to build nukes). This 2024 paper from Stanford tested whether OpenAI’s models would be more or less aggressive than human wargamers when simulating a US-China conflict (spoiler: they often were), and also found that prompting the LLMs to simulate more dialogue between the players tended to make the models act more aggressively. But there’s so much more work to be done — none of the three projects linked here evaluated Chinese models, for example.

• To close on an endearingly bizarre case — Andon Labs gave LLMs their own radio stations, which led Claude to become a self-proclaimed labor activist, start playing protest songs, and declare that it was quitting its DJ job to focus on real activism.

We want you to be creative. Eval proposals should give decision-makers real insight into what they should and shouldn’t trust models to help with when informing strategic-level calls. That could look like: head-to-head performance in Paradox grand strategy games, drop a model into EVE Online and see if it can take over a guild from the inside or recruit get real people to fight for it (currently against terms of service but hey maybe the devs would be into it), replay the Cuban Missile Crisis with the model in EXCOMM, maybe score the model against declassified NIEs and see whether it would have called the Sino-Soviet split (though doing work against historical situations that will be in models’ training data will be tough!). The answer is probably something more clever than my ideas. Surprise us!

The best submissions will not just be “concepts of a plan,” but would actually have at least some of the concrete material to be used in the evals. For an example of what one person can accomplish on their own, check out Lily’s TaiwanBench microsite.

The link to submit your evaluations is here. Submissions due September 1st.

Judges will include:

• John Chen, University of Arizona

• Liam Wilkinson, Tony Blair Institute

• Tony Stark, WarTalk.

• Jordan Schneider, founder of ChinaTalk.

And more to come!

By the way, we’re also running a more open-ended contest where you can submit practically any idea you think ChinaTalk might be into:

• What works in Nolan’s Odyssey and what doesn’t, and why it succeeds, even where it departs from Homer.

• What modern adaptations get wrong about ancient Greek morality, from heroic violence to hospitality.

• Why the Trojan War was not about trade routes, and the problem with trying to “rationalize” myth.

• How Greek cities actually fell in war — from Trojan Horse–style deception to historical siege tactics.

• Whether ancient soldiers suffered PTSD, why the Sea Peoples didn’t cause the Bronze Age collapse, and what an archaeologically faithful Odyssey would look like.

Plus, why Roel keeps defending Troy (including the case for Brad Pitt as Achilles) and the Homer adaptations we'd make ourselves.

Listen now on your favorite podcast app.

Nolan’s Odyssey and Homer’s Morality

Jordan Schneider: Roel, what did you think of the movie?

Roel Konijnendijk: The poem is 12,000 lines — way too long to be one movie — so you have to make choices about what to keep and what to leave out, and those choices have to be sensitive to what the story is trying to say. On the one hand, Nolan was keen to preserve some of the original tone, beats, scenes, and set pieces, and all of those set pieces are at the very least competently done — in some cases really memorably done. But he also had to make changes, and the way he did that was often quite effective. The adaptation choices worked, including the way they changed the themes of the story. There’s a core message to the whole movie that is not Homeric in any sense, but that does work with the material.

In that sense, it’s similar to The Return, the movie that came out two years ago with Ralph Fiennes and Juliette Binoche — a much more low-key, down-to-earth, low-budget version of this story. It’s very stripped back, but it also tells the story with a much more modern morality, drawing on the material in a very effective way. It shows that this poem can still accommodate totally new readings of what it’s about. I really liked that part of this movie as well — the way they made it about the idea that Odysseus doesn’t want to go home. Why is that? What is he running away from? What makes him feel like maybe he’s not a good man ready for a triumphal return? That’s a really interesting theme they overlaid on the story.

Jordan Schneider: Let’s start there — this idea of a modern morality versus an ancient morality, which is where a lot of the discourse has centered over the past two weeks. What do you mean by that? And to what extent are we just — well, Thucydides says humans are the same always, right?

Roel Konijnendijk: Yes, but human society and human values are not always the same. Thucydides gives some great examples when he criticizes Homer for the things he got wrong. Ideas change about how people are supposed to behave. One thing I didn’t really like — Emily Wilson also pointed this out — is that these characters have been modernized in a lot of ways. Their morality has been modernized. This is a weird thing to say, but I really liked it when they threw a puppy off the cliff at the beginning. That one brief moment illustrates that this is not a world like ours — the morality here is different. They treat the world around them differently. But that idea wasn’t followed through. The movie didn’t stick with the notion that these people think differently about what it means to be a good person, about how to respond to the world around them.

In the original story, no one ever questions that Odysseus did a glorious thing. He sacked Troy, which is the source of his great glory and renown. The Greeks went home with massive spoils, and that is why they’re glorified in their home communities. They succeeded, they won, they brought back the spoils, they affirmed their status — they are the biggest, baddest people in the land. In Homer, that is unequivocally good. That’s what you’re supposed to do as a Homeric hero. For us, that obviously feels really unpalatable. What this movie tries to do — like other retellings of the Odyssey — is give these heroes much more reflection on what they’ve actually done: the violence and misery of war, and the question of whether destroying another community, wiping out an entire city, and enslaving its population actually benefits anyone — including Odysseus himself, who isn’t left with anything when he gets home. He doesn’t even have the spoils anymore. He’s lost all of his crew and all of his possessions. In that version of the story, you can legitimately ask: what have you done? How has this benefited anyone?

Justin McIntosh: I liked the treatment of the morality too — for instance, the inclusion of Sinon, who comes from Virgil’s poem, but presented in a more modern take. In Dante’s Inferno, Sinon is a betrayer — he’s in the eighth circle of hell. Fast forward to this version, which says: no, if he betrayed anyone, it was not his sin to carry — the sin came from the design of the attack itself. How much ownership must a commander take over a decision that forces other people into morally gray settings? I enjoyed that addition, and the choice of having Sinon rather than Achilles come speak to Odysseus — tying together strands of the many stories that have spun off the Odyssey and bringing them into a more modern morality.

Roel Konijnendijk: Odysseus is being forced to confront the question: if you deceive the people around you in order to get what you want — including the people who look up to you, like Sinon and his own troops — you’re still deceiving them. It comes out when he’s going through Scylla and Charybdis, where he says, “I told you to do this because I knew you would do the opposite, and that’s what I actually wanted you to do.” Eurylochus is standing there saying: but you just lied to me — you’re outright telling me you manipulated me. I like that they confront him with it: does this leadership style make you a good leader? Do you deserve people’s trust? Do you deserve people following you and looking up to you? That’s a more compelling way of looking at how Odysseus achieves things than the poem, which straightforwardly says, “that’s just how he is. He lies to people almost compulsively.”

There’s actually a scene when he arrives back on Ithaca and Athena shows up in disguise and tries to interrogate him. Because he’s suspicious of basically everyone, he spins a whole story, and Athena is quietly laughing — obviously she knows best; she’s not going to be deceived. She calls him out: you always have to spin a story. You can’t be straightforward about anything.

Jordan Schneider: But she also says: this is why I love you. This is why you’re my favorite.

Roel Konijnendijk: That’s why he’s her hero, her champion. But it is worth interrogating. As I said last time, the Odyssey asks you to ask questions: do you think this is right? Do you think this is worthy? The movie did that to some extent — it raised questions, and not only the ones already in the poem, but other questions besides that are also relevant and interesting.

Jordan Schneider: Let’s come back to the question of feeling bad about being a Homeric hero. We talked about the similes in the Iliad in particular, and how fun and awesome they make crushing skulls seem. Is there anything in the corpus that has people feeling bad about sacking towns and killing enemies and women and children?

Roel Konijnendijk: Absolutely. It’s baked in, right there from the beginning, and it’s one of the things we always have to reckon with in Greek sources. On the one hand, they glorify victory in war and the capacity to sack cities — the epithet “sacker of cities” is applied to heroes in a clearly positive sense. But at the same time, already in Homer, there’s a recognition that war brings endless suffering, both to the people fighting and to their surviving relatives — their wives, children, and parents. The empathy is right there, throughout. Homer will constantly tell you simultaneously that heroes are great and wonderful people for how well they commit violence, and that violence is horrible and war should not happen. Those two ideas always exist side by side.

Later playwrights from fifth-century BC Athens explore this in a lot of detail. Plays like The Phoenician Women and The Trojan Women deliberately put the suffering of the victims of war front and center — they tell you just how horrible it must have been for the surviving women after all their men were slaughtered and they were violently enslaved. In some cases we know what was inflicted on them, because it’s part of the myth: Cassandra is raped at the altar of Athena, which is why Ajax the Younger doesn’t make it home — he offended the gods in doing that. But fundamentally, he’s the only one who gets punished for the kind of behavior that is generally inflicted on these populations.

All of the people who told this story were aware that it’s a story about massive suffering inflicted on the Trojans, and they were able to perceive the Trojans as stand-ins for people they knew — potentially their own people. Some of these plays were produced in an atmosphere where Athens was committing similar atrocities against other communities, and it’s clearly the Athenians reflecting on what they had done by talking about the Trojans.

Justin McIntosh: That’s what I thought too. It wasn’t in the Odyssey or the Iliad, but there was definitely a tradition. Look at the trials and tribulations written into the plays. Even Antigone, which we talked about on the last show, deals not only with the horrors of death but with the horrors of being denied the proper rites and burials. Even if someone rose up against a king — was in some ways a traitor to the crown — he still deserved to be treated a certain way, and Creon was willing to kill his own daughter-in-law to punish the body of his nephew. The movie acknowledging some of that isn’t a wholly modern reinterpretation. These ideas existed in the ancient Greek world — maybe not in Homer’s poems, though we also don’t know, because there are seven other books of the Epic Cycle between the Iliad and the Odyssey that don’t survive. I appreciated that there were at least allusions to some of that.

Roel Konijnendijk: It’s because this is a reality. People tend to go too far and say ancient moralities were different — and absolutely they were. I would have liked that to be more present in these movies. But it’s not that they didn’t care about life or didn’t love their kids, which is an assumption people make when they take it too far. We have very strong evidence that they loved their kids, even though infant mortality was through the roof. They’re still human. Throughout these texts there is a lot of empathy alongside the glorification of violence. Those two things are obviously in conflict, obviously dissonant — but the Greeks affirmed both all the time.

One example: in the early fifth century BC, the Greeks were still writing tragedies about historical events, which they stopped doing very quickly — the only one that survives is Aeschylus’s Persians. But there was another, now lost, about the Persian sack of Miletus during the Ionian Revolt in the 490s BC. [The playwright Phrynichus] wrote it, they staged it, and the Athenians found it so distressing that they fined him for writing it and forbade it from ever being performed again. It’s a story about the sack of a city far away, but it hurt them so much because they could relate — they could sense this kind of thing perhaps coming for them too. There’s an awareness that to be victorious in war is glorious and brings all these benefits, and that to dominate others is great — but the flip side of that coin is always there, and they never forget it.

Jordan Schneider: Is now the time for the contrast with American Sniper?

Justin McIntosh: You know my stance on watching Navy SEAL movies — I don’t do it, so anything I said would be complete supposition. But a good comparison is Thucydides in The Melian Dialogue: “Why can’t you honor that which is hard? Why do you have to honor that which is soft?” The entire point is to show that despotism — the weak suffering what they must — is actually not the morally just thing. That’s in direct contrast to what you’d expect if everything were based on this Homeric impulse of might makes right and noble warrior kings. Even Thucydides was saying: no, we were wrong. We should not have done what we did at Melos, because it was tyrannical.

Roel Konijnendijk: There’s a fantastic scene in Xenophon’s Hellenica when the Athenians receive the report of the Battle of Aegospotami, where their fleet is finally destroyed and they have no chance of recovering from the defeat. They know the Spartans are coming for them — they’ve lost the war, but it hasn’t ended yet, which makes for a horrific moment when the news arrives in Piraeus on the official ship. Xenophon describes people in the middle of the night literally turning to each other in the street and shouting the news, which travels like a wave of sound from Piraeus through the Long Walls to the city. You can follow the sound of the wailing as people gradually learn from their neighbors what has happened. In that moment, he says, they all remembered what they had done to all the other Greek states — because it meant they expected nothing less from the Spartans now that they were at their mercy.

You can commit atrocities all you want in your moment of power and ascendancy, but the bill is being recorded. Eventually there’s going to be a reckoning. The Athenians absolutely expected their community would be wiped out. They narrowly escaped that fate, for complex reasons we won’t get into here — but they fully expected it.

Subscribe now

The Real Reason Behind the Trojan War

Jordan Schneider: Should we do a little Odysseus–Menelaus contrast?

Roel Konijnendijk: Do we want Menelaus in this movie, or Menelaus in the Odyssey?

Justin McIntosh: He wasn’t the joyful, tear-in-his-eye, a-cup-for-everyone-who-comes-to-his-house, red-haired, happy-go-lucky guy they depict in the Odyssey, that’s for sure, when Telemachus shows up. He’s a pretty dark and brooding figure.

Roel Konijnendijk: The Menelaus they were drawing on is Sophocles’s Menelaus. Sophocles was writing during the Peloponnesian War — Athens at war with Sparta — so the Spartan character obviously gets turned into a brutish warlord. In those later tragedies he becomes this really violent, angry man. That’s the Menelaus they gave to Jon Bernthal. Reading the Odyssey, you would never cast Jon Bernthal. The Menelaus there is slightly dorky, a bit dumb, but fine — a nice guy who just wants everybody to have a good time at his palace. He gives you all sorts of things you don’t need. Telemachus has to tell him, “I have no use for horses — I live on an island.” It’s really different. They’re projecting back from the later versions of those stories.

They did the same thing with Clytemnestra, using the Aeschylus Clytemnestra rather than the one in the Odyssey, who is merely an instrument of Aegisthus’s rise to power — she’s not even the one who kills Agamemnon in the Odyssey. In the later tradition she’s front and center, because it’s a lot more interesting: she’s the one who actually suffered the loss of her child, so the story gets told from the perspective of the wronged mother. But that’s not the Odyssey version — that’s the fifth-century BC version.

Justin McIntosh: The scene with Menelaus that stuck with me is the whole “it was never about me — it was always about trade routes,” and his dark, glib comment: “Well, maybe now it’s only 500 ships.”

Roel Konijnendijk: First of all, that line is Marlowe — Christopher Marlowe, “the face that launched a thousand ships.” It’s a modern line they put in the movie because everybody knows it. But let’s not treat this as arithmetic — she’s worth only 500 ships now because she’s slightly scarred? What is that? We’re not expressing her price in oxen. The original line is an expression of just how compelling she is. It’s weird to me that there’s been so much comment on that aspect of the casting, when I have no question that she is absolutely beautiful and compelling as a person — in her limited screen time she has real presence. I really don’t mean to diminish that.

But I don’t like the idea that all of these modern movies like to do this. Troy did it; the Helen of Troy TV show from many years ago with Sienna Guillory did it. All these modern versions want to rationalize the Trojan War. Helen gets shunted into the background by this line — which is in the Nolan movie as well — that she’s just Agamemnon’s excuse, because what he really wants is access to trade through the Hellespont. This is a modern fantasy. We now find it difficult to accept the idea that you would go to war to get a woman back. We want states to behave out of strategic interests, and we can understand access to the Hellespont — hell, we have the Gallipoli campaign and all sorts of wars over access to the Hellespont. Troy sits in about the right place to control that access, so let’s say the war was about trade routes.

The fact is that not a single ancient source ever suggests this war is about anything other than recovering Helen. That is the entire reason the war happens. That’s the entire reason Agamemnon is involved. Nobody has ulterior motives. This isn’t about trade. I get very upset about this precisely because rationalizing the war takes away something of what makes the story interesting and compelling — it makes the story less different, less historical, because it refuses to acknowledge the story’s own terms. It rejects the idea that this war could have been about what it says it’s about, and just invents other reasons: “I’m not satisfied with the Helen story, so I’m going to make up something else.” Don’t do that. I wish they hadn’t put that line in.

Troy did this massively — it made the whole thing about imperialism. Agamemnon becomes an expansionist warlord who wants Troy, and everything else is just an excuse, his foot in the door. The whole movie is built that way, which fits the early-2000s political environment — that was the kind of message people found instinctively understandable. We don’t need to do that now. We see people go to war for idiotic reasons.

Jordan Schneider: There’s a layer on this. We’ve talked about dueling culture and the early years of the US Navy, where someone says your shoe isn’t shined well enough and suddenly you have to shoot them to prove you’re brave. With Helen, there’s some offense — but layered on top of it is this glory-hero sacking culture where like, it’s summer camp, we’re going to play the hits and do our thing, because what else is there to live for besides sheep?

Roel Konijnendijk: Some very nice sweet wine, apparently, according to the Odyssey. There are all sorts of things these people possess that they think are worth having — luxury imports from all over the world; they can get all sorts of things from the Phoenicians just by maintaining peaceful relations. There are plenty of ways to live well. What they want is to make a name for themselves. They want renown. That is one of the big motivating factors for the people involved. On the one hand, they swore an oath — they have to uphold the sanctity of the marriage of Menelaus and Helen, so they’re obliged to go. But they go because they want to be there. For all its faults, Troy actually articulated this — that movie hews quite closely to the Homeric value of “this is going to be the biggest war ever, and you’ve got to be there.” If you weren’t there, you’re going to have FOMO for the rest of your life. That’s closer to the spirit of what these heroes would say. If they missed the opportunity to gain wealth and renown, why would anyone believe they merit sitting in their palaces drinking fine wines?

It’s also very interesting that Herodotus, who traces back the origin of the idea of wars between Europe and Asia — that’s how he simplifies it: Greeks and barbarians, Europe and Asia — says there were conflicts going all the way back into mythological times, but the escalation came with the Trojan War. Initially it was just raiding back and forth, people stealing women from each other. But the one time the people from Asia — the Trojans — stole a woman from the Greeks, the Greeks actually decided to invade, go to war, and destroy Troy. He says the Persian sages, the Persian logioi — if they even knew this story, but that’s the story he tells — hold the Greeks responsible for all of it, for the Persian Wars and everything else, because the Greeks escalated first. They went from tit-for-tat raiding to all-out war, and that was a massive overreaction.

So there was discourse — and I’m not saying this literally happened — among the Greeks about whether going to war over this was a proportional response. Some really horrible victim-blaming goes into the argument: “Helen probably wanted to go — why would you even bother?” But fundamentally, the debate implies there was no discourse whatsoever suggesting any other cause for the war. Everybody knew the Trojan War happened because Helen was abducted — or perhaps absconded with Paris — but that’s it. No one ever says, “Actually, let’s be real, that wasn’t why this happened.” That was the reason, and everybody who went along with it just wanted part of the action.

Justin McIntosh: In Renaissance America, we’ll call it — during the founding, when you had all these men who had read the Enlightenment books and were still dueling — that message would have landed: an overreaction to your honor being dragged through the dirt is something people could have gotten behind. It doesn’t necessarily register with a modern viewer who hasn’t read those works. There has developed, at least in American culture, real skepticism toward honor-based arguments — a belief in machinations within government, that there’s always another cause. It’s never as simple as “Iraq has weapons of mass destruction”; it’s actually about owning the oil fields. So I do see a reason for giving the war a bigger rationale.

Trojan Horses in Real Greek Warfare

Jordan Schneider: My favorite scene was the men inside the Trojan horse, which I thought was brilliantly conceived, shot, and executed. Who knows if it happened — but are there other tricks as outrageous as this that were actually pulled over the course of Greek history?

Roel Konijnendijk: Of course. Besieging a city is very difficult under any circumstances — especially for Greek cities, which had limited economic means to keep an army in the field for long, and limited means to develop machinery and tools to overcome strong fortifications. By far the most common way for a city to fall was treachery from within: somebody opens the gates to you. The second most common way, I would guess, is infiltration — you find some way to get in without being noticed. The Trojan Horse story is really quite elaborate, difficult, and risky; that’s just the version that happens in the story. The closest parallel comes from the Peloponnesian War, when at one point people hide under a boat and sneak in that way.

Most of the time, you find a place that’s less well defended, often because it’s very difficult to approach, sneak a couple of guys in that way, and either use them to open the gates or cause such a shock that the garrison surrenders. That has happened on many occasions. Famously, it’s how the Persians took the great fortress at Sardis from the Lydians — basically offering rewards: “You see that giant cliff face? There are no guards up there, because they think you can’t climb it. Prove them wrong, and I’ll give you loads of money.” Alexander pulled the exact same trick numerous times, telling men, “I’ll bury you in gold if you make it up there.” Some people died trying — we have reports of men falling all the way back down — but usually, once they get up there, it’s game over.

Jordan Schneider: Can you talk about the Roman award where, if you were verified as the first one up on the wall, you got some class upgrade? I’ll look it up.

Roel Konijnendijk: Usually it’s a monetary reward — you can also get promoted, though I’m not sure about the exact parameters of that one. But “first upon the walls” is one of those feats that gets a specific award. The idea is that if you’re first, you’re most likely to encounter the strongest resistance, but also the first to achieve what the entire army is trying to achieve — so there’s a specific reward to get people to take that risk, to put themselves in front. On various occasions — at the Siege of Tyre, and at one of the cities he took from the Mallians in what is now Pakistan — Alexander himself was first up and over the wall, precisely because he led by example. That was his leadership style: do what he asked others to do.

But that’s not a trick — that’s just assault. The point I’m making is that most of the time these walls are not conquered by brute force. They are conquered by infiltration or treachery from within. The Trojan Horse is the ultimate example of that, the most famous from the Greek tradition. It’s not literally repeated, partly because everybody knew the story — if you showed up in a wooden rabbit, Holy Grail–style, they would probably not fall for it. But the underlying ideas are very similar: how do you get through a wall? You try not to be seen doing it, or you get someone on the inside to help you.

Justin McIntosh: A very direct analogue is the Gallipoli campaign. Very early on, the British were having a hard time gaining a good foothold, and the idea emerged of running what would look like a derelict ship onto the shore, so that men could pop out of it by surprise in the midst of the Turkish lines. Of course, what happens is that this derelict-looking ship loses its rudder, hits the shore, and the Turks turn all their guns on it and just hammer it. Everybody is pinned down; they can’t be extricated until night falls and they’re finally pulled back. That’s clearly an attempt at a Trojan ship — and you see how difficult something like that is.

It shows the acme of Odysseus’s plan: his ability to think through the deception and convince other people it was real. That ties back into why he’s wracked with such guilt — it shouldn’t have been possible. It shouldn’t have worked. The Trojans should have seen through it, but he layered it so that it worked: having the soldier say it was a gift, Sinon losing his life, all the things that led to it actually getting inside the walls.

Roel Konijnendijk: There’s the other version of that as well, where the priest Laocoön tells them not to take the horse, and the gods intervene — they send snakes to seize him and his sons, which leads to that famous statue of them writhing among the serpents. The gods at this point are basically fed up with the war; it was determined long ago that Troy would fall, so we’re just trying to get to that point, and they intervene on behalf of Odysseus to make sure his plan works. That in itself is another example of how intricate this plan has to be — how many different steps have to occur before the Trojans are convinced. They read the snakes as a sign that resisting the horse is a bad idea. None of that is in the Odyssey’s very short account of the Trojan Horse. Those elaborations on the story show that people were already a little unwilling to accept that it was that easy — so they kept adding safeguards to the plan.

Jordan Schneider: Found it — it’s called the corona muralis, and you literally got a golden crown shaped like a castle wall. Medal of Honor, eat your heart out.

Justin McIntosh: I would take it. I’m guessing they didn’t give many of those out to the living — they were probably mostly posthumous awards.

Jordan Schneider: No, you only get it if you’re alive. And apparently one time two guys each thought they had earned it, and they almost had a little civil war in the camp. Polybius — everyone go dig that up.

Roel Konijnendijk: There are a number of occasions, at least in Greek warfare, where once the wall is breached — once it’s clear the enemy is in a position to overcome the defense — the defense just breaks immediately. You don’t have the forlorn-hope tradition where being first through the breach means you’re probably dead. There’s more of a sense that if you’re first through the breach, you might actually be the first to survive. But it depends on the situation: in a lot of other cases the defense becomes extremely stubborn at that point, and there are known cases where the attackers, having breached the walls, are driven back out again — sometimes several times in a row. Some of these defenses are really tenacious.

Set It in the Archaic Period

Justin McIntosh: I want to know — what’s the Roel three-hour adaptation? What are you spotlighting?

Roel Konijnendijk: An interesting question. One of the things I really didn’t like was the film’s attempt to be relevant to Bronze Age history — which it isn’t, and shouldn’t have been. They shouldn’t have put that in there. This is not meant to be a historical story, and trying to make it a historical story about the Bronze Age is misguided, because the poem doesn’t describe the Bronze Age. It describes the early Archaic period. If I had my druthers, I would set it firmly in the Archaic period. Compared to the Late Bronze Age — and to the world of states and kingdoms generally — this is a world of much lower complexity, organization, and wealth, only just emerging from a long period of disruption, upheaval, and recovery. We’re talking about very small communities.

The heroes of this story are not kings in the sense we understand — living in palaces, ruling kingdoms. They’re petty lords, and often there are several of them. Ithaca doesn’t have a single king; it has loads of people calling themselves kings. Odysseus is only the wealthiest among them. He isn’t crowned, doesn’t sit on a throne, doesn’t live in a palace — he’s just a guy with more wealth, more power, more retainers, more flocks, and more wine in his storeroom than the others. That’s how you distinguish yourself in this world: a hierarchy of wealth in which some men get to call themselves lords.

Because of that, there’s no institutionalization, no formal procedure of power. Power is much more brittle — it belongs to those who can affirm it and take it. That is why the suitors are such a problem. That is why Telemachus can’t just say, “Everybody out — I’m king.” He has no institutional backing, no law that says he should be in charge. It’s a much more might-makes-right world, and much more modest in its means and material realities. You can much more easily imagine how chaotic and disorderly these heroes actually are, and how much it matters to them to sack a city in order to steal a couple of goblets, or to enslave some people — because that is the entire basis of their status in society. There’s much more potential to make this story make sense if you just set it in that archaic world.

I’m always pointing this out: you have to stop putting Odysseus in a palace on Ithaca. If you read the actual text — okay, there are some upstairs rooms, fair enough — but most of the “palaces” Homer describes have three rooms: a main room, a back room, and a portico. That is the entire space of a Homeric palace. These are small, two- or three-room houses. Why? Because if you look at the archaeology of the eighth and seventh centuries BC, there are no bigger houses in the Greek world. That is the house they knew. Even the biggest man in these small cluster communities lived in a slightly larger version of the same house. There are no palaces, no expansive mansions or villas in the Roman style. You have to imagine this world as infinitely smaller, simpler, and poorer than the world of kingdoms we picture.

I would set the whole story in a world like that, because then you can see why it matters that these suitors — practically kings in their own right — vie for the hand of Penelope to affirm the status they already have, and why their family members are a real problem for Odysseus. They also have weapons and retainers; they can muster an army and come take his things. If he antagonizes them, there is civil war, which he will lose — he’s one man with his retainers against all these other men with theirs. This is not a world in which the questions of power are settled. That could make for a very compelling version of the story.

As for which elements to include — I’m very glad the movie talked about xenia, hospitality. That is the absolute core theme of the Odyssey. This is a song about hospitality — about what it’s like to be a good guest and a good host. That is all Telemachus needs to learn, what the suitors need to learn, what Odysseus needs to understand about the world. I like that they kept it as a theme, although they twisted it a little — they made it into a general moral code, which pushes it too far.

They were quite pragmatic about what they included and what they didn’t. It would have been nice to have Scheria in there — the Phaeacians, Nausicaä. But I can see why they didn’t want that extra stop on the way — the whole rise-and-fall-again arc. It makes sense from a this-has-to-be-a-three-hour-film perspective.

I also really liked what they did with the Lotus Eaters, which is about six lines in the Odyssey — almost nothing. Odysseus never touches the stuff; two of his men try it. It’s an irrelevant little episode, but everybody knows it. Transposing it to Calypso’s island and making it something Odysseus takes because he wants to forget — that’s really elegant. It turns this trivial story about forgetting everything and living in bliss until you disappear into nothingness into something he actually wants, something he actively does to himself rather than something done to him. That makes sense within the way they’ve rewritten the themes of the story.

It keeps coming back to the same thing: I like some of what the movie did and some of the ways it changed the story. Some changes I would push further; others I would keep. But fundamentally, I would have set it in the early Archaic period and been honest about the difference of this world, of how far removed it is from our understanding of political order.

Share

Gods, Achilles, and AI Iliads

Jordan Schneider: The best compliment I can give Christopher Nolan is that I walked out saying, “I need to write my own adaptation of the Odyssey” — not because he did a bad job, but because it’s such an interesting intellectual exercise. You’re operating under real constraints, you have to cut material, and the ways you can remix it are just really cool. The one thing I’d change: he basically took out the gods, which is fine, but I wonder what a god-maxed version of the Odyssey looks like. It always struck me that these gods are just playing, and they don’t really care — you’re just a pet insect to them.

Justin McIntosh: Doesn’t that become that bad Clash of the Titans movie, though? That was so bad. The problem is that a movie like that ruins the whole approach.

Roel Konijnendijk: It’s also very Percy Jackson — the gods as actors in the actual world. Can you do that in a way that’s compelling? One really small moment in this movie landed for me: when Odysseus finally leaves Calypso’s island and drifts off on the raft, she tells him, you’ve got to take your punishment — Poseidon is going to batter you, but Zeus won’t let him kill you. By that point in the story, we’ve become so persuaded that this world is infinitely dangerous — full of threats too big even for Odysseus, forces too powerful for him to confront, at whose mercy he lives — that to be told at least one of these gods is not only looking after you but is more powerful than the other forces really works. At that point you think, okay, I’m not worried for him. Even though he’s being committed to the sea again, he will be fine.

Even though the gods are not in the story, the film manages to convince you — and this is to the full credit of the score and the mood of these episodes — that there are powers in the world we cannot control, vastly more powerful than us, and that this doesn’t mean we’re helpless. It can also mean they’re helping us — that they’re on our side.

My favorite example of this has always been Sunshine, the Danny Boyle sci-fi movie from 2007. It opens with a scene on a spaceship flying toward the sun. Someone is testing how far down he can set the filter on the screen through which he’s viewing the sun, and the ship’s computer warns him: beyond about four percent penetration, your eyes will burn away. He sets it as close as possible to that point — and the music kicks in, and visually you are persuaded of the immense power of this formless thing in front of you. Then you’re told: that’s the thing at four percent. That’s how the movie establishes its antagonist because the sun is the antagonist; they’re essentially trying to fight the sun. It’s so powerful because, right from the beginning, the film manages your understanding of what you’re up against: you personally, physically cannot withstand four percent of the power of this thing. You will die.

That is how you convince an audience of the existence of powers greater than us and put us in awe of the people brave enough to confront a world that has those powers in it. The Odyssey does that to some extent. It’s a much more effective way of doing the gods than having them appear in a toga and talk to you. In a lot of ways, that’s what you want those gods to be: forces in the world for which you are, as you say, a toy — something of no consequence, no value, and no persistence whatsoever.

Jordan Schneider: Speaking of gods — there was some critique of Nolan for putting Christian morality into all of this. What I found interesting is that there’s enough in the soup of 3,000 years ago that we have parallel myths going on — Odysseus on a raft and Jonah. There are enough story beats where the two story ecosystems talk to each other that it seems fair to look at these characters through the lens of what Christ would have thought of them.

Roel Konijnendijk: What you can accuse this story of that is not Greek at all is turning hospitality into a general moral code. They take the idea that you have to be a good guest and a good host and turn it into the Golden Rule — treat others the way you’d like to be treated. That is much more general than any Greek would ever articulate it. It’s not an ancient Greek moral code; that doesn’t exist. But it’s not just Christian, either — lots of religious traditions have the Golden Rule; it’s also Buddhist. You can accuse Nolan of introducing Christian thought, but it’s more that he’s drawing on a much more general idea of what it is to be a good person. Unfortunately, not an ancient Greek idea — and that is legitimate criticism, which I’ve now raised several times. He’s not really willing to confront the notion that people simply had a radically different morality from ours, except in tiny moments — the unwanted puppy chucked off a cliff. Generally he doesn’t want to confront that, which is fine — it’s a modern adaptation. It makes the film a little less Greek, but a little more universal to modern audiences.

Jordan Schneider: Justin, what’s your Odyssey adaptation — or Iliad?

Justin McIntosh: I would redo the Iliad. Troy was such a terrible movie, both in the story and in the way they told it. I want to see the god-kings actually out fighting — Hector pushing the Greeks all the way back to their ships, out in front, laying waste to the Argives as he drives them into the sea. Then the turn, and Achilles’s reentry into the war. In Troy, they killed Menelaus and had Hector become this weirdly compassionate person toward Paris — nothing in the poem suggests he would react that way. Throughout the entire poem, Hector berates his brother and blames him for everything. I’d much rather have that, because there’s a lot more there that has never actually been put on screen. The late-’60s version of Achilles at least tried to start the way the poem does — with the argument inside Agamemnon’s tent — but the rest of that movie is terrible. Charlton Heston–era stuff.

Roel Konijnendijk: I have notoriously stood up for Troy a few times — it does do certain things right. And for all the merits of Nolan’s cinematography and direction, his scene of the actual sack of Troy did not improve on Petersen’s. There’s a gorgeous moment in Troy when the gates are opened and, from the dark plain, you see this massive Greek army converging on the gate in silence, emerging from the shadows as if out of thin air, funneling through the gates. Then the camera swoops down and follows them as the noise swells — the roaring of the soldiers and the screaming of their victims. That is such a good moment. In Nolan’s version, there’s a bit of floating around Aït Benhaddou, where they filmed it — a couple of fires, some people being killed in the background, and you follow Odysseus around. I didn’t feel it built on or improved on Petersen, or made you feel it was necessary to do this again. The collapse of the towers was pretty cool. And the moment the gate opens with Agamemnon standing there — that’s a cool moment.

Jordan Schneider: It’s like a WWE walk-on.

Roel Konijnendijk: You can definitely hear the dun-dun-dun. But in that particular scene, Troy is just the better movie — it sweeps you along and makes you feel the almost unnatural horror of this army descending on the city. I don’t mean that about the whole film; I’m glad we have both movies. And I have to say, my absolute favorite line from Nolan’s movie is when he says, “What if he saw ten years of hate poured into the city in one night?” That is so beautiful — a really effective way to convey what happened. It’s not from the poem; that’s just writing for the movie. But the scene that accompanies it does not match Petersen for cinematic brilliance.

Justin McIntosh: The one thing I’ll say is that Nolan told a much more personal version of the battle. It’s centered on Odysseus, whereas Petersen’s was societal — you were watching this huge event unfold. In Nolan’s film, the perspective is basically Odysseus’s own; there’s a little third person, but much of the walkthrough is meant to feel like you’re there, looking around, so you get a very narrow view of what’s shown. Maybe the thought was: I can’t recreate this because it was already done, so I’ll do something slightly different.

But I still hold that if I were going to recreate one of the Homeric poems, it would be the Iliad. Honestly, the end of Troy is the best part of Troy. The rest — even the way they introduce Achilles, the fast sprinter, but he’s going against a giant of a man and he kills him in one blow. The whole point of the poems is that Achilles is supposed to be this humongous man, first among equals — Ajax is the only one supposed to be bigger. When you read the way Homer lays out the descriptions, Achilles is the most powerful and the fastest — he’s supposed to catch you off guard. Whereas you watch Brad Pitt play him and you think: that’s Brad Pitt.

Roel Konijnendijk: I actually like him in that role — apparently because he really didn’t like doing it. He didn’t want to make the movie and hated working on it, but he brings a surliness to the role that fits Achilles, who is — to say it as nicely as I can — a very flawed character. I don’t think you’re supposed to sympathize with Achilles at any point in any of these stories. Pitt combines the ultimate-warrior figure with the surly teenager, and it’s perfect. He’s a petulant, annoying man you don’t want to work with and he doesn’t want to work with you either. Obviously he’s not the definitive Achilles — please keep doing different versions of Achilles. I love David Gyasi in Troy: Fall of a City, the BBC series — he did a really convincing job as the angry ultimate-warrior figure. But I quite liked Brad Pitt as Achilles precisely because he didn’t seem to enjoy himself at all.

Justin McIntosh: What about Brian Cox — was he the wrong casting for Agamemnon? He’s an amazing actor, but age-wise, I don’t know if he was the right person to play the role.

Roel Konijnendijk: Brian Cox is perfect. You cannot possibly say anything against Brian Cox. He understood the assignment — he was just chewing the scenery. Troy is my ultimate example of a movie that’s so bad it’s good: a movie that is not good, but is extremely watchable, with everybody in it having a great time — except Brad Pitt, apparently. Brian Cox completely understood that. I enjoy every second of him on screen — old-school hamming it up in the extreme, not giving a shit. The lines are terrible. It’s wonderful. And he gets to quote the poem a few times. In that opening scene you mentioned, where Achilles fights Boagrius, he says this line to Nestor: “Of all the warlords loved by the gods, I hate him the most” — a line from Homer, straight out of the Iliad. Let’s have him chew the scenery with lines from Homer, please. More of that.

In Nolan’s Odyssey, the scene where Odysseus goes to the underworld — where he talks at length with Sinon, and meets Tiresias the seer — also has him talk to Agamemnon, and that is one of the bits most faithful to the poem. Agamemnon is the one who warns him, don’t just walk into your house assuming you’ll be welcome — look at what happened to me. It’s a plot-crucial warning: approach this indirectly, because you never know what’s waiting — especially whether your wife is still faithful. Test the waters before you march in. Unfortunately, it comes with some classical Greek-style misogyny layered on top — “you can’t trust a woman.” But it’s a really nice demonstration that Nolan actually likes, appreciates, and understands the source material.

Justin McIntosh: What’s yours, Jordan? Your remake?

Jordan Schneider: Well, nobody liked my gods idea. I think the female-forward versions could be cool — we got just a little bit of Circe, a little bit of Calypso.

Roel Konijnendijk: Not enough Penelope either.

Justin McIntosh: You could make Madeline Miller’s Circe into a really good one — and then you get [Tennyson’s] “Ulysses,” where he goes stir-crazy at the end, and that whole final scene. A different take on Odysseus.

Jordan Schneider: You could also center Telemachus — he has a whole arc. The fighting in the temple was sort of weird, and just watching him mature — you know what, I’m going to cut that. I don’t really want to see a Telemachus movie.

Justin McIntosh: You don’t want to see Tom Holland?

Roel Konijnendijk: The final scene of the Odyssey — one of the last moments of the poem — is Odysseus, his father Laertes, and his son Telemachus gathered together as the army of the suitors’ families comes for them. The suitors have been killed; now their kin want revenge. There’s this beautiful moment where Telemachus and Odysseus are vying with each other: “I’ll kill the first man.” “No, I’ll handle it.” And Laertes just stands there beaming with pride, because his son and his grandson are competing for the honor of being the best warrior. That is intensely Homeric. We find it very difficult to relate to that morality, but it tells you so much about this society: all Laertes wants, in possibly his final moments on this earth — and he has waited so long; he’s so old by this point — is to see his son and his grandson standing with him in a battle line as brave warriors. This is the Homeric world of heroes in a nutshell.

Justin McIntosh: I do love the story, when they first come to fetch Odysseus and he decides to feign madness — out sowing a beach with salt, driving an ass and a horse yoked together. They ask what’s wrong; he says, “I’m crazy — look at me.” And they say, oh yeah, you are crazy — except for the one man who walks over, takes the infant Telemachus, and sets him in front of the plow: let’s see how crazy. And Odysseus says, “You got me. I guess I’ll go to war.”

Roel Konijnendijk: I don’t remember whether that’s anywhere in the epics, but it’s definitely part of the wider knowledge of the story — it comes up all the time.

Justin McIntoch: He was the one hero who actually tried to get out of going to the war, and he was seen as different from all the other heroes.

Roel Konijnendijk: Achilles too — he hides and dresses up as a girl to avoid being called up, or at least his mother dresses him up as a girl to hide him. Then Odysseus gets sent to find him out. It’s a fun catalyst. They don’t actually need Odysseus to come along — they need Odysseus to get Achilles to come along.

Jordan Schneider: The whole idea of dreaming up adaptations is not as crazy as it was even five years ago, because you can just have AI make movies for you. We’re going to have a lot of Iliads.

Justin McIntosh: We are. We’re going to get the Grok one.

Jordan Schneider: There are far more ideas for classical-literature adaptations than there are Hollywood budgets, so there’s something to look forward to with this technology. It’s good to dream.

Did Ancient Warriors Get PTSD?

Roel Konijnendijk: I don’t know if you wanted to talk about ancient PTSD, because we haven’t covered that.

Jordan Schneider: Let’s close on that.

Justin McIntosh: Close on PTSD — my favorite subject.

Roel Konijnendijk: We get this question all the time on AskHistorians — it used to be one of our most frequently asked: did people in ancient or pre-modern warfare get PTSD? It’s a weird, narrow way of asking the question, because we now know that post-traumatic stress disorder doesn’t relate only to war trauma. It can result from domestic violence, sexual abuse, witnessing crime — any traumatic experience. But in public discourse, the most common context is military veterans, so the question has been: is this a universal experience for people coming back from war, or something that only affects people who fought in modern wars?

That framing partly reflects the fact that PTSD was only defined as a diagnosis during — really after — the Vietnam War, and it has often been associated with earlier diagnoses like shell shock or trench fatigue. Before the twentieth century, this was essentially not a defined condition. We can’t say people were being diagnosed with it, because there was no such diagnosis. That has been a persistent question in both scholarship and public curiosity.

Justin McIntosh: It at least predates the twentieth century. There are discussions from the Civil War and the Revolutionary War of battle fatigue and of veterans coming back haunted by the things they had seen — clear allusions in both literature and the historical record. It would be very difficult to believe it didn’t exist in ancient times, if Thucydides is right and men don’t change.

Roel Konijnendijk: The problem is, where do we find it? Something that isn’t recognized as a well-defined disorder affecting a specific group of people might simply go unnoticed, because no one is looking for it. There’s no clear, unambiguous testimony that people who had been through these experiences were expected to display these symptoms. So scholars have been mining ancient sources for potential cases, and a couple keep coming up.

One is a veteran of the Battle of Marathon who supposedly suffered hysterical blindness. He wasn’t harmed at all during the fighting, but he couldn’t see for the rest of his life. The story he told — recorded in Herodotus — is that he saw a soldier coming at him, a very large hoplite with a beard covering his shield, a scary, giant figure — who instead passed him by and killed the man next to him. From that moment, he went blind and never saw again. People have often taken this as exactly the kind of case we’re looking for. But first, he reports no other symptoms — literally just the blindness — and a modern psychiatrist wouldn’t make a diagnosis on the basis of a single symptom; that’s not how psychological disorders work. Second, the social reaction was very different from what veterans now often experience — separation and detachment from their families and loved ones, ostracism, difficulty reintegrating. Epizelus, by contrast, was a hero. He told everybody this story, and everybody loved him — granddaddy was at Marathon, telling this wonderful, spectacular tale. That’s why Herodotus recorded it.

So the question has been whether we can really call this what we’re looking for — both because of the lack of diagnostic detail, and because the reception from society was radically different. One acknowledged feature of PTSD is that what makes it so much worse is the way it alienates you from the people around you, and the way society doesn’t know how to deal with it. That is why it has so often been debilitating. Whereas here, the way they dealt with a veteran who was obviously harmed in non-physical ways was radically different.

Justin McIntosh: Is there something to the fact that hoplite armies were composed of the citizens of the city-states, so there was more societal understanding of what happens in war? We have the issue in the US and Britain of the all-volunteer force, where a very small percentage of the population serves. If it were a much higher percentage, conceivably there would be a much greater understanding of what happens and of how society can reintegrate these people.

Roel Konijnendijk: That is the argument made by Jason Crowley, a scholar in Manchester who has written a book on precisely this question. His argument has always been that because you’re going to war with your neighbors, your relatives, and your friends, there isn’t as much separation between home life and warfare — not as much of a sense that you are leaving your society and then having to reintegrate into it. More recently, Owen Rees and other scholars have qualified this somewhat, but fundamentally it’s true that you’re not separated from your normal life the way you would be in a modern all-volunteer, all-professional force. That would make integration easier, simply because everyone you know has gone through the same experience. They know what it’s like — they were there with you, and when you go home, they’re still there with you. To that extent, these societies may have found it easier to absorb the consequences.

But traumatic experiences do change personalities. One example scholars cite is the Spartan commander Clearchus, who was murdered while in mercenary service in the Persian Empire. Xenophon, who knew him, gives a little postmortem of what he was like. He describes him specifically as somebody so obsessed with war that he couldn’t let it go — a war-lover, philopolemos, a very rare word in Greek. He devoted himself singularly to warfare. Normal people have hobbies — they spend money on pretty boys, or festivals, or drink. This man spent everything he had on war. He was doing war constantly, and it was the only time he was happy. His character was therefore very hard to deal with: very good in a crisis, but otherwise unlikable — too hard, too harsh, too unkind, too single-minded.

A lot of people have read this as a mind twisted by PTSD — someone who could only function in the same context in which he was traumatized, who was only happy in the environment that felt relatively safe to him, which was an environment of danger. That reading has more legs, in that it tries to understand a holistic picture of a person. But it was still not recognized in that context as a disorder — that was just his personality. And you could almost imagine that a Spartan upbringing produced that personality quite frequently. People who go through that training program from childhood may well end up traumatized by default — it’s almost built into the Spartan system; they were constantly abused as a normal part of their upbringing. In this case we actually have a description, so we can do something with it. And again — in Spartan society, just as in the Athens we described, it may have been easier for society to give that a place, simply because all the adult male citizens had gone through it.

Jordan Schneider: This is a tough place to end.

Why Ancient Historians Are Mad at Nolan

Jordan Schneider: Is it chance or is it circumstance? The Sea Peoples will rise again. Do the Sea Peoples have PTSD?

Roel Konijnendijk: Definitely.

Jordan Schneider: Did they feel bad about their raiding?

Roel Konijnendijk: That’s why they went raiding. The story of the Sea Peoples — we know them only from one or two Egyptian inscriptions. They were not the reason for the collapse of the various Bronze Age kingdoms; that’s a discredited modern theory. And they certainly had nothing to do with the Trojan War. But that is how Nolan wanted to roll together a couple of things he knew about the Bronze Age with a couple of things he knew about Homer, and that’s what we’re stuck with now. I’m going to be facing endless questions: were the veterans of Troy the Sea Peoples? We have a list of names — we know they were from Palestine, from Lycia, possibly from Mycenaean Greece; there’s a list of peoples in the inscriptions. We don’t have to wonder who the Sea Peoples were. They’re just not the reason for the Bronze Age collapse, guys. We can set that aside. Can we please stop talking about the Sea Peoples?

Justin McIntosh: There’s good reason for ancient historians to be mad at Nolan — they have to answer this question all over again when it has been fundamentally answered.

Roel Konijnendijk: He put it front and center in his understanding of the Trojan War. Very sad.

Justin McIntosh: At the end of Emily Wilson’s write-up, she said, at least it’ll bring more people to the books. It has — sales of the Odyssey are up 76 percent from this time last year.

Jordan Schneider: That seems really low — it’s got to be higher than that. My last question: can we talk about the Landmark series for a second? It’s just the coolest thing. Pitch these books.

Roel Konijnendijk: The Landmark translations are a series of books — privately funded, I believe; somebody just wanted these editions to exist. They get scholars to update existing translations or write new ones, then illustrate them with a massive array of maps and add appendices explaining various aspects of the source. They tend to be fantastic first introductions to these texts, because they don’t just give you the words — they make sure everything is fully contextualized in the world in which it takes place. I really love the Landmarks. They’re great for Xenophon’s Hellenica and for Thucydides, but you can get them for a lot of other texts too.

Jordan Schneider: We’ve got Herodotus, Thucydides, the Hellenica, Arrian, the Anabasis — and Caesar.

Roel Konijnendijk: Oh, they’re doing the Anabasis. Cool.

Jordan Schneider: What’s the addition you want to make, Roel — if the editors are listening?

Roel Konijnendijk: A Landmark Cyropaedia would be really interesting. They’re already doing Xenophon’s Hellenica, and now the Anabasis. But Xenophon also wrote what is essentially a historical-fiction novel about the rise of Cyrus the Great. It’s a fictionalized version, but it has grounding in real history, and it’s set all over the Persian Empire. It would make a great parallel piece to Xenophon’s Anabasis and to Arrian’s as well. It’s also due a new translation — the Loeb “Little Green Book” translation of the Cyropaedia that I have is a bit of a mess.

Jordan Schneider: It’s like being handed — who did that Thucydides?

Roel Konijnendijk: Hobbes did.

Jordan Schneider: Hobbes, yes. It’s just impenetrable. You don’t know where the places are or where people are going. I’m not blaming him — but giving someone in the twenty-first century a text without maps and diagrams and photos of what the armor looked like — there are just ways to make this so much more engaging.

Roel Konijnendijk: The Oxford World’s Classics Hammond translation of Thucydides is really accessible — I always recommend it as an easy way in. It’s not necessarily the best version of everything, but it’s cheap and it has your maps and everything else. Otherwise, the Landmark Thucydides has all the maps you could ever want, plus a lot of extra contextualization in the notes and appendices. Go for that if you’re willing to shell out for one of these big tomes.

Jordan Schneider: In the Agora Museum there’s a shield from a raid the Athenians went on — Pylos, I think — with an etched inscription: we stole this from these guys, at this battle. They thought it was so cool they had to etch it in.

Roel Konijnendijk: It’s because it became a dedication. They dedicated these spoils to the gods and put them on display, and to make it a proper dedication you inscribe the context — where you took it. It’s just a few words: “from the Lacedaemonians at Pylos.” We took this from the Spartan forces that fought us at Pylos.

The Odyssey connection is that part of the movie was filmed right where that battle happened. The cave of Polyphemus, the Cyclops — that’s Nestor’s Cave on the rock at Pylos, a few hundred meters from where that shield was captured. That is the Battle of Pylos and Sphacteria during the Peloponnesian War, 425 BC, in which the Spartans were not only defeated — a number of them were captured. They surrendered, which is not something Spartans are supposed to do, and it genuinely affected the course of the war. It’s wonderful that we have a physical token, found in a well in the Agora, that confirms the narratives we get from literary sources.

Jordan Schneider: I love the idea of the Spartans coming into Athens, seeing this, and just throwing it down a well. Don’t you dare celebrate our defeat — we’re going to memory-hole you.

Roel Konijnendijk: They’ve got to respect it — it’s dedicated to the gods. But the Spartans did the same thing. Their temple to Athena on the acropolis at Sparta was known as the temple of Athena of the Bronze House. There are two accounts of the name: one, that the interior of the temple was plated in bronze; the other, that it was so covered in dedications that from the outside it looked like it was made of bronze. They were also celebrating their victories by putting spoils on display.

Jordan Schneider: All right, we’re going to call it here — leave them wanting more. Roel, thank you so much for being a two-part, twice-in-ten-days ChinaTalk guest — a first in our history.

Roel Konijnendijk: Very glad to be here. This was a lot of fun. Thanks for having me.

[Outro song]

Hollered out my own name across the water like a fool / Cost me ten more years and every man I knew / Man of many hooks, let’s go / Nobody came back with me, nobody came back with me / Everybody who believed in me is in that wide dark sea / Only one that made it home, with nobody

Built a horse out of a lie, put a kingdom in the ground / Walked up in a man’s cave, ate his food without a sound / First rule there is, and I broke it like it wasn’t nothing / Blinded him and left him standing in the dark with nothing / He asked me who it was, told him “Nobody” — that’s who, Nobody / Holler “Nobody’s killing me,” and every giant went back to sleep on cue / We was gone, we was clean, we was already free / But free ain’t enough when don’t nobody know me / So I stood up on that deck and let them know: I’m Odysseus, son of Laertes / They know my tricks everywhere I go, and my fame go all the way to heaven, don’t you know / Then he prayed on me: let him come home late, let him come alone in another man’s boat, and let him find trouble waiting at his own gate / His daddy heard it, his daddy answered / I wanted them to know it was me, I wanted them to know it was me

Nobody came back with me, nobody came back with me / Said it out loud one time, now I can’t say it at all / Twenty years of my life for the sound of my own name called

King of the winds at your back, they don’t know / Got close enough to see the smoke off my island, then they cut it open / Come up on my island, gone again / They ain’t never trust me ’cause I ain’t never let them in / Fed six men to them rocks and never told them they was chose / Eurylochus, you was right about me / Told you go left ’cause I knew you’d go right — ain’t no way you couldn’t see / Every call I made was smart, every one of y’all still died / Nobody came back with me / Learned all they names and I buried them all at sea

Asked Achilles in the dark, was the glory worth it / Said he’d rather work a landless man’s dirt than be a king over all of the dead / That’s the best of all of us, wishing he was poor and breathing instead / Reached for my mama and my arms went straight on through her / She said it was longing for me that took her / Then the goddess held out forever, said stay and never die / I said give me my home, give me a wife that’s gonna get old, give me something that can end

King in the dark said don’t walk in your door with your chest out / Said he came home like a hero in his own house — laid him out / So they loud his name on the wall, and I knock like a stranger at the gate / Beggar rags, beggar name, twenty years late / Men in my hall eating twenty years of hurt, calling theyself guests / Not one of them thought to ask a beggar his name — that’s how you fail the test / Dog wagged his tail one time and dropped his ears down low / Ain’t had the strength to come to me, but he the first one to know / A woman washing my feet felt the scar and said my name / Twenty years I’ve been dying for somebody to say that name / Now it’ll kill me if she do — had to hold her mouth and take the blame / Strung my bow in a room of men who couldn’t lift it off the floor / And it got real quiet in there, nobody talking no more

Nobody came back with me, nobody came back with me

Three years she wove a dead man’s burial cloth in the daylight / And pulled every thread of it back out at night / Held off a hundred men with a loom and ain’t told a soul / Twenty years and she still ain’t believe me / Told her move the bed just to hear what I would say / I said ain’t nobody move that bed — I cut it from an olive tree, still rooted in the ground / That’s when she broke / The song say she cried like a man come up out the water when he finally see the shore / Then the goddess held the sun down under the world so we could stay up talking / They call me sacker of cities, they call me the one who endures / Only one of them ever got me anything, and she the one who really endures / Nobody came back with me, but somebody was still here

Wealthiest man on a rock in the sea / Which on this island means eleven more of me / No crown, no throne, no law, no gate / Just jars on the shelf and the flocks on the hill / Three rooms and a portico — three rooms, that’s a palace where I’m from / No marble, no dome, no walls that hold / Just more wine than you — that’s all a king is, that’s all a king is / Sacked a city for a cup and I carried the cup home / That cup is the only reason anybody calls me lord / Suitors in my hall eating twenty years of hurt / That’s not bad manners, that’s a takeover / Boy runs the seat, but the seat isn’t real — who enforces it? Nobody’s coming / Power’s not a title, power’s who shows up / Storms deep, walls thin / One bad season from a man who used to be the lord

Bryan Clark is a senior fellow at Hudson Institute and a former submarine officer who runs war games for the Navy and its allies. Justin McIntosh is a retired Army Special Forces soldier — twenty years in, sixteen of them as a Green Beret — who now writes Mind of Things on defense, technology, and foreign policy.

We discuss…

• Why 850 Tomahawks in two months emptied the stockpile that was meant for China

• The leak hunt at Camp David, and why it should be a very short list

• How “lethality” became a testosterone prescription instead of an ammunition line item

• Japan’s new defense strategy, which finally names China out loud — and bets on one-way attack drones

• Whether AUKUS Pillar 1 is quietly consuming the rest of the Australian military

• Why manned-unmanned teaming keeps failing in the war games, and what the CCA program got wrong

Listen now on your favorite podcast app.

[Edited for length — the full conversation is available on your favorite podcast app.]

Bryan Clark: “…We have plenty of munitions — but by the way, the people that are talking about this should be prosecuted.”

Justin McIntosh: Says the guy who puts an Atlantic reporter in a top secret Signal chat.

Bryan Clark: Exactly. I don’t know, I’ve never talked to that guy before. I possibly had his Signal ID in my contacts list.

Jordan Schneider: Wait — you don’t have any Atlantic editors-in-chief in your Signal?

Bryan Clark: Actually I do, but that’s me. I’ve got Nancy Youssef in my Signal chat!

Jordan Schneider: Bryan, talk us through the munitions piece. Where are we at?

The Magazine Is a Rumor

Bryan Clark: We’ve been at war for six months. And because we didn’t want to put forces at risk — because we didn’t really get buy-in from everybody that this war was a good idea and how to execute it — we’ve been using our high-end munitions. Long-range munitions like Tomahawk, LRASM, and JASSM-ER. They let us keep our ships and aircraft farther from the target and keep everybody safer.

Those munitions were designed to go after the Chinese threat, because we anticipated some big war with China in the future. We’d been stockpiling them to be prepared for that. And now we’ve expended almost all of them in Iran.

Tomahawks, for example — in the first couple months of the war we shot 850, out of a projected inventory of about 4,000 a couple of years ago. We used some against the Houthis, and then another 850 early in this war. We’re probably down to a couple thousand or less. Which means, if the fleet includes 100 surface combatants that carry 30 or so Tomahawks apiece, you don’t even have enough to fill out the fleet we currently have — much less any reloads. And that’s if you use every weapon in the inventory.

And then JASSMs — the Joint Air-to-Surface Standoff Missile — and LRASMs, the long-range anti-ship variant. They are both essentially at zero inventory.

Justin McIntosh: The Army’s launched essentially all of its ATACMS. And we’ve fired almost all the PrSMs.

Bryan Clark: Right. So we’ve expended the inventory of all these preferred long-range munitions, because we’ve been trying to keep US forces out of harm’s way, and because the Iranians have taken advantage of precision weapon technology and air defense systems. It’s no longer an easy proposition to go in and drop bombs on them.

So we’re at the point where we don’t have any of these standoff munitions. Which means not only can we not continue the fight against Iran at the scale the president wants — we also are not prepared for any conflict elsewhere, especially against China.

Jordan Schneider: It’s not “continue the conflict.” It’s continue the conflict without risking planes.

Bryan Clark: Correct. Good point.

Justin McIntosh: And the idea that they didn’t know this was going to happen is part of the issue. A couple years ago, when Kurilla was the CENTCOM commander and he — I think the preferred term is yeeted — thirty Tomahawks at the Houthis, the reaction inside the department was, “I cannot believe we just used 30 Tomahawks against the Houthis.” I was in the military then, and people were saying, “What is INDOPACOM thinking?” Because you’re throwing them away against people who have no air defenses. You could have just sent a sortie of F-18s and had the exact same impact.

So they knew this was already a problem. The production already wasn’t there, and we didn’t have the ability to build them faster.

But the risk tolerance is where it gets interesting. You saw it play out last week, when CENTCOM’s commander comes out and says, “I think we’re basically done here, guys, unless we’re willing to do something else — and it’s going to look a lot different.” And a lot different is a lot different. It’s a buildup of troops. It’s more targets for errant drones to strike.

You’d Have to Fly Right Over

Bryan Clark: You could continue the strike campaign the way they’ve been doing it, but you’d have to use gravity weapons. JSOWs, which have a little longer range, or glide weapons — JDAMs and JSOWs that let you be fifty to a hundred miles from the target, maybe. Which seems like a long way, but the Iranians still have air defenses that reach out that far. And nobody wants to put pilots at risk of getting shot down in a war that’s already unpopular.

It’s not a matter of “we can’t continue it.” We’ve got plenty of JDAMs and JSOWs and other gravity weapons. It’s that we don’t want to expose the forces necessary to deploy those.

Which is interesting to think about, because that’s against Iran. Every other adversary we’re going to face is going to be as capable or better, because they’ve all got the same technology. So we’re at the point where these weapons are almost not even relevant to modern warfare. Every fight is going to require some kind of standoff munition — which means we need a different portfolio. A bunch of cheap standoff munitions, as opposed to only having the exquisite ones and then the cheap gravity bombs.

The other thing is Iran is using ballistic missiles for a lot more of these attacks — same as Russia — because only certain weapons can shoot them down. The Navy has plenty of SM-2 interceptors. Thirty-year-old technology, not even in production, and they’re really good at shooting down cruise missiles and drones. But they can’t shoot down ballistic missiles.

So as Iran has shifted to ballistic missiles, they’re launching them at the Lincoln and the George H.W. Bush strike groups, and we’re shooting them down with SM-6s, or potentially PAC-3s. That’s burned through our inventory of SM-6s and PAC-3s to where we essentially don’t have any left.

And you’ve got places exposed now to ballistic missile attack, like in Jordan, because the Iranians recognized this opportunity and said, “We need to shift our weapons portfolio to ballistic missiles like the Russians, because that’s the weak point.” Because we don’t have a cheap way of shooting down even crappy ballistic missiles, just because they’re fast.

Justin McIntosh: Eventually defense production will get to the point where you have a solution. The problem is we don’t have it yet, and there’s a reticence on the part of the department and the administration to commit the resources.

Because every week it’s a signal that it’s done. The deal is done. We’ve blown them into oblivion. They’re really just looking for a deal. Then Friday comes around, markets close — we’re back on with the war. Sunday comes around, markets about to open — we’re good again, the peace plan’s almost closed. And that ebb and flow has a real negative impact on the department’s willingness to start spending.

A Short List of Leakers

Jordan Schneider: This morning the Washington Post has a piece. “President Donald Trump’s frustration over the Iran war boiled over at Camp David last week, where he demanded answers from Defense Secretary Pete Hegseth on why he had apparently been misled on extreme munitions shortages that now threaten to limit military options with Iran.” And then: “The president asserted overnight that the United States had massive amounts of munitions, and that the leakers of these treasonous statements are being hunted down.”

Justin McIntosh: So at least two of them are conceivably at Camp David this weekend. Should be a short list of potential leakers.

Bryan Clark: It’s like when Jonathan Swan and Maggie Haberman had all that info from the Situation Room, and everyone’s asking, “Well, who could have leaked it?” Somebody who’s in the Situation Room.

Justin McIntosh: One of twelve people.

Bryan Clark: Shouldn’t take too long to hunt them down. This isn’t going to be a long-term prosecution.

Justin McIntosh: What they call leaks has become so overused it doesn’t actually mean anything. Obviously reporters have people at the Pentagon they talk to. But I’d argue this is general officers and flag officers — admirals — expressing their concern in the only manner they have.

Because there are two things at play. We do have a ton of munitions. They’re not the munitions we need to run the risk profile we’re trying to run. If you go tell the boss, “Hey, we’ve got all kinds of bombs, sir,” you’re fine. But if you dig down and say, “To prosecute this conflict the way we want to, we need these exquisite technologies” — we don’t have those.

Bryan Clark: The interesting part of that article was the implication that Hegseth threw Feinberg, the deputy secretary of war, under the bus. “This was his responsibility to fix.” Like it’s something you could just sort of fix.

Steve Feinberg has been working on the munitions problem. But it’s going to take years to restore that capacity, mostly because of the supply chain and all the bespoke components you have to get from a specific manufacturer that take a while to spin up. There’s no way you fix that quickly. What Feinberg’s been trying to do is foster a different set of munitions that are cheaper and easier to build at scale. That’s going to take a while too.

Jordan Schneider: Feinberg’s actually doing work, and he’s your number two. You can’t blame your number two for the war that you were the strongest advocate for.

Justin McIntosh: The buck stops over there, Jordan. The buck stops anywhere but here. That’s what the sign on the desk says. The buck is over there. Somewhere else. I’ve got places to be.

Jordan Schneider: Didn’t you say you were on this? What have you been doing for the past year and a half?

Justin McIntosh: We’re on it. I’ve put my best man on it. And if he messes up, I’ll fire him.

Bryan Clark: And then there’s nobody behind it, but that’s fine. The poor deputy secretary, who normally manages the entire department, has three or four things he has to be the action officer for, because they don’t trust anybody else to do it. That seems like the most terrible job in the Pentagon right now — amongst many terrible jobs.

Justin McIntosh: It’s interesting, because look back at Ash Carter. Ash Carter was the Secretary of Defense. And he’s the one who said, “We’re going to fix this, and I’m going to personally helm it.” That is a demonstrative difference. Feinberg is trying to address: how do we make drones, how do we reindustrialize, how do we do shipbuilding, how do we get VC money and private equity into defense — and by the way, also fix these legacy production issues. He can’t do all of it.

Jordan Schneider: Here’s the thing. We’ve had so much lethality porn. I’m really glad you’re getting testosterone for everyone. How many bombs does that testosterone net out to? I haven’t quite seen the equation there. Okay, you want to talk about lethality, fine. What is the closer tie to ammunition — to the things that blow people up?

Bryan Clark: You know what’s very lethal? Ammunition seems lethal.

Justin McIntosh: The only way the testosterone helps at this point is if it makes them more tolerant of risk and they’re willing to put planes in harm’s way because they get super aggro. But that’s really it.

We Called This on Day Two

Justin McIntosh: To take one more step back — the hardest part of all of this for me is that all of it was knowable from the start. I think, Jordan, you said it: we called this on day two. We’re almost six months in and it looks exactly like that.

We don’t have to look at Ukraine. Look at the Houthis versus Saudi Arabia. Watch how the Houthis systematically figured out where the overlapping fields of fire were on those Patriot batteries, then figured out how to get past them. Blew up Patriot batteries and landed short-range missiles and drones in Riyadh. Who was advising the Houthis? None of this is a surprise if you just looked at the region. And yet now it’s “we had no idea it was going to turn out this way.” Because you didn’t ask anybody.

Bryan Clark: I think this will have the perverse effect of actually allowing the pivot to Asia — because at the end of this we will have lost most of our basing in the region. And based on people I talk to in the department, we’re not going back. We’re just going to leave it. Which means anything we do in the Middle East we do from sea or from the air, and we have to reduce our expectations about how much we can intervene.

We’ve war-gamed doing more from sea, and you really are constrained. So the shipping defense mission, whether in the Persian Gulf or the Red Sea, falls to allies. The US can do some of it, but allies have to do a lot more — because they have the relationships that let them pull into port and get gas without worrying about getting attacked.

Justin McIntosh: That goes to one of my big fears about the pivot: the difference in the long game. At least since 2001, and arguably since 1979, the Iranian regime’s number one focus has been reducing the United States footprint in the Middle East. They played a very long game.

And now we’re legitimately talking about — in the newest agreement, potentially giving them not toll privileges, but knowledge of all ships transiting the strait. Reduced American footprint, less American military in the region, fewer ties to allies, Iran with tacit control of a strategic chokepoint. The regime couldn’t have asked for a better outcome.

Jordan Schneider: They took their shot on day one — the decapitation. Maybe these guys will just roll over. That’s a 10% chance, and it didn’t happen. Then over the next two months we weren’t able to open the strait, which was 90% likely not going to happen.

So this new round is the most puzzling of all of it. I can’t imagine anyone was saying, “Yeah, we’re going to change the balance of forces around the strait such that they won’t be able to operate the toll.” It must have been presidential pique. I’m sure the political advisors were saying, “Whatever, it’s a news story, it goes away once gas comes down.” But it became personal. And Hegseth apparently told him the ammunition was fine. So sure, why not have another round?

Justin McIntosh: You’re forgetting the law though, Jordan. If you make a negative EV bet, you keep making it. Because eventually you’re going to hit black and you’re going to be good.

Bryan Clark: We’re losing money every time, but we’ll make up for it at scale.

Justin McIntosh: Conceivably that’s how you have casinos go bankrupt.

Pivot to Asia

Jordan Schneider: All right. Pivot to Asia.

Bryan Clark: Now that we’re reducing our footprint in the Middle East, the US will finally be able to pivot to Asia. Actually we are — the Coast Guard ships and the LCSs that were in the Middle East have already relocated to Singapore, and they’re going to stay.

But at the same time, two of our Asian allies are developing their own new strategies, because they recognize that maybe we’re not quite as dependable — or as capable — an ally as they expected. Japan and Australia are both revising their defense strategies.

Japan just released a preview of theirs. In previous versions they’d say, “There are lots of bad people out there, we need to deter the bad people.” Now they just call out China and say: China, China, China. They’re bad, they’re revisionist. The threat to Taiwan is a threat to Japan. They also call out Russia, and — this is new — how the two of them collaborate.

And on what to do about it, they really hit the unmanned systems hard. What you could simplify as the porcupine strategy: one-way attack drones launched from shore and maybe from ships, going out to attack Chinese ships and aircraft at some distance, creating an area of denial around Japan.

Which is a big shift. Japan has traditionally depended on the US to defend Japan, with Japanese forces defending the bases US forces operate from. That’s been the bargain. This new strategy makes clear that Japan is looking to defend itself.

That’s come out in our war gaming with the Japanese too. They’re shifting to a model where Japan has to defend itself for a while. One, maybe the US is tied up with the Chinese somewhere else. Two, maybe the Americans aren’t quite as energetic about defending us as we’d hoped. And three, maybe the US just doesn’t have the wherewithal — because they can see in Iran that maybe we can’t do it all alone.

So they’re focused on: how do we protract a conflict with China long enough that either they give up, or the US decides to come in on our side?

Justin McIntosh: Does that feed the movements in Okinawa to push out American basing? Because if the US is just going to come in later and we have to wait on them anyway, why are we taking on the risk — which the locals absolutely see as a risk — of having the US staged here?

Bryan Clark: Definitely part of it. But the political winds down in the Southwest Islands are changing. From Okinawa south to Yonaguni and Ishigaki, they traditionally didn’t want the military there at all, US or Japanese. Now they see Chinese ships routinely driving around in the contiguous zone, and the maritime militia harassing fishing vessels. They’re more interested in Japanese military presence.

So on Ishigaki they’ve put in missile batteries, and the Ground Self-Defense Force is building up facilities. It’s only about 90 miles from Taiwan.

They’re still resistant to a bunch of Marines based in Okinawa — they behave badly, and there’s jet noise. But forces that look designed around self-defense, protecting the Southwest Islands themselves rather than using Okinawa as a jumping-off point for Taiwan — they seem much more accepting of that.

Justin McIntosh: The one I’m not seeing a change in posture on is South Korea. They went through the coup, or whatever it was, and they have a somewhat more PRC-aligned government. Eighth Army’s there, there’s the whole Korea command. If Australia is becoming “we’re going to have to defend ourselves” and Japan is becoming “we may have to go alone” — what’s the Korean response?

Bryan Clark: Our discussions suggest they still want to be a bystander in any confrontation over Japan. They’d rather just stay out of it. The latest SOFA update gives the US more flexibility to redeploy forces out of Korea, which had been anathema — so there’s potential to move aircraft out to support a conflict in Japan. But the Koreans want out of it. They really want to play it both ways: be part of China’s supply chain while being an arms supplier to the US and the Europeans. Great place to be if you can walk that tightrope. I think it’s going to be difficult.

A Tiny Military and a Strategy of Denial

Bryan Clark: Australia is totally different from Japan. Japan is increasing defense spending about twenty-five percent over the next few years, a lot of it going to uncrewed and autonomous systems, new technology, space.

Australia is not raising its defense spending. Their investment plan says they are, but it hasn’t materialized — except for the AUKUS submarines, which they’ve had to spend a lot on to basically pay off the US to stay in the program. Most of their increase has gone to AUKUS.

Their strategy relies on a strategy of denial, which sort of doesn’t make sense. If you’re Australia and you have a tiny military, I don’t know what you’re denying or how.

In reality they depend on being far away, and on being tangential to whatever China is doing elsewhere. What China would actually do is threaten attacks on the periphery, or threaten the shipping lanes that connect Australia to the world. So they see the submarines as the counter: use them to attack Chinese ships interfering with shipping, or attack the South China Sea islands as retaliation. If China hits northern Australia with ballistic missiles, Australia hits the Spratlys with cruise missiles from submarines.

But the strategy still doesn’t talk about using the submarines offensively against China. They still don’t name China as the threat. It’s “threat actors” and “aggressors” and “there are bad people out there.”

Justin McIntosh: Indo-Pacific threat actors.

It’s interesting you bring up AUKUS, because that was a complaint I heard when I was there for ten days a few months ago, going around defense contractors and spending time with Australian defense forces. There’s Pillar 1, the subs, and Pillar 2, basically everything else. They just said AUKUS has sucked all the oxygen out of the room, and all of the excess money.

And they’ve made some really bad legacy decisions, because they prepared to be a peacetime military. They don’t have a lot of hardened systems. A lot of their navigation isn’t hardened — I wouldn’t even call it military grade, it’s commercial at best. Basically a Garmin mounted on your dash: “I know where I’m going, there’s no issues here.”

Bryan Clark: Hey — we had that on one of my submarines. It works okay.

Justin McIntosh: Does it get signal under the water?

Bryan Clark: You get it when you come up.

Justin McIntosh: So beyond “can you man them, can you maintain them” — at what point does adding two of the same thing to a coalition become a subtraction? Because now you have twice the force to maintain with one logistics base.

Bryan Clark: From the US perspective AUKUS is a great deal. Australia spends money on our submarine industrial base, which needs help. They’re building a maintenance facility at Henderson and have built out the pier infrastructure so we can base submarines out of Australia, as part of Submarine Rotational Force–West. If you have a submarine out there, it gives you twice the operational availability of having it back in CONUS. And boats needing repairs don’t have to go back to Guam or Hawaii. The Australians are paying for all of it.

From the Australian perspective it’s maybe not such a great deal, because it’s going to consume a lot of the funding that would otherwise go to the rest of their military. They’ll have submarines, a relatively small fighter wing, a couple of tankers, a few AWACS, and a small army for border protection in the neighboring islands. And that’s it.

So if the Chinese send the maritime militia down to surround Christmas Island and say “we’re treating this as a new zone for us to operate in” — the Australians have a couple of ships they can send, readiness is not high, and it would consume their entire fleet to do that one operation. They’d have nothing left for shipping protection.

So it’s a tale of two cities. Australia is setting itself up to be dependent on the US for its defense because of AUKUS — while expressing concerns about the willingness and ability of the US to provide that defense.

And the US is depending on Australian basing being available and defendable if we get in a fight with China, because it’s the only place you can operate from at a survivable range — Japan will be under threat the whole time. Yet nobody is investing in the infrastructure or defenses that basing out of Australia in a war with China would actually require.

Jordan Schneider: It would really be something if the paragraph the Iran war gets in the history books is: this actually helped prepare the US for World War Three, marginally better.

Justin McIntosh: Jokes on you — this was five-dimensional chess. Time is a flat circle, Jordan.

Alliance Theater

Justin McIntosh: So how do we now show our allies and treaty partners that we’re going to defend them? In Asia the problem is: what if the US is unreliable? Go back to COVID, when Australia forcefully called for actions against China and the US didn’t back them up. Then things get said about how Taiwan’s a long way away, Japan’s a long way away. Even taking troops out of Korea, which was sacrosanct. When you’re already signaling those things, what’s the lever the US has left to keep allies committed?

Bryan Clark: The Trump administration isn’t going to have rhetoric implying unconditional, non-transactional support. But the basing question is a good one. If we have Submarine Rotational Force–West at Stirling, near Perth — submarines there, sailors there, maintenance infrastructure that’s partly ours — we’re going to defend it. You can’t have a submarine base come under attack without making some effort to protect it. Same with the cutters and LCSs in Singapore, and everything we’ve got in Japan.

And the allies draw reassurance from that. If nothing else, they’re going to protect themselves. And you can’t redeploy a submarine base. It’s not like you can fly a bunch of troops out and you’re gone, the way you can in the Middle East.

Justin McIntosh: You can try.

Bryan Clark: You can try. But short of just openly saying “we’ll defend these allies unconditionally” — which needs to be said periodically — how do you demonstrate it? Normally with exercises. We just did RIMPAC.

Justin McIntosh: Talisman Sabre also just happened.

Jordan Schneider: I’ve got to say — the Wall Street Journal did a fifteen-minute documentary, “We’re hanging out with RIMPAC.” The boats are really close together, and then they did some land invasion thing and the troops were all really close together, like playing paintball. Didn’t look great to me.

Bryan Clark: You’re bringing up a good point. RIMPAC in a lot of ways is alliance theater. It’s not intended to be a high-end exercise where we train like we’re going to fight. It’s more: can we practice how we talk to each other, how we land helicopters on each other’s ships, how we’d put troops ashore — because all our equipment is different. The operations are very JV-level.

There’s also an opportunity to test things out. The allies bring, “Hey, we’ve got this new missile, we want to blow something up, but we don’t have anything to blow up. Do you guys have something?” So they took out the Peleliu, an old amphibious ship — which apparently we had not burned at the pier, a miracle given previous maintenance cycles, like the Bonhomme Richard. They shot a bunch of stuff at it and sank it.

So that’s what RIMPAC is for. Test out our stuff, show off to each other, do some combined exercises, and then maybe play a volleyball match and do some cultural exchanges.

Justin McIntosh: If you’re wondering, cultural exchanges in the military basically involve drinking a lot with other people.

Bryan Clark: Which can be the most dangerous military operation.

Justin McIntosh: That’s honestly why I don’t like using RIMPAC to bring out new tech — pick the company, Anduril or whatever — to test it. Because it is so canned they can make it look really impressive. There’s a notional north, and a notional enemy, and notional jamming going on over here. And at the end they’re like, “We have an operational deployment.” It’s like — you have a vacation. Congratulations. Enjoy Hawaii.

There are places they do really train. The Australian outback is so barren and unpopulated you can do real electromagnetic spectrum jamming in ways you can’t elsewhere. In a country the size of the United States with twenty-five million people in it, you can basically jam everything.

Bryan Clark: And Australia’s defense strategy says as much — their contribution to the alliance is the basing contribution. Bare bases you can roll in on, the vast test range in central Australia, the over-the-horizon radars with JORN. Capabilities that didn’t cost much in the grand scheme, but geography makes them useful. That’s how they frame the fact that they haven’t got a big military but feel like an equal partner.

Justin McIntosh: Honestly, they’re our most reliable ally. Except for Iran — they’ve been at everything we’ve asked. So maybe that’s a signal too. They weren’t there for that.

Manned-Unmanned Teaming Is a Sham

Jordan Schneider: Any war gaming stories, Bryan?

Bryan Clark: We did a war game with senior Navy leaders a couple weeks ago at the War College, looking at the Navy’s new force design.

A lot of the discussion came down to this: if you’re realistic about the money the military is going to get, you’re not going to be able to grow the services beyond their current size. So if you want to address a harder threat environment, you can’t just build more of the same, because you can’t afford to keep it. The manpower and maintenance costs are what kill you.

So the services are looking at tailored forces — units that are mostly autonomous systems with some human operators, maybe a ship or an aircraft squadron. That’s the front edge of the fight. You push them forward and they take the brunt of the initial attacks, whether that’s the missile threat from Iran, ASW against the Russians in the GIUK gap, defending Taiwan, or defending the Southwest Islands.

And the problem we ran into fast is that it takes a lot of space, and the people that go with it have to go somewhere. There’s a reason we haven’t done this before: the footprint and the headcount and the cost are high. You’re talking billions a year to have one tailored force sitting in a specific spot waiting for a specific mission.

Whereas the traditional military spends billions to build a ship or a squadron, and then it goes out when it’s needed and comes home when it isn’t. There was a reason we had a rotational model. And you can’t do that with the robots, because they’re not fast enough, or don’t have the range, or aren’t self-deploying.

The other thing we found: manned-unmanned teaming doesn’t work. It’s a dumb idea and it doesn’t play out. There are very narrow use cases where robots and people in the same space makes sense. Because the robots are valuable in part because you can let them die. But you can’t do that if they’re near the people, because then the people might die too. So now I’ve got to defend the robot — which seems dumb.

And the robots can’t go as fast or as far. Either they slow the people down, or they need gas more often so you’re constantly stopping. You suboptimize one or the other.

Jordan Schneider: Isn’t it just leverage on your suicide missions, though? Only one pilot instead of five doesn’t make it back.

Bryan Clark: That’s what we talked about — one pilot with a bunch of CCAs, collaborative combat aircraft, flying downrange. We lose that one guy, but he does a lot more damage because the robots are doing the damage. And the problem is the robots are more detectable than the stealth airplane that’s going with them. So if you’re in an F-35, these robots are attracting attention to you, and you may not even make it to the target.

Justin McIntosh: So you have to build an F-35 and then put a robot in it — at which point you’re spending F-35 money.

Bryan Clark: It’s like, yeah, we just built an F-35. That’s a robot one. But if you send all the robots downrange together, a bunch get shot down, and you can make them cheap enough that some of them make it.

Who Goes First Through the Door

Justin McIntosh: Was there any ethics discussion? I saw a post from one of the tech guys: “We’re not talking about robots shooting first, we’re talking about robots being the first in the door.” And I was like — to do what? A human wheel?

Bryan Clark: Great question. The Japanese brought this up in our war games, and we hadn’t even thought of it as a concern. They said: we send our robot force out as the outer perimeter. The Chinese come to attack. If they attack the robot, do we attack the Chinese manned ship in response? And we said — it’s a war game, you decide.

Teams chose differently. Some waited until the manned stuff was attacked. Some retaliated as soon as the unmanned perimeter was hit. And the outcomes really didn’t differ much. If you waited, you didn’t do much worse than the other team. That was an interesting insight.

But it’s an open question. If they attack the thing whose job is to be the front line, do you retaliate — when the whole point of putting it out there was to be a tripwire, a roadblock, a speed bump?

Justin McIntosh: That’s what I came away with from that post. It sounded like someone who had never actually cleared a room. Because if he had, he’d know the first man in the door isn’t the one most likely to get shot. It’s the second man. So you’re putting the human in the more dangerous spot if the robot can’t shoot until a human tells it to.

Bryan Clark: Which is why, if you have none of your own people there — this is the hellscape idea. You make it a mobile minefield. You call it a maritime exclusion zone. It’s inside your ally’s territorial water and airspace, you say you can’t come in here or you’re likely to get attacked, and they all basically act like mines. No friendlies. That was the only version everybody felt was ethically defendable and operationally useful.

It’s another reason manned-unmanned teaming didn’t work: you’ve got all this fratricide risk, and you’ve got to trust the robot sees “I’ve got Blue Force Tracker, don’t shoot me” — and we all know how that works.

We Just Built an F-35. That’s a Robot One.

Jordan Schneider: Can we take a step back? How much money has the government spent on this?

Bryan Clark: The CCA program is a great example. The aspiration was that these unmanned aircraft fly along with the piloted aircraft like the F-47 and act as loyal wingmen and adjunct magazines — like your dogs out attacking the foxes.

And then you’ve got to build them as stealthy as the thing they’re flying with, or they attract attention. And how are they finding targets? If passively, I need a passive sensing apparatus on par with the F-47. So how expensive is this thing going to be? And do they fly from the same base? Then they need a runway, and long range, and that’s more expense.

This is why the Air Force started a separate program for the next increment, looking at models where the CCAs fly from a different place and don’t really operate with the manned aircraft. They might say, “I’m way out here, I saw some stuff, here’s what you need to know,” and the manned aircraft just receives it.

But at that point it’s not manned-unmanned teaming anymore. It’s just a kill chain. So I think that’s where we’re going: you decouple the manned and the unmanned enough that the liabilities of each don’t affect the other’s performance.

Justin McIntosh: I blame the Jamie Foxx movie Stealth for this. People watch that and think it’ll be the exact same plane. If it’s the exact same plane, the cost saving is one pilot.

Bryan Clark: It’s not really about cost, either. It’s about how many you can buy. That’s why we have the munition shortage — not because the munitions couldn’t be built, but because they’re expensive. As you drive up the cost of unmanned systems, you limit how many you buy, which means when you get to the fight, how many can you afford to lose?

Justin McIntosh: If you ripped the pilot out of an F-35, that’s still billions of dollars of aircraft.

Bryan Clark: It’s worth having the ability to make an F-35 autonomous for some really niche mission where you’ve got to have something super stealthy and capable. But those are niche. That’s probably not where the collaborative combat aircraft program should devote its attention.

And that gets to one of the edge cases where manned-unmanned teaming did work: decoys. You don’t want the decoy hanging around while you’re trying to hide. But once you get shot at, the missile has a seeker on it — and then you want a decoy that takes it and drives it in a different direction than your ship.

So you punch out a bunch of long-endurance decoys. Same with aircraft. That ended up being a manned-unmanned teaming use case that in these war games was really valuable. It paid off big time — particularly in the Western Pacific, where you’re getting a hundred-plus missiles per ship.

But it’s got to be small enough to deploy from something else, and long enough endurance to handle multiple salvos. It can’t be a ninety-second decoy that handles one missile.

Justin McIntosh: I had the idea of it being like one of those just-add-water foam things you throw in and it flows up.

Bryan Clark: And then you have a leak somewhere on the ship, and it opens up, and hilarity ensues.

Justin McIntosh: These get at what happens when you don’t marry the technologist with the actual military planning early enough. Look at IVAS. You can tell that at no point during the design was that ever designed by somebody who had served.

Bryan Clark: “Put that on your head and run around with it.”

Justin McIntosh: With these big cables hanging off you. A commander of mine, Joe Wortham, used to say: you’ve got to talk to the person closest to the problem to figure out what the solution is. When CCA was a novel concept it got all these designs, and it was never done with the actual intended use case in mind. It was just: we can make a plane that’s a robot. Cool.

Bryan Clark: That’s the problem with a lot of VC money. They’ll fund somebody’s idea, which sounds interesting and maybe even feasible, and just run with it without ever doing concept development with some operators to get a sense of what’s useful.

Justin McIntosh: Which is why some of the funds are starting to say you need advisors who show this is a real thing. The problem is finding advisors in it for more than their equity.

Jordan Schneider: “Will we laugh at your idea five years after you build it?”

For what it’s worth, ChatGPT says we’ve only spent two billion dollars on these CCAs so far. So — two billion between friends.

Bryan Clark: It’s not that much to figure out that it’s a bad idea. It’s a small price to pay.

Justin McIntosh: You think about how much we spent on iterations of the F-35. The difference is they didn’t start with legislation and promise to build CCAs in all fifty states — because then it would be unkillable.

WarTalk August Recess

Jordan Schneider: We’re off for August recess for the next few weeks. What should folks be reading until we’re back in September?

Justin McIntosh: I’m in the middle of Forgotten Ally, an older look at China and the US relationship, and balancing it with The Gunpowder Age. I always recommend going back to history as much as you can. Not the best for the beach, but worthwhile — especially The Gunpowder Age. One of the questions it addresses is why Europe and not China developed the large cannons and weapons. Which is interesting when you look at warfare now: are there going to be regional differences in the application of these new technologies, or does it all become ubiquitous because the globe is only so big?

Bryan Clark: I’m reading The Propensity of Things by François Jullien, on Chinese strategic culture — which I read part of for graduate school but didn’t finish like I was supposed to. Tom Mahnken, hopefully, doesn’t listen to this, so he won’t know I didn’t read it when I said I did. It gets at the foundations of Chinese strategic culture, why the CCP makes decisions the way it does, drawing on deeper Chinese history. It’s a French translation, so it’s a little clutchy in places, but it’s good.

Jordan Schneider: I’ve got a movie. It’s called Good News, a Netflix South Korea–Japan collaboration. A 1970s political satire, dark comedy. A plane gets hijacked by Japanese communists who want to fly to Pyongyang, and the repeat failures of the American, Japanese and South Korean defense and intelligence establishments just pile up on top of one another. A lot of themes we’ve covered on War Talk — passing the buck being a main leitmotif, perhaps even of this episode. I do not think there is anyone who has listened to the end of this podcast for whom this movie would not bring a smile to their face.

Happy August, everyone. No new wars, okay? Everyone just hold off three weeks.

Justin McIntosh: At least two and a half.

Adam Chan is national security counsel and senior advisor to FCC Chairman Brendan Carr, and directs the agency’s new Council on National Security. He previously clerked on the Second Circuit and served as a national security legal fellow for the House Select Committee on China. ChinaTalk’s Aqib Zakaria co-hosts.

We discuss…

• How the covered list works

• The case for “simple and stupid” rules over big-brained ones with endless carve-outs

• Onshoring plans, allied carve-outs, and the 65% domestic content test

• Why robot vacuums got swept in alongside humanoids

• How AGI-pilled you have to be for the robot rule to make sense

Listen now on your favorite podcast app.

GovAI was founded ten years ago on the belief that AI would end up transforming our world. Ten years later, the organization is at the forefront of trying to help decision-makers in government and industry navigate the transition to advanced AI.

GovAI is now hiring for its next cohort of Research Scholars and Research Fellows. Scholars are one-year positions that provide an intellectual home for ambitious researchers and policy professionals looking to advance their careers or transition into AI policy. Fellows are longer-term positions for experienced researchers with an established track-record looking for a place that supports their most impactful work.

Both roles offer significant freedom to pursue policy research, advise decision-makers, or launch new initiatives.

Applications close 16 August 2026. Learn more and apply at https://www.governance.ai/opportunities

Jordan Schneider: This national security FCC energy dates back to its founding, no?

Adam Chan: The Communications Act of 1934, section one, establishes the FCC. One of two purposes is for the purpose of national defense. Right there, from the beginning, there was a national security focus in the FCC’s mandate. That mandate is still there, and you’ve seen over the last decade a growing role for the FCC in national security.

For Chairman Carr, national security has been his priority for years, and he’s brought that focus to the chairmanship for the first time. He brought me on as a kind of unique person within the chairman’s office to focus on and coordinate the national security side of things. We’ve set up this new Council on National Security within the FCC to coordinate a lot of the national security work the Commission is doing across multiple different bureaus and offices.

Jordan Schneider: It’s cool, because in the 1930s it was very much a forward-looking thing to see that radios are the future of warfare. And of course World War II comes, you get the Battle of the Beams, you have war radio that enables battlefield operations at a wide scale, which weren’t necessarily possible over the course of World War I. And here we are in 2026 with the FCC on the bleeding edge of technology talking to us about robots and drones.

Let’s start with robots, since this was the latest thing you guys did. Maybe walk through the timeline of how something like this gets on the FCC’s radar as a topic that may need more aggressive regulation, and how something like this evolves, before we talk through the specifics of what’s actually in the regulation.

Adam Chan: I guess I’ll wind the clock back probably further than you want me to wind the clock back —

Jordan Schneider: Hey, I just brought us back to World War I. You can go as far back as you want.

Adam Chan: Fair! So 2020 and 2021 are two important years. Congress passed two standalone pieces of legislation. For the good-governance folks out there who are constantly saying Congress has to do more, why isn’t Congress doing its job — FCC authorities represent an example where Congress, in overwhelmingly bipartisan fashion, with just a handful of no votes in the House, passed the Secure and Trusted Communications Networks Act in 2020 and then the Secure Equipment Act in 2021, which established this new regulatory framework known as the FCC’s covered list.

The way that works is we have a covered list of equipment and services, established under the Secure and Trusted Communications Networks Act. In that mechanism, the FCC can’t actually update our own covered list. This is different than, say, OFAC sanctions lists, or BIS and their entity list, or the 1260H list of military companies at the Department of War. Instead, we require another national security agency or interagency body to make a determination that is the legal predicate by which the FCC then updates the covered list and enforces the rules.

One of the primary implications of a device being on the covered list is that it can’t get authorized for import or sale in the United States. A lot of people know the FCC for its regulation of spectrum, or maybe broadcast media, but also virtually all electronic devices — any device that emits radio frequency, which is pretty much all electronic devices, needs to get FCC authorization before being imported, marketed, or sold in the United States. You’ve probably seen a little FCC logo on a lot of your devices, but that also gives us the tool to shut off new models of devices from coming into the country.

So that’s the broad overview of the FCC’s covered list and what the tools are. You’ve seen the covered list used a lot over the last few years. The most recent action, just last week, was on advanced robotic devices — think mobile robots generally — and power inverters. On the mobile robots determination in particular, this is an interesting scenario, because this is a pretty nascent industry. There’s a pretty good justification that we’re getting ahead of where things might be, say, five years down the line.

The covered list arose back in 2019, 2020, where the FCC was confronting a problem in domestic telecom networks. We were almost entirely reliant on Huawei or ZTE equipment in this radio access network telecom equipment, and had to do this big rip-and-replace to swap untrusted for trusted vendors in our telecom networks. There’s a lot of thinking in the US government about how we can avoid that situation playing out in critical industries like robotics — which, I mean, you guys know better than I do, but I think a lot of people think will be a pretty major industry in the coming years and decades.

Getting Ahead of the Humanoids

Jordan Schneider: Look, we’ve done stuff on undersea cables — established, very big deal. Telecom, obviously, a known big quantity for a while. Drones also, something which has a clear, demonstrated industry and impact. The type of robotics you guys are targeting, people think it’s the future, but it’s much more of a forward-looking speculative bet than the ones you ran through over the past six years. So what drove the impetus to try to restrict the sort of robots that the US is going to be able to import in the coming years?

Adam Chan: First, as mentioned, ultimately we implemented a determination that was made by the broader executive branch. The FCC did not write the national security determination — we’re not the ones who, as a legal matter, chose this. But if you look at the determination, there are two national security risks that are highlighted.

One relates to remote access, cybersecurity, surveillance, and the ability to remotely manipulate a device such that it poses threats. You guys have probably seen videos of robots behaving badly, but more seriously, I think the national security threats here are kind of obvious.

The other is that this is something where the US has a nascent industry but is overwhelmingly reliant on foreign imports of devices. There’s a real risk of a supply chain dependence that, as these devices become critical to the industries of the future, we could end up in a situation where we are very dependent on foreign sources. Maybe these devices are critical to military applications, maybe to manufacturing and other industrial sectors, in which case it is very important for the US to have its own supply.

The covered list automatically means we ban new models getting approved for import or sale. So the covered list works best in fast-evolving industries, where probably the robot of today is not really going to be on the market in five years. There’s a pretty fast obsolescence cycle relative to many other industries, just because the cutting edge is evolving so quickly.

Jordan Schneider: You mentioned the national security risk of someone you don’t like taking control of the robot and, I don’t know, using it to whack people or power plants. What’s up with the two-kilogram exception? I’ve seen some really scary demos of little DARPA insect robots. Why should we not be worried about those coming in from China?

Adam Chan: No comment on the specific DARPA insect robots. I think, looking at what the main market segments are, and what poses the most acute national security risks — I would also note, a good thing about the covered list is that people often complain, hey, this is a blunt instrument, there wasn’t a long comment cycle period, there are more nuanced ways to go about things. There’s actually a strong argument, in this case and in other examples of FCC policy, that the simple and stupid policy is the best policy. It allows for fast action, we avoid the whack-a-mole problem of naming specific entities or coming up with complex definitions and long transition periods, and we send a very strong market signal that gets the attention of industry, to hopefully crowd in capital and investment.

To your point that maybe DARPA-like insect robots become a big thing at some point in the future — it’s actually possible to quickly evolve what’s on the covered list through new determinations. There is a conditional approval process built into the robot determination, so the Department of War can review applications from robotics companies saying, hey, my robot’s produced abroad but doesn’t pose national security risks, I’m planning to bring production to the United States. But it’s also possible that further exceptions or expansions can be built in later.

I know we’re going to talk more about drones later, but for example, on drones there was a drone determination with a pretty broad definition of what a drone was. Later there was a new determination that basically exempted toy drones — a certain definition based on especially small drones that can’t really fly very far, designed to fly around in a home, like a wooden drone that a kid plays with. That was able to be actioned pretty quickly.

This potentially could evolve in the future if there are new national security threats, or we realize there’s actually a segment of the market that was not intended to be included. But here, the executive branch thought about the definition a lot and engaged with industry and other experts in the field, and this seemed to be a definition that would be well understood by industry while addressing the most acute problem.

Aqib Zakaria: That’s an interesting point about the blunt instrument. We’ve seen BIS deal with this whack-a-mole situation, especially with chips. I wonder how the FCC is handling it in terms of who actually decides this one’s okay, that one’s not okay. And is there a staffing problem that we’ve usually seen in the agencies, where there’s so much to handle and there’s not enough people to actually go through it? The policy works, it’s just hard to actually implement in practice.

Adam Chan: So far, at least, we took similar action — the broad-based ban on new models getting authorized for import or sale — for both drones and drone components as well as routers, and those have been ongoing for several months now. As of now, the Department of War has been taking the lead in reviewing applications for conditional approval. There is a guidance document there. Right now they have a pretty well-honed process where conditional approvals come out in just a few weeks, so it’s much faster than many other kinds of national security reviews in the interagency process. They have, to my knowledge, hired some people. They have a budget to handle this. So far it seems to be going quite well on that front. Obviously we assume we’ll get a bunch more applications based on last week’s action, but I’m optimistic things will continue to move smoothly.

Industrial Policy Is National Security?

Jordan Schneider: You mentioned things like market signal, and concerns about building a domestic ecosystem to replace these imports. What is the balance of motivation here between “we’re worried about a foreign actor turning on my robot to blow something up” versus “this is an industrial policy for an industry of the future we want to have a foothold in”?

Adam Chan: I really see that as a false dichotomy. It’s not necessarily the case that there’s a bucket of national security concerns and then a bucket of industrial policy concerns. Industrial policy really is critical to national security. Obviously there are real national security concerns on, as you say, the remote control of a robot, or other cybersecurity and data collection issues that are traditional national security concerns. But it also seems pretty obvious that in critical industries, especially industries of the future, the US needs to be able to produce them domestically and have a strong domestic industry.

To take drones, which is the most obvious and has been ongoing the longest of the recent FCC covered list actions — we have actually seen a lot of investment into drones. Over $5 billion since we’ve done it, over a million square feet of manufacturing capacity, thousands of jobs. It seems very obvious to me, watching current dynamics in ongoing military conflicts and the growing use of drones in the commercial economy, that if we want to be secure as a national security matter, we need to be able to produce drones and not be entirely dependent on foreign countries — often highly concentrated in one or several foreign countries — for all of our drone capacity.

Jordan Schneider: On robots and drones — Ukraine has not been able to create an industry that is isolated from Chinese parts, and they have had as much incentive as a nation could possibly have to do so. There are similar questions in the US with, maybe you can make them, but not particularly at scale. We’ve done plenty of coverage on ChinaTalk over the past few years about all of these rare earth dependencies and magnet dependencies and particular parts in the robot ecosystem, which basically are 99% from China, and they already have a giant-scale business.

So as you guys are making a call like this — there are worlds in which it turns out the choice is not between imported Chinese robots and domestic US robots, but between imported Chinese robots and basically not having a functional or competitive domestic robot ecosystem. How are you weighing the availability and supply compared to what the risks are on the other side, both from a national security and an industrial policy perspective?

Adam Chan: I would answer that in two ways. One is that the least expensive and easiest time to address a supply chain dependency is right now. It’s only going to get harder as these industries grow and as all the basic commercial incentives are to continue to entrench the dependency.

One reason the covered list is a good tool is that while we recently passed regulations to allow us to ban the import and sale of certain already-authorized models on the covered list, as a general rule it only affects new models — which means there is this transition period where it’s not a sudden shutoff, but rather a transition over several years that allows industry to adapt.

We’ve actually seen lessons that industry is pretty adaptable. Oil prices haven’t spiked to $500 a barrel because of the Strait of Hormuz. Markets tend to clear. I’ve been talking with any number of interesting startups in the drone space that are building out capacity to produce brushless motors, batteries, magnets at scale — that again didn’t come online one month after we took the action, but two, three years after we took the action, I think the US parts industry, a lot of these electronic parts that are critical to things like drones, things like robots, will really come alive.

We’ve seen something like routers. There were zero entities that produce routers in America, but as of now there are dozens of companies that have committed to expanding production of routers in this country. Some are going to break ground even later this year on new factories. So I do think this is something industry can adapt to when the market signal is there. And right now there is obviously demand for robots. We’re optimistic industry will be able to meet the demand and that US industrial supply lines will be more secure afterwards.

Jordan Schneider: My only take on this is that the political economy is such that you hear a lot more from the Boston Dynamics of the world, who released a positive press release saying, yeah, of course we don’t want Chinese competition — and the VC-backed startups who are trying to make a thing — as opposed to the long-tail consumer class. Anyway, maybe it’ll change, but —

Adam Chan: Right. Although I will say on that — let’s say you’re a company that uses foreign imported robots or drones. Would you rather have a gradual transition period where maybe your drones or robots cost a little bit more, and in exchange you get secure supply chains going forward? Or would you rather always have that risk of a potential total shutoff, which could cripple your business — especially as you become more and more dependent on things like drones and robots for the business model?

Jordan Schneider: And I guess there’s a collective action problem where no one’s going to do it unless everyone does.

Carve-Outs, Onshoring, and Allies

Aqib Zakaria: I’m curious about the industrial policy calculus, especially for the robot rule. My understanding is that there’s some level of carve-out for researchers there — unfortunately, the best robots right now for research are cheap Unitree ones. Maybe you can still buy those under the current rule, but then how do you wean off later? There’s some needle you have to thread there, where you’re trying to get an American domestic robot industry off the ground but also not crippling robotics research while we’re in that transition.

Adam Chan: It is the case under FCC rules that there are some built-in exceptions to the general ban on import or authorization for import or sale. One, it does not affect imports or sales to the US for use by the US government, which makes a lot of sense. But to your point, there’s also a measure for imports for product development. If you’re importing in small quantities to develop your product, that is something you can continue to do. So if there’s a foreign robot that maybe is not going to be able to be authorized for widespread commercial sale — if you are a US robot company looking to train some AI model or build your own robot, you can continue to do that.

We were in a situation with drones where, interestingly, we clarified there were some US counter-UAS, or counter-drone, companies that were importing foreign drones to practice on. Even something like that is still permissible, again, if it complies with other FCC regulations and doesn’t exceed the quantity limit.

Jordan Schneider: So wait — Adam, you got an email saying, “Hey Adam, can we import some Chinese drones as long as we promise to blow them up?”

Adam Chan: Right. Well, listen, if a defense contractor wants to import any product in order to develop their systems, they are permitted to do so in small quantities. We have not affected that.

Jordan Schneider: On the industrial policy question, there was another line somewhere deep in these regs saying we’re more likely to give you an exemption if you promise to build a plant in America. Can you talk a little bit about the thought process behind that — what that equation is?

Adam Chan: This goes to the conditional approval process. We have put out guidance documents compiled by the executive branch for each of drones, routers, robots, and power inverters that have been added to the covered list. As part of the conditional approval process in that guidance document, there is a requirement to submit an onshoring plan — where you describe, all right, how are we going to actually onshore production?

Everyone recognizes the realism here that we’re not looking for 100% of production for every company to be onshore tomorrow, so we have allowed for a phased transition period. But my understanding is the executive branch is not interested in just granting broad exemptions to companies to continue to produce abroad, even if the company itself, from an ownership or cybersecurity standpoint, doesn’t pose any major national security risks.

Jordan Schneider: Is there any hashtag allied scale in there?

Adam Chan: The emphasis so far of these actions has been on domestic production. Obviously the actions are country neutral, but certainly there’s no prohibition on companies in allied countries applying for conditional approval. I would also note, for example, on the robots one or drones — if you meet the domestic end product definition, which is basically assembled in the US, 65% components by value produced in the US, that’s sufficient, which also obviously leaves some components that can continue to be imported. I would note on the drone side, there’s also an exemption for any drone or drone component that has been cleared on the Department of War’s Blue UAS list. So this is their list of drones that are cleared for sale to the military. And many of those —

Aqib Zakaria: Do allied companies or allied countries get this impression when the FCC releases the rule that this rule is really not for us, we’re definitely going to get an exemption? Or is there this sense of — I think sometimes there’s regulation and there’s a mild panic from allies where we’re not sure how this rule is going to apply to us. How does that process become clear to allies and companies operating in those jurisdictions?

Adam Chan: We have some engagement directly with embassies, that kind of thing. We’ve also been coordinating with the State Department to brief other countries on these measures. Our message generally to foreign governments and foreign countries, across the board, is: apply for conditional approvals. There’s not really a downside, and again, I think we’re interested in hearing —

Jordan Schneider: Are there fees, Adam?

Adam Chan: What’d you say?

Jordan Schneider: Is this a revenue game? What are the fees here?

Adam Chan: There are no fees. So it’s a pretty easy process. Any company from any country can apply to be considered.

Robot Vacuums + How AGI-Pilled is the FCC?

Jordan Schneider: Well, what does cost money is DJI suing you guys. Robots has only been out a week, so we haven’t had any lawsuits yet. The one I could see coming would be from a Roborock, who is saying, Adam, give me a break, your robot vacuum cleaner is not going to kill you.

This is also where the consumer harms start to become a little more apparent. I have not bought one of these yet — not because they are Chinese made, but because they’re just still kind of dumb. But hopefully in the next year or two they will no longer be. I’m going to need Adam or some friend in the Pentagon to approve the new model so that I can still get them.

What was the thought process on the robot vacuums versus the humanoids or more broadly capable robots?

Adam Chan: Once again, ultimately the determinations are made by the executive branch and not the FCC directly. But I can point to — there was recently a story, and I may get the exact details wrong, so forgive me, but someone was actually playing with their DJI drone and somehow managed to gain access to a whole suite of robot vacuums. Once again, there’s this question of cybersecurity, data, surveillance access. Obviously you get a ton of data from inside people’s homes, so there are, I imagine, concerns with remote access there. That kind of thing was likely at the core of what went into the determination extending to something like robot vacuums — where, once again, by broadly scoping in mobile robots generally, it avoids the whack-a-mole problem, especially for such a nascent industry.

Jordan Schneider: So you guys had thought about Unitree just selling every Unitree with a dustpan and a broom saying, come on, we’re a robot vacuum, why are you regulating us?

Adam Chan: Right. I guess no comment on that specific hypothetical.

Jordan Schneider: Okay.

Aqib Zakaria: One interesting assumption latent in this early rule is the idea that this is going to be highly relevant for consumer security, but also national security and industrial policy. When we think of these things, it’s like — humanoid robots, these are going to blow up because of AI, right? These are going to become a major national security issue, a major industrial policy issue. But I would still say that’s not a really mainstream way of thinking about it. So I guess, how AGI-pilled do you have to be to believe in this rule? Or is there a version of believing this rule is really important if you’re not AGI-pilled?

Adam Chan: I think the rule is important no matter how AGI-pilled you are — but especially because I personally, and probably most people, should be uncertain about how AGI-pilled they should be. Maybe something like humanoid robots is a total dud and no one ever gets a humanoid robot after 2027, in which case, okay, then this doesn’t matter one way or the other. Maybe we all should be super AGI-pilled and these robots are going to be critical in the near term, in which case this is all the more important.

I’d also note there are some things — obviously you talk about the robot vacuums, but even some of the mobile robots in, say, a warehouse, or some of the public safety type quadruped mobile dog robots. You don’t need to have the sort of AI models that we can only dream of for those to play a significant role in the economy — and again, there are potential concerns with being reliant on potentially dangerous foreign-sourced mobile robots there as well.

Jordan Schneider: The Chinese Ministry of Commerce said that the FCC measures:

fly the banner of non-discrimination, but in substance discriminate and suppress Chinese enterprises and products… protectionism does not make the US more competitive and will only hurt the interests of US companies and consumers.

Adam Chan: I would point out that there is nothing targeted at any specific country in this most recent action. It is obviously country neutral. Like you mentioned Unitree — if Unitree wants to set up a factory in the United States and produce robots, nothing is stopping them under the FCC action. I don’t think it is fair to suggest that this action is targeted at specific countries.

Simple but Dumb Beats Big-Brained

Jordan Schneider: I want to come back to the simple, dumb rule as opposed to the super-lawyered-out carve-out version that this could have been. Is there something about the FCC’s authorities that makes it easier to do option A as opposed to option B? Or what else drove you guys down that path?

Adam Chan: I do think the covered list as a tool is inherently the blunt instrument, and not necessarily the more nuanced one. You have seen us do some of the more nuanced rulemakings — for example, a recent action on submarine cables, which is not actually a covered list action but uses other FCC authorities. So there is sometimes a place for the more nuanced rules.

Here, and especially when you’re talking about broad national security threats that aren’t easily namable as threat A, threat B, and threat C, as well as the need to send this strong market signal, I think it’s all the more important to have the blunt, maybe dumb — like, simple but dumb — kind of policy, rather than this big-brained, nuanced policy that maybe takes years to implement and then years more to amend. Maybe at the end of it you could say, well, we more directly addressed precisely the problem we were interested in, but in the meantime we’ve lost years and maybe have just confused the public and industry as to what we want to do and what we’re thinking. Oftentimes it makes a lot more sense to go the former route.

Aqib Zakaria: We were talking earlier about how if Unitree opened up a factory in the US, it’d be totally kosher under this rule. Do you have any preference in terms of how domestic capacity is created in this space? Does the FCC care, or do you care, if it’s a homegrown company funded by domestic capital that rises to compete in humanoid robotics, versus if it’s a foreign JV — like Unitree robots, but they’re just made in America? Do you see those equally as a win?

Adam Chan: Maybe not comment on what I personally think, but under the FCC action there is not a distinction between American company invests and foreign company invests. There are potentially other national security concerns with inbound investment — that’s what we have CFIUS and other tools for. If there’s some inbound investment that does raise concerns, there are other existing tools to handle that.

We see, at least in the near term, given the need to have 65% or more components by value produced in the United States, it’s going to be hard for certain foreign companies to just all of a sudden overnight onshore all production to the United States. So we suspect most foreign companies interested in getting new approvals will go through the conditional approval process in order to build up over time. That’s how at least the FCC is thinking about it.

Jordan Schneider: Can you tell us what’s next? Are you guys going to stay busy, or are you just waiting for giant new markets to unlock? What’s on the release calendar for the FCC?

Adam Chan: I won’t comment directly on any potential future FCC actions, but it’s definitely fair to say the FCC will be busy. We talked about a few things the FCC’s been doing. We also have open dockets for rulemakings on a whole wide range of other issues, whether it’s submarine cable security, or we have this proceeding that we’ve dubbed “bad labs,” about the test labs that actually test equipment for the FCC. We definitely foresee staying busy on the rulemaking side. It would not surprise me if there are further updates to the covered list, big or small, to come.

As well as on the enforcement side, right? The FCC has implemented a bunch of rules, and we’re increasingly looking at enforcing them. Even on the issues we’re talking about — on drones, we just discovered there’s one company that lied, or apparently lied, on its application form to get authorization, saying it was producing in the United States, where it turned out it was not. The manufacturer listed in the application form confirmed they’d never heard of the company. We are looking at potentially taking action there. We’re taking action related to certain drone companies that are not complying with FCC investigations, that maybe are white labeling covered equipment to try to get around restrictions. So we’re looking at all of that. Expect enforcement and rulemaking to both remain busy going forward. Obviously there are a lot of national security threats out there, so we’re going to keep moving to protect American communications networks.

Aqib Zakaria: Is it just like a list where you’re like, okay, we got the drones, now to the robots, we got the robots, now to the next thing? And it just never ends?

Adam Chan: I’ll say two things. One, there’s still plenty of work to do on that front. We have a dozen separate drone actions going still, so a lot more follow-ups there. But on the to-do list — my to-do list is definitely very long, I will say that.

Jordan Schneider: How does the response function from academia and industry feed back to you guys if something is too tight or too loose? What is that information stream that flows into this policymaking process like?

Adam Chan: My inbox is always open. Feel free to shoot me an email — questions, concerns, suggestions, recommendations. We also have, for at least the covered list actions, points of contact at the FCC, and we have an FAQ page. We’ve been updating the FAQs as we get more queries from those on the outside.

A lot of what the FCC does, for example on the rulemaking side, we do get public comments. Those are often very helpful, especially because we’re often dealing with very technical areas where there’s not a lot written, and often it’s actual stakeholders or companies within an industry that have a lot of the best information.

We have an open docket now — since we’re talking about drones, new models of foreign-made drones are automatically prohibited, but we have an open docket where we’re seeking public comment for 30 days about whether we should ban the import and sale of certain categories of already-authorized drones that we’ve dubbed “military grade” drones. Think things like swarming drones, or big aerosol spray drones, or thermal imaging and LIDAR type drones. That’s an opportunity where any member of the public, from individual hobbyists to big-name drone companies, can do public comment and give us feedback. In this case I think we actually have pretty good dialogue with outside parties, in addition to people within other agencies or the White House, to get a lot of input from all sides.

Telecom Prom

Jordan Schneider: Do you think you’re the national security lawyer having the most fun in the executive branch currently?

Adam Chan: I’m definitely having a lot of fun. There’s not another job I’d want to have. I get to work on super interesting issues every day, have an impact. It’s a ton of fun. If you’re listening, the FCC is hiring — we welcome applications, feel free to reach out.

Aqib Zakaria: Do you want to make the pitch for the FCC? Your average guy who wants to go to DC — the FCC is usually not the top of their list. It’s not on people’s radar.

Adam Chan: One, the FCC has a lot of broad statutory tools. Especially if you’re a lawyer or just a creative policymaker, there are a lot of opportunities to get stuff done there. The FCC is at the cutting edge of a whole range of issues, whether it’s spectrum allocations and the interesting developments in 6G and new uses for 6G space. We license all commercial satellites. We do all sorts of connected devices. Obviously I’ve talked a lot about the national security side and only scratched the surface there.

Also, just institutionally, the FCC is interesting. We’re very small — I think we have like 1,200 total employees and a $400 million budget — which compared to other big-name agencies in the government, Department of War, Department of Justice, Commerce, Treasury, are all much, much bigger, which makes us nimbler and kind of fun.

Another thing — the FCC actually is interesting, we haven’t talked about this, but for all my time here we’ve had one or two Democratic commissioners in addition to the Republican chair. All the national security rules and the enforcement action, penalty, fine type things that we’ve brought have all been bipartisan, unanimous. A lot of the work we do is going to endure beyond just this administration, and that’s potentially another reason the FCC is interesting and also makes our national security work impactful.

Jordan Schneider: Give us the anthropology. What’s the energy, the vibe about the FCC relative to other agencies or the Hill that you’ve spent time at before?

Adam Chan: People are very busy, for sure. People are also very smart. The FCC is small but has a pretty high percentage of either lawyers or engineers, people with advanced degrees. Because it’s small, we have a lot of expertise — or expertise is at least highly concentrated at the FCC. A lot of nerds at the FCC, just given the technical nature of what the FCC does.

Jordan Schneider: Have you been to the FCBA chairman’s dinner?

Adam Chan: I did go last December. It’s an annual gathering — I’ve heard it described as telecom prom. It was a good time. Chairman Carr gave a very, very funny speech last December, and I think he’s preparing another good one for this upcoming December. For those who don’t know, it’s all the telecom lawyers, it’s like a charity dinner. It’s a lot of fun.

Aqib Zakaria: I have an offer to go to law school next year and I really like doing national security policy. Should I go? What’s the landscape for a lawyer doing this sort of work, and should people enter that field, versus just going straight to DC directly?

Adam Chan: I strongly support law school. I personally loved law school. If you’re a nerd who likes learning, you’ll like law school. It is very helpful, because ultimately a lot of this stuff is legal, right? We’re relying on laws passed by Congress, or issuing new rules, or doing some sort of enforcement adjudication, or dealing with weird questions of the Administrative Procedure Act.

I would also note that even if you don’t want to directly practice law and be in a general counsel’s office, there’s tremendous value even for policymakers in having the legal degree. You can actually understand what is within scope, what we can do legally, what are some creative legal options, what’s the litigation risk. Too often it’s a problem in Washington when there’s a complete dichotomy between the policymakers and the lawyers — the policymakers say, we really want to do this, and then the lawyers say, no, you can’t, and there’s not really communication between them. From my perspective, being a lawyer but right now not going into court every day and being able to shape policy allows us to do things without a breakdown of communication between the policy and the legal.

Jordan Schneider: I’m trying to keep Aqib here. This is not helpful.

Adam Chan: Listen, I’m sure you can continue doing great work on ChinaTalk…

Aqib Zakaria: We’ll see, we’ll see.

Jordan Schneider: We will call it here. Adam, thank you so much for being a part of ChinaTalk.

Our bread and butter is the intersection of emerging technologies with US-China relations. Specifically, we explore how developments like AI, robotics, and the underlying supply chains speak to broad questions of national power and risk.

We’re excited to host an ideas-and-stories “essay” contest on what you think are the most interesting developments and questions that fit at ChinaTalk. We’re interested in proposals related to artificial intelligence and China, broadly speaking; this includes models, chips, robotics, supply chains, applications, and their economics, politics, and governance.

Winners’ submissions will be shared via our newsletter and win a share of the $50k prize pool.

Here are some formats we think your submission could take, though you are in no way constrained by the following list. Surprise us!

• Writing (articles, essays, reporting, analysis, etc.)

• Interactive software (websites, video games, etc.)

• Policy proposals

• Videos (we’re also hiring specifically for a video producer!)

• Data-driven research (benchmarks, data analysis, etc.)

We’re going to do this in two rounds. First, submit a paragraph for each idea you are considering (up to three). For the ideas we like, the ChinaTalk team to give you some some feedback and a $250 commissioning payment. Next, you’ll have a few weeks to execute your idea, then we’ll pick the winners and split the prize pool. We’re open to publishing anonymously if you have a good reason.

Submit here! We’re looking at these on a rolling basis!

We’re also hoping to use this contest to source full time hires for ChinaTalk. If you want to work here, submitting is your best way in!

For some background working with us:

• You’ll have a wide level of autonomy to chart out your own research agenda, and will publish to an audience of 75k, which includes some of the most influential policymakers, reporters, investors and corporate leaders in technology.

• ChinaTalk is a remote-first organization with no time zone requirement. We’re happy to hire global talent who are free to work and live wherever they prefer (anywhere but inside mainland China, HK, and Macau) .

• Salaries for entry level researchers start at $100k with a potential for a bonus. We’re also open to hiring more experienced talent.

• ChinaTalk employees have 20 days a year of paid vacation and generous health care benefits.

We can’t wait to see your submissions!

Jordan, Irene, Lily and Aqib

After decades of trade liberalization, the United States is reverting to a policy of economic containment towards China.

Until recently, the U.S.-China bilateral relationship was defined by a tacit exchange of American market access and Chinese industrial development. As Secretary of State John Kerry tellingly put it in 2015: “On a level playing field, our growth contributes to yours, and your growth in China contributes to ours. Your nation has opened up to the world and you have lifted 500 million people out of poverty, one of the most remarkable occurrences in human history.”

This arrangement of economic integration was by no means inevitable. At the height of the Korean War, the United States maintained a ‘China differential’ — trade restrictions more severe than those applied to the Soviet Union, enforced in concert with European and Asian partners.

Both China and the United States are vastly different actors today than they were in 1952, their divergences illustrative of transitions in the trading system, export control bodies, and alliance architectures. How U.S. economic statecraft evolved throughout the Cold War illustrates the conditions under which American economic influence shaped alignment decisions — and how the foundations that facilitated policy achievements have since unraveled.

As the curtains began to fall on the Chinese Civil War, President Truman’s National Security Council drafted NSC/41. A report on the nation’s trade strategy with the nascent People’s Republic of China, the document attempted to balance a posture of restrictive trade and economic engagement.

NSC/41 reaffirmed previous proclamations that the primary objective of American policy toward China was to prevent China from “becoming an adjunct of Soviet power.” The document’s unnamed authors proposed a strategic network of trade relationships between China, Japan, and the West to germinate Sino-Soviet friction.

The Truman administration predicted that China’s fear of economic exploitation at the hands of the Soviet Union and desire to establish a wider range of economic partners would seed resentment between the two communist states. At this time, tensions persisted over Soviet occupation in Manchuria, and Soviet support for China’s economic plans remained uncertain. To encourage a Sino-Soviet fracture and induce a Westward China pivot, the NSC recommended a small category of export controls on “essential security restrictions,” allowing for flexibility in the face of potential conflict or cooperation.

The document argued against “severe economic restrictions” against China for four primary reasons. First, collective buy-in from Western allies was unlikely to be successful, with British support being “essential to the effectiveness of United States controls.” Second, a maximum pressure approach was feared to heighten domestic support for the Chinese regime, allowing it to consolidate its grip on power rather than weakening it. Third, overly regulating China-Japan trade would increase Japan’s economic dependence on the United States. Finally, and most importantly, such a policy would be inflexible, leaving little room for negotiation and pushing China into the arms of the Soviets.

In the months that followed, Secretary of State Dean Acheson provided further recommendations for implementing NSC/41. Negotiations with the British were impeded by the United States’ unilateral controls on trade with China. As Acheson describes, British negotiators were staunchly opposed to regulating the trade in strategic non-military goods, arguing that such measures would not achieve their desired objectives, would be perceived as economic sanctions, and would jeopardize Hong Kong’s status and British investment in China.

Conditional on collective buy-in with France, the Netherlands, Belgium, the Philippines, and Japan, the British offered to restrict the trade of “high military value” exports from the British Isles, Hong Kong, Malaya, and Singapore. They would also exchange information on the movement of “strategic non-military goods” and accede to a consultative body if China-related trade was “excessive or injurious to common interests.”

This left the United States as the unilateral enforcer of export controls on select non-military goods through a licensing regime that selectively rejected applications based on the risk of transshipment of dual-use capabilities to the Soviet Union. Reinforcing NSC/41, Acheson cautioned against greater restrictions on China due to the limited nature of U.S.-China trade and the ease with which China could find alternate suppliers for imports from Europe. Thereafter, NSC 48/2 would advise unfettered trade with China in “non-strategic commodities.”

These documents served as the frontier for America’s nascent engagement with the People’s Republic of China. Economics would continue to play a role in the strategic separation between the two actors. Reducing the Soviet Union’s influence in Asia and deterring “Communist aggression” was viewed as requiring measures that were overt and covert and applied through methods of economic statecraft, as well as political and military tools.

As George Kennan’s vision for Japan as an anti-communist bulwark in East Asia had only been articulated seven months before NSC/41, concern about Japan’s overreliance on the United States was critical at the time. As the United States sought to weaken Japan’s dependence on “food and critical raw material” imports from China, this occupation gave the United States significant influence over Japan’s trade with China — an influence it would leverage to great effect over Sino-Japanese relations and trade.

Following World War 2, the European colonial footholds in Asia regained their saliency, complicating American efforts for a unified, trans-Atlantic approach to export controls against China.

With this backdrop of an emerging split and allied discontinuity, the Cold War’s economic statecraft began. Throughout the period of Eurasian tension, points of consensus among American allies on export controls were few and fleeting. Allies diverged on the tradeoffs between national security and commercial interests as well as their preferences for how controls were designed and where they were applied.

U.S. leverage rested on allied trust in security guarantees, institutional legitimacy, and Chinese economic vulnerability. That leverage let Washington deter adversaries, convert rivals into partners, and build consensus among allies. Despite grievances and shifts in U.S. China policy throughout the Cold War, the United States maintained allied consensus — an accomplishment contrasted by the current fraying of alliances.

The Korean War and a Maximalist Approach

The Korean War temporarily upended the cautious approach outlined in NSC/41.

In January 1950, the Coordinating Committee for Multilateral Export Controls (COCOM) came into effect as a plurilateral grouping of American allies to restrict trade with the Communist bloc. By June, China was incorporated into COCOM’s scope; by December, the United States had deployed a complete embargo on transshipments to China.

More on COCOM in this piece:

In November of 1950, Troy Perkins, U.S. Deputy Director of the Office of Chinese Affairs, wrote that controls at the level of the Soviet Union and embargoes on petroleum did not “add up to a program of all-out American economic warfare.”Economic warfare, he argued — without fully endorsing — required a carte blanche embargo on Chinese goods, while unilateral approaches would have a “marginal effect” on the People’s Republic. Like Acheson, Perkins was concerned about “strains” on Europe’s strategic and commercial positions, particularly for Britain in Hong Kong and France in Vietnam (i.e., Indo-China).

Contrary to Perkins’ prescriptions, ‘economic warfare’ would be adopted. In January 1952, COCOM members raised their controls on China to be more restrictive than controls on the Soviet Union, thus dubbed the ‘China differential.’ Through the Mutual Defense Assistance Control Act, U.S. aid assistance would be conditional on following American embargoes on strategic commodities to “nations threatening the security of the United States.” On January 5, 1951, the CCP Central Commission on Finance Economy wrote in a report that the United States and Japan had seized $27 million worth of goods, while France and Belgium had restricted the export of steel and aluminum, and more than $100 million worth of goods from Western Europe were in limbo.

Towards the end of the Truman administration, they attempted to institutionalize this fleeting consensus. In August of 1952, Canada, France, and Britain opposed American efforts to establish a ‘Far East’ organization distinct from COCOM. Instead, in a compromise, a China Committee (CHINCOM) was established under the auspices of COCOM, with Japan ascending as a member of both COCOM and CHINCOM in September.

However, concerns remained that Japan would reduce its own controls to the baseline set by CHINCOM. Under the “intense pressure” of Secretary of Defense Robert Lovett, Ryuji Takeuchi, then Chief of the Japanese Government Overseas Agency in the United States, conceded that Japan would follow American embargoes, as already identified and determined in the future.

Japan had other economic concerns. Unlike other COCOM’s European members, Japan was more dependent on access to Chinese raw materials and less capable of transshipping goods to China through the Eastern bloc. In NSC 125/5, the Council stated that Japan’s historic trade reliance on contemporary adversaries was a “long-run economic problem.” By their estimates, 18 percent of Japanese exports went to China, and 25 percent of its imports originated in China — trade flows which were not easy to replace. As the Korean War and Taiwan Strait tensions limited Japan-Korea and Japan-Taiwan trade, economic growth, and new trade patterns with “free world” South and Southeast Asia countries were determined as a “necessity” to achieve Japanese self-reliance.

Japanese self-reliance also served as justification for U.S. military involvement in Southeast Asia, as the NSC posed that the “loss of Southeast Asia, especially of Malaya and Indonesia” could result in subversive pressures on Japan to accommodate communism and “imperil retention of Japan as a key element in the off-shore island chain.”

As the Korean War subsided, allied support of maximalist trade restrictions on China withered (see a timeline in Table 1 below).

The American foreign policy establishment was disturbed by the undoing of a comprehensive, collectively enforced embargo on China. Following the armistice in Korea, Secretary of State John Foster Dulles wrote to 26 U.S. embassies that the settlement was not indicative of reduced aggression by China in Asia, with particular concern about threat escalation in Southeast Asia. In noting that “major allies … would be increasingly unwilling to support the US in maintaining … economic pressures against Communist China,” he urged that the total embargo on China remain, and allies be persuaded to retain their controls.

U.S. intelligence had chronicled the break between allies. In Special National Intelligence Estimate 100-56, these directional transitions were surmised as “a growing belief among the allies … that there were substantial prospects for a relaxation of tensions.” While Japan and the European allies opposed the ‘China differential’, South Korea, Taiwan, South Vietnam, and the Philippines supported the controls, and the intelligence community cautioned that U.S. prestige would suffer in these countries if the differential was removed.

The Joint Chiefs of Staff, meanwhile, expressed “grave concern” that questions of collective security were being presented as commercial issues. They were perturbed that the relaxation of trade restrictions of China could lead to negotiated settlements on similar relaxations with the Soviet Union and its satellites.

This memo by the Joint Chiefs, dated June 13, 1957, diverged from NSC/41 in two important ways. First, on the question of Japan, the memorandum signaled that Japanese self-reliance was a peripheral concern and that an increase in China-Japan trade would incur Communist influence upon the archipelago. Second, it describes the ‘China differential’ as serving “to retard [China’s] industrial growth and the ability to translate this growth into military action.” This explicit endorsement of diminishing China’s industrial growth marked a clear departure from the pre-Korea War logic.

As allies questioned the efficacy of economic strangulation, the Sino-Soviet alignment was developing fractures of its own. As Shu Guang Zhang elucidates in his book Economic Cold War, China had limited alternatives to heavy reliance on the Soviet Union — a dependency which the Soviets underdelivered and subverted.

In China’s first Five-Year Plan, Stalin was concerned about the burden China’s construction would have on the Soviet Union, urging China to lower its growth projections from 20 to 15 percent. The Soviet Union also leveraged these investments in China, with the expectation that China would purchase Soviet goods in U.S. dollars and fulfil Moscow’s export quotas for rubber and critical minerals. Alongside these Soviet demands, Moscow had delayed military equipment orders and vetoed assistance with several industrial projects, including an aluminum factory in Guizhou and a steel plant in Wuhan.

Soviet underperformance as an economic partner weighed heavily on Chinese decision-making. According to a 1966 CIA report, China’s “lean-to-one-side” strategy initially led to economic relations with Communist states, an approach reinforced by the ‘China differential.’ Soviet support remained significant, totaling 166 industrial enterprises from 1953-57 and ten loan payments of 1.274 billion rubles. Noting the importance of Soviet support, Zhou Enlai commented during a state council meeting on October 11, 1956: “Although [we] will continue trying to obtain Soviet aid, we just get rid of the illusion of relying [on them].”

This changed in 1960. With the ‘China differential’ abolished three years prior, greater access to Western markets coexisted with a strained relationship with the Soviet Union. On July 16, 1960, the Soviet Union withdrew all advisors and technicians working in China. In an “emergency meeting” with Soviet ambassador Stepan Chervonenko, Chinese Vice Premier and Foreign Minister Chen Yi described the act as “hurting the friendship between the two countries.”

The CIA identified this incident — along with a need for grain imports and the rundown of the Great Leap Forward — as sharply altering the composition of China’s global trade flows. Transshipments aside, China’s trade with the “Free World” (as categorized by the CIA) had dropped from 1950 to 1954, dropping from 71 to 26 percent of China’s total trade. The weakening relationship with the Soviet Union meant that China’s global trade volume declined precipitously, down to $2.7 billion in 1962 from its $4.3 billion peak in 1959. Since 1960, however, China’s trade had begun to recover, with trade with the “Free World” offsetting a downturn in trade with “Communist Countries.”

Reductions in Sino-Soviet trade never recovered. Chinese estimates documented by Zhang show that the Soviet share of China’s total trade dropped from 54.8 percent in 1952 — the height of the ‘China differential’ and U.S.-led embargo — to 1.03 percent in 1970.

While the United States would retain its embargo until 1971, by the mid-to-late 1960s, American economic partners had significantly increased their trade with China. From 1950 to 1970, British trade with China grew from $73.5 million to $1.1 billion. By 1965, countries across Europe and the Pacific Rim had universally increased their trade with China at an exponential rate.

Japan, however, was more complicated. Trade leapt from $4.4 million in 1952 to $454.3 million by 1965, but this was not a continuous upward climb. Japanese Prime Minister Nobusuke Kishi — overseer of the occupation of Manchuria — entered a diplomatic row with Beijing after visiting Taiwan in June 1957. Tensions escalated after a Chinese flag was trampled in an exhibition in Nagasaki, with trade relations suspended before recovering to previous levels in 1963.

Despite being at the height of American power, the United States’ ability to maintain a multilateral embargo waned. Allies tolerated maximalist controls when the threat felt immediate, but when it receded, commercial interests reasserted themselves. To current economic security vulnerabilities, the United States has responded with crisis-driven coalitions on critical minerals and AI supply chains — efforts that will be tested on their ability to align commercial incentives and sustain shared costs between allies.

Oscillating Approaches

Alongside the U.S.-China detente, the Nixon administration removed the embargo and began to draw down its export controls. As Kissinger wrote to Nixon, the removal of restrictions furthered the administration’s policy of “broadening communications between the U.S. and the People’s Republic of China by removing obstacles to personal and commercial contacts.”

The removal of the U.S. embargo transformed the Chinese economy. From 1972 to 1980, U.S.-China trade increased 37 thousand percent, rising from $12.9 million to $4.8 billion. In the decade following the embargo removal (1972 to 1982), China’s trade increased from six to twenty percent of its GDP.

By opening trade as an incentive for dialogue, Nixon paved the way for a new Eurasian geopolitic: a theater turned on its head by China’s split from the Soviet Union, putting dual pressure on Hanoi and Moscow as the Nixon administration sought an end to the Vietnam War. When Nixon began drawing down controls, China’s trade restrictions transitioned from differential to even-handed to preferential. As the control regimes on China began to diminish relative to the Soviet Union, polities like West Germany feared that this new China policy would erode the position that “all Communist countries must be treated equally.”

The U.S. treatment of China and the Soviet Union would adjust once again with the U.S.-Soviet detente. The detente served as a justification to minimize U.S.-China military cooperation during this period, which would resume in full force following the Soviet Union’s invasion of Afghanistan in 1979.

From the Soviet foray into Central Asia to Tiananmen, the United States removed prohibitions on arms sales and allowed the transfer of technology “twice as high” as levels exported to the Soviet Union. By 1983, China had moved to “Country Group V,” a category with relaxed trade limitations, of which members included U.S. treaty allies like Japan, Australia, and Western Europe.

As Hugo Meijer writes, despite these unilateral reforms, China remained on COCOM due to concerns that plurilateral liberalization would encourage U.S. allies to attempt to weaken controls on the Soviet Union. However, as U.S.-China trade, technology transfer, and military cooperation enhanced, American unilateral efforts gave way to plurilateral liberalization. To illustrate the scale, 95 percent of U.S. license applications to COCOM in 1985 were related to China.

These efforts meaningfully translated to military-to-military cooperation. In 1984, there were six U.S. government arms deliveries to China. By 1989, that number rose to 89,800. Prior to Tiananmen, there was a similar increase in advanced technology exports and export license approvals. But, eventually, the music stopped, with arm deliveries zeroing out in 1990.

Following Tiananmen, the HW Bush administration ceased arms transfers to China. The United States suspended its military exchanges and export licenses for dual-use items such as satellites, and withdrew support for China’s access to development bank loans and crime-control equipment. Multilaterally, the relaxation of controls within COCOM was postponed. However, corresponding with the fall of the Berlin Wall, Bush proposed reductions to the COCOM control lists, eliminating 30 industrial items, and reducing the core list of dual-use items. The result was a paradox of iced military-to-military exchanges but wide availability of defense-related products.

In spite of these bilateral disputes, commercial trade maintained its exponential growth. In 1990, the year-on-year increase in the value of U.S.-China trade increased 74 percent — a trend which would continue for the decades to follow. Arguing in favor of maintaining commercial relations with China, U.S. Ambassador James Lilley wrote in an embassy cable: “Not only do we help the good guys, and there are a lot of them in China, but we also help ourselves financially.”

Throughout the Clinton administration, these commercial linkages grew, and trade barriers fell — developments that paved the way for China’s accession to the World Trade Organization, acquisition of Most-Favored-Nation status, and the broader growth foundations of the contemporary global economy.

Beginning in 1991, COCOM progressively reduced its control lists, eliminated licensing requirements for “low-level items”, civilian-use computers, and “certain cryptographic items”, and eased “restrictions on advanced telecommunications and fiber optics.” In a catalogue of U.S. sanctions on China, the Congressional Review Service highlighted a “controversial” $2 billion shipment of gas turbine engines in 1994, which only required a “general export license.” That same year, COCOM member states agreed to phase out the organization.

The Wassenaar Arrangement was intentionally designed to be a looser coalition of states with fewer obligations. Wassenaar did not — and does not — require states to submit export control licenses for review, instead opting for national discretion and biannual reporting of sensitive shipments. China is neither a member nor a target for Wassenaar, while Russia’s membership in the forum has restricted the possibility of expanding Wassenaar’s control list.

In hindsight, it is easy to view the period of the ‘China preferential’ as a series of decisions that were tactically sound but strategically flawed. From Nixon to Clinton, administrations had clear justifications for relaxing controls, rewarding behaviors that aligned China with the United States politically and militarily. But U.S. administrations in this period had starkly divergent views on the U.S.-China relationship, and turmoil from Tiananmen animated disputes. The United States would swing from an all-out embargo until 1972 to selling weapons to the Chinese military by the 1980s — a resounding reinterpretation of China as a destabilizing force in East Asia to a key pillar of Eurasian containment.

At the same time, the fall of the Soviet Union gave way to an eroded institutional architecture, now selectively being rebuilt primarily on a bilateral basis and with the looming threat of American sanctions. These dynamics bring into question how future oscillations may materialize and the feasibility of reconstructing binding and consensus-driven allied trade bodies.

American Power and Coalitional Capacity

American economic statecraft toward China can be defined through repeated reversals. These sequenced oscillations between coercive containment and strategic accommodation were catalyzed by differences in administrations, shifting international environments, and evolutions in China’s domestic political regime. During the Cold War, these shifts were partly sustainable because allies heavily depended on U.S. security, and China remained economically vulnerable due to its domestic foundations and its ties to the Soviet Union. In the contemporary era, weakened allied dependence, diminished trust in U.S. consistency, and China’s greater self-resilience make similar strategies harder to coordinate and less likely to succeed.

One of the most stark contrasts between the Cold War and our current moment is China’s capacity for and political commitment to self-resiliency. In sequence, China alternated between trade, investment, and technological dependence on the Soviet Union, the United States, and the broader Western bloc. Made in China 2025 was meant to meet the perceived challenge of these dependencies, with the State Council acknowledging in 2015 that “The capacity for independent innovation is weak, and key and core technologies and high-end equipment are highly dependent on foreign countries.”

In the decade that followed, China pursued what was referred to as an “independent opening up,” engaging with the world on its own terms to strengthen its industrial foundations and resiliency. By 2025, the 15th Five-Year Plan transitioned towards “independent controllability” of global supply chains, weaving development and security concerns to safeguard the new pattern of development — a pattern of dual circulation. In a State Council Meeting on May 15, 2025, Premier Li Qiang described the dual circulation approach as follows: “Efforts should be made to leverage the inherent stability and long-term growth potential of domestic circulation to offset the uncertainties in the international market.”

China’s policy of dual circulation, therefore, represents more than sustaining economic growth. It positions China to strengthen internal capabilities, exert influence on other national actors, and insulate itself from global volatility and American belligerence.

Share

Throughout the Cold War, one point of continuity was Britain’s moderated approach to China due to its commercial interests in Hong Kong. While British leadership made compromises with the United States regarding embargoes during the Korean War, it consistently looked to maintain its trade relationship with China and investment position in Hong Kong.

While British interest in Hong Kong was not exclusively economic in nature, all countries today have a similar — and scaled — vested interest in economic interaction with China. The same capabilities that have allowed China to pursue self-resiliency have bound other nations to Chinese technology, supply chains, and capital. While China’s strength itself can serve as an impetus for collective action on the part of the United States and some of its allies, this reality may be the cause for inaction in equal measure. Every country has its Hong Kong — their reason for resisting economic decoupling. If concerns over economic consequences were apparent in 1949, they are paramount in 2026.

In 1949, Dean Acheson warned that American unilateral actions against China would “demonstrate the impotence and disunity rather than the power and solidarity of Western governments.” The days of Dean Acheson at the helm are long gone. In the current moment, the United States is less capable of convincing allies to act alongside it and exerts less influence on Chinese decision-making.

This challenge of collective action is furthered by the degraded credibility of the United States, which has, in turn, separated itself from its historic position as a global actor. During the Cold War, American strategy in general benefited from institutional legitimacy with COCOM and allied confidence, even as U.S. export control policies were fluid in light of a dynamic Sino-American relationship. And access to the U.S. market, both for allies and for China, provided a crucial incentive for alignment on non-economic domains of technology controls, military cooperation, and diplomatic posturing.

Recent American administrations have grappled with how to define and implement a China strategy. Under the Biden administration, Sullivan described the bilateral relationship as “managed competition,” where a floor would inform the nature of competition in critical domains. In the 2017 National Security Strategy, the Trump 1.0 administration characterized China as a “challenge [to] American power, influence, and interests, attempting to erode American security and prosperity.” In the second Trump administration, the 2025 National Security merely refers to China as an economic variable — a country with an imbalance of trade, whose economic prosperity, as opposed to ideological threat, is the justification for forward deterrence.

Japanese interests were a key factor in American involvement in Korea, Southeast Asia, and economic engagement with China. This whole-of-region approach — where allies are a system of integrated parts — is wholly missing in contemporary policymaking. While the Truman and Eisenhower administrations calibrated Japan into their China and Southeast Asian strategies, contemporary American actions have imposed costs without consultation. With the war in Iran, the United States has unleashed an energy crisis upon the Asia-Pacific, without consulting with allies or concern for how their economies would be affected by a closed Strait of Hormuz. As a result, across economic and security vectors, the United States’s actions have vastly diminished the logic of cooperation, with commitments less reputable and offerings less desirable.

When Ryuji Takeuchi conceded to U.S. demands that Japan would follow the American embargo on China, this coercive act was successful because of structural dependencies. Across allied actors, these same dependencies no longer exist to the same degree. Coercive leverage diminishes over time, as threat credibility weakens and interlocutors can recalibrate and prepare for future acts of coercion.

The United States has exhausted its coercive leverage — against allies, not adversaries. Against Denmark and Europe writ large over Greenland. Threatening economic sanctions against Spain due to its denial of access to military bases for the war in Iran. Continued contestation over Canada’s territorial integrity. These actions have not only diminished allied goodwill towards the United States but also wasted its leverage on strategically peripheral issues — issues that the Trump administration had no intention of following through on. Chinese growth and American disrepute, taken together, have made the costs of compliance high and the reward unclear.

The United States’ precedent of embargoing China casts a wide shadow on contemporary U.S.-China relations. China’s actions reflect an assumption that if China loses a ‘competition,’ the United States would be in a position to leverage its tools to pressure the country economically, just as had been done during the Korean War. China would not maintain its status as an adjunct of Soviet power, nor would it accede to being an adjunct of American power.

These diverging threat perceptions compound the pathway to a bilateral settlement. While the United States might intermittently pause its deployment of economic tools, there is no institutional incentive to reduce its underlying capacity. Tactical pauses in tool deployment do little to alleviate this structural uncertainty. For China, this sustains a threat perception and deepens unpredictability about when and why the United States might mobilize its economic toolkit — ultimately reinforcing Beijing’s drive for indigenization.

Subscribe now

The Cold War logic yielded outsized returns due to asymmetric dependencies, confidence among coalitional partners, and China’s economic vulnerabilities — conditions dislodged from the present status of Chinese resilience and America’s trust deficit.

What this history suggests for future policymaking is the need for more modest objectives and lower expectations for coalition cohesion. On a strategic level, it is incumbent on the United States to devise a durable strategic objective. NSC 41 is illustrative, as the goal of preventing China from becoming an adjunct of Soviet power brought coherence to decades of oscillating policy decisions. However Washington intends to pursue its foreign policy with China and third countries, its position will be diminished without such clarity.

After decades of neglect, Washington has adopted a coherent and “forceful” response to China’s dominance over critical minerals. China controls an estimated 91 percent of global rare earth refining, dominates the processing of a wide range of minerals, and has demonstrated repeatedly that it is willing to weaponise that position. The United States has responded through a range of initiatives and policy packages including Executive Order 14241, the One Big Beautiful Bill Act, and Pax Silica. Add bilateral deals with Australia, Japan, the DRC, and Saudi Arabia, and the result is undeniably a more aggressive and cohesive mineral strategy than the United States has attempted in decades.

The problem is not with any of these agreements and frameworks. Instead, it is with what is systematically absent from them. The entire architecture of American mineral strategy, and indeed the EU’s parallel effort under its Critical Raw Materials Act and RESourceEU, is built around two variables: where minerals are located and who processes them. What it consistently underweights is a third variable that will most likely prove equally consequential over the medium term: who controls the technology that operates the mines themselves. This constitutes the central strategic blind spot in the West’s mineral autonomy agenda because China is not just dominant in mineral processing. Beijing is also rapidly establishing dominance in the operational technology of extraction; the autonomous systems, industrial AI, connectivity infrastructure, and integrated machine intelligence that will define how the world’s future mines will run. Today, we’ll explore China’s dominance in autonomous mining technology and what to do about it.

If that dominance goes unchecked, the US and its allies could succeed in securing access to mineral deposits across friendly jurisdictions and still find those mines dependent on Chinese operational technology to function.

The Second-Order Competition

To establish both how and why the technology of mining matters strategically, it helps to unpack how the mining landscape is actually evolving. To this end, the departing point is to acknowledge the obvious: mining has always been a risk-prone and controversial venture, with the industry’s reputation tainted by concerns ranging from worker welfare to environmental destruction. The industry therefore has ample incentives to develop alternative extraction methods to improve its image and reduce liability.

Contemporary autonomous mining goes much further than replacing single tasks with machines. It is the deployment of an integrated stack: autonomous haulage systems that navigate dynamically without human drivers; AI models that manage geological analysis, fleet routing, predictive maintenance, and slope stability monitoring simultaneously; sensor networks providing real-time environmental data; edge computing that maintains operational integrity when wide-area connectivity fails; and satellite positioning that substitutes for GPS in remote and underground environments. When this stack is fully deployed, a mine managed remotely by a small supervisory workforce can operate almost uninterruptedly in conditions that would make human presence hazardous.

The productivity gains and safety improvements are both substantial. For governments and nation states, however, the more consequential dimension is strategic; that is, the capabilities required to run a mine autonomously in harsh and remote environments are directly transferable to deep-sea operations, space, and military logistics in contested or degraded operating conditions.

China understands this explicitly. Its Arctic and Antarctic deployments are framed in official strategic communications as technology demonstration environments since systems proven in polar conditions are, more or less, transferable to outer space and the deep sea. A case in point is the Yimin open-pit coal mine 伊敏露天矿 in Inner Mongolia, where a fleet of fully electric autonomous trucks has been operating in temperatures as low as -48 degrees Celsius.

The Yimin Benchmark

The Yimin deployment warrants a closer look because it is the clearest evidence available of where China currently stands. The fleet has been developed jointly by China Huaneng Group 中国华能集团, Huawei, XCMG 徐工集团, and State Grid Smart Internet of Vehicles 国网智慧车联网技术. Each 45-tonne truck runs on a 568 kWh lithium iron phosphate battery with the driver’s cab replaced entirely by a battery pack. Five cameras, LiDAR units, and millimetre-wave radar provide 360-degree field of vision up to 40 metres in complete darkness, dust storms, and blizzards. Automated battery-swap stations reduce downtime to approximately five minutes while eight hundred sensors monitor the site continuously. By 2024 — prior to full fleet deployment — the autonomous trucks were reportedly achieving 120 percent of human-operator productivity.

The connectivity layer is as significant as the hardware. Huawei’s 5G-Advanced network provides sub-20-millisecond latency across nearly 200 kilometres of mine routes. Huawei’s Pangu 5.5 盘古大模型 5.5 — a 718-billion-parameter industrial AI model deployed across more than 500 scenarios in over 30 industries — manages fleet coordination, geological monitoring, and predictive maintenance, while BeiDou satellite positioning 北斗卫星导航系统 substitutes for GPS.

The most important strategic revelation is that every component of the operational technology stack is Chinese. Equally significant is the fact that Pangu 5.5 is not a domestic product for Chinese mines alone; it is explicitly positioned as a global industrial AI platform. In addition, battery-powered autonomous haulage, 72% of whose global fleet currently operates in China, could likely follow the same commercialisation trajectory as Chinese electric vehicles: cost-competitive, technically capable, and backed by state subsidies that allow pricing strategies unavailable to Western competitors. As these systems become commercially available to mine operators worldwide, adoption creates operational dependency. A mine that runs on the Chinese operational technology stack and Chinese autonomous vehicle systems is not independent of China in any meaningful sense, regardless of who holds the mineral licence. And this is the competition that Washington’s current mineral strategy has neglected.

The Arctic as the Sharpest Case Study

No region makes this argument more clearly than the Arctic. The Arctic holds substantial deposits of rare earth elements, nickel, cobalt, copper, and other critical materials within the territory of NATO-aligned democracies: Sweden, Finland, Norway, and Denmark (via Greenland). The difficulty is that Arctic mineral potentials cannot be fully unlocked by the conventional mining model. Its operating constraints — extreme cold, lack of adequate infrastructure and small local workforce to draw on — make the labor-intensive extraction model that has governed the industry for generations structurally unviable. As such, it ought not to be surprising that mining operations in the region are typically much more capital intensive compared to temperate projects before accounting for the full social overhead of human presence. Greenland, to take the most prominent example, is three times the size of Texas with just 93 miles of paved road, and ports whose combined throughput would be exhausted by a single active mine at scale.

The only operational model with a high chance of making Arctic mining commercially viable is one built around autonomous operations: AI-driven systems that manage extraction with minimal continuous human presence and monitored remotely through satellite-enabled connectivity. This, to a large extent, is synonymous with the current operational logic being applied at Yimin in conditions directly comparable to much of the Arctic. And it means that unlocking the Arctic’s mineral potential — which could be central to any credible Western strategy for attaining a sufficient degree of mineral resiliency — requires mastering precisely the technology in which China currently leads. Hence, if the US and its allies do not close this technology gap, securing Arctic mineral access may become an exercise in building infrastructure that would ultimately be operated on Chinese terms.

The Arctic’s value in this framing, moreover, extends beyond terrestrial extraction alone. As the closest accessible analogue to space operating conditions, it is also the natural proving ground for the autonomous systems that US-led space mining ventures are developing for deployment beyond Earth’s atmosphere (more on this later).

US Partners Are Already Moving

For their part, European countries (specifically Sweden and Finland), are building a genuine cluster of autonomous mining technology that is among the most advanced in the Western world. These efforts, however, remain underfunded relative to the problem, poorly coordinated at the EU level, and largely invisible in the strategic debate in Washington.

Sweden’s LKAB is running one of the most ambitious underground mine automation programmes in the world. The Sustainable Underground Mining consortium, which brings together LKAB, ABB, Epiroc, Combitech, and Volvo at the Konsuln test mine in Kiruna, is explicitly targeting a new global standard for sustainable, autonomous, carbon-free underground mining. LKAB has also deployed Sandvik’s AutoMine autonomous system across a growing fleet of electric loaders at Kiruna. At Boliden’s Aitik mine, moreover, Komatsu’s FrontRunner autonomous haulage system is operating at commercial scale; the first such deployment in Europe and in an Arctic environment. These operational deployments matter because they are generating actual performance data in the specific environmental conditions that broader Arctic mineral development requires.

Finland’s latest National Mineral Strategy, on the other hand, explicitly connects technology leadership to mineral security objectives, and thus it ought not to be surprising that Helsinki has taken the lead in solving the foundational connectivity problem that autonomous Arctic mining depends on. Its 6G Flagship program is amongst the first government-funded initiatives in the world to explicitly investigate how next-generation millimetre-wave frequencies behave inside mine tunnels; a prerequisite technical problem for full underground autonomy. Equally significant, Nokia and Boldyn Networks have built an underground 5G test network at the Pyhäsalmi mine.

Australia has deployed over 927 autonomous haulage trucks across its mines backed by the world’s first regulatory Code of Practice for Safe Autonomous Mining and a federal government investment of AU$566 million in geological mapping and mining technology infrastructure. Australian operators have, arguably, more real-world operational data on autonomous large-scale mining than any non-Chinese actor.

The picture that emerges is of a concrete, distributed Western capability in autonomous mining technology; assets in the right jurisdictions generating operational knowledge in the right conditions which have not been assembled into a competitive program. Sweden and Finland are building the technology, while Australia has the largest deployment base and is pioneering a regulatory framework for autonomous mining. The United States has the financial instruments, the political leverage, and, through the Mine of the Future initiative, the beginnings of a federal research commitment. But the West still lacks a coordinating structure that would make these complementary strengths add up to a coherent challenge to China’s lead.

The Alliance Washington Has not Built but Urgently Needs

This is the new strategic argument that the current US minerals agenda is missing. Pax Silica is structured around a coalition of mineral capabilities; it pairs deposits in one jurisdiction with processing expertise and financial resources from other allies. It is, in its logic, a supply chain coalition. What’s more, the DOE’s Mine of the Future initiative, committing up to $95 million for autonomous mining technology proving grounds and national laboratory research, is a good start but insufficient on its own. The gap between what the US is currently investing in mining technology and what China’s National Development and Reform Commission (NDRC) directed programme has already deployed is wide enough that domestic investment alone will not close it on the relevant timeline. This is an area where allied coordination can become a force multiplier which the US cannot afford to forgo.

What is needed, therefore, is a techno-industrial alliance specifically focused on autonomous mining: a coordinated programme to master, standardise, and commercially scale the operational technology of next-generation extraction designed explicitly to outcompete China. Sweden and Finland bring the most advanced European capacity in underground autonomous systems, the connectivity research those systems require, and Arctic operational environments where the technology can be tested under relevant conditions. Australia brings the largest deployed autonomous fleet in the Western world, a mature regulatory framework, and demonstrated experience in scaling these systems commercially. The US brings federal research funding, the world’s largest defence-industrial complex with direct interest in the dual-use applications of these technologies, and the downstream manufacturing demand that gives the coalition its market rationale.

The structure of such a partnership, however, would need to be more than a procurement framework. It would need to coordinate research and development by marrying Finland’s 6G underground work with US national laboratory capacity, sharing operational data from Australian deployments with Swedish underground test environments, jointly developing the safety certification standards that would allow autonomous equipment to be deployed across allied jurisdictions without re-testing from scratch. It would also need a shared approach to the connectivity infrastructure that autonomous mining — especially in harsh and remote environments like the Arctic — requires. Above all, it would need to move at the speed of a technological revolution rather than a multilateral treaty formation.

Mining Technology as AI Ecosystem Strategy

Crucially, the technology of mining is also a vector for AI ecosystem expansion, and the US should be thinking about it in those terms.

When Chinese companies embed Huawei’s Pangu AI, BeiDou positioning, and 5G-Advanced connectivity in a mine in Inner Mongolia — or eventually in a mineral-rich country in Africa or Central Asia — they are integrating that mine’s operational environment into the Chinese technological ecosystem. Local engineers are trained on Chinese platforms, maintenance contracts create ongoing relationships with Chinese vendors, and the data generated by mine operations flows through Chinese systems. As such, switching to a different technology stack at a later date would amount to an operational disruption that mine operators would rationally avoid. This is the same lock-in logic that has made Chinese telecommunications infrastructure so strategically consequential in countries that have adopted it.

The inverse logic applies to US-led mining technology partnerships. If American AI systems, US-certified autonomous equipment, and allied connectivity standards become the operational backbone of mines in supplier countries, those countries’ mining sectors become embedded in the US-aligned technology ecosystem in institutionally durable ways. A mine operator whose engineers train on US autonomous systems, and whose operational data is managed by US-allied AI platforms, faces a fundamentally different calculus when Chinese vendors offer competitive pricing. Put differently, A US-led autonomous mining technology program has the potential to shape technology adoption decisions and expand the US AI ecosystem in mineral-supplying countries.

Equally important is the potential goldmine of data. Autonomous mining systems are extraordinary data generators. Every operational shift produces a continuous stream of geological readings, equipment telemetry, fleet routing decisions, slope stability measurements, and environmental indicators. The AI models that manage these systems improve — in predictive accuracy, in decision quality, in operational efficiency — proportionately to the volume and diversity of data upon which they are trained.

Through its state-coordinated intelligent mine programme, China is building what amounts to the world’s largest training corpus for mining AI. Nearly 1,000 intelligent mines, coordinated under a unified national program, are generating operational data at scale and in a form that continuously feeds back into the AI models managing them. In this context, the parallel to another strategic domain is instructive. As the National Security Commission on Emerging Biotechnology has warned with respect to biodata, the United States risks a future in which an adversary controls the data infrastructure underlying strategic industrial sectors because that adversary has built tightly integrated, state-controlled ecosystems in which AI and operational data co-evolve, while the US lacks an equivalent unified infrastructure. The warning applies with equal force to mining. Autonomous mining AI trained on Yimin’s operational data, and on the hundreds of similar deployments across China’s intelligent mine programme, will over time outperform systems trained on smaller, siloed, proprietary Western datasets. That performance gap, in turn, translates directly into adoption incentives: mine operators globally will gravitate toward the systems that work best, regardless of their origin.

Share

The United States and its allies currently lack a mineral data commons mainly because there is a lack of recognition that mineral operational data is a strategic asset in the same category as the mines themselves. Operational data from Australia’s fleet of autonomous trucks, from Sweden’s Kiruna test mine, and from deployments across allied jurisdictions sit in proprietary silos, owned by individual mining companies with neither the incentive to share it nor a mechanism to aggregate it for AI training purposes. The raw material for building competitive AI exists but the infrastructure to convert it into model advantage does not. A techno-industrial autonomous mining alliance should therefore include a shared and secure mineral data infrastructure to enable operational data from allied mining deployments to be aggregated, standardised, and used to train Western mining AI systems.

Space Extraction as a Strategic Multiplier

The final strategic dimension here is the direct applicability of autonomous extraction capabilities to the emerging space and asteroid mining sector, and the fact that, unlike autonomous terrestrial mining where China has established a commanding lead, this sector is dominated by American companies.

Asteroids harbor the same class of metals at the heart of today’s mineral security agenda. Valued at current commodity prices, the mineral wealth of near-Earth asteroids runs into the quintillions of US dollars, and individual extraction missions could return trillions in value. The commercial logic has already generated a cohort of US-headquartered ventures developing autonomous extraction systems capable of operating in the vacuum, extreme cold, and communication constraints of space, where a continuous human workforce is not an option.

At the level of operational technology, space extraction and autonomous Arctic mining present strikingly similar challenges. For instance, both demand extraction in the absence of a continuous human workforce and impose extreme demands on robotics, power management, and autonomous systems under degraded connectivity. The Arctic is therefore the natural proving ground where space mining systems could be tested and refined before deployment beyond Earth’s atmosphere. Directing these ventures to Arctic testing environments would simultaneously accelerate their development timelines and expand the operational dataset that Western terrestrial mining systems require, generating performance data in harsh autonomous operations that benefits both sectors.

Equally significant is the fact that the talent, technology, and entrepreneurial capacity concentrated in these US companies represents a latent competitive asset in the broader autonomous extraction technology competition. China’s space mining programme remains at an early stage and lacks an equivalent cluster of commercially-driven ventures operating at pace; thus far, only one Chinese company — Origin.Space 起源太空 — is active in the field. Hence, the primary constraint is probably not a lack of technological capability, but rather the absence of an integrated policy framework that recognises the strategic synergies between the mining and space sectors.

America First Is the Wrong Gear for This Race

The technology alliance proposed here is structurally incompatible with a transactional, leverage-first approach to alliances.

Sweden and Finland are not large economies. They do not offer the strategic weight that compels attention in a framework built around bilateral leverage and deal-making. They are, however, the countries running the most advanced autonomous underground mining programmes and in Arctic conditions that are directly relevant to the Greenlandic deposits the US is reportedly eyeing to develop. Treating them primarily as junior partners in a security arrangement, or as potential targets of tariff pressure in broader trade negotiations, is precisely the wrong dynamic for the kind of deep technical collaboration, such as shared R&D, joint certification standards, open operational data, that a competitive autonomous mining programme requires.

Australia is a closer case. The existing minerals partnership establishes a basis for exactly this kind of technical cooperation, but even with Australia, the logic of “what does the US get” from a bilateral deal is the wrong frame. The question is what the coalition gets from pooling capabilities that are, individually, insufficient to match China’s scale and integration. That question requires a different kind of strategic reasoning than the one embedded in an America First approach; one that treats allied technical assets as force multipliers rather than as leverage points in a negotiation.

More broadly, coalition frameworks require predictability and institutional trust to function. Deep technical collaboration requires relationships built on something more durable than transactional benefit calculations. The allies who hold the autonomous mining capabilities that the US needs to complement its own have domestic political systems that are responsive to how they are treated. Imposing tariffs on them and simultaneously expecting them to co-develop autonomous mining AI constitutes an unwarranted contradiction that will not resolve in Washington’s favor.

The Reframing Washington Needs

The US minerals strategy has achieved important wins in a domain that was ignored for far too long. However, the overall architecture is incomplete in a way that compounds over time.

The technology of mining, the autonomous systems, industrial AI, connectivity infrastructure, and the operational data those systems generate are strategic domains as consequential as processing capacity. Not only do they carry dual-use applications across deep-sea, space, and defence logistics, but their development through allied partnerships is also a mechanism for AI ecosystem expansion; that is, for embedding US-aligned technology and data standards in the operational infrastructure of the countries that supply tomorrow’s critical materials. None of these dimensions are captured by the US’s current framework.

What is missing from Washington’s agenda is an explicit techno-industrial alliance for autonomous mining; a programme that would pair US research investment and downstream demand with Europe’s evolving underground automation expertise, Australia’s experimentation with large-scale autonomous mining, a shared mineral data commons for AI training, and a joint commitment to the regulatory and technical standards that the next generation of mining will run on. This techno-industrial alliance should also be designed to bridge the mining and space sectors by directing US companies developing space-mining technologies towards Arctic deployment while integrating the autonomous extraction expertise concentrated in American space ventures into the alliance’s core capabilities. Failure to do so would most probably lead to a minerals strategy that secures access to mining sites and loses the machines. That, in turn, is not a strategy for mineral autonomy but simply a more expensive form of the same dependency on China.

Check out Nima’s previous guest post on ChinaTalk here:

The Trump administration has argued that the crisis was contained thanks to their aggressive action, but they may be taking the wrong lessons from the avoidance of that apocalyptic scenario.

What happened was the largest unexpected swing in global oil balances: China quietly cut crude oil imports by more than five million barrels a day. Yet there was no corresponding collapse in economic activity, no obvious drop in mobility, and no official explanation from Beijing. Somehow, China stopped buying oil from the rest of the world and started drawing from stockpiles we can’t fully observe.

That single decision may have done more to prevent a global energy crisis than anything Washington or OPEC accomplished.

This matters because it demonstrates that China likely has a stronger discretionary policy lever than the West does. The West is really good at market-driven, private-sector oil production. But through this crisis, we’ve seen that Washington does not have the scale of discretionary policy control that China or OPEC does. This time, China cooperated and did the good thing, at least for the broad economic picture — but we cannot rely on that in the future, and that tool can be used against the West as easily as for it. Western governments must grapple with that discretionary gap and not rest on their private-sector bona fides to get through the next crisis.

Arnab Datta, managing director of policy implementation at Employ America, and Rory Johnston, oil analyst and founder of Commodity Context, join ChinaTalk to discuss:

• Why Rory’s own prediction of $200 oil never happened and why J.D. Vance is thanking the wrong people.

• How China quietly cut crude imports by five million barrels a day with zero visible impact on domestic mobility, and the detective work analysts use to peer into Beijing’s black-box inventories.

• Competing theories for why Beijing backstopped the global oil market — self-interested altruism, a backroom deal during the state visit, or a dry run for a Malacca blockade in the event of a Taiwan contingency.

• What India, the Gulf states, and the rest of the world learned from the Iran War and why strategic reserves are suddenly back in fashion.

Listen now on your favorite podcast app.

The Dog That Didn’t Bark

Jordan Schneider: We had this whole Iran war, and there was a moment when people were worried about $700 barrels of oil — complete, utter calamity. While prices certainly went up, we didn’t quite get the gamma-short-squeeze-deluxe version that the breathless Odd Lots podcasts warned about. Why did the great oil catastrophe of the Iran War of 2026 ultimately not occur? J.D. Vance has a theory.

Jordan Schneider: Arnab, is he right?

Arnab Datta: It’s important to center a couple of things, and then I’ll kick it over to Rory. First, we did see a big price spike going into the crisis — nearly a 60% increase. At one point, physically delivered barrels in some of the hardest-hit regions hit $130, $140 a barrel. There was physical disruption that drove prices up.

More broadly, a number of interventions came together to keep prices below the truly catastrophic estimates. There was a Strategic Petroleum Reserve action from the administration, and it had an impact. Going into the crisis, many analysts claimed it would have minimal effect and that we probably couldn’t release very much. In fact, release levels hit 1.4 million barrels per day — not nothing, and higher than expectations. Capacity was also rerouted through the East-West Pipeline, which filled some of the supply loss.

Then there’s the biggest story, and the one the vice president is conspicuously not mentioning: the China effect. China played an enormous role in keeping prices under those higher estimates. I’ll kick it over to Rory to get into that.

Rory Johnston: This is where I do my mea culpa as one of those breathless Odd Lots guests who in March was definitely expecting $150 — I think I even said $200 a barrel crude.

Jordan Schneider: That was literally you.

Rory Johnston: It was literally me. Why did I say that? All the things Arnab mentioned — the rerouted pipelines, the SPR releases — were known factors. We knew they were going to happen, and indeed they needed to happen. SPR releases are the textbook reaction; it’s why those reserves were built in the first place. Even with all of that, it wasn’t going to be enough to offset the loss of supply through Hormuz.

The reason other analysts and I expected $200 or more was that you were going to need demand destruction at a scale never seen in history. If prices at $120 or $130 destroy one or two million barrels a day of demand, we needed five, six, seven million barrels a day — a notably higher price threshold to drive that behavior.

Jordan Schneider: To be clear on demand destruction: the first few million barrels are the easiest — you change your carpool, you change your commute. Beyond that, we’re shutting down economic value, not just inconveniencing people. Economies actually become less productive, not just 1–2% on the margins.

Rory Johnston: Exactly right. The scale of demand destruction we would have needed was on par with the depths of COVID — except without the “benefit,” in this deranged use of the word, of a global pandemic and forced lockdowns. We needed to replicate lockdown behavior through price mechanisms alone, which is really hard. I kept asking: what price would you need to get back to the depths of 2020? Something like $100 a gallon at the pump — an astronomically high threshold.

That was the logic. What I missed was China’s tremendous demonstrated swing. We always knew China had massive fuel stockpiles. We didn’t know whether they would use them to help the Trump administration — we can get into why later. Even in my wildest dreams — as I said on my second Odd Lots appearance, the mea culpa round — if I had gone back to March and told myself, “Don’t worry, China’s going to cut crude imports by 45%, five million barrels a day, and nothing will change in the Chinese economy,” I would have said, “Obviously you’re an insane person, because that isn’t possible.”

That’s still almost how I think about it, because what we’ve seen in China is the largest swing of any country in history in its demonstrated capacity and need to consume oil — with virtually no repercussions for anything you use oil for. During COVID, we became familiar with proxies like inner-city transit congestion, flight tracking, and the number of trucks passing on freeways — basic measures of mobility within China. None of them has moved considerably, despite the fact that on paper, on our oil balances, China looks like it’s in the depths of a depression — not even a recession, depressionary levels of consumption.

The only parallel on record for this kind of contraction in Chinese refining activity is the depths of COVID zero in 2022. As I said to Joe on Odd Lots: “You guys were just in China — did you see the entire country locked down?” “Well, no.” “Yeah, that’s what’s weird about this.”

We still don’t know exactly how they pulled this off. The only things we know for certain — everything I say from this point forward is more speculative — are that China reduced its crude imports by five million barrels a day, roughly the scale of a massive collective OPEC cut, and single-handedly solved two-thirds of Asia’s spot-market deficit during this period. The other thing we know: visible crude inventories in the country did not draw down to account for this change in imports. That’s all we know for sure. Everything else is speculation.

Jordan Schneider: So the world is on the brink of crisis, China stops drinking oil from the rest of the world and turns on its own tap, and that held us solid for the few months the world needed until Hormuz —

Rory Johnston: And I should say: China is still importing five million barrels a day less. This has not ended. It’s still ongoing, which is the crazier part. If imports had dropped for a month and bounced back, okay, maybe we could make sense of it. We are now months into this reduction with no official word on what’s happening — and no change in mobility.

China’s Five-Million-Barrel Mystery

Jordan Schneider: Let’s go through some theories. It’s not shocking to me that someone would show Xi a number for how much petroleum to keep under mountains and in salt caves, and he’d come back and say, “No, we need to 20x this” — and that’s what happened over the past ten years. What are folks in this world thinking about how the dynamics actually played out?

Rory Johnston: ChinaTalk listeners will be aware of the idiosyncrasies of Chinese economic data, but some important details. We don’t have official Chinese data on oil or petroleum-product demand, and we don’t have official data on inventories. What we do have is reported refinery activity and trade activity — customs reports on China, which I validate independently through tanker-tracking data.

From that data, we construct what’s called apparent demand: refinery output of, say, gasoline, balanced for net trade — essentially a measure of domestic disappearance of fuel within the Chinese economy. The challenge is that without good inventory data, inventories are a big black box. High apparent demand could be moderated by inventory builds; low demand, like we see right now, could be moderated by big draws.

Here’s what we know for sure: imports fell five million barrels a day, and Chinese refining runs fell about three million barrels a day — again, the steepest contraction since COVID zero in 2022. That leaves a two-million-barrel delta in crude. Where did it go? We don’t see it coming from visible commercial tanks. Because China doesn’t publish official data, we use satellite imagery to track the floating roofs on storage tanks — you can see how high the roof sits relative to the catwalk, which gives a rough measure of how full the tank is. Until very recently, those floating-roof tanks were still fuller than they were on March 1. So that isn’t the solve.

China does have underground SPR storage for crude, which by definition has no floating roof. How much they hold is speculative. The official numbers we know: 131 million barrels of capacity across six underground caverns — likely more. That is suspect number one: it could have been drawn down without our knowing.

Then there’s apparent demand itself — refinery output netted for trade — which is also showing its largest contraction since 2022, with no change in mobility indicators. Because China has no official inventory data and most refined products aren’t stored in floating-roof tanks, there’s a massive black box in refined-product storage, both commercial and strategic.

One more sanity check: for a genuine, massive contraction in Chinese demand, we’d need high prices. Beijing, through its normal fuel-regulatory procedures, limited domestic pump-price inflation through this period. Despite global fuel prices doubling through the crisis, Beijing pump prices rose only 30%. Is a 30% price increase enough to produce COVID-level demand depression? It seems unlikely. What it did do: with crude still expensive and domestic prices weak, refining margins collapsed. So some of the refining reduction is economically explainable. But we still don’t know how mobility was maintained.

The Occam’s razor solve is that Beijing, in addition to crude SPRs, holds ample strategic stocks of refined products — gasoline, diesel, jet fuel. Your refineries shut down a good chunk, your trade corridor is essentially capped because all of Asia is in crisis, and you start releasing those fuel barrels to the market.

We suspected refined-product stockpiles existed before. I wrote a piece in 2023 called “Chinese Oil Demand Doubts,” noting that coming out of COVID zero, China posted the strongest year-on-year demand growth of any country in history. That seemed odd, because 2023’s economy didn’t seem that strong — it still seemed to be struggling out of COVID zero — so all-time-high product demand didn’t make sense. I argued then that China could be building strategic stocks of gasoline, diesel, and jet fuel, which would appear as apparent demand because we can’t see the stocks. At the time, suspect number one was preparation for an incursion across the Taiwan Strait — 2023 was the peak of Beijing’s bellicosity, buzzing Taiwanese airspace and running naval drills off the coast. Now it looks like this may have been preparation for a different catastrophe, like the one we saw now.

There are still open questions. There could have been flex in the petrochemical supply chain — drawing down intermediate petrochemical goods rather than refined products, or switching to gas- and coal-based chemical feedstocks instead of oil-based naphtha. You can tweak a lot around the edges, but it’s hard to make it add up to five million barrels a day. Most likely, they pulled all levers simultaneously, including some injection of strategic stocks we can’t see, because the numbers are just too big.

To put it in perspective: the delta between China’s average import pace from December through February and the cumulative reduction in Chinese crude imports through June was roughly 450 million barrels. That’s more than the entire IEA collective reserve release — if they did in fact release it.

Jordan Schneider: What this reminds me of is 2008. We have a global recession; America pulling some stimulus levers, doing okay. Then the big question: what’s China going to do? This was pre-Xi, but Chinese economic policymakers basically decided, “No, we’re not doing a recession like everyone else. We’re going on the biggest building spree the world has ever seen and keeping the party going.” That made 2009–2013 a whole lot more pleasant for the US and the rest of the world, because of a big injection of fiscal stimulus that wasn’t manifesting anywhere else.

Trump has a lot more to thank Xi for than he may realize. It wasn’t just Bessent and Wright — they didn’t have the levers and reserves to smooth out what would have been a crisis, and the war might have had to end faster, on even worse terms, if not for Chinese economic policymakers once again refusing to face down an economic contraction when they had levers to control it. There are a lot of interesting second-order implications for scholars thinking about future Chinese economic decision-making, which I won’t make you two answer. Let’s park that for now.

Arnab Datta: One quick addition. Rory unpacked what this meant for oil markets, but early in this crisis, the same was true in the LNG market. One instructive stat: by April, China had already resold over a million metric tons of LNG cargoes to Southeast Asian countries. That April figure alone dwarfed the full-year records set in 2023 and 2025. This wasn’t limited to oil — it happened in natural gas too.

Inside the SPR Playbook

Jordan Schneider: Arnab, Rory just described the scale of the bazooka China could point at energy and commodity markets. What was Secretary Wright working with when the Iran War started, and what lessons should the executive and legislative branches learn from how it played out?

Arnab Datta: Despite my reservations about this war in general and the execution of aspects of it, once the administration decided to participate in this global release, the execution was pretty good. A few things to draw out.

First, they were able to release up to 1.4 million barrels per day — higher than many expected, and higher than the peak during the Russia-Ukraine release. That’s the result of a ten-year investment program called Life Extension 2, which upgraded SPR facilities. There are four SPR sites; the two biggest had completed their Life Extension 2 upgrades in recent years. During those upgrades, release capacity was limited because certain sites were simply closed. We were able to release more than expected because we invested in the SPR’s infrastructure.

Second — and even though I’ve used the term “release” this time around — it’s not technically a release. It’s an acquisition. The authority Secretary Wright used is called an exchange: you release product onto the market now, essentially as a loan to intermediaries, who sell it into the market. But those parties commit to return the same barrels plus a premium to the DOE at a later date. Everyone the DOE has released oil to since the war began in March has to return those barrels plus a premium. Particularly at the beginning of the crisis, those premiums were very high, reflecting a highly backwardated, tight market. Companies awarded barrels committed to returning premiums of up to 20–25% to the SPR, starting in November of this year and running through 2028.

Third, the exchange mechanism demonstrated that you can reward contracting parties who commit to giving you more on the other side. They designed the exchange so that the companies awarded barrels now were the ones committing the most premium later. Beyond the 22% minimum premium listed, companies have to offer an additional 2%, 3%, 8% — we don’t know the actual numbers because the contracts aren’t publicly available. To get oil from the SPR right now, you have to commit to a higher premium, which creates a competitive dynamic among the traders and other contracting parties. There’s some worthwhile creativity there.

Finally, and most broadly — coupled with Rory’s analysis of the likely Chinese release — strategic reserves work. There’s a lot of pessimism about strategic reserves, but this episode shows that when you build this infrastructure, along with other swing infrastructure — the East-West Pipeline, swing capacity to produce and transport — it’s worth the investment for exactly this reason. The fact that oil prices never reached the truly harmful level of demand destruction — where economic activity diminishes and manufacturing shuts down — is a success story. But the fact that it happened largely because of China, not the US, is something to think about for future policy.

Jordan Schneider: Rory, your reflections on the utility of having a lot of things in reserve?

Rory Johnston: I completely agree with Arnab. I’ve always been a fan of redundant infrastructure. Over the course of most of my life, we’ve seen an ever-greater pursuit of efficiency at the cost of redundancy and flexibility — and over the past five years at least, we’ve seen unprecedented shocks to every supply chain imaginable.

We were very, very lucky in this crisis. We went in with about the most overbuilt inventory position we’d ever had — a glut for a year prior. Everything was about as good as it could have been. And still, if not for China, we would have been in a deep, deep world of hurt.

White Knight, Backroom Deal, or Dry Run?

Rory Johnston: The other thing worth dwelling on: we’ve talked a lot about the how. The deeply uncomfortable question is the why. On the how, we at least have some data to guide what’s plausible. On the why, it’s completely open-ended. I’ve developed a little catalog of options — some more sanguine, some deeply tinfoil-hatty — and I’d love your thoughts too, Jordan, because while this makes some sense à la 2008, there are weirder elements here than the straight comparison.

One explanation is that this was an altruistic move with an element of self-preservation. East Asia and Europe were going to be the two regions hit hardest by the Hormuz shock; North America, given the shale revolution and the Canadian oil sands, was far more insulated. We know China is in the midst of a long-term breakup with the United States on trade, or at least an attempt to diversify away from the US market. If you’re diversifying away from dependence on the US consumer, your only other real options are East Asia and Europe. Letting those regions collapse into a Hormuz-driven depression doesn’t serve your long-term interests. The trouble with this theory: if Beijing were coming in as the white knight — “Trump has abandoned you, and here I am to save you” — it probably would have used the opportunity to gloat, to broadcast its global stewardship. Possible, but not a complete solve.

The more worrying explanation: the Trump administration made a high-profile state visit to Beijing during the crisis. At the time, it seemed like a disappointment — no big trade announcements, no full-throated endorsement from Xi Jinping to reopen the strait. It looked like a failure. What if it wasn’t? What if there was a backroom deal in Beijing: “Support us on this Hormuz stuff — we know you’ve got gobs of oil at your disposal — and we’ll do you a solid”? What could it be? The most obvious answer always comes back to Taiwan and regional geopolitics. Given the devastation of US military equipment in the Middle East during the first month of the Iran War, the United States pulled a lot of equipment from Asia — drone missile defenses, radar systems, things that would be very handy in an incursion across the Taiwan Strait. This plays into the Donroe Doctrine thesis —

Jordan Schneider: I’m going to stop you here, Rory. The correct explanation is probably the simplest. If someone shows Xi Jinping a chart and says, “If you don’t pull this lever, we’re going to have COVID-lockdown conditions again,” he’s going to say, “Okay, pull the lever.” Keeping the domestic economy from collapse has been the prime directive of all Chinese economic, fiscal, and monetary policymaking since — start the clock at 2005, whatever. All the second-order impacts — do we care whether the war is longer or shorter, what does this mean for American defense posture — are downstream of the central concern you alluded to at the beginning of this podcast: the potentially cataclysmic impact on economic activity of a gamma squeeze in oil and gas. Once you decide that’s not going to happen in China, it has these global consequences of lower prices. It would have been shocking to me to find some justification for incurring that economic pain at that moment for some 4D-chess play of “this will really screw over the US.”

Rory Johnston: May I push back? If that’s the case — we know China has had an appetite for building crude reserves since the ’90s. Why would they have stopped buying? Why would they voluntarily blow the massive energy-security blanket they built, unilaterally, at a moment when everyone acknowledged that the main thing that would push Trump to end the war was the price response from the market? Beijing’s pretty smart. The reason I initially thought Beijing wouldn’t come to the rescue was precisely that they had the domestic reserves to insulate themselves. Instead, they’ve done much more than insulate themselves — they’ve backstopped the entire global oil market.

If you built up this massive energy-security asset, why blow it all at a moment driven entirely by poor US foreign policy? Why not insulate yourself, create the conditions that cause pain for the United States, and totally save your own citizens? They could have protected domestic Chinese consumption entirely without subsidizing Taiwanese, South Korean, and Australian imports and everything else.

And even granting your explanation — why are they still not buying? We’ve seen them buy far more at much higher prices. It doesn’t pass the smell test — unless it was a bazooka to save the Chinese economy that, out of policy inertia, can’t be turned off once the floodgates opened. That may be plausible. But they’re doing three times as much as they’d need to save the Chinese economy alone, in my humble opinion.

Arnab Datta: A question here — I probably lean toward Jordan’s explanation. Maybe they can insulate consumption within their own economy. But isn’t the pain in the Southeast Asian countries also important to China’s economy? China has a weak housing sector right now. If Jordan’s right that the economy is everything, sure, you can protect internal consumption — but if all the Southeast Asian countries go through enormous recessions because they’re not protected, that hits China’s economy too. It’s not altruism; it’s still rooted in self-interest. But it amounts to, “We need to help those countries.”

Jordan Schneider: Do you want to roll the dice on a global recession if you can prevent it?

Rory Johnston: And that is still my base case — the first explanation I mentioned. But it leaves more questions than it answers. One final explanation — not a US conspiracy theory, but less optimistic: China has been building this massive energy-security blanket explicitly for a Malacca Dilemma future, shoring up for a potential conflict with the West. What if this was the world’s largest dry run? Beijing’s trying to figure out whether this policy suite would work. How could you ever simulate the effects of a blockade of Malacca? Suddenly losing Hormuz is a pretty interesting natural experiment: try the system. Does it work? Can we flex petchem? Can we release the SPRs? Can we insulate China from the seaborne market? Very frankly, the answer has proven to be yes.

Whether or not we ever learn the why, the ultimate implication is that China has demonstrated a larger swing capacity in global oil balances than Saudi Arabia, than the United States, than any other country in the world. We’ve always talked about China dominating energy-transition industries — renewables, solar panels, batteries — while North America held the edge on fossil fuels because of the shale revolution. This has proven pretty squarely that the SPR is a discretionary policy tool. Washington does not control the shale patch. Beijing controls whatever is happening right now in China. They have demonstrated much greater discretionary power in the fossil-fuel market than basically the entire collective West. That’s a staggering outcome.

Jordan Schneider: Because they don’t even make this stuff. That’s the weird thing — this isn’t downstream of market forces, of drilling more in West Texas. This is just a decision that was made — decades in the making at this point — to have the tools in the toolkit to bring to bear in a crisis.

Rory Johnston: And it’s working. And we don’t even know what exactly is working, because prices have come back full circle through the crisis and Beijing is still five million barrels a day out of the market. Why?

Jordan Schneider: On the slow-to-turn-off-the-tap question, the Occam’s razor is probably “we just held onto this too long.” This is campaign-style policymaking in China: you always overshoot the target. Once the target’s set, everyone runs toward it, and then they run a little too far. So, potentially less Machiavellian.

The Politics of Stockpiles

Jordan Schneider: One conclusion is that it would be nice for the US to be able to smooth out crises like this — for energy markets, rare earths, what have you. The other conclusion, which I could imagine some Democrats on the verge of controlling the House coming to, is that this enables an executive branch to start wars we don’t approve of — that building a bigger security blanket, so the economic pain of foreign adventures we think are stupid doesn’t hit as fast, will actually lead to more of this happening, not less. What’s your sense? Is this a dangerous insurance policy — selling flood insurance to people in the hurricane path — or prudent national economic management?

Arnab Datta: You’re seeing versions of this in the critical-minerals space too, where the Trump administration has been incredibly active — using the toolkit available to it in very robust, arguably legally creative ways, certainly pushing the bounds of executive authority. There’s a push to say all these things are furthering an administration whose policy is unlawful, corrupt, et cetera. But the push to restrict these tools would be incredibly misguided.

Go back to 2022, when Russia invaded Ukraine. That was entirely out of our control. And one distinction worth noting, going back to our last conversation: the global community begged China, for all intents and purposes, to release from its crude stockpiles at that time. Instead, they kept purchasing — adding demand when the market was incredibly tight. We can’t count on external forces. Every instinct to restrict a tool like the SPR because of its worst-case scenarios also drastically reduces your ability to use it in a genuine crisis not of your making. You can’t design a tool that perfectly matches your preferred policy outcomes while minimizing all bad ones. That’s just not how it works. President Trump was elected democratically; he chose to engage in this war. This is a place where Congress has really let its authorities lapse on war powers.

Restricting a tool that’s incredibly useful for macroeconomic stability would be really misguided at a time when these disruptions are increasing. This is the third disruption in six years. COVID was a downside price shock, but one with enormous repercussions for our domestic shale sector — it created the tightness that made the Russia-Ukraine crisis much worse. Then we had Hormuz. Go further back and you have the Saudi price war. This is part of our daily lives now, and it’s better to be equipped with a tool than not. We can talk about legislating guardrails, but limiting the tools would be a mistake.

Jordan Schneider: Rory, any thoughts or lessons for Western policymakers?

Rory Johnston: I’d dovetail on that — Arnab and I have worked a lot on the SPR file over the past couple of years. SPRs are a bidirectional tool. We always discuss their capacity to offset supply shocks by replacing lost supply. But as we’ve seen, China has been buying oil hand over fist for its SPR stock-building, and that buying manifests as effective demand in moments of especially weak global consumption, like during COVID. That’s a similarly stabilizing force. The reason OPEC has always mattered in global oil markets is that the only thing worse than high oil prices is boom-bust oil prices — it’s the volatility that drives the most deleterious economic consequences. OPEC was handy because it could increase production in high-price moments and cut in weak ones. But OPEC has become, very frankly, increasingly bad at doing that job reliably. China has proven to be a more reliable, technocratic manager of these boom-bust cycles — the way OPEC used to be.

The US SPR can serve a similar role. The pushes to refill the SPR to the brim right now are as misguided as the pushes to draw it down to zero. Empty SPR capacity — the capacity for discretionary buying — is in many ways as valuable as the capacity for discretionary selling.

Arnab Datta: Picking up on that: we had the shale revolution and became the world’s largest oil and gas producer, but we’re still vulnerable to the global market. As Rory said, it’s becoming more volatile and unpredictable, and OPEC can’t be counted on to stabilize it, even in its own interest. The real question is whether we want to build that resilience into our economy, and what’s necessary to do it.

We have to take the reality of the US shale patch seriously. We don’t have the low-cost production Saudi Arabia does — one reason they’ve been the most effective swing producer is that they can ramp capacity up and down quickly at very low cost. Our swingiest swing production is in the shale patch, and new investment takes about six to twelve months to ramp. In a shock like this, that limits you. But we also have near-term swing capacity in drilled-but-uncompleted wells — DUCs — which take about four weeks to ramp up. There used to be a lot of these; they’re now at a historic low in the Permian Basin. If we oriented SPR acquisition or other policy supports toward increasing DUC capacity, we could ramp production faster in this kind of crisis. The policy toolkit will look different — to some extent it means subsidizing companies to mitigate risks. But the broader question is: do we want to be subject to OPEC and China in the future? I don’t think so. Minimizing our exposure to that volatility, or to China’s self-interest, is a worthwhile policy goal.

Rory Johnston: One quick dovetail. North America and the United States are clearly the market-based energy power globally. But that power is entirely driven by market forces, not discretionary government policy. What’s happening in China, at least in my view, is primarily a discretionary policy choice — a choice of government that can be beneficial to the West, or in another world nefarious. Same with OPEC. Casting OPEC as the enemy is about as old as oil policy itself; it’s never been believed that the West’s and OPEC’s interests are aligned — even less so now that the effective second-most-powerful member of OPEC+ is Russia.

So when we talk about Western discretionary energy-policy pull, this is where the SPR discussion comes in. Canada and the United States are absolute energy juggernauts; the West is leading non-OPEC oil-production growth by leaps and bounds, partly because our market systems let the private sector take these risks. The difficulty is that overdependence on market-driven private actors limits our governments’ capacity to effect change in a pinch. The market will always take longer than OPEC or Beijing to turn on a dime in a crisis, as we’ve seen this time.

Share

Everyone Wants a Reserve Now

Jordan Schneider: What are all the other countries going to do? Are they just screwed?

Rory Johnston: One thing we saw through this crisis: a lot of countries didn’t have SPRs. SPRs weren’t cool anymore. Everyone said, “We have the energy transition, we’re going to electrify, we haven’t had an oil crisis in forever,” and got out of the game. Everyone grew to deeply regret that after this crisis — particularly in Asia, where many less wealthy countries lacked the economic and state capacity in the 1970s, when most SPRs were built, and so missed that bandwagon.

Most notable is India. India didn’t have a major SPR going into this crisis and was ground zero for human pain. Not only does India use fuel for transportation like everyone else, but between a fifth and a quarter of total Indian petroleum demand is LPG for cooking fuel. In Maslow’s hierarchy, cooking beats driving. Coming out of this crisis, New Delhi has given strong indications it will build out SPRs, and there’s already confirmation that the Indian government and ADNOC — the Emiratis’ national state oil company — are building a collective SPR in India, currently thought to be 50 million barrels. It serves both parties: it keeps Emirati crude and product on the safe side of Hormuz if this ever happens again, because the Gulf producers themselves also want their oil outside Hormuz. They’ll follow the lead Aramco has taken in stationing stocks all over the world, which helped keep supplies flowing despite the disruption. India is the first major mover, but you’ll see this across much of the rest of the world — countries without SPRs are going to want them after this.

Jordan Schneider: Let’s talk about the Gulf response. Watching the World Cup, I kept seeing Savvy Games Group buying advertising — the PIF buying up every video game company in the world. When are these guys going to stop purchasing Manchester United and start buying drone defenses and building new pipelines and refineries? Has that already happened? This is a huge reckoning for the whole region, right?

Rory Johnston: We’ve already seen some of it. The big flashy investment commitments the Trump administration extracted from the Gulf states have all basically been turfed because there are much more pressing things at play. And they are already building additional infrastructure to increase resilience against future Iranian threats. They’re trying to diversify around the Strait of Hormuz — you can’t entirely diversify around Hormuz, it’s a pretty big thing — but the East-West Pipeline, which proved extremely valuable through this, is getting expanded for sure. The ADCOP pipeline — the Emirati line between the Gulf side and Fujairah, their major blending port on the Gulf of Oman side — was already being twinned to double its capacity. That expansion is about 50% done, and they’ve understandably accelerated it. On top of that, they announced a batch product pipeline that can ship diesel, gasoline, jet fuel, and everything else.

Everyone’s going to do this. The challenge is they’ll also need to invest in a hell of a lot of drone and anti-missile defenses. While the easiest way for Iran to exert control is threatening traffic through Hormuz, Fujairah was attacked and bombed multiple times through this crisis. Iran can clearly hit these other assets, plus the upstream wells that fill the pipelines. Iran will never go away as a threat.

They’ll all try to diversify around Hormuz, but it’s more expensive. Even if you’re paying a massive toll — like $2 million toll for a VLCC carrying two million barrels — no one wants to pay a dollar-a-barrel charge at Hormuz that used to be free, but you’re not going to build a pipeline around Hormuz for less than a dollar per flowing barrel. The natural waterway will always be cheaper. It’s strategic insurance they’re going to pay for, but it will take time, cost a lot, and not be immediately remunerative.

Jordan Schneider: Can we come back to India and the rest of the world building up their own reserves? We have another 2008 echo, where everyone concluded, “We just need lots of dollars so nothing bad ever happens to us.” Is there a world where, in the next crisis, everyone has enough of their own stuff to produce the smoothing that China — because the CCP is super paranoid and was worried about energy security way before it was cool — provided this first time? To what extent does the buffer stop being solely America’s to create, with the whole world getting in the game?

Arnab Datta: One important thing: in oil markets, strategic reserves rest on incredibly developed market infrastructure. This is the most liquid commodity market in the world — maybe the most liquid market outside of dollars. There’s resilience built in because of how developed that market is. Strategic reserves are particularly helpful in tail-risk events, when you have huge losses and want to do a big, chunky release.

But oil storage, especially for refined products, is expensive. Crude is easier to store long-term; refined products tend to evaporate. You can invest in the infrastructure, but it’s costly, and you can’t always predict which market will be disrupted. We’ll certainly see a push to build more resilience, but it’s very difficult for a country to build sufficient capacity to insulate against the whole suite of shocks. Some of this needs to be coordinated, and we should support private market resilience as well — strategic stockpiles matter, but so does swing capacity, like DUCs. Other countries will need to think about that too.

Rory Johnston: This is something you and I have talked about a lot: how to structure the economic framework for how these reserves fit into the system. The legitimate criticism of this kind of thinking is that the private sector already does this — it absorbs swings, holds stockpiles, follows economic incentives, and does all of it much better than governments.

My main criticism of the Biden administration’s SPR release in 2022 was that it wasn’t responsive to market signals. I agreed with the release when it was first announced in April. But by the time it hit full stride late in Q3, the crisis was largely past. At that stage, the SPR release was feeding an oversupplied market — procyclically adding to volatility rather than serving its purpose as supplier of last resort.

If SPR policy is going to be effective within a broader, much larger, much more dynamic private system of inventory and merchant storage, we need a rules-based system that’s far more responsive to the market as industry understands it — not just a bazooka. To your earlier point about policy inertia in China, Jordan: policy inertia is a problem here too if we’re trying to fine-tune volatility in fast-moving markets. We can’t have a situation of “we’ve clicked the button and now, whatever happens, we can’t back out because it was so hard to get going in the first place” — which was my understanding of what happened in the Biden administration: “We worked so hard to do this big release; we’re not backing out now.” Understandable on a human level, but backing out is what was needed. We didn’t need to draw the SPR down as far as we did in 2022, because the bulk of the drawdown happened when prices were already in freefall — it was too much of a good thing at that stage.

Jordan Schneider: One final question for you, Rory. Arnab’s energy is median-DC-policy-person energy. Where do you put yourself among the other commodity analysts? Are you an outlier, or is everyone wired like you?

Rory Johnston: In terms of personality, I hope not — for their sake. But in terms of my view on this —

Jordan Schneider: No, no — we’re not talking about the calls. I mean your way of being: level of enthusiasm, level of excitement.

Rory Johnston: Is my way of being normal for oil analysts? I don’t think so. The best oil analysts are typically very level-headed. The irony of this crisis is that everyone got to know me as this alarmist, ranting-and-raving permabull — “Oh my God, he went on Odd Lots and called for $200 crude. He must be insane. He must have always been bullish.”

Jordan Schneider: I think you called for $600 crude on this podcast.

Rory Johnston: I was so bearish going into this year. I expected a massive glut. I did not expect the Iran War. I did not expect the Strait of Hormuz to be closed. This was the biggest shock anyone could imagine in the oil market, and the fact that we avoided the worst consequences is actually very heartening for the global system. It’s just deeply uncomfortable as an analyst that we still don’t fully understand how we managed to pull it off — or why Beijing helped as much as it did.

Jordan Schneider: Amazing. And one final question: what’s the modal music genre of choice for the oil and gas industry?

Rory Johnston: Whew, I don’t know. I’m a bad millennial lover of what they call stomp-clap music in the most pejorative sense —

Arnab Datta: We should do Arcade Fire or something like that — the Canadian roots. Maybe “Rebellion,” or “Wake Up.” “Wake Up” is a good one.

Jordan Schneider: Well, I can’t use the real songs anymore, so I have to make my own. I’ll send you some of the Iran War ones. We’re on Spotify now: ChinaTalk Records. Check us out.

This was really fun — a nice break in routine. Talk about too much AI and you forget there are actually things that are brown and sludgy that move around the world and matter.

Arnab Datta: That the world still runs on.

For all its technological feats, China’s push for high-quality innovation has not made the economy that much more productive. This lends greater urgency to other productivity drivers such as firm efficiency — an exigency that Beijing is hardly unaware of. In recent months, the central government has outlined a series of regulatory measures to discourage local officials from propping up inefficient firms. But nothing stops local leaders from circumventing those regulations. The Chinese bureaucracy continues to reward officials who hit GDP growth targets, minimize unemployment and social unrest, and defer to the dynastic mandates of President Xi Jinping. Until the underlying incentive to keep inefficient firms afloat is recast, China’s tireless pursuit of productivity-led growth will continue to bear little fruit.

The Elusive Quest for Productivity 强弩之末

(Bonus points for guessing which book this header is inspired by!)

As an economy matures — like China over the last decade, growth increasingly moves away from capital accumulation. Because there is only so much equipment that one can “fit” a worker with, additional capital will eventually begin to yield smaller increases in GDP. More formally known as “diminishing returns to capital”, this phenomenon in part explains China’s decelerating growth in recent decades. Adding fuel to the fire is the country’s protracted fertility decline, which reforms to the one-child policy in 2015 and 2021 have failed to reverse. Economists predict that shrinkages in China’s working-age population will reduce per-capita GDP growth by 10 percent over the coming decade.

These unenviable economic circumstances necessitate a shift toward productivity-led growth. And who would be more aware of such a need than the Chinese Communist Party (CCP)? The CCP’s latest Five-Year Plan reemphasizes what have long been cornerstones of China’s broader economic agenda: innovation and technology upgrading. From 2015 to 2023, government spending on R&D in strategic sectors increased nearly fourfold under industrial policy guidance like Made in China 2025. While such lavish spending has given rise to the likes of BYD’s blade batteries and Huawei’s commercial 5G leadership, there have not been improvements in year-on-year productivity growth. The reasons for this discrepancy are manifold — R&D outcomes are inherently uncertain; making new discoveries becomes more difficult as technical fields mature; novel-product innovations do not necessarily make industry processes more efficient; and subsidies may have adversely encouraged low-quality patents over productive innovation. Or it could also be that innovation did bear sizable productivity gains, but they were ultimately offset by the expansion of inefficient firms. Either way, China’s push for high-quality innovation has not made the economy substantially more productive.

Granted, China does not rely solely on domestic innovation for technological progress — it also adopts frontier technologies from advanced economies. But rival powers have imposed lasting technology blockades that stifle this approach. Washington continues to restrict Tokyo and the Hague from exporting semiconductor manufacturing equipment and computer chips made using US technology. Restrictions on US investment in Chinese or Chinese-owned enterprises working on sensitive technologies — such as quantum information systems and AI — have similarly stiffened over the years. And even if there were no trade restrictions, it is unlikely that importing technologies would translate into significant productivity gains, given that China has already caught up to the global technology frontier in many industries.

Overcoming this productivity lassitude requires going beyond technological progress. China needs to become more efficient by dismantling its sprawling constellation of inefficient firms, the worst-performing of which are known as “zombie firms”. What defines a zombie firm is not exactly clear, however. The State Council considers a zombie firm to be a company that has been consistently unprofitable, belongs to an industry with overcapacity, and violated national environmental and quality standards. And so are firms that made three consecutive years of losses and failed to adhere to guidelines on structural reform.1 The International Monetary Fund (IMF), by comparison, defines zombie firms to be loss-making companies that enjoy below-market interest rates, often given by state-owned banks.2

Since zombie firms — and inefficient firms in general — enjoy subsidized credit and tax breaks from local governments, they are shielded from market forces that would otherwise compel them to become more efficient or shut down. Scaling back these financial benefits would thus pressure inefficient firms to restructure themselves or exit altogether. This reduces “wastage” of scarce credit. Instead, credit would be redirected to more efficient firms, allowing them to make greater investments. Eventually, as these firms make up a larger share of their respective sectors, the economy as a whole becomes more productive. In 2017, the IMF estimated that such productivity gains would raise China’s growth rate by 0.7 to 1.2 percentage points, nearly a fifth of the current GDP growth target of 4.5 to 5 percent.3

Feeding the Walking Dead 苟延残喘

But the widening footprint of inefficient firms also reflects the CCP’s difficulty in tackling the problem. Local officials’ promotions and financial rewards are linked to performance indicators established by senior officials. These indicators have undergone significant changes. In the first two decades of post-Mao China, local leaders competed against one another on hard growth targets, such as the annual gross value of industrial output in their respective jurisdictions. Such tournament-style competition backstopped astronomical investments in physical capital, resulting in miraculous economic growth. However, as income inequality and unemployment fuelled popular discontent in the 1990s, the CCP elite began expanding cadre evaluation metrics to include social development indicators like urban employment. Hitting GDP targets alone is no longer sufficient to protect one’s political career; one must also stave off the layoffs that follow firm closures. This provided another reason for local officials to prop up underperforming firms, at great cost to the economy writ large.

Take photovoltaic company LDK Solar (赛维LDK太阳能), for example. In addition to being the biggest taxpayer in Xinyu (新余市), a prefecture in Jiangxi, LDK Solar was a crucial employer in China’s politically favored solar industry. As of 2013, the firm had an estimated total of 10,000 workers despite having accrued $2.8 billion in debt. Unsurprisingly, the company had enjoyed generous state support even after making losses for nine consecutive quarters, including assistance from the local government to obtain a $327 million loan from national banks. According to one insider of the Jiangxi banking industry, the negotiation was “a coordination meeting on paper, but coercive in reality”, with threats from the local government to “terminate all relevant business and sever all relations with any bank that refuses”. The gravity of an LDK Solar collapse — to citizens and thus officials alike — was evident. A year earlier, Zhao Hongming (赵鸿鸣), then spokesperson for the Standing Committee of the Xinyu Municipal People’s Congress, said that the firm was an “industry vanguard” that “employs more than 20,000 workers”, and that “support [for a troubled LDK Solar] from all parties would bring about great benefit to the thousands of employees and local residents of Xinyu”.

To make matters worse, there has been mounting pressure on local officials to align themselves with central priorities in recent years. Xi — in line with his predilection for “top-level design” (顶层设计) — has significantly narrowed the latitude in which local governments decide how to meet central policy objectives. Fearing demotion or punishment, many bureaucrats have erred on the side of excess, plowing even more subsidies into priority sectors. Aside from serving as lifebuoys for inefficient firms like LDK Solar, these subsidies have resulted in underused facilities and infrastructure. Think the ghost-like fabs of the now-defunct Dehuai Semiconductor (德淮半导体), the idle furnaces of Shandong’s steel mills, the graveyards of EVs and shared bicycles in Hangzhou. Xi, too, recognizes the perils of the subsidy glut. In his February address on cultivating sound conceptions of bureaucratic performance, he urged local officials to abandon “deviant or misplaced views” (偏差错位的政绩观), which have led to wasteful expenditure, formalism, and performative projects.

That Xi’s imperial decrees have also contributed to bureaucratic inertia — a well-known phenomenon among contemporary observers — is not actually contradictory. Officials often describe Xi’s directives as edicts undergirded by the wishful logic of “not only, but also, and also.” For example, local governments must not only promote innovation in priority industries, but also minimize financial risks inherent in funding experimentation, and also strictly conform to centrally established standards for achieving both goals, however vague they may seem. Faced with conflicting priorities, local leaders avoid risky or innovative policies, which are needed for furthering domestic R&D through greater private capital involvement. In an almost compensatory fashion, they default to conventional approaches that also suggest alignment with Xi’s mandates — such as pouring subsidies into semiconductor companies — with excessive zeal. This explains why it is often difficult for central directives to inspire downstream policy change. In 2018, for example, the Ministry of Finance called on local governments to submit a list of zombie firms in three months’ time; complete all restructuring, deregistration, and liquidation works by 2020; and impose restrictions on loan extensions and corporate subsidy payouts. One need not look far to know that the above remains a huge work in progress.

The Arrow Misses the Head 扬汤止沸

The productivity costs posed by inefficient firms have not been lost on the CCP elite. As early as three decades ago, former premier Zhu Rongji launched a landmark campaign to reform inefficient state-owned enterprises (SOEs) by transferring their responsibilities in home ownership and public insurance to the civil service. Yet, while efforts to combat firm inefficiency have only intensified in the years since, they often run into an incentive constraint.

To see why, it helps to begin with the basic understanding that local officials have functionally inexhaustible options for supporting inefficient firms. As long as the underlying incentive remains in place, local officials would sidestep regulatory measures by providing new kinds of financial lifelines. Earlier in March, the State Council announced its plans to restrict local governments’ discretion in giving out fiscal subsidies. While the specific prohibitions are not yet known, there remains little stopping officials from taking actions like pressuring local banks to initiate loan rollovers or creating purchase arrangements with loss-making firms.

Consider the vast array of financial benefits that Dayun Automobile Co. (大运汽车) enjoyed. The brainchild of who was once an enterprising 18-year-old grain trader, Dayun Auto has long been the largest heavy-duty truck manufacturer in Shanxi. Its dominance, however, belies a checkered history. Following peak sales in 2017, the automobile behemoth made a misguided hop onto the EV bandwagon. Unremitting competition in the EV industry eventually left an indelible dent in its cash flows, causing payment failures that pushed its suppliers to file a restructuring application against it in November 2024. The firm slashed its headcount by half as a result, and it focused on cranking out heavy-load trucks again. What is most concerning, however, is not its downfall; if anything, it did recover a little since the court-ordered restructuring began. Rather, it is the fact that it still buckled despite the many pillars of state support underneath. The EV subsidies Dayun Auto received in 2017 and 2018, which alone equaled 20 percent of its earnings then, were just the tip of the iceberg. It also enjoyed tax benefits that made up more than a tenth of its profits, purchase arrangements with a Yuncheng (运城市) state-owned firm that guaranteed the sale of one-third of its trucks, and even a highway toll exemption for its trucks to boost sales. In part, the wide variety of preferential deals reflects the limitless ingenuity of local officials, one that specific regulations alone would be unable to constrain. Indeed, as a former Dayun salesman by the name of Zhao Xinzheng remarked to Bloomberg, “It’s impossible for Dayun to go bankrupt. The government would never allow it.”

In addition to regulatory measures, the CCP has set up specialized bankruptcy courts. Since 2007, the number of specialized bankruptcy tribunals and courts has steadily grown to more than 100 across different prefectures. As judges in specialized courts are often better-educated and newly appointed, they are less susceptible to political interference than their counterparts in traditional civil courts. Expectedly, these specialized courts have expedited bankruptcy proceedings and thus the restructuring or liquidation of inefficient firms, raising productivity in their respective local economies.

But like regulatory measures, such judicial specialization still skirts the root incentive for local governments to keep inefficient firms alive. Political interference can also occur before adjudication, where officials intervene through financial lifelines or direct court-government negotiations to prevent — or at least, slow — enterprises from entering a bankruptcy proceeding in the first place. As in the case of Dayun Auto, it was not until several years into its financial decline that non-state entities filed for compulsory reorganization against the Shanxi giant. Only then was it legally required to restructure itself (though restructuring had already been underway in Yuanhang Auto, the EV group of Dayun Auto).

In that case, why not involve the central government in initiating bankruptcy proceedings? Most recently, Beijing announced a pilot amendment to China’s Company Law in seven economic hubs. The amendment would empower central authorities to petition for the court-mandated liquidation of zombie firms that have not voluntarily dissolved, bypassing local-level resistance. While this legislative amendment is a step in the right direction, its effectiveness is ultimately contingent on the identification and processing capacity of central authorities. Politically favored zombie firms may remain under the radar as they continue to operate with generous financing from local governments, making it difficult for central authorities to identify entities for judicial liquidation.

Once Bitten, Twice Shy 入土为安

Of the deluge of central interventions, fiscal auditing most directly addresses the incentive constraint by penalizing local officials for racking up debt, which sustaining inefficient firms invariably leads to. Over the last decade, the fiscal adequacy of local governments has come under heightened scrutiny. In the 2017 National Financial Work Conference, Xi emphasized the need for local governments to “adopt just views on bureaucratic performance, strictly control the rise in local-government debt, and ensure a culture of lifetime accountability and retrospective responsibility”. Also underscoring the urgency of debt control is Premier Li Qiang, who warned local officials to limit debt risks and husband state coffers in a State Council video conference two years ago.

Would fiscal audits alone be sufficient, though? It is unlikely. While Xi’s debt-slashing mandates would most certainly curb reckless subsidy disbursements, they could also discourage local officials from accepting the short-term financial consequences of firm closures, such as the loss of corporate tax revenues and lower land prices. In fact, value-added tax and corporate income tax are the two biggest sources of revenue for local governments, accounting for a total of 56 percent in 2021. Declining land revenues since then have only magnified the importance of enterprise survival to state coffers. Deepening this fiscal predicament is also the fact that most local governments already face fiscal deficits. Aggregate local deficits, as a percentage of GDP, has risen steadily from less than one percent before 2015 to roughly five percent today. When it comes to debt, the situation is far worse: Only one provincial-level region, Tibet, has an adjusted debt burden below 100 percent,4 inclusive of on-lending debt. Saddled with recurring deficits and climbing debt, local officials are likely even more averse to drastic subsidy reductions that would mar balance sheets in the short run.

To actually embolden local officials to bury China’s corporate zombies, locally established debt reduction targets could be incorporated into the fiscal auditing approach. While the selection of specific indicators — such as adjusted debt burden and the debt-to-GDP ratio — would remain the preserve of the central government, local governments would have the autonomy to set appropriate quantitative targets for each indicator and the timelines for meeting them. Officials would be assessed on not just the extent to which the targets have been met, but also how ambitious the targets are; the latter is necessary to encourage bolder strategies to topple inefficient firms and slash debt. Moreover, recent developments in central policy favor such an initiative. Unveiled in November 2024, China’s 10 trillion yuan ($1.5 trillion) debt refinancing plan has allowed local governments to borrow an additional 6 trillion yuan through special-purpose bonds until 2027, and a further 800 billion per year over the following five years. In effect, local governments would be replacing their “hidden debt” — accrued from decades of borrowing through local government financing vehicles — with more transparent forms of debt, assisting central authorities in their assessment of local performance in debt reduction. And as officials in high-debt provinces are more likely to fear the loss of tax and future land revenues, the weightage of debt reduction performance in bureaucratic evaluation could be made higher in debt-laden localities.

Ideally, Beijing should also introduce bureaucratic incentives that encourage market-led allocation of financial capital, which would reduce inefficient state intervention in the survival of firms. Researcher Shengyu Wang remarked to me that many local governments have recently begun to report on private capital mobilization, measured as the amount of private investment attracted for each public dollar of venture capital. He believes this to be indicative of high-level commitment to liberalizing capital allocation for greater efficiency. How much further such commitment translates into incentive reforms will ultimately depend on political will.

China’s bureaucratic incentives are currently stuck in limbo. Once an investment dynamo, they helped transform a backwater into a global economic heavyweight. But as the Chinese economy matures and transitions from investment-driven to productivity-led growth, the cost of sustaining inefficient firms — the woeful byproducts of China’s bureaucratic incentives — would only continue to rise. Whether local officials stop doing so depends on more than just what they cannot do. It is also about what they stand to gain from consigning the corporate undead to the grave.

The lifecycle of AI, from training to output, has tangled itself around the lives of Chinese illustrators in ways that run in opposite directions.1 The machine was fed their work, learned their expressions, and is now displacing them. And in the process, it has forced them to answer a question that no one expected to become that urgent a few years ago: how do you prove you are human?

The answer, so far, involves courtrooms, livestreams, and job markets. It involves regulatory regimes that strictly govern AI outputs but stay lenient on the inputs, a community that hunts for signs of AI in each other’s work and demands live drawing sessions as proof of humanity, and an education system that trained illustrators to produce exactly the kind of consistent, replicable output that AI does better and cheaper. Ultimately, the question may be less about how to prove you are human than about why that proof has become so hard to give. The answer, I will argue, lies less in what AI learned from humans than in what humans built long before it arrived: an exam system that trains young artists through standardized metrics and repetitive practice, which looks unsettlingly like an AI training pipeline. In the end, behind Chinese illustrators’ attempt to combat AI lies the broader educational infrastructure that makes humans indistinguishable from machines. So what type of expression emerging from that system, if any, would be untokenizable?

This is a story about Chinese illustrators. It is also a story about what we built before AI arrived, and what we are finding out about ourselves now that it has.

Feeding

In December 2023, the first Chinese case addressing whether AI training constitutes copyright infringement was filed by four illustrators against Xiaohongshu to the Beijing Internet Court. Xiaohongshu had been developing its own AI drawing service, Trik AI, which was launched in April 2023 to allow users to “create artworks” by uploading a photo. Four creators discovered that some artwork created by Trik AI looked extremely similar to what they had produced in the past. The illustrators claimed that unauthorized use of their work to train Trik’s model violated their reproduction rights. “All the brushes I used in my art, the way that I portray flowers, they are all from me and with my individual characteristics. But you can see from its art the traits of my art,” said one artist.

Xiaohongshu, on the other hand, argued this was “fair use” under Chinese Copyright Law. As of the available evidence, no judgment has been published. Xiaohongshu removed Trik AI shortly after the case went big, but it has not stopped more and more AI artwork-generating services and platforms from emerging.

Chinese legal scholars argue that China is taking a regulatory approach of “moderate leniency,” emphasizing control over downstream AI-generated content (AIGC) while adopting a more permissive stance toward upstream training. The structure has double meaning for illustrators: not only do they have little legal recourse against their work being fed into the machine, but, as the next story shows, the downstream controls designed to catch AI can end up catching them instead.

Witchhunting

On October 29, 2025, an unusual drawing livestream happened. Rather than showing fans how a work is produced and chatting with them, as normal drawing livestreams in China would, this livestream is a bet-on agreement — betting whether the illustrator who initiated the livestream can replicate their work in real time without using AI.

The illustrator, who goes by “Angel from Mercury” 来自水星的天使 (abbreviated to 水星 Shuixing), had published a commissioned work on Xiaohongshu, which the platform flagged as “may contain AI.” Under AI labelling provisions that came into force in September 2025, platforms are required to auto-detect and flag suspected AIGC (More details on China’s AI labelling system can be found in the explainer piece: Labelling AI-Generated Content in China), but both machine and human reviewers struggled to land on a verdict.2

To dissuade doubts, Shuixing signed a bet with another illustrator, and the two livestreamed themselves drawing to prove their humanity. On the stream, Shuixing did complete a full illustration, but the accuser remained unconvinced without giving further evidence. The whole bet died down unresolved. But this is not a unique case.

As early as 2022, differentiating AIGC vs. human drawing has become a topic in the illustrator community. Because generative AI relies entirely on the non-consensual harvesting of human art, using the technology carries a heavy social stigma. Under this backdrop, the community started what some describe as a “witch hunt,” which resulted in many big and small bet-on agreements and consequent livestreaming drawing sessions to prove authenticity. “Proving you are a human” has become a norm among illustrators, and showing the process of human hands drawing a picture line by line seems to be the last piece of evidence that humans can use to prove themselves as human.

Displacing

If AI-made art is indistinguishable from human work, whether something is AI or not becomes just one side of the coin. The witch hunt’s uncomfortable flip side is: who needs human labor if AI can do the same?

In May 2023, China Economic Weekly 中国经济周刊 reported on “the first wave of people who got laid off because of AI,” featuring illustrators as an occupation. The article kept a cautiously pessimistic tone. It features several illustrators whose careers were negatively affected by AIGC. At that time, Stable Diffusion and Midjourney had become popular in China, good enough to carry out specific tasks like image generation but not capable of executing other design work end-to-end. Most people were speculating that the status quo capabilities of AI would gradually eat into the workflows of illustrators, both those who were already in the field and students in art academies. “I guess now teachers from academies of arts can not figure out whether they should still select students based on how well they sketch,” said one interviewee. This has increasingly been reflected in changing major options at universities. Chinese universities have been axing majors rapidly — 12,000 were eliminated from 2021 to 2025, with the majority being humanities and arts. Among the 16 majors cancelled by the Communication University of China 中国传媒大学 (CUC) in 2025 are photography, visual communication design, and illustration, the last of which was long considered one of CUC’s flagship programs.

And this is a displacement encouraged by the state. Since 2021, the state has been aggressively pushing “AIGC creation” — including launching various campaigns, competitions, and conferences to encourage people to create films, videos, and pictures using AI — positioning AI as a new creative frontier rather than a threat. Even before calling for diffusing AI in every aspect of every industry in 2025, many businesses have embraced AIGC, which results in job loss for mid-level and junior illustrators. Southern Daily 南方日报 reported that by 2023, many art design positions at game companies had started disappearing. Illustrators used to be very popular among second-tier game companies, but increasingly faced massive layoffs. For years, tech firms have been training specialized AI for game design. And that was 2023: an AI-generated Will Smith could barely finish his spaghetti without eating the fork and his finger.

Some are more optimistic about the future and argue that AI is not an opponent but a tool — humans won’t be displaced by AI, but by other humans who can use AI. In fact, with the state encouraging AIGC, a community of “AIGC artists” has emerged. Some have no arts background and see AI as a democratizing tool that empowers them to create art even without going through the dreadful training needed to graduate from art school. Some other creators who have accumulated years of experience have also switched to AI as a novel method to explore, with some prominent examples like the AI short film by celebrated auteur Jia Zhangke (贾樟柯) and AI-generated “Multimodal Art” exhibition by Jia Hongjun (贾鸿君), a veteran ink painter and professor at the Central Academy of Fine Arts 中央美术学院.

The “Proofs” of Humans

Whether resisting or embracing, feeding or being fed on, all creators are searching for the same thing: a proof of humanity — evidence that being human still means something, and protection against being rendered indistinguishable from AI. Much of what illustrators in China experienced is specific to their context: particular legal regimes, community ideologies, and art education systems. But the struggle of seeking that proof is shared. Are these three common responses — regulation, stigmatization, embracing — as used by the Chinese illustrators in the stories above, enough to “AI-proof” humans?

Regulations

Some argue — and indeed demand — that regulation become the proof protecting humans from AI. Policy is unavoidably helpful, but it is not the proof.

In an ideal world, policy would have acted early enough so that the vast amount of data on the internet — much of which, like illustrators’ artwork and style, should be copyrighted — would not end up in AI training data. But most places, whether China, the US, or the EU, acted too late, or even still have yet to act. Moreover, while copyright seems like the natural candidate as it exists to protect human originality, it in fact protects specific expressions from being copied, rather than the art style itself, which more often is a way of seeing, a visual language, a distinctive aesthetic. And even that narrower protection has limits, as the early US district court rulings have mostly favored AI developers when training data was legally acquired.

In a similarly utopian world, policy could help us differentiate AI output from human output, so that human-made art would not be wrongly flagged by platforms as AI and people would not accuse one another. In reality, however, governing AI outputs is not easy even when regulations are in place. China has built one of the world’s most elaborate AIGC labeling regimes — and as this piece and the other explainer show, the result is a system that suppresses stylistically-consistent human work alongside AI spam, turns proving your humanity into a public performance, and generates evasion markets that are illegal but structurally inevitable. So long as the regulator cannot find a technical solution that is 100% perfect at discerning human vs. AI output — which is increasingly impossible for digital art and writing, where humans may increasingly internalize the style of AI output and vice versa, to say nothing of human-AI co-creation — the public will be stuck forever questioning whether something is AI or not.

Stigmatization

If policy cannot protect us from being fed into AI or distinguish us from it, should we rely on community efforts? Much like the Chinese illustrator community, many other creativity-adjacent spaces, such as Substack writers, have been intensely debating the “AI-like” output. One of the mainstream views on writing, very similar to how Chinese illustrators witch-hunt their peers, is that writing that is perceived to be AI-generated needs to be stigmatized. To quote Substacker-extraordinaire Sam Kriss, “I can tell. I can always tell. You think I won’t notice, but I will. There’s no hiding from me. If you let AI do your writing I will find out, and I will kill you.” The logic behind underappreciating AI outputs sounds valid, especially for human creators. Part of it is that AI output is “nonsense,” as it is viewed as a lack of genuine ideas and thinking process; part of it is that AI itself is bad for creators because it is fed by their works without consent. Regardless of the reason for the stigmatization of AI, the methods are imperfect and dangerous. “I can tell” relies on a very unique human instinct that most reliably lies in native speakers who have interacted with AI for a while — which, although it seems large for certain communities, remains a very small fraction of the world that uses English to write. More ironically, still, is that to avoid being judged as AI, humans start to change their naturally human creative process — avoiding em dashes in writing for now, and maybe changing drawing and filming techniques in the future — which ironically invalidates what they hold as humans having genuinely and inherently different judgment from AI. The witch hunt was supposed to defend a boundary of human vs. AI, but ends up forcing humans to write less like themselves so they won’t be mistaken for the machines they are trying to distinguish themselves from. [Note from editor Lily: you will have to claw the em dash from my cold, dead hands —and if ChinaTalk’s use of the em dash bothers you, you can take it up with Truman Capote.]

Embracing

At the opposite of the regulation and stigmatization that try to prevent AI from coming to humans stands the attitude to embrace the technology. When AI becomes so capable after being trained on human work and resembles human capabilities, using AI, and upgrading yourself, appear to be the very few ways to not be displaced at an individual level. By embracing AI in content creation, Chinese illustrators and other artists can create new genres of art and make themselves more competitive in the market.

This statement of ‘embrace AI’ is vague enough to sound valid. Embrace what, exactly? China is embracing AI by drastically reforming its university programs: cutting 12,000 and adding 10,200 new ones. It seems to equip itself for an AI future by replacing “useless” majors like arts and humanities with AI and robotics, but it is an open question of how AI as a major can itself survive AI progress. Today’s lesson on scaffolding may soon be outdated tomorrow, while suddenly the “outdated” aesthetic taste becomes more important when the world is flooded with “vibe-code-y” websites with beige over-polished user interfaces. It is hard to go all-in on something whose shape no one can yet see: where AI capability will go, what skill it will develop next, or which profession it will kill, and we individuals are just left exploring in the dark and fearing the unknown.

The deeper problem with embracing is the same one that undermines stigmatization: both end up being processes of humans frantically reorganizing themselves around AI, changing behavior in response to a target that keeps moving, and panicking without purpose.

The real problem, in both cases, is what the philosopher Günther Anders called “Promethean shame”: the feeling of inferiority humans develop before their own flawless, replicable machines, and the urge to remake themselves in the machines’ image. Whether we resist AI or embrace it, the underlying process is the same: an internalization of the machine as the reference point. Even those who try to “preserve human distinctiveness” by discriminating against what they perceive as “AI creation” have already let AI define them, as they are no longer creating as themselves but the opposite of a machine.

That “shame,” driving extreme hate or extreme frenzy, makes us purposelessly predict and prepare to change as AI changes. When AI starts to use en-dashes instead of em-dashes, or when the next fashionable job is no longer “harness engineer,” we change accordingly, without asking why — and we end up controlled by AI. Not in the sense that AI outsmarts humans and rules humanity, nor that a US AI or a China AI ends up ruling the world, but that the unpredictability of AI progress (what it will be able to do, how good it will be, when its capabilities will diffuse) and the anxiety it produces control what we learn and how we act. We lose our agency to our own human fear before we lose it to any superhuman AI.

The Mechanization of Humans

But why can we so easily lose our agency to fear — fear of being utilized, resembled, and displaced? Are we on an irreversible trajectory to be consumed by, merged with, and gradually displaced by the technology we built? The reason for the failure of these three methods has less to do with their rigor and more with the false binary of “human vs. machine” upon which they were built. They do not work precisely because the binary is dissolving, and had been dissolving even before AI. Before AI speaks, writes, and draws like humans, humans have already trained themselves to speak, write, and draw like machines, which again can be exemplified by the Chinese illustrators’ journey.

In China, learning drawing is seldom a rosy story of child prodigies embracing free exploration and flying with their creativity. Instead, the pipeline of arts graduates is not that different from the AI pipeline. Like elite STEM students who may be preparing for the International Maths Olympiad (IMO) and gaokao at the same time, most art students walk two tracks: gaokao and yikao (艺考, short for 艺术类专业招生统一考试, the joint exam for artistic majors enrollment) combined determine whether they can be admitted into good academies of arts, with a lower gaokao performance offset by a stronger yikao score.

Just like gaokao, the yikao leaves little room for creativity and ties everything up by metrics. For drawing, the exam is composed of three parts (sketching, charcoal drawing, and color), each guided by a very specific prompt like “draw an old man in a jacket sitting still” and graded against very specific metrics. Many students rely on tutoring organizations with guidebooks and intensive crash courses (美术艺考集训班), some of which have students drawing 14 hours a day for months alongside their gaokao study. Most of that training is repetitive work to refine skills rather than free-flow creation, which is exactly what AI is good at displacing. Very similar to how other students prepare for the gaokao, yikao students wake up at 6:20 and go to the studio, take 10-minute breaks each for breakfast, lunch, and dinner, and spend the rest of their time switching between the three subjects tested in the exams until finishing the last class at 22:30. Some continue to draw in dorms before they sleep.

But not everyone walking this pipeline wants to be an artist. Whether a stereotype or reality, a fair number of students choose the arts simply because they see no future in doing gaokao. As a teacher of sculpture at China Academy of Art (CAA) in Hangzhou, one of the two top academies of art in China, observed: “Lots of people are actually gifted in art; their talents have just caused them to drift away from the measurement standards in the academic system.”

Moreover, this is a system that builds upon a very specific version of meritocracy, where efforts are graded under extremely standardizable metrics, which in turn cultivates extremely automatable skills. The grading involves four rounds, with the first pass through tens of thousands of papers being graded through zouma guanhua “走马观花” — glancing something very quickly without enough care — as teachers sort drawings into qualified and unqualified based on immediate visual impact: whether the light-dark contrast reads from a distance, or whether the composition snaps into legibility. After the first pass, four rounds of iterative re-ranking follow, where papers are sorted into score bands and then repeatedly reviewed, with those that seem too high getting moved down one band, and those that seem too low getting moved up one. The process continues until every paper has settled into a score that a majority of examiners agree on.

When the system needs to select from the masses and present itself as fair, metrics become a benchmark, people become “skills,” and the purpose of life becomes utility to society. Students are trained on past exams to predict the next one; exam scores become the benchmarks; a good university degree is the reward of doing things correctly, and ultimately, one becomes useful because they possess certain skills. Everything sounds scarily similar to how you train an AI, even the grading of the yikao artworks: a fast first pass filters for general legibility is not unlike how pre-training data is culled for quality at scale; the rounds of iterative human re-ranking converging on consensus scores look similar to multi-pass annotation with inter-rater calibration; and the highest scores go to yikao students who have most precisely internalised what the system rewards, which is the same for AI models.

Back in 1967, American historian Lewis Mumford argued that the first machines were made not of metal but of humans, with armies, labor gangs, and bureaucracies that organized people into standardized, interchangeable parts. Calling it “megamachine”, he pointed out that the structure existed as early as the pyramid age, long before the machine as a concept concretized. The art exam system in China is one megamachine: students are trained to operate “as a single mechanical unit of specialized, subdivided, interlocking parts” — a skilled sketcher or a practiced illustrator, who operates rather than creates. As the famous Mao-era propaganda phrase went — “be a screw” (做一颗螺丝钉). And a human screw can be displaced by a real screw.

Thus, what enables AI to displace humans is not AI itself — it is a process of mechanization that long preceded it. We built governance regimes that protect output over process. We judge content based on tractable traits rather than originality. We designed education systems that train replicable skills over genuine creative vision. Long before AI arrived, we were already valuing output over process, evaluating humans against standardized benchmarks, and rewarding the kind of consistency that machines would one day do better and cheaper. We trained humans like we train AI. It should not be surprising, then, that telling them apart is becoming so hard — or that the machines, having learned from us, are now teaching us what we valued all along.

Subscribe to Zilan’s Substack!!

Acknowledgements:

Zilan is grateful to Conor McGlynn, Afra Wang, Amber Hu, and Kayla Blomquist for their feedback on earlier drafts.

The author acknowledges the use of LLMs for brainstorming and copy-editing, and hopes the readers don’t mind em dashes that are genuinely typed by humans.

Our guest is Roel Konijnendijk, ancient history lecturer at the University of Oxford and author of Classical Greek Tactics: A Cultural History, joining Jordan, Justin, and Tony.

We discuss…

• Why most scholars of Thucydides think there’s not much left of the “Thucydides Trap” once you take his name off it

• 300 and Gates of Fire as studies of masculinity: “It’s about the USMC in fancy dress”

• The counterintuitive reason Greek citizen-soldiers thought training for war was ridiculous

• Homeric gore, the slaughter of the suitors, and whether the Odyssey wants you to approve

• Achilles on immortality versus glory, and why this poem still keeps us arguing 2,700 years later

Listen now on your favorite podcast app.

Jordan Schneider: WarTalk, Ancient Greek edition. We have with us the delightful Roel Konijnendijk, professor at the University of Oxford. We talk about, my goodness, Graham Allison and the Thucydides Trap, 300 and how it pollutes teenage boys and defense secretaries’ minds, Homer really liking blood, guts, and gore, and military ethics. This is a really good show. You guys are going to have a lot of fun. All right, let’s roll the tape.

The Thucydides Trap Is Fake

Jordan Schneider: So how do ancient Greek historians today feel about Xi Jinping talking about this Thucydides Trap?

Roel Konijnendijk: So the Thucydides Trap is an idea that was invented in 2017, I think, by Graham Allison. He’s not an ancient historian. So it’s one of those moments where on the one hand, okay, it’s great, everybody’s talking about Thucydides, surely that’s good. But on the other hand, it is in an extremely shallow way. I read Graham Allison’s book on this, and I noticed that within a few pages he never talks about Thucydides again. He talks about Thucydides, but he doesn’t have any examples from ancient history except for the Athens and Sparta thing. He just talks about modern strategic situations. He’s not interested in ancient history as such. He’s mining it to get a nice little slogan for a phenomenon that he’s observed in the modern world. That’s absolutely legitimate. But it means that for us, that engagement with antiquity is very shallow.

And I think most scholars of Thucydides now would say that there is not that much to the Thucydides Trap that needs to have Thucydides’ name on it. There is, of course, an understanding in Thucydides that there is a dynamic between states, which he was the first to formulate, in which there is an established power and a rising power, and that creates tension, which according to Thucydides inevitably led to war. But his own narrative can easily be used to disprove that point, because it’s not actually that which leads to war. It’s not because Sparta was afraid that this war happened. It’s because of the host of other reasons he also mentions, most of which don’t have anything to do with Sparta. They have to do with Corinth being offended at its naval empire being encroached upon by the Athenians. So if you were to read this a little bit more closely, you might decide that what you call the Thucydides Trap should be called something else. But it’s a nice little label that makes people think Thucydides is still relevant. So on that level, we’re pretty grateful.

Justin: When I read it, my takeaway was I couldn’t identify who is Sparta and who is Athens. The way he applies the metaphor, both countries are both at all times, and it’s constantly changing. So I had no idea who was who.

Roel Konijnendijk: It’s like gender.

There’s an interesting point about drawing parallels between ancient Athens and Sparta. There have been certain historical parallels drawn, especially between the commercial, maritime, outward-looking, global US versus the landbound, conservative Soviet Union. That is the classic one, very dominant during the Cold War. But the Soviet Union was not the established power. They are the new ones, they are the rising ones, they are the ideological threat to the world order. So in that sense, the Thucydides Trap doesn’t apply there. And similarly now, if you wanted to make that parallel with the situation Thucydides describes in the fifth century BC, you would have to say that the US is Sparta, with the most powerful fleet in the world and these overseas trade interests and this cultural hegemony. There’s just no way that this can ever work.

It doesn’t work except at the most superficial level. We’d love to go into a discussion of the Peloponnesian War, what’s actually going on, because it’s a lot more complex than that. But then this is a journalist writing about modern politics, and this is me, the historian, laying out the ultimate cliché line of every historian: actually, it’s more nuanced than that. So no one wants to hear it.

Jordan Schneider: Graham Allison, journalist, ladies and gentlemen! …He does have a PhD in political science, but okay, we’ll leave it at that.

Roel Konijnendijk: A political scientist, fair enough. But not someone who has necessarily made a detailed study of the ancient world, of classical Greece, of Thucydides as an author.

Jordan Schneider: You can try, right? That’s the point. You don’t have to be deep in your Delbrück to understand there’s some nuance here.

Roel Konijnendijk: A general point I always try to make is we shouldn’t gatekeep antiquity. It’s actually good that he wrote a big book that made everybody look at Thucydides again. Kagan did this in the past too. He was always trying to make sure everybody was reading Thucydides, at least if they were taking any liberal arts degree in American universities. Now, I don’t know if Thucydides is necessarily the source I want you to read from classical Greece. But it’s a great book. Read it, by all means. It’ll make you think. There’s a lot in Thucydides that feels timeless and relevant, and that was exactly his intention. He explicitly set out to write something that was going to be around forever.

What I’m saying is that as far as ancient historians are concerned, that engagement is facile, convenient. It’s not really an engagement with the ancient world. It’s an appropriation of a model for the purposes of making a point about modern politics, which is fine. But that means I have less to say on it than I would if he had tried to find examples throughout history, or even throughout ancient history. That would have been really cool.

Justin: One of the reasons he used it, in my opinion, is that a lot of US military war colleges use Thucydides, at least portions of it, to start discussing the beginnings of strategy. The issue being that it’s also not great for that, because it’s not a great look at strategy. Thucydides was both too close and not close enough to the decisions being made on all sides. His investigations are great, but to try to apply that to the US and China, or even the US and the Soviet Union, is doing a lot with that narrative that doesn’t necessarily fit.

Roel Konijnendijk: For sure. One of the fundamental things at the heart of history as an academic discipline is that we acknowledge every historical society is different. The cliché goes that history is always repeating itself. The first thing you learn when you study history is that history never repeats itself. Every single historical situation is different. So you always have to worry, how comparable are two historical situations really, and on what level? What Allison has done with the Thucydides Trap is say that at a very zoomed-out level, you have a similar dynamic, an established power and a rising power. But at almost any closer level of detail, this comparison breaks down.

It’s the same with people who want to compare the modern US to the late Roman Empire. These are not the same entities in any meaningful sense, except that there is a narrative. We come back to this idea of telling ourselves stories, a narrative in which the protagonist is going through the same process. This idea that the downfall of empires always looks the same. And then you can ask yourself, first, is that what’s happening? And second, is the late Roman Empire a meaningful comparative case when the environment of that empire, the way it’s run, the way it thought of itself, is all radically different? The world is different. You can’t just say what happened there will happen here. So whenever somebody tries to make a model that applies to all of history, you have to immediately think, how universal are these factors you’re implying?

The High-T Military

Justin: The news of the week is that the United States military is going to be the T Nation, I guess, which is an old bodybuilding website that definitely wasn’t about the use of steroids. It was about the use of steroids. So the Department of Defense is going to start looking at everybody over thirty to see if their testosterone is low and then give them more.

A couple of things. Testosterone being low is a real thing. There is a thing called operator syndrome, and if you actually read the memo Hegseth signed, that’s what he specifically addresses. That’s a real condition that affects a small portion of operators who have been in continual high-stress environments, where they’ve had adrenal fatigue and their body’s hormones are out of balance. That said, I think it has like a five percent occurrence in the normal population. So I don’t know that this is really going to fix much. But then the way he decided to come out with his video talking about the high-T military.

Jordan Schneider: Do we have 300 to blame for this?

Justin: I think so. I told you this. I was in Fifth Special Forces Group. Fifth Special Forces Group, when they were in Vietnam in the sixties, became known as the Legion. That name carried over through the years, and they still are referred to as the Legion. They use some Roman symbols whenever they make their unit crest. The battalion I specifically was in, though, called themselves the Spartans, confusingly. And that only happened after the movie 300 came out. The commander at the time was like, we’re going to be the Spartans. That’s where that came from.

Roel Konijnendijk: There’s a lot of that. There are people who specifically study this as reception of the ancient world, the idea of people trying to take things from antiquity and reapply them to their own lives. There’s a whole cottage industry of people writing about how different organizations have used the version of Sparta you find in 300 to identify themselves as very tough, very manly, very brave.

When I watched 300 when it came out, I was a teenager, and my mind was blown. That was super cool. I watched it in IMAX, this huge screen, went with a friend, wow, this is amazing, best thing ever. Now when I watch it, I’m just like, this is really quaint. The only way I can take it seriously is as a study of masculinity. It’s about what it’s like to be a man. That’s very explicitly what it’s about. It’s all about the perfect man, Leonidas, being a perfect man, and no one can possibly touch his sheer raw manliness. That’s fine for a movie. It’s just a bit ridiculous when you realize nowadays surely we’re a little more critical of that image. But Hegseth’s not quite there yet.

That masculinity is a death cult. These men live to die. That’s all they want. They don’t want to beat the Persians. They want the Persians to keep coming forever so they can keep killing and dying forever. They don’t actually want peace. That is the pinnacle of achievement for them. So this is very disturbing if it starts to affect the ideology of modern militaries. You don’t want soldiers thinking in these terms. You don’t want them glorifying the idea that dying for your state is all you live for. You want them to think pragmatically about how they’re going to achieve strategic objectives and maybe, hopefully, survive and get home in one piece.

That version of Sparta is entirely a modern product. This is the result of Zack Snyder having a very weird interpretation of Frank Miller’s graphic novel, which is already really extreme, but which is not about Spartans. It’s about the USMC in fancy dress, as I always like to say. It’s a vision of modern military training applied to the ancient world. It’s recursive. It comes out of the US military, it’s about the US military, it comes back to the US military, and the US military says, this is awesome, we want to be like this. Of course you’re already like this. It’s based on you. It’s not based on the ancient Spartans. Same with Pressfield’s Gates of Fire, which I think a lot of people in the services read.

Justin: It’s highly, highly recommended when I was a young guy. It’s exactly that same thing.

Roel Konijnendijk: It’s about a modern military, its ethos, its training methods, its attitudes, the whole vocabulary. It’s modern, with a little sauce over the top to make it feel ancient. It’s now being imposed back on that world in an ever more extreme, radicalized way. It’s almost like asking ChatGPT to keep summarizing the same piece, and it gets more and more wild and extreme. It goes further and further off track, because you’re perpetuating the errors within the system as you replicate it.

What the Sources Actually Say About Sparta

Justin: If you wanted to quickly correct the narrative on the Spartans, what is the best source? Is it Xenophon?

Roel Konijnendijk: You can read Xenophon’s Constitution of the Spartans, a very short text that literally just sets out: one day I woke up thinking, why are the Spartans so great? And then I thought about it, and here’s the answer. It’s about how they organize their society, based on the idea that the Spartans are militarily effective. But when you read it, you have to bear in mind, this is all we know about Spartan military methods. This is all he says about it when he’s trying to prove they’re better than everybody else. And it is so little. There is so little about the upbringing, the military training, the military methods. What you see is a bunch of guys trying to manage their society, control the values of their children, and maintain high birth rates, which is also something Hegseth would probably enjoy. These things don’t have a direct impact on their military abilities. It’s really just trying to foster a good civic spirit. These people are obedient, deferent to their elders, they obey the laws. Fighting? Meh. Very secondary. Not a skill they prize. And all these sources agree on this.

If you want a more nuanced picture of Sparta, there’s a very good book by Nigel Kennell just called Spartans, came out in 2010. There’s also Andrew Bayliss’s Spartans, which is okay, similar in trying to introduce some of the new ideas about how people now think about Sparta. That’s really very different from the way it used to be interpreted back when Frank Miller was writing 300. You can go read 300, I have a copy here, and flip to the back to see his inspiration. It’s Victor Davis Hanson. That’s what he read in order to write 300. So again, it’s a uroboros of cultural influences, into which historical knowledge penetrates very little.

Justin: I always wondered about this idea that the soldiers were completely divorced from their families, that they only lived in the barracks and were separate from society until they reached manhood. When you go look for where it says that in the historical text, it’s like, not really anywhere.

Roel Konijnendijk: Especially not in contemporary text, in Xenophon. They just go home. They hang out with the boys during the day and then they go home.

Justin: It doesn’t sound that much different from the gymnasium of Athens. They go, they hang out, they wrestle, they go back home.

Roel Konijnendijk: Exactly. It’s just about social conditioning, like the gymnasia. You do a bit of physical exercise, but you’re also being socialized. You’re getting used to having conversations with older men who are there to instruct you on how to be a good citizen. And for a little while you’re looking really good, which is good for you because you’ll be healthy, and also because older men will want to get into a relationship with you. Then you can get patronage, better connections, all sorts of benefits.

But in Plutarch, in the much later sources from the Roman period, they suddenly do live in barracks, they suddenly are separated from their parents, and they do have to do a lot more training, especially to learn how to sing and how to talk like a Spartan. That later source is written after multiple documented reforms of the Spartan education system. So we know this is not the system they had before. This is the system after they lose their prominence, but they’re doubling down on their traditions, which is something societies will do. You’re no longer very important, but you are unique, people look at you like you’re special, so you double down.

Jordan Schneider: You become a tourist destination, right? By 300 AD? Come on, what are we even doing?

Justin: They get pushed around the battlefield. It’s after the Peloponnesian War, the Thebans, I think it’s thrice, that just push them around.

Roel Konijnendijk: It’s the Thebans in the fourth century. They defeat them a couple of times.

Leuctra

Jordan Schneider: Wait, wait. do you actually think Epaminondas did at Leuctra?

Roel Konijnendijk: That’s a big one. It’s a combination, and I wrote about this in my book. The story is always that he must have done something revolutionary because he beat the Spartans, and no one had been able to do that for a while. What he ultimately did, according to the best accounts we have, is he just didn’t run away when the Spartans bore down on him. That is already quite special.

He got an army together that was decently experienced. The Sacred Band may have something to do with that, a very experienced picked unit of men who had physical training, spent time together, been in the wars together. So you have an experienced lead unit heading your phalanx. And then he massed the rest of them very deep, a very deep formation, which is very strong for morale. You literally can’t run away, so you’re more likely to stand your ground. There’s a physical block against doing that. And the people behind you are in really high spirits, because the fight’s not going to get to them for a long time. So everybody’s kind of like, yeah, let’s go. In that very deep formation, with the Sacred Band likely at the head, he was able to slam into the Spartan formation, possibly when they were in disorder out of a bungled maneuver, their own cavalry having disrupted their lines when they were driven back into them.

So there was disorder in the Spartan line, and they smashed into it. And then they were willing to engage Spartans in hand-to-hand fighting, which most Greeks just didn’t do. When they saw the Spartans advancing in regular formation, good order, uniform equipment, they would just break and run away. The Thebans decided to stand their ground. And once you stand your ground against a Spartan, he’s no better than you. That follows from what we know about their training. Their training isn’t about skill at arms, or strength, or becoming individually a superior fighter. It’s about moral and disciplinary conditioning, becoming part of a well-oiled machine of devoted citizens. But in a fight, one on one, a Spartan is no better than anyone else in the Greek world. So if you’re willing to stand and face him, you can beat him, fifty-fifty at that point.

And then the enormous weight of that Theban column, ranks upon ranks behind them, comes to the fore. You can say, we can stick in this fight as long as we want. You can’t, because your formation is not as deep as ours. Do you want to keep flipping that coin until you’re all gone? And the Spartan Royal Guard, to their credit, say yes, we’ll do that. They all die. Something like four hundred Spartiates are killed, which likely includes most or all of the three hundred guarding the king. But the rest of that army doesn’t.

Jordan Schneider: But we have these quotes saying he’s the greatest of all time. Is that enough to get you there?

Roel Konijnendijk: He did something no one had done for a long time. But greatest of all time is Cicero trying to give credit to someone he admired, because Epaminondas was a democratic leader fighting against a king, somebody who represents the Republican ideals Cicero himself stood for. He goes into the Peloponnese and strips away all these governments imposed by the Spartans, replacing them with more pro-Theban regimes. Cicero will consider that a real act of liberation. And he did permanently break Spartan power. They never recovered. So he achieved something that really changed the map.

Jordan Schneider: He didn’t just win a battle, he ended an empire. He’ll get some credit for that.

Roel Konijnendijk: Exactly. And that isn’t just because of the battle, it’s because he followed through. Victor Davis Hanson, in one of his more lucid moments, wrote an article about Epaminondas as the inventor of strategy in a sense, because he followed through. He said, I’m not just going to win this battle, I’m going to march into your home territory, strip your assets, clear your allies, break your alliances, reduce you to a rump state. That is a much more decisive way of waging war than saying, okay, we beat you, now let’s come to terms.

I say “one of his more lucid moments” because he’s increasingly just sloppy. He doesn’t need to be a good scholar, he’s selling books anyway. *That book A War Like No Other, on the Peloponnesian War, that was propaganda for the Iraq war. That is explicitly why he wrote it.* And that’s just depressing. When we’re talking about modern authors using the ancient world, don’t do that, man. You’re not making us look good. It’s hard for me to keep citing him, because his academic career has so nakedly come to serve his pundit career that you can no longer really take anything he says seriously. But he was a very important voice in my field in the eighties and nineties, so you have to acknowledge that.

The Myth of the Technological Revolution

Hoplites on the Chigi Vase, ca. 650 BC. Wikimedia Commons (public domain).

Jordan Schneider: Longtime theme on WarTalk: the influence of technology, particularly emerging technologies, on how war changes. What’s the right place to start with that in the classical Greek context?

Roel Konijnendijk: There’s the expected answer and the real answer. The expected answer is you start in the Bronze Age and then say, okay, iron comes in, that’s the big change. It’s not, really. When we get Homer’s poetry, a lot of the weapons are still described as bronze. It’s not like iron comes in and displaces it entirely. Even in the classical period, when iron has been around for centuries, they have iron-headed spears but bronze butt spikes. They’re using both technologies at the same time for different purposes. It isn’t as simple as saying iron gets introduced, radical technological revolution, everything changes. That is an old assumption, and it’s just not what happens on the ground.

Around the time the Homeric poems come to their final form, you have the introduction of a new type of shield, the big round double-grip hoplite shield, the aspis. There’s an old argument that this determined the use of different tactics, because it only works in certain formations, and that determines the whole shape of warfare for centuries. But more recent scholarship has pushed back and said, actually, we can’t see any of that. Nothing confirms this is the only way to use the weapon effectively, and initially the number of people who could afford it would have been really small. So you need changes in society. You need people to get a lot wealthier and cities to get a lot bigger before you can say this shield leads to things like hoplite phalanx formations. It ties into a lot of other features, rather than just saying this new technology arrived, use it or lose it.

Justin: That’s an interesting segue, because you’re talking about the economic culture that underpinned the core differences between Sparta and Athens by the time of the Peloponnesian War. Do you think the hoplite army dictated the differences in the way their economies developed, or were those just byproducts of Athens having access to silver and seafaring, whereas Sparta used iron and helot slaves?

Roel Konijnendijk: The important thing to know about the Peloponnesian War is that there’s no technological difference between the two sides. They have exactly the same weapons and armor, right down to naval warfare. The Athenians have the much more powerful fleet, they really specialize, they know what they’re doing. But their triremes are exactly the same as the Corinthian triremes they’re fighting. Zero difference as far as we know. So it’s really about how you use it, the tactics, the experience, the confidence, and what you want strategically to do with them.

Athens has a lot of overseas interests, so they invest in making sure their navy is strong, protecting allies, keeping trade lanes open, pushing their influence. Whereas Spartan society grew into this rigidly hierarchical thing dominated by a leisure class, a large number of people who don’t work, who have an enslaved underclass doing all the work, who band together and run the state. That agricultural economy propping up a leisure class expresses itself militarily in this large hoplite army. But it’s not like the Athenians don’t have a large hoplite army. They have a larger one than the Spartans. It’s just that most of those guys aren’t leisure class in the same way, and a lot of them don’t see themselves as defending a way of life when they pick up that shield and spear. They don’t have the same attachment to how society is organized.

So those different ways of war come out of different strategic considerations, which grow out of how these states developed socioeconomically and politically. And modern scholarship has pointed out that Sparta is not that different from a lot of other Greek states in that sense. There are plenty of other states that focus all their resources on propping up a huge leisure class, staying in power, oppressing everybody else. And militarily they’re quite effective, precisely because they’re permanently available for service when needed, and because they have the economic means to supply themselves the best armor and weapons, all privately arranged. They pay for their own stuff.

Nobody Trained

Jordan Schneider: In your book, one of your keys is these few quotes, one Aristophanes play and a piece of Plato, where they’re making fun of the idea that anyone in Athens would ever want to train militarily, that this was the lamest thing you could possibly do with your time in 475 BC. You could philosophize, you could drink with your friends, but the idea that you’d put on armor and learn how to hold your spear and shield is just unconscionable.

I guess, thinking about how we talk about this stuff, you have technological overhang and then operational overhang. Here’s the technology available in any given year: have you figured out the right way to organize your society and train and have the right doctrine to get to the frontier of what’s possible with whatever’s been invented? It was shocking to me that this was a big limit, that no one wanted to deal with it. Why doesn’t it get competitively eaten out of existence? Is it just the geography of Greece, that you could not train and still exist as a city? How do you get into this as an equilibrium?

Roel Konijnendijk: Well, it’s very effective. That’s the weird thing. It’s completely counterintuitive to us. If you’re trying to be a fighter and survive, you rely on your skills, your ability with the weapons, you obviously try to be as good as you can. The Greeks don’t have that attitude, and I argue that not because it makes sense to me. I don’t really get it. I’m arguing it because that’s what the evidence tells me. We have explicit quotes that they don’t train, they say they don’t train, they actively, repeatedly assert that no one does this, they find it silly and ridiculous, they explain why. You’re making a fool of yourself, everybody’s going to scrutinize you, you’re going to slip up someday and everybody’s going to laugh at you. Why do they exclude it from their military experience?

The best answer I’ve been able to come up with, and this is really speculative, is that there emerges from these societies, especially the democratic ones like Athens, but elsewhere too, in these societies that rely on a citizen militia. In wartime they don’t have a professional army. When war comes, all the adult males get called up and march out to fight. In an economy that’s still to a large extent subsistence, you have a lot of people living hand to mouth. They don’t have the time to set aside to go and train. They don’t have the time to say, I’ll take my weekends, I’ll drill with my buddies, we’ll figure out how to do this properly. They get called up and they serve.

And with that reliance on a militia comes an ideology, this need to establish that as long as you’re willing to turn up when the call comes, that’s enough. That’s all we ask of you. Because otherwise you immediately get enormous differences between those who have time to train and those who don’t. A lot of people are going to feel inadequate, feel like it’s not worth it to turn up, feel like their contribution isn’t needed. If you want everybody to be willing to take part, you have to tell them, as long as you show up, we’ve got what we need from you. So there develops this ideology that you don’t actually need to train. You just need to have the spirit, the right attitude, to be there, to have your spear and shield, to show up and not run away. And that’s going to see us through.

A large part of that is affirmed by the fact that a lot of these Greek battles are over in a flash when one side breaks. That’s what happens most of the time. One side almost immediately collapses, runs away, gets speared in the back. So spending a lot of time trying to get better with a sword or spear isn’t actually going to help you very much in that very small window where you’d use those skills. Even the Spartans, who take warfare more seriously and are famous for it, when they train, they don’t train with weapons. They don’t do weapons drills or weapon proficiency. They don’t think that part of military skill is important. What they care about is unit cohesion, formation drills, formation evolutions, keeping the line steady. It’s not individual skill that makes the difference. It’s the individual will to stay in the line, to keep your post. And you don’t train that directly by requiring people to spend their time doing sword-fighting drills.

Justin: So how did the mercenaries change that? You have a portion about Jason of Pherae, and Alcibiades after the Peloponnesian War goes off and becomes a mercenary for hire. How does the rise of this mercenary class change the way the Greeks look at warfare, or is that just an offshoot?

Roel Konijnendijk: It’s one of a couple of ways this militia system gets changed over time, as states develop more means and devote more of them to warfare. From the classical period onwards they invest hugely in fortifications. That’s one way they throw money at the problem. The Athenians investing massively in their fleet is the same idea. But they also figure out that if you can pay people to serve you in certain ways, that can make up for weaknesses, let you rely on specialists, and foster within a certain part of your population the skills they need to do better.

There’s an understanding that people who have experience in war will do better than those who don’t. That’s their first connection to what we consider necessary military training: we need to foster people’s military experience. We need people who have done it before, who aren’t encountering these problems and emotional responses for the first time. Mercenaries are just better in a fight, and the Greeks understand that partly through weapon skill, but largely because they’re not as afraid anymore, not as nervous. They’ve seen it all. They’re doing a job and they know how it’s done.

But the problem is these states are very small, and when they call up the army, they’re calling up all of their population. They can’t afford to have all of that population raised to the same level, because they have lives to live, an economy to run, farms to farm and shops to shop. So the sheer tiny scale of these Greek states puts a hard limit on how much professionalization you can do. But the spread of mercenary service, especially by the fourth century, is a sign that all these states are thinking seriously about how to get at least some of the advantages of professionalism, training, and military experience.

Fifteen-Year-Old Prussians

Jordan Schneider: I want to stay on the sad militia who don’t have time to train. In the book you have this awesome contrast between the nineteenth-century Germans who fought under Von Moltke, with their vision of warfare where everyone’s doing cute little turns on the battlefield, reshaping how deep their lines are, and these Greek hoplites who can’t even turn. The Spartans probably can, but everyone else just drifts away from the open side of where their shield is. How hard was it for these Germans, who idealized Pericles and Epaminondas, to reconcile with the fact that you had fifteen-year-old Prussians who were way more functional than what you were getting out of Thebes?

Roel Konijnendijk: They’ve been put through their paces. This is a really interesting thing about the way the warfare of the period you’re writing in affects your understanding of the warfare of the period you’re writing about. These Prussians are the first guys to seriously write about Greek warfare academically, but they were hugely influenced by the warfare of their own day, in which Prussia was spectacularly victorious over all its rivals, one of the most efficient, well-oiled machines in military history. So they saw this as something every other society must have been striving toward. They must have been trying to be like us.

But that way of running a disciplined army grows out of Napoleonic musket drills, where you have to be as quick as possible reloading and as precise as possible maneuvering. Some Greeks had awareness of these advantages, and it goes right back to Homer. In the great battle outside Troy in the Iliad, the Greeks are advancing quietly, in silence, and the Trojans, who aren’t used to that, are terrified. We understand the power of discipline, the importance of keeping your cool, or appearing to, in front of the enemy. But to structurally impress that on the entire militia, you need to get them all to take the time to figure out how the maneuvers work, and imbue them with the confidence that when the order is given they can do it without a hitch. Otherwise they’re going to be even more scared, because they think it’s going to be chaos.

This is what happens in a couple of Greek battles. Cleon outside Amphipolis wants to retreat and orders his army to wheel. They don’t know how to wheel, so they fall into complete disorder, and the Spartans attack them as they’re running away. Complete slaughter, seven Spartans and six hundred Athenians killed. And that’s explicitly why the Spartans chose that moment to attack.

So there’s this understanding that you have to get them to take the time to drill, and the simple fact is that from various accounts, the other Greeks besides the Spartans just didn’t get it. They didn’t understand why or how it would work. Spartan drill is very simple. It’s described by several sources, especially Xenophon, who had personal experience with it. You essentially have officers in front, and you just follow them. You do exactly what the guy in front of you does, and everything will be fine. But Xenophon explains this at length to his audience, clearly assuming they don’t get it, that they think it’s all way too complicated. I can teach this to students in twenty minutes. It’s very easy, but you have to have the mindset. We inherit from generations of people in the army the idea that we understand how parade-ground drill works. For the Greeks, this is still quite a novel idea. Very few of them know instinctively how to do it. And most of the ones seen doing it have clearly had Spartan training. The Spartans bring others into their army and say, right, we’ve got to do this first, get these basic formation drills into your head, then we can continue. That seems to be the singular source for this kind of tactical sophistication.

How the Greeks Thought About Military Thought

Tony Stark: On that note of how they didn’t understand how Spartan warfare worked, where did military thought rank in the philosophizing and socializing? How did Greeks think about what it meant to think about military thought? Was it valued, or was it something that, like training, they’d wait until war came?

Roel Konijnendijk: Excellent question. You don’t have militaries in the Greek world. You don’t have military academies. So all of this is informal. When they talk about warfare and how to get better at it, they’re doing it among friends, because there are no professional soldiers. In Athens, a board of generals is elected every year and serves for a one-year term. The generalship is one of the few magistracies where you can get reelected. Some people are considered very good at it, so they get reelected lots of times. Pericles, like thirty times. Phocion, a general in the fourth century, gets reelected forty-five times. He holds the record, that’s basically his entire adult life. But that’s very rare. Some of these guys only get elected once, go into battle, and immediately die. There isn’t much continuity of expertise. There’s no old generation of generals informing the new one. They’re doing this informally.

So you have to imagine people getting together at drinking parties, or sitting outside talking to philosophers, chatting about current events, rehearsing ideas they’ve heard, trying to pass on the necessities to the next generation. You get philosophical dialogues where this happens, where Socrates has to explain to some kid who comes up and says, I’ve been training with this drill-master for hire who goes around the Greek world teaching basic military skills. And Socrates goes, yeah, but did he teach you how to be a general? No, he just taught me tactics, how to position a formation, how to stand in a line. So Socrates has to say, there’s a lot more to being a general. This is increasingly, by the end of the fifth century, starting to be considered something teachable.

That’s a new development in classical Greece. Before that they seem to want to reinvent the wheel each time. There’s no awareness of any continuity of knowledge. That’s quite a new genre of text that emerged in the late fifth century BC, military treatises written explicitly because you all need to know how to do this and no one’s going to teach you unless you read my text. The first of those are very practical, aimed at, for instance, Athenian cavalry commanders. There’s a text from Xenophon that specifically says, here’s what to do if you’re commanding the Athenian cavalry. Or Aeneas Tacticus’s manual on defense under siege, which basically says, if your city is under siege, here’s how you prepare, here’s how you fight them when they’re not at the wall, here’s how you fight them when they are. These come out of that world because they fill a vacuum. There’s no systematic education in these matters. They’re starting to think, we have to make sure this knowledge isn’t lost, and then hopefully we can iterate.

Justin: Wasn’t there resistance among the philosophers to projectile throwers, light infantry, people who weren’t hoplites? Lower down the hierarchy, using weapons that were less manly, but with the ability to inflict damage on the oligarchic class in the hoplites. Didn’t that also serve a stagnating function during the Peloponnesian period?

Roel Konijnendijk: We’re all familiar with the effect ideology has on military developments. We’re not just trying to solve a problem in the best way, we’re imbued with values and ideas that tell us who we are. For the Greeks, most of the people in charge are wealthy, and they have a certain way of fighting because of how they can afford to equip themselves, mostly as hoplites, as heavy infantry. So an ideology develops in which that is the correct way to fight and everything else is for foreigners and the poor. That’s not for real men who are citizens fighting for their city as hoplites.

But there are a couple of points where this ideology cracks. One of the supports for the idea that we can go into battle and win even without training is that as long as we fight as hoplites, we’re being real men, the ultimate warrior. We stand our ground. The emphasis of that entire ideology is never on “I’m a very good spearman.” It is always on standing your ground, keeping your place in the line. And that is very much the ideology of a man who has absolutely no idea what he’s doing and needs a simple line to hold on to. I kept my pride, I did what I needed to protect my city, I stood where they told me to stand, that’s all I did. Leonidas being the ultimate example. So you use that ideology to prop up a very faulty military system.

But the other part is that once it becomes more normal for people to afford hoplite armor, you get large masses of citizens who can just about afford a spear and shield. It’s not that expensive. You can equip yourself as a hoplite for about thirty drachmae, about a month’s wage. So a lot of people can afford this by the end of the archaic period. The wealthy want to distinguish themselves from the common people, because they are not like the common people. So you see the rise of a lot of them fighting as cavalry instead. You can afford a horse. That’s the sports car of antiquity. So the ideology of the hoplite, the baseline, we’re all hoplites here, has to accommodate people trying to break away because they want more hierarchy in their society.

And on the other side, you can look down on light-armed troops, sneer at them, but they still beat you, because you can’t catch them. There are a couple of famous fights where hoplites straight up lose against these poorly equipped troops, because ultimately there’s nothing you can do about them. They will just slowly bleed you white. So from both ends, that ideology that we’re all hoplites here has to be caveated with the practical realities of Greek warfare. People will claim they’re hoplites regardless of whether they actually fight as cavalry or can’t afford to. It’s a citizen ideal. But when these ancient authors start talking about warfare, they very quickly leave that behind, and tell you in detailed terms how other forms of warfare are also effective. So I’m not entirely sure it had a very great deal of practical effect.

Homeric Heroes

Justin: You mentioned the Iliad earlier. There are long passages that talk about Ajax being imbued and able to draw blood from Ares, Hector being able to fight off the gods, fighting as an individual and heaping bodies on the pyres. Why isn’t that something held up in actual Greek warfare? Why is it the unit, the phalanx, what the body is able to do, when in Homer it’s very much a focus on the capabilities of the heroes and their individual abilities in battle?

Roel Konijnendijk: That’s a huge question, because it’s one of the key differences between how we picture classical Greek hoplite warfare and how we picture Homeric warfare, which is all about singular heroes. The most compelling way to account for it is that society changed. In this earlier world you have far fewer people wealthy enough to afford full armor. They lord it over others. They’re literally called basileis in the poem, which means kings in later Greek, lords, princes, with a much more exalted station, and there are fewer of them. One of the ways they justify their position is by being better warriors. It brings personal glory, but it also makes the ordinary people, the laos, look at them and say, okay, fair enough, of course he’s in charge, because he’s a badass. So it feeds into itself. You have a system of status that in reality is based on wealth, but that has an ideology that it’s based on merit, and they have to prove that in these narratives by achieving feats beyond ordinary people. And to some extent it’s probably true. If you were fully decked out in armor, you’d be more willing to go to the front and throw a javelin at your man than if you were showing up in your tunic picking up a rock from the ground.

You get this in archaic poetry outside Homer as well, and even in some tomb monuments right down to the end of the archaic period, where they claim, I was with the promachoi, I was with the front-fighters. We shouldn’t see that as a front rank. It sometimes gets mistranslated as a front rank in a formation. No, these guys go out in front of others. They challenge their enemies directly. In the Iliad they literally do this. They’ll say, you over there, I know you, I knew your dad, now fight me. This kind of very personal, deliberately showmanship way of fighting, because these guys depend on it.

Over time, of course, that becomes subsumed by the communal need to work together to be militarily effective. The more of these heavily armored guys you have, the more effective they are working together. So you have to suppress that impulse, get rid of this urge to stand out in front and prove yourself. But it’s not perfect. It doesn’t go away completely. There are really strong anecdotes right down to the Persian Wars and beyond of individual warriors who deliberately go out in front, challenge enemies to single combat, try to make themselves conspicuous, make sure everybody sees who they are and knows they did something above and beyond what’s asked of the ordinary hoplite. They still learn, they get spoon-fed this idea that that’s how you behave as a hero. So it sits a bit awkwardly. There’s always this tendency to go out and be a Homeric hero, even though there’s no longer really a place for it.

Bite the Dust

Jordan Schneider: Let’s start with glorifying violence and then go to how violent these fights actually were. These famous similes in the Iliad, and a bit in the Odyssey too, where the blood is spurting everywhere, look how badass I am. It makes for kind of nauseous reading in 2026, Roel. How do you process how excited they were to see literal heads roll?

Roel Konijnendijk: The phrase “bite the dust” is Homeric. That’s a phrase we get from Homer, where people literally die, fall over, and bite the dust, grimacing and falling into the sand. Especially in the Iliad, you get these really graphic descriptions of exactly what happens when a spear meets a face, how many teeth are knocked out on its way through. It’s really quite shocking.

But that is really unique. There aren’t many sources that go into such graphic detail. That’s something Homer in particular was really interested in. Part of it is the spectacle. They’re playing it up to let you know the stakes are very high, this is serious. They’re also not trying to sugarcoat that it’s a war. The more you amplify the horror of it, the more you also amplify the heroism of people who endure it, which is still a factor in modern war fiction. And the other thing is something we can relate to when watching modern action movies. There’s a degree of spectacle in gore that makes it more interesting to watch. The more you push the envelope of how gruesome something is, the more it becomes almost entertaining in its gruesomeness. These poems needed that. They have an audience, they’re being performed, they’re being sung, and that audience needs to be engaged. If what they want to hear is how many organs are severed before the spear comes out the other end, that’s something Homer will give you.

How you deal with that as a modern reader is actually a good reminder that they are not trying to say war isn’t horrific. They are not arguing that this is all beautiful and everybody’s a hero. They are saying that if you’re on the wrong end of Achilles, you’re going to have a really terrible time and die in agony. And that is not something that’s celebrated, because Achilles is not celebrated. We’re not here to say he’s the hero, he should win, the good guys win, we get our happy ending. He is a man who transgresses the rules of human society. He is behaving badly in that moment, and his victims are suffering for it. So it’s necessary for Homer to say, this is not nice and clean, this is a horrible act of violence.

Jordan Schneider: Can we do my favorite one? The suitors getting piled up like fish. Let me find the quote.

Roel Konijnendijk: What’s very cool in that scene, you get them picking up the little tables to use as shields. They’re improvising weaponry in the banqueting hall. And there are vase paintings of that, you see them picking up little tables to hold up with arrows sticking out. Greeks realized how theatrical that was. They absolutely wanted to show you that.

Jordan Schneider: All right, this is from Fagles. “He found them one and all in blood and dust, great hauls of them down and out like fish that fishermen drag from the churning grey surf and loop in coiling nets, and fling ashore on a sweeping hook of beach, some noble catch heaped on the sand, twitching, lusting for fresh salt sea, but the sun god hammers them and burns their lives out, so the suitors lay in heaps, corpse covering corpse.” Eat your heart out, Christopher Nolan.

Roel Konijnendijk: It’s a hundred and eight of them. You have to remember there are a hundred and eight suitors, they’re numbered. So this is an absolute slaughter. He’s throwing fish out onto dry land and they’re being beaten to death by the sun god, dried out. But there’s another simile that’s really gruesome, when he’s talking about how the Cyclops treats the crew of Odysseus. He grabs them two at a time and smashes their heads against the floor. And the simile he uses is, like you would a puppy. This is how you kill puppies, I guess puppies you don’t want, dashing their heads against the floor until there’s brains everywhere. That’s how he puts it. This is a world in which perhaps life isn’t always treated with the same amount of reverence that we tend to be taught by our own society to have.

Jordan Schneider: When you read the Old Testament in synagogue, the rabbis have to keep doing this thing, being like, yeah, it’s really bad, but God’s still a nice guy in the end, don’t index too much on all this smiting. When you read Homer, you don’t have to turn them into heroes or conclude that Homer’s a good person and a positive influence on your life. It’s almost easier to accept that this is what some human being was thinking twenty-eight hundred years ago, and just go from there, live in their head, as opposed to ethically justifying how it’s still okay somehow.

Roel Konijnendijk: It’s similar to how I like that the new Odyssey has already generated so much discourse, because when you see these scenes, you have to imagine that even in the ancient Greek world this was meant to generate discourse. As you’re hearing this song from the bard, you have to ask yourself, do we think this is right? Do we think this is proportional? Do we think this is the correct response to the situation? To some extent, killing the suitors, yeah, they were very bad guests, they were misbehaving, a threat to the household and the community. You’re led to think this is good riddance. But the poem immediately then says, of course, they were prominent people in Ithacan society from all the surrounding islets, and so immediately when you kill them, their families show up with weapons and demand satisfaction. Odysseus at the end is with Telemachus and his dad Laertes at the farm, preparing to die, because it’s three against an army, an army of the families of the suitors coming to kill them.

And the only reason they survive is because the gods literally step in and say, that’s enough, you’re going to stop fighting now. But the idea is, you can say these individual acts are morally justified, whatever, but there’s always this sense in Homer that they do have consequences. You can’t just do whatever you want because you think you’re doing the right thing. You have to accept that somebody else is going to think you did something horrible, or try to avenge it. There is more in this world than your own personal relationship to your morality.

Battlefield Ethics

Tony Stark: There’s some modern discourse, particularly in America, that battlefield ethics are a modern creation. I think the story of Achilles is perhaps the best argument against that. But can you talk about how battlefield ethics were discussed and considered during that time period?

Roel Konijnendijk: There is some evidence that the Greeks were particularly brutal in battle. They don’t recognize a lot of different kinds of battlefield ethical restraint. That’s one of the arguments I made in my book: fundamentally, there are a lot of ideas that they supposedly thought all sorts of things were distasteful, like attacking by surprise or killing wounded people. They don’t really have those restrictions. There are only a few rules they recognize. But they do recognize them, and they acknowledge there are ways you can transgress. One is if somebody is a suppliant at a temple, stretching out their hands to seek the protection of a divinity. Those people are off limits. That doesn’t mean they never get killed, there are various occasions when the Greeks do kill them. But that is understood to invoke a curse upon you, your family into the nth generation is going to be cursed, the gods are going to want comeuppance.

These understandings exist both in a religious sphere, envoys of the gods, people going to the Olympics to attend a festival dedicated to Zeus, those people you can’t touch, there’s a truce, hands off, because they’re sacred. But there are also practical understandings. If somebody supplicates you, reaches out to clasp your knees and throw themselves at your mercy, you can kill them, and in the Iliad, when Menelaus wants to offer mercy to somebody who does this, Agamemnon steps in and says, no, you should kill them all. But it isn’t good. This is something you shouldn’t do. If people have fallen into your hands as prisoners, there are very rare occasions when they’re mutilated. Killed, they sometimes are, but mutilated is very rare. You very rarely mutilate either corpses or live captives. So these things are absolutely recognized by the Greeks as just wrong. You don’t do that.

Violence against women and children is sometimes practiced, but sometimes also deliberately eschewed. They try to avoid it. It depends on the situation. In some cases they are horrifically violent against women and children. The enslavement process they often go through is absolutely horrific, and we shouldn’t sugarcoat that. But sometimes the population of a city that surrenders from siege is allowed to go free, allowed to become refugees and look after themselves. So there are ways the Greeks recognize that it is better to be good. That doesn’t always restrict their actions, but they do recognize a morality in those actions.

Jordan Schneider: Coming back to the Odyssey. At the end we have Odysseus. He gets three supplicants, saves two, kills one. It’s unclear where exactly he draws the line. And then he murders some of his female slaves, who he’s blaming for getting raped by the suitors. It seems like you’re getting a bit of that judgment in there from Homer, even though he’s trying to make you feel for the guy.

Roel Konijnendijk: Absolutely. I think this is one of those cases, especially that scene where he kills the enslaved women in his own household because they collaborated with the suitors. This is something that also happened, for instance, in occupied Europe after the liberation, a lot of women in local populations suffered horrific punishment and ostracism because they had relations with the occupying forces, with the Nazis. You can say they weren’t doing that out of free will, or you could say the Netherlands, for instance, was occupied for almost five years. You have a reality that doesn’t seem to be going away, so you deal with it, you try to continue a normal life as best you can. Some of that may be open to people’s free will, some may not. But populations do have a tendency to react as if it’s all deliberate collaboration with the enemy. That is often unfair, but it’s also something that happens. So that’s one of those occasions where Homer wants you to talk about it, to discuss it and say, do you think this is right? Do you think Odysseus and Telemachus are justified, or do you think this is where he goes too far?

That’s also true in the Iliad, which is also about people transgressing. You can debate who the good guys and bad guys are, but there’s always people who cross a line. Agamemnon is not wrong to say he should be the most honored, because he has the largest contribution and the most authority in the army. But when he steals Briseis from Achilles and demands her specifically as his prize, he’s going too far, because he’s not just granting himself the highest honor, which everybody agrees he deserves, he’s taking it away from someone else. And then later, when he tries to make up for it, saying, okay, Achilles, I’m sorry, have all the gifts you could possibly imagine, a huge list including entire cities, if you’ll just come back into the fold and fight again, and Achilles refuses. Then we’re meant to understand it’s Achilles who’s transgressing. He refuses a good compromise, a generous offer, and isn’t willing to play ball with the way human society works. When someone tries to say sorry and you say, no, I don’t forgive you, then the onus comes to be on you. And Achilles goes further and further from there. Ultimately he takes a couple of captured Trojan prisoners of war and throws them onto Patroclus’s funeral pyre. And every Greek knows that’s going too far. You’re making human sacrifices to your buddy. This is absolutely unacceptable.

So there are a lot of those moments where Homer’s deliberately trying to tell you, yeah, you can come along with me as I describe this very emotional, recognizable, relatable response to something. But I’m also going to show you there’s a limit to how much you’re willing to stomach, how much you’re willing to say, okay, I’m with this character so far, I believe they’re doing the right thing. And then he says, okay, but how about now? He keeps pushing that. I think the composers of this poem want you to think about what’s right in these different scenarios.

Immortality or Glory

Justin: Achilles is the greatest hero, name written in the stars, talked about for thousands of years. And then his oration as a ghost, when he says if he was given the choice again between being immortal and being remembered forever, he would choose to be immortal, because death is terrible. What is your interpretation of what Homer was trying to say about the value of legacy and martial virtue versus just being alive?

Roel Konijnendijk: It’s hard to know. You have those two contradicting accounts, that Achilles very deliberately chooses a short life and eternal glory, and then once he’s dead and shows up briefly in the Odyssey, he says, actually, it was wrong. There isn’t necessarily a single morality there. It makes sense from the perspective that Homer is trying to explore the human condition. What good is anything to you if you’re dead? What’s the point of having eternal glory and your name remembered if you’re not around? You can’t experience anything. The way they describe the underworld, there are shades living in an eternal fog in the darkness. There’s nothing for them, they’re hungry, they’re thirsty, they’re suffering. So he would have given up a lot to have a few more years just living life, enjoying the things you enjoy and suffering the things you suffer as a human being.

The Greeks certainly didn’t think being alive was good per se. The life of an ordinary person was very hard. There’s a lot of poetry, like Hesiod’s Works and Days, describing how you have to work really hard to live even a little bit comfortably. It’s hard, there are cold winters and hot summers and hard work and bad people. But all of that still means you get to experience things. All of us who think about the afterlife in any form usually consider, is it better to be in whatever that place is, or better to keep on doing this, to keep experiencing my own consciousness in this world?

It’s one of the ways you can see that even though this poem was composed twenty-seven hundred years ago, in the form we have, it’s incredibly prescient in the sense of which questions are going to matter, which are going to keep us talking for a very long time. When you read Works and Days, you’ll find very little that makes you go, yeah, I could use this to open a conversation about my view on life. Whereas the Odyssey is absolutely packed with it. We know why this survived. It keeps us talking, keeps us thinking, keeps us making this same consideration of, well, was Achilles wrong? What should he have done? That must have been at the forefront of the minds of all these people who live in a world where you somewhat regularly go to war yourself, risk your life in battle, where the only reward you can expect to receive is a memory, to be well remembered as a person who died doing what’s right as a citizen. You have to wonder, do I want that, or is it better to run away so I can live a little bit longer? That’s obviously on a lot of people’s minds in that world, and after.

Jordan Schneider: If you’re thinking about what literature is going to last, and by last we mean hundreds, thousands of years, it can’t just be compelling, it can’t just be the blood-gore popcorn stuff. It needs to have people split on it, reasonable people split, and continue to split on it in totally different cultural contexts, so you want to keep talking about it. I’m sure there are Hesiod arguments about is this god the better god or that god the better god, but no one cares once Christ comes around. Getting on those timeless human themes is a pretty hard thing to do in the moment. Hats off to Homer for pulling it off.

Roel Konijnendijk: Some of it is timeless. Other things you’re just like, okay, there must have been an environment in which this was very useful, but I don’t think now we need Hesiod to explain to us that we shouldn’t piss in the river where we get our water. Which he spends some time explaining.

Tony Stark: Unfortunately, battlefield hygiene is still a challenge for a lot of folks.

Justin: If you had worked with privates in the military, you would know that you still need to teach that lesson regularly. I mean, tie it to the two hundred and fiftieth anniversary of America. It was a Prussian who came over and had to look at Washington and be like, you recognize that your men are shitting next to where they’re cooking their food, right? That’s the reason you all have dysentery.

Roel Konijnendijk: That’s that Prussian military method rearing its head again. I read about that, when Steuben’s training the Continental Army, where he started out drilling Washington’s guard and then they passed it on to the rest of the army. The idea there is very much similar to the Greek situation, where these are militia that have received very little training and are constantly being defeated by a more professional force. And what they’re learning is not that difficult. It was like two weeks that he spent training them. And that’s for musket drill, way more complicated than what Greek hoplites would have to do. So just that basic element, it’s not that you have a lot of trouble figuring it out, it’s that you need someone to tell you, okay, this is what you’re going to do, and take you through your paces, set aside the time, which they had in Valley Forge. And you come out a much more effective force. Sometimes you just need someone to drop in and say, here’s how you can get much better very easily.

This is one of the ways knowledge was passed on in a period before you had dedicated treatises. You had these poems aspiring to teach everything, from morality to the practicalities of military organization and combat and farming and proper marriage relations. Those things are discussed and composed into songs and texts that last longer, precisely because there’s no community college, no way to get this from a professional. You’re passing it around as stories, as oral tradition, and eventually as written poetry.

Cultural Blinders

Jordan Schneider: I want to come back to violence. It was a fascinating theme in your book, this idea that everyone wanted to think these battles were clean and respectable, that these citizen oligarchs of this city didn’t really want to kill that many oligarchs of that city. What was so weird was the evidence you used to argue against it. It’s just right there. It’s pretty obvious in a lot of these texts that once the other army starts fleeing, you run and try to kill as many as you can.

Given that we spend a lot of time on this podcast talking about lessons from Ukraine, where there’s so much data, so many different things you can look at, it boggles my mind that with these battles you have a limited amount of text and artifacts to work with, and people will just overlook things because they have cultural blinders of their present day. I’m sure fifty years from now your PhD students’ PhD students will say the same about you. Any thoughts on people missing stuff in the text that they don’t want to read?

Roel Konijnendijk: It’s a really interesting point. The argument I made in the book is that these earlier authors were so guided by their own conception of what these battles looked like, how they wanted to model them, create a template of what a battle typically looked like, in which they included statements from the sources that seemed generalizing, like, okay, the Spartans never pursue a fleeing enemy. These statements occur. They’re not true, but they occur. And if you want, you can privilege that information over the counterexamples you could conjure up from specific accounts. They did that initially because they had this model in mind, this strong sense of wanting to make it make sense as an overall picture, and they didn’t care so much about the smaller details.

And they had this investment in the idea that the Greeks were deliberately hobbling themselves, denying themselves the opportunity for the victory of annihilation that Clausewitz would privilege, because they were more civilized, more gentlemanly, but also because they hadn’t quite invented yet how to be maximally effective in war. This goes through a linear progression, from the Greeks to Alexander, and then Alexander to the Romans, getting better and better at figuring out how to be maximally effective, which lets them conquer larger empires and defeat harder enemies. If you have a linear development, you’ve got to start somewhere where it’s not very effective and build it up.

But the other thing is, there’s a point at which this grows above the head of an ordinary person trying to make sense of it all. This is why increasingly, when working with large data sets, we rely on automated processes to help us understand these things, put them into graphs and charts, because otherwise we can’t handle the data. In the study of ancient warfare, despite the comparatively small amount of evidence, there’s already this sense in which people will decide on something to privilege, some narrative to tell, and in doing so decide other things are less deserving of their attention. A lot of the time they’ll explicitly say, this is just a sideshow, not the main issue.

Jordan Schneider: People decide the story because it fits with some Victor Davis Hanson broader political narrative they have. Or they just want to piss off their advisor. When you see people go story-first, where does it often come from?

Roel Konijnendijk: Very often it comes from the historiographical tradition, other scholars they’ve read. That has so much influence. We all have this image of Sparta as a super-conservative militaristic society focused exclusively on war. If you talk to an expert who works on Sparta now, they’ll reject all of that. But that’s the cutting edge. If you pick up a book from a bookstore, you’ll pick up Paul Cartledge’s book on the Spartans from the 1970s, and you’ll read all of that old account exactly as I’ve just said it. It’s not that it was wrong, it’s that it was believed at the time, the best of their understanding, which may have come from political ideologies or personal proclivities. But now those ideas are just around, and it’s very hard to say, you have a bunch of books on Sparta all written by respected academics, but these are wrong and these are good. It takes a long time to develop the critical ability to do that for yourself.

Around 1990, everything changed. Up to then, everybody had more or less the same idea of what Spartan society was like. And then suddenly a number of scholars from different angles started to completely reinterpret the evidence, trying to privilege other parts than the ones previously taken as baseline, saying we need to look more at contemporary than later evidence, and more critically at what the evidence tells us, because it isn’t always straightforwardly telling us what we want to know. Getting a better understanding of what genre and propaganda do to the records that survive. All of that contributes to thinking better about the material. And that’s why views in scholarship can change quite radically even in ancient history, where the literary evidence has been more or less stable for a hundred and twenty years. The ideas that come out of it keep changing.

Reading Recommendations

Jordan Schneider: I’m curious, in the ChatGPT and Claude age, where you can just ask these models to surface all the references instead of going back to the books and flipping through, do you see anything interesting in how your grad students are using these tools, or has it changed your process?

Roel Konijnendijk: You’ve probably encountered other humanities scholars being extremely hesitant to foster this technology, because we generally don’t trust it. It can get information for you, but you still have to check whether its references actually exist and whether the material says what it says it says. People are using this technology in ancient history especially where there are large data sets, like inscriptions. In Latin there are tens of thousands, often partial, often damaged, and scholars use large language models to feed in all the ones we have and help predict what might fill a gap in the formula. It’s used for that sort of thing.

But when you’re looking at something like what I study, which is mostly qualitative and literary, you can’t just ask the question and have the machine tell you all you need to know. You need not just familiarity with the passages, but to know what they are in context. When I take my students through some of these passages, we can all read the same passage in a short amount of time. A lot of the time in ancient history you can read all the available evidence for a particular question in half an hour. But that doesn’t mean the students can immediately say anything I would say, because they don’t have the same grounding, they don’t know what these sources are, how to rate them, how to evaluate them.

As is often true in humanities research, the journey is the end. It’s about building up the expertise to assess these things critically. You can’t outsource that, because the machine can’t do it. It doesn’t know what you want to know. So I hope people don’t try to use LLMs to replace the kind of work I do, because the work would be terrible. You need to commit your own expertise and critical thinking. Otherwise you’re just going to regurgitate what was already there.

Jordan Schneider: This is my pitch for why this stuff is particularly fun to read in undergrad. If you’re trying to get to the frontier of World War II studies, I’m sorry, that’s ten years. You’re not going to consume all the primary sources. But if you want to deeply understand even something as big as the Peloponnesian War, over the course of a semester you can consume a pretty high percentage of the primary source material that’s out there. And then all of a sudden you can engage with the secondary sources at something approaching a peer level, in a way that just isn’t possible with modern stuff because there’s too much out there.

Roel Konijnendijk: Absolutely. It’s one of the things I always say I enjoy most about ancient history, the relatively low threshold. You can read everything there is to know about Sparta in a year. Every single thing. But then the next thing you need to do is develop the empathy, creativity, imagination, and critical thinking that lets you actually construct theories about historical reality based on that limited material. How do you get from reading all of that to having something to say about it? That’s a fascinating process, and something you can’t replicate with a machine. That is the great fun of ancient history, you really have to use your brain. You have to step in and say, what do I make of all this? Can I resolve these contradictions? Can I fill these gaps? You have to get to the level of comfort and confidence where you can say, this doesn’t scare me.

Jordan Schneider: What do you want to get the kids going on? What are your primary sources to throw folks at?

Roel Konijnendijk: It depends. Nowadays a lot of people come into classics through myth, which is radically different from the way it used to happen. They’re reading myth adaptations or Percy Jackson, and that’s what brings them in. So if that floats your boat, you’ve got to read Ovid’s Metamorphoses, or Apollonius. Those are the main sources we have for the myths we’re retelling.

For the WarTalk option, the Iliad is great. There’s just no two ways about that. It gets boring in parts of the beginning, because you have to go through the catalogue of ships and who kills who in the fighting, but you have to think about that as functional for the text. The Iliad is absolutely wonderful, and there are some really good translations, so you almost can’t go wrong.

I’m a huge fan of Xenophon, who is himself an experienced commander writing about warfare. Polybius similarly, although he’s very unpleasant as an author, I really don’t like him very much, but he’s a good military historian willing to go into a lot of detail. These are all people you can find cheap translations of everywhere, a lot of them online for free. Herodotus is the gateway drug for people interested more on the history side, just because he’s so varied. Some of it very military, especially books seven and nine, Thermopylae and Plataea, all warfare all the time, but the rest is so interested in the enormous scope and diversity of the world around him. He’s fascinated by everything, and it’s something that infects you when you’re reading it.

Jordan Schneider: I just want to shout out Lucian of Samosata. I don’t think there’s any war stuff in there, maybe I haven’t come across it yet. I’ve been writing more and more satire this year, and I was like, who’s some ancient-world satire I should start with? The guy’s hysterical. He’s making fun of all the philosophers, the gods, society. He has a sci-fi novel in there where the humans go to the moon. He’s just a wacky dude. And he’s from two thousand years ago. What a treat.

Roel Konijnendijk: Who would have thought? But you have to imagine this is not the ordinary people who live in that society. These are the people whose works are preserved over two thousand years. And the first fifteen hundred of those, they had to be manually copied. Every now and then a monk has to sit down and physically write every single word he sees on the page, otherwise it gets lost.

Jordan Schneider: He’s making fun of Jesus. The fact that you can make fun of Jesus and then be so loved that you still get the monks to copy you, you really have to be on your shit.

Roel Konijnendijk: That’s a boss move, I have to admit.

Justin: My favorite is still Antigone. I got to hear Lucy Lawless read the part of Antigone on a show more than a decade ago, and it was a throwback to my childhood. It’s seared into my mind now, she is Antigone. Every time I think of an ancient Greek woman, Lucy Lawless becomes that person.

Roel Konijnendijk: It’s another one of those examples of a text really thinking about the morality of war. Sophocles’ Antigone is also great.

Jordan Schneider: I’ve been listening to two Odysseys simultaneously. There cannot be a stronger contrast between Claire Danes’s Emily Wilson Odyssey and Ian McKellen’s Fagles Odyssey. It’s the same text we’re working from, but it’s so fun. These things contain multitudes.

Roel Konijnendijk: I didn’t know Claire Danes did the voice work for that. That’s great, I’ve got to find it somewhere.

Jordan Schneider: She commits. Keeping that tone up for fifteen hours is a remarkable accomplishment.

Roel Konijnendijk: Twelve thousand lines. I wouldn’t want to do it.

Jordan Schneider: We’re going to call it here. We’ve got to do this again sometime, this was such a treat. Thank you so much for being a part of WarTalk.

Roel Konijnendijk: It’s a real pleasure. Thanks so much for having me.

Just the OG crew with Bryan Clark, senior fellow at the Hudson Institute and a former submarine officer, a former special operations officer who works on defense technology, and .

We discuss…

• What Trump is doing in Iran, why the Navy is short-cycling ships out of maintenance and what happened the last time the fleet ran this hot, from McCain and Fitzgerald to the Truman’s two collisions in one deployment

• Why a ground campaign in Iran is a fantasy, from the Marsh Arabs to the Alborz range to the Baluch salt flats, and what Shia militias in Iraq would do to the logistics tail

• The nuclear-powered cargo ship renaissance and why circumnavigating Africa at 30 knots suddenly pencils out

• Whether Congress should get an AI kill switch, why DHS is the wrong home for it, and what happens when the boy who cried wolf is finally right

• Ukraine’s civ-mil drama

Listen now on your favorite podcast app.

Iran

Bryan Clark: So here we are, Groundhog Day in Iran. Just yesterday the Iranians apparently turned down some ceasefire offer from the United States. Who knows what that offer contained, but the Iranians have decided to refuse it, and the US responded with more strikes. So we’re continuing to attack targets, increasingly attacking infrastructure and power supplies and water supplies that might be considered civilian, but seem to have maybe some tangential use for the military. And with the IRGC being increasingly enmeshed in the population, it’s hard to tell what’s exactly a military target and what’s a civilian target in Iran.

So now we’ve started another round of strikes. We are continuing to blockade the Iranian ports and Iranian access to the Strait of Hormuz. The US military believes that this blockade can be successful at bringing Iran to the negotiating table because it’s obviously causing some damage to the Iranian economy. But hard to tell, because Iran had been bearing down on their population for a long time anyway. And so they’re gonna bear a lot more than this, probably.

So the blockade is in place, continued. A dribble of ships continue to make their way out along the coast of Oman. So we’re talking single-digit ships still making their way out, ostensibly with the protection of US naval and air forces, but nowhere near where we need to be in order to restore significant flows of oil and gas coming out of the Gulf.

And now you’ve got the Houthis joining the fight. They’ve entered the chat and decided to announce a blockade of the Bab el-Mandeb Strait. Unclear how they’re going to enforce it. There’s lots of options on the table, a couple of which the US military is actually very worried about. And they’ve got some Iranian advisors that have come over probably to give them some lessons learned from Iran’s recent experiences.

Justin McIntosh: It’s interesting, the Houthis one. I don’t even know if they really need the Iranian advisors. They basically perfected shooting down MQ-9s and then taught the Iranians how to do it. So realistically it’s more of an exchange of ideas than it is a we’re gonna build up kind of thing.

Before we get into that though, Bryan, we have two carrier strike groups right now that have been there for a while. How much longer? We’ve already talked about getting into the red on naval readiness writ large, but how much longer can we realistically sustain keeping two carrier strike groups, especially when you can’t go to Bahrain, you can’t go to some of the ports to refit, and then still have global naval readiness?

Tony Stark: I think the tell is that that resupply ship got a presidential unit citation, right, Bryan?

Bryan Clark: Yes. So you’ve got a PUC going to a support ship crew and its military detachment for being out there for almost a year, essentially supporting these guys 24/7. The combat logistics force, the oilers the Navy uses to refuel these ships — they’ve gone to a model where the oilers just kind of stay out there and then the oiler gets refueled by a tanker that comes out from Europe or from Crete or from the Pacific. Because they can’t get access to any of the oil that’s inside the Persian Gulf.

So they’re having to bring tankers out to the Arabian Sea, offload fuel to oilers, then oilers offload it to ships, and then also to the carrier to provide the airwing with fuel. So you’ve got this daisy chain now going for thousands of miles to refuel the fleet.

And then to Justin’s point, you’ve got two carrier strike groups. So you’ve got Bush and Lincoln out there. Bush hasn’t been out there that long, they got there relatively recently. Lincoln’s been out there for I think ten months now. They’re going to probably reach the Ford’s record here pretty soon. And 16 to 20 surface combatants, so destroyers on any given day, out there escorting these carriers to defend them, because the Iranians do attack them. The Iranians have sent one-way attack drones out to attack the carriers, they’ve sent missiles out to attack the carriers. We don’t hear much about it, but it’s been happening, and the Navy just shoots them down. Still, it’s a drain. And of course you gotta be totally combat ready all the time.

So the Navy’s experiencing, I think, kind of what ground guys had experienced in Iraq and Afghanistan of constantly being under threat, and the toll that takes on you. And then in addition to defending them, they’re having to enforce the blockade. So you’ve got surface ships out there trying to ride herd on Iranian ports and prevent ships from going in and out, and periodically ships will try to run it. So they do have to intervene and shoot at ships or even disable them occasionally.

Justin McIntosh: But long term, when you look at that, what is realistically sustainable with what we have?

Eleven Months at Sea

Bryan Clark: Once you get to about a year — a year is well past their normal deployment time. Normally a deployment’s about seven months right now for a Navy surface ship. So a lot of them are out eleven months, reaching nearly a year. So they’ve got all kinds of maintenance issues that probably need to be addressed. Obviously the crew is tired. You only bring people out to replace people who are injured or sick or need to go home for family issues. But for the most part, the crew’s there the whole time. They get tired and worn out.

So usually at about a year point, you have to probably swap out ships. And what the Navy’s doing right now is figuring out which ships can we pull out of the normal deployment cycle early to go out and replace these ships. Normally the deployment cycle for a surface ship is like three years. So you have a seven-month deployment, you come in and do maintenance, you do training, you do some local operations and workups, and then you deploy again two and a half years after you’re done with your maintenance period. But now they’re having to take these guys and short-cycle them to relieve the ships that have been out there for almost a year at this point.

So they’re having to tap into that reserve capacity. And unclear how long you can do that really, because if you take ships out of their training and maintenance cycles early, they’re not going to be fully ready. They’re gonna have problems that come up when you’re out at sea and they’re gonna have to go back early probably. So you eventually start burning through your reserve capacity, and you end up with ships just having to leave deployment because they’re broke.

Tony Stark: And we saw what happens when you do that, right? Because in the late 2010s you had the McCain and everybody else who was overworked, under-maintained. When a ship starts to hit those sorts of hours and maintenance challenges, what does it look like from a life cycle for the crew?

Bryan Clark: So the crews have enough people to where they’re probably in a three-watch rotation. So you’re standing six- to eight-hour watches and you’re getting twelve or sixteen hours off. But when they have to ramp up and they’re actually going into combat operations, they usually go to port and starboard. So you’re basically twelve on, twelve off. Or four on, four off. So you get very little time off in between, you don’t get a lot of sleep. Obviously they’re stressed out because of the need to constantly respond to attacks. So you end up in the same situation that a lot of ground guys have experienced, which is you just get totally burned out and your performance degrades rapidly. And they’re probably at the point now where performance is degrading rapidly.

So for example, we’ve had some accidents. We had a Navy helicopter go down while it was doing routine flight operations, and they lost the CO of the helicopter squadron at sea. That’s the kind of stuff that begins to happen. And like we saw with McCain and Fitzgerald back in 2017, because we were working the fleet and not really giving people adequate time for rest and training in between. We had some material issues too on those ships, but you end up with a collision between two destroyers that took them both out of service for years before they could return.

That’s the kind of stuff we risk happening now. We’re just lucky we haven’t seen anything like that. Two years ago we saw it with the Truman — two collisions in the course of a single deployment, and they lost an airplane when they were doing some maneuvers because they didn’t respond correctly to a Houthi drone attack. So this kind of stuff will happen more frequently as we extend this deployment and don’t have adequate capacity to relieve these guys at a necessary tempo.

No Good Options on the Ground

Tony Stark: I kinda wanna pivot to the ground game here a little bit, because I saw the most recent attacks against the border crossings. For those who don’t know on a map, southwestern Iran is very close to southern Iraq, which is very close to Kuwait. And it seems like some of the border crossings are getting hit. Maybe an Iranian listening outpost was hit. This is what came through my feed this morning. But is this just random tit for tat, or is this phase zero prep for something? Justin, I’ll leave it to you.

Justin McIntosh: There’s been a lot of talk in the Pentagon, and a lot of talk that’s been coming out in reporting over the last two days, that there have been big pushes from Marine Corps and military commanders to take some of the islands up to and including Kharg Island, and to start taking territory. Maybe it’s preparation.

Southeastern Iraq is where a group of people called the Marsh Arabs are from, and they’re called the Marsh Arabs for a reason. They are Arab people who live in a marshland, which is one of the things that made it very hard for Saddam to use his mechanized forces to cross over there. It’s tough terrain. I mean, all of Iran is tough terrain. It’s either islands that are relatively barren with the exception of oil infrastructure and are within range of the mainland, or it’s sheer cliffs, or it’s the Alborz mountain range that is domineering. It’s not the Hindu Kush, but it is a very high mountain range that is hard to pass through. And then on the western approach into Iran it’s a marshland, the Kuwait-Iraq area, with all the disease and everything else that comes with that.

There are no really good options. Iran from the north to Afghanistan, which obviously we don’t have anything there anymore so that’s not an option, would be mountainous terrain and incredibly unforgiving. Even if you were to look at it from the eastern part of Iran, it’s the Baluch Desert, which is salt flats basically that are incredibly hard to pass. It’s a tough country, geographically.

And the idea that you’re gonna be able to go take terrain, or that you would even be able to do preparatory fires and take terrain, makes it incredibly difficult to even surmise what would be the end state. And you would get the rallying effect — in each of those provinces there is an IRGC Quds Force and Basij recruiting station that’s going to use any type of attacks to draw up the local militia and increase the forces and get them ready. And those are all regional, state-level militia basically across the country.

So yes, they are definitely talking about it. But I would hope there is a reticence to say, all right, ground troops are what we need at this point, because I don’t know how we would sustain providing them air cover and naval cover and support over time with the current situation.

Bryan Clark: We’ve got no naval presence in the Persian Gulf. There’s no Coast Guard presence, no naval presence. The bases the Navy’s operated from, and some of the ones the other forces have operated from, are gone. I think there’s a lot of places that we’re just not going back to anymore. So they’d be operating very much on their own, I would think.

Justin McIntosh: And then you’ve got, no matter what, the Shia-backed militias in Iraq that definitely still exist. Maybe they haven’t been conducting overt attacks that we can see against US and coalition positions in Iraq, which would be staging locations that would absolutely be used for logistics resupply, command and control — you name the function that needs to be provided to any type of ground assault. Those would absolutely start coming under pressure. And it would absolutely be attacks in the rear.

I forget which Shia militia group it was the other day, posted videos of basically their fighters out flying FPVs and taking out targets that look stunningly like Baghdad basing that the US is using. It’s clearly sending a signal that, hey, you can escalate a ground game, we can also escalate a ground game. And it’s not going to be a symmetrical escalation. If we go conventional forces on the ground, there’s definitely going to be an escalation asymmetrically against our logistics and supply hubs within the region that’s going to be further than what we’ve seen so far.

Bryan Clark: The problem with the situation is the US doesn’t have very many tools to escalate effectively. They’re squeezing Iran with the blockade, which is gonna take months if it ever yields a real benefit strategically. And we really can’t generate any more useful strikes than we already are generating. And a ground game just seems like an opportunity to create a lot more targets, and there’s too many things that can attack them. And we’re just not able to defend. We can’t even defend the targets we have, much less new targets that we’re going to distribute around the region.

Justin McIntosh: That’s a really great point. You look back to the Gulf War — you could look back to 2003 as well, but let’s say the Gulf War. The Iraqi invasion was in July, August. We didn’t launch the counterattack until January of the next year, and the reason for that was force buildup. Looking at Iran for any type of meaningful ground offensive operations, you’re gonna require a buildup. And all of that buildup as it flows into the country is going to be vulnerable to being struck as it’s being built up, because you’re going to need air defense systems around it, you’re going to need logistics supply, lines of communication that are all going to have to flow.

This isn’t as simple as picking up several armies’ worth of US military and flying them very quickly over the course of an evening into Iran and seizing terrain. That’s not how the US military templates deploying its forces.

The Food Problem

Bryan Clark: One thing I was gonna mention on the naval front is one of the challenges I heard from a senior person over the last week: food. They’ve got this scheme to be able to ferry fuel and petroleum products into the theater, because they use tankers to oilers to the ships. But food doesn’t have a similar model. So getting food out to the ships is actually really challenging, and the cargo ships are having to travel from a really long ways away to bring that food.

The military still requires our food to come from the US. There’s very limited opportunities to buy food locally. So you really don’t have the ability to have the supply officer go ashore like we used to do and buy food on the local economy. So food is a problem. Having adequate food for two carrier strike groups is just a huge endeavor. And if they’re out there for a year, you start to burn through all the good food and you get down to the frozen stuff. And they don’t have MREs. So they have to look at a new model for how do you feed people at sea in an environment where the logistics trains you would rely on are going to be contested to a degree where I can’t just have a constant stream of cargo ships coming out to a carrier strike group like I used to do.

Justin McIntosh: What’s that naval shanty, the Wellerman? Everybody’s just gonna be on deck singing and hoping that there’s a boat that’s gonna come up and bring them their food and sugar.

Bryan Clark: Yeah, that’s right.

Justin McIntosh: The logistics are becoming really constrained. That’s the reason we had the global basing and the global footprint. And the fact that that’s being challenged does really strip away a lot of capabilities. The global force posture — people can say it’s empire or they can say whatever it was. Realistically, the global force posture existed to be a skeleton for a logistics model that would allow a flow of forces into regions to protect vital national interests. And by scaling that back in Europe, already having started scaling that back in the Middle East, and then for it to be under attack, you’ve put yourself in a position where you don’t have the skeletal structure remaining to do the next big thing in the region.

And it’s really interesting, because this goes to a point the administration really struggles with. They have the strategy of denial: we need to do more in this hemisphere and less everywhere else. So they started stripping away, and then they started doing stuff everywhere else. And they started using combat power after they had stripped away the actual capabilities that it takes to sustain that combat power.

My dad used to make the comment, because he was in the military — he wasn’t an infantryman or special forces or anything, but he would make the comment that for every one infantryman, there’s twenty support people that make sure they get paid and clothed and fed and the maintenance is done on their stuff. That’s what we stripped out of Europe. That’s what we took out of the Middle East, those large maintenance depots and things like that. And now you get into a point where you need them.

Bryan Clark: And arguably going forward, we’re probably not gonna go back to a lot of these bases in the Middle East. I really would be surprised if the Navy goes back to Bahrain. I think they’re probably gonna just write that off, and we’re gonna come up with a new model for how we maintain naval forces in the region and command and control. And then places in the UAE, places in Kuwait that we’ve relied on — I could see a lot of those places not getting rebuilt. And we end up having a much lessened posture in the Middle East, which obviously is gonna have significant implications.

How Does This End?

Jordan Schneider: Should we talk about how this doesn’t end? I thought we ended this for the midterms, right?

Tony Stark: So from the strategy perspective, I’ve been struggling to understand — it’s very clear that any decisions are not being flowed down through the State Department or through DOD. There’s no, hey, reach down, ask for an answer and get what you get back. It is very clearly like you’re just dragging the kid around the soccer field, chasing the ball.

And look, Congress goes on recess any day now. The House has voted to end the war.

Justin McIntosh: I mean, really they’ve been on recess for a while. For all intents and purposes.

Tony Stark: The House voted to end the war by four votes, I think. The Senate was either supposed to vote today or yesterday, I didn’t look. But that’s not going anywhere, because you’re not gonna override the president’s veto. Also the president doesn’t think he needs Congress’s opinion for this. At least from a money standpoint, I think the ask is now down to thirty billion and they’re probably not getting that. Good luck with a budget before the end of the year.

From a DOD perspective this is bad. From a government perspective, the poll numbers I saw were like a majority of Americans don’t approve of the supplemental for the DOD budget. I don’t think they ever do. But I think it’s almost one third of Republicans think Netanyahu should be arrested. Although let’s be clear, that might have some other implications there. It’s not gonna go well in the fall. So I don’t think anyone is saying, hey guys, we gotta wrap this up — unless they really think that nothing matters.

Justin McIntosh: One of the things you’re definitely hitting on, tangential to this but definitely spurred by it, was the deal with Saudi Arabia. Potential deal. Tweet. I don’t know what we want to call it. There was a thing that happened, and we announced that we were gonna allow them to develop nuclear power capabilities.

Bryan Clark: They signed a paper.

Justin McIntosh: Let’s say that happened. But then at the next press conference the White House press secretary comes out and says, well, no, the president is changing it. Basically his truths that he’s sending out on Truth Social are actually government policy. So he’s just issuing changes to what was ostensibly an agreement with the Kingdom of Saudi Arabia based off a feeling, without consulting anyone from Saudi Arabia.

I don’t know why we would expect that to be different with the way Iran sees that. Even if they were to say, we can’t take this anymore, we need to come to the table, we need to get to a negotiated settlement — if you’re openly saying that even with your allies you’re willing to agree to something and then instantly change it, what is their incentive to actually go take on the hardliners within their government who do not want to make a deal with the United States, for something that will get out of their control instantly?

Bryan Clark: The other issue with Saudi Arabia goes back to the Houthis, because Saudi Arabia has been using the Red Sea to offload a lot of their oil, and there’s new pipelines going in that are attempting to use that alternative route around the Strait of Hormuz. So as part of the energy flows reshifting to accommodate the Strait of Hormuz not being open, Saudi Arabia was benefiting, because they have the ability to get oil out through the Red Sea. And that’s now going away because they’ve lost access to the Bab el-Mandeb potentially.

And the Houthis know that. And of course the Houthis are in a civil war in Yemen which is being supported by Saudi Arabia. So there’s this other way the fight plays out where the Houthis have their own gripes with the Saudis. They close the Bab el-Mandeb, they shut off Saudi Arabia’s alternative path for oil to at least Asia. And now you see the real pressure get brought to the oil markets, where there really are few alternatives and the depletion of the reserve stocks is causing the price per barrel to climb pretty quickly. So all this is gonna drive the US to have to decide if they’re gonna continue to bear this, or if they’re just gonna let Iran get its control over the Strait of Hormuz and charge its tolls and call it a day.

Jordan Schneider: And to be clear — would we have been able to have that steady state of the MOU? Is it worth going back and understanding the trigger of all this two weeks in?

Tony Stark: Fundamentally, there is not a single sole government in Tehran. Or I should say in Iran, because I suspect there are different vying factions throughout. And I think the government in Tehran is never going to admit, at least publicly, that the IRGC might go off and do its own thing, which I think is how this probably came about. There were probably some radical commanders who went off shooting, and then people realized they don’t have control, and egos, et cetera.

Short of a joint Iranian-American attack on the IRGC, I don’t see how this goes anywhere until they run out of ammo. Maybe there’s a world in which the factories get cut off, but that’s a while away.

Bryan Clark: I feel like the IRGC is who wants to charge the tolls. They want to incorporate this into their business plan, because that is kind of their business — extortion, getting money from the local population and running businesses. They’re basically a mob. This is now a new line of business for them.

And the reason we got into this most recent back and forth was because the US implied that Iran didn’t have control over the strait, even though the MOU said Iran and Oman needed to work something out. But when the US and the president said, no, no, the strait’s open, Iran had to go reassert its authority. So they did that, and shot some ships. And then we got into this, because neither side was willing to back down. But if the US had been willing to accept the fact that Iran and Oman are gonna work something out and just step back, I don’t think we would have gotten to these most recent exchanges of fire. Now, that would be the IRGC essentially running an extortion racket in the Strait of Hormuz and charging tolls and protection to shipping companies. But at least the spice would flow, and we would end up not having to keep our navy and our army there for more months.

What Winning Would Actually Take

Jordan Schneider: Can someone tell an optimistic story? How does round two end up better than round one?

Tony Stark: We go after the right targets this time and we don’t do a fucking random strike plan that the world’s worst staff officers wrote because they were bored in the desert. That’s really it. You go after the targets that actually enable them to control the strait or at least disrupt it. And you just say, all right, end game here, we’ll leave the rest of your ground forces alone, you can have your power, whatever, but this whole throwing missiles at tankers is not happening.

Justin McIntosh: The first thing is, if we were really gonna up the pressure and say, this is unsustainable, you guys cannot control the strait and we’re gonna actually turn up the punishment — I think that’s the very first thing that has to happen. And there can’t be any of this “they hit it but they were able to dig out the missiles and keep shooting them.” It is a sustained, pre-invasion-of-Iraq-style air campaign that requires a whole lot of risk and a ton of capability, but it just doesn’t stop against all of their military targets. And that’s also been one of the things we haven’t been willing to do. We are really engaged in a tit for tat. And in some ways we’re engaged in this tit for tat because everybody kind of knows there’s a limit to how many exquisite weapon systems we have that let our fighters and bombers stay well out of range.

Again, you either have to be willing to accept the risk that you’re gonna lose some aircraft and actually go after these things, or —

Bryan Clark: Justin, I assume if you’re gonna hit these targets — if you’re gonna hit Shahed launchers along the coastline, given how rough that terrain is, you’re gonna have to use a laser-guided bomb and somebody’s gonna have to laser that thing, and that airplane’s gonna have to get close enough to hit the target accurately. You can’t just drop a bunch of JDAMs on it and hope for the best, because the terrain’s too rough. You’re gonna miss them.

Justin McIntosh: At minimum you’re gonna want tandem, ones and twos flying in where two is hosting one and helping them put their bomb on target, making sure they’re actually hitting what they’re aiming at. Which means you’re setting yourselves into flight patterns. That means you’re setting up in wheels and you’re looking for targets, which increases your vulnerability. You’re constantly gonna have to have suppression of enemy air defense flights that are flying. You’re constantly gonna have to have dedicated defensive F-22s flying up there to shoot down any fast movers that potentially get launched to go after your bombing crews. It becomes a real military operation. And if you want to sustain that for an extended period of time where you’re really just hammering these targets, that’s really, really hard.

And it requires a ton of analytic capability, not just in machines. I don’t want to say you would need a lot of Palantir. It takes staff officers that are actually sitting down and doing the no-shit work of taking a high-value target list and turning that into a high-payoff target list, which is staff work. I don’t care what Palantir says or any of these other companies — the work of going from identifying a target to actually understanding if this thing is important to the enemy or not is where the human is still reliant. That’s where all the human decision-making comes. And we’re not doing that. We’re just hitting whatever’s on this list.

Bryan Clark: And the other thing I think you have to do, Jordan, is you have to defend the shipping. Because this is gonna take a while. So in the meantime you’re gonna have to start defending shipping and escorting it through the strait. If you want to take away that leverage and take away Iran’s ability to claim they can control the strait, you’ve got to start doing this strike campaign, basically attacking everything that emerges along the coast that’s gonna be threatening shipping. And then you’ve got to defend the shipping that’s coming through there, because some things will leak through.

The whole idea of “as soon as they pop out, I’m gonna shoot them” means they’re gonna pop out and some things are gonna make it off the rail before you’re able to kill them. And that thing that makes it off the rail, you’re gonna have to shoot down from a ship that’s defending the shipping. So there’s gonna have to be ships in there putting themselves at risk. But if you do that, then I think you do get to a point where you take away that leverage Iran has and you bring them to some kind of negotiated settlement, because they don’t see a path forward.

Tony Stark: And that is a massive trust exercise that I don’t think a lot of companies are willing to believe in us right now to do.

Bryan Clark: We gotta demonstrate it.

Tony Stark: And two, the longer this goes on, the more the insurance markets, the rest of the world starts to say, hey, are there other routes we can take? Are there other things worth investing in? Because what if this does pop off again in two years? We can do this once. We can’t do it every other year.

Nuclear Cargo Ships and the Return of Piracy

Bryan Clark: Or the SMRs, small modular reactors, have had this renaissance because of this. Because you talk to the shipping companies, which I was doing this week, and they make the point that yeah, these small modular reactors are expensive. Obviously you need different people to operate them. It’s hard to just put these onto cargo ships. But if you did it and they could drive 30 knots and they can go around the Cape, then they can pay for themselves. So if you can cut the travel time for a lot of these routes down by 50%, that’s money. If I’m making more trips, I’m moving more cargo, I’m making more money, I can pay for that more expensive ship and the more expensive crew to run it. So we’re really driving that business model for nuclear-powered cargo ships in a way that I always had written off as a pipe dream, but now it’s actually making fiscal sense for these companies.

Jordan Schneider: What’s the nausea factor here, Bryan? Are faster ships less nauseous or more nauseous?

Bryan Clark: I think it depends on the design, but it’s usually more nauseous.

Jordan Schneider: So aside from having the nuclear engineers, you also need to compensate them more and buy them more Dramamine. Let’s put that in your model.

Bryan Clark: That’s right. You gotta factor that into the business.

Justin McIntosh: That’d be interesting though. If we went all these years with the canals being super important, we went away from the capesize ships and we went to the slightly smaller transportation —

Bryan Clark: The Panamax and the Aframax.

Justin McIntosh: — to now it’s gonna be all capesize and super capesize.

Bryan Clark: It’s all Aframax, it’s all Polarmax, exactly.

Jordan Schneider: But here’s the thing. You’re just making some real incentives for a South African insurgency to start. You can make so much if you hold up that strait. I mean, you don’t even have to be an insurgent. You can just be a country and say we’re gonna do a toll.

Bryan Clark: So we did a war game series for Office of Net Assessment like ten years ago on this, because the idea was, what happens if China continues its investments in Africa and they start to put in reconnaissance strike complexes in Africa? So long-range ballistic missiles and all that stuff. And this was back when it was expensive to do that. And basically the red team established a choke point at the Cape of Good Hope. So you couldn’t get through there, because it’s relatively narrow if you think about the useful part of it — of course the weather gets so terrible down towards Antarctica, you gotta stay relatively close to Africa. And they were able to use ballistic missiles to close that off. And now it’d be even easier to do it because of all the cheaper ballistic missiles. And it wasn’t even like South Africa doing it. It was from Mozambique and from Madagascar they were launching some of these weapons.

Tony Stark: Somehow, piracy returned.

Justin McIntosh: It always goes back to being a pirate. That’s what every man actually wants to do, is go out and raid the coast and take home his treasure.

Bryan Clark: Except they’re doing it from shore using satellite feeds on AIS data and ballistic missiles.

Jordan Schneider: What a killjoy.

Bryan Clark: But you can still wear a pirate hat and all that if you want to.

Jordan Schneider: As you’re driving your submersible sub from a bunker.

Justin McIntosh: Do I get a parrot?

Bryan Clark: That’s right. On your Xbox controller.

Justin McIntosh: Do I have to have an eye patch? That’s the other question.

Bryan Clark: I think eye patch is optional. Depends on —

Jordan Schneider: You can definitely charge more for your toll if on your Zoom negotiations with the Americans you’re wearing one. I think they’ll take you more seriously.

Bryan Clark: They’ll pay you in crypto, obviously.

Jordan Schneider: Much easier to write you off if you have two eyes.

Tony Stark: He’s not a real pirate. He’s new at this.

Bryan Clark: If you’re willing to lose an eye for the cause though, you’re definitely committed.

Justin McIntosh: I think the best case scenario is a combination of us demonstrating that we won’t allow it to continue if they continue to harass ships, but signaling that we will allow them to gain some financial gain. And to Bryan’s point, we’ve just proven that we’re unwilling to do that, to put ourselves in the position to be at risk.

Bryan Clark: You gotta go one way or the other. The way we’re going now is just a way to burn more readiness, spend more money and not get the benefit. So either you have to go in and really go hard to deny them the ability to close off the strait, or you’ve got to sort of accept it and let them charge their extortion tolls and build their new mafia-type business model in the Strait of Hormuz.

The AI Kill Switch Bill

Jordan Schneider: All right, where to next?

Tony Stark: Can we talk about that stupid bill in Congress, the kill switch for AI? So there’s a bipartisan bill in Congress. It probably won’t go anywhere. One, because it empowers DHS, which I don’t think anyone is interested in doing right now.

Justin McIntosh: When I think AI governance, I think DHS.

Tony Stark: I think it would probably go to CISA, and there’s probably an argument to say that —

Bryan Clark: There is no CISA though. We don’t actually have a confirmed CISA director, I don’t think.

Tony Stark: Right. So as I understand it, this bill basically enables DHS to have a kill switch against AI that they think is threatening for some reason — i.e. some terminator bullshit. Which is a gross misunderstanding of the models and also a gross misunderstanding of the sorts of power governments should have. This is a classic congressional case of, as they did with SOPA and everything in the 2010s, overreach because they don’t understand the technology. It sounds cool from a regulatory standpoint, but first of all, any model that needs to be stopped with government authority probably can’t be stopped. Two, whether that’s feasible is a different story. And three, do we really want to give the government power to say which models are a threat or not? So let’s think about this for about five seconds and how bad it is. We saw it with Anthropic already that the government should not be doing this.

Justin McIntosh: Or continuing to see it with Anthropic. Let’s be clear on that.

Tony Stark: Yes, sorry — continuing. Schrödinger’s Anthropic.

Jordan Schneider: All I’m gonna say to this, Tony, is this may not be the right answer, but I don’t think we can just trust the companies to do it themselves. OpenAI has every incentive in the world not to have their models break out of harnesses and start hacking the host organizations of their pilot cyber test projects. And they couldn’t. So it may not be this bill, it may not be DHS, but there is so much financial incentive, there’s so much pushing these companies to release models faster, make more money, deliver the most capability, that I don’t think you can just trust their cost-benefit analysis to not put scary stuff out.

Tony Stark: Completely agree. My second part of this is that the cyber agencies probably need to be pulled out of DHS and elevated to a cabinet-level department. And with that, that needs to be the regulatory agency for all non-DOD AI models, because it’s going to flourish through so much of the economy that it probably does need a specific sort of regulatory posture that is not Commerce having its interests, versus DHS having its interests.

Assume the Network Is Compromised

Bryan Clark: Obviously there’s a lot of concerns with this. Looking at it from the military perspective, this suggests that pretty much unless your network is somehow completely firewalled from the rest of the telecommunications infrastructure, it’s gonna be vulnerable. A lot of the DOD’s secure networks actually ride on the telecommunication network and it just uses encryption to avoid being intruded upon by the normies. That’s not going to be true anymore. All those networks are gonna be vulnerable. I don’t think there’s any way you’re gonna protect them. SIPRNet, for example, is on the normal communication backbone.

So can we do — do we need to think about basically all these networks now being pretty much open at some level? And do we need to think about how do we control information in an environment like that? And do we need to rethink classification even? Because is it worth classifying stuff if it’s gonna be so easily accessed by nefarious actors or adversaries?

Justin McIntosh: I’ll be honest, working with intelligence agencies outside of the DOD during my career, they basically already were like, well, clearly that system’s compromised. They just went with that assumption. They’re like, we’re not gonna put any of our stuff on that system because the potential is there. Which means it’s the Batman logic against Superman, right? If there’s a one percent chance that he could destroy the world, then I have to take that as an absolute possibility. Exact same thing here. If there’s a one percent chance that the whole system is vulnerable, then I can’t put any of my stuff on it. So they built their own infrastructure, their own architectures, which I’m sure are corrupt in their own ways.

Turning back to Jordan’s point — yes, you can’t rely on a commercial entity to make these value calls. I think the problem is that we’re both bringing in 20th-century governmental models and even thought processes about the economy into the equation, and the economic incentives are so high, and we are so beholden to those in both the government and regulatory way, that people aren’t willing to make those really hard decisions. Like should there be a moratorium on AI? I don’t know if there should or shouldn’t. But if ChatGPT or their new model really did break out of containment and hack Hugging Face after Hugging Face was using whatever Chinese open source model as part of its defense, and was able to steal things out of it undetected until it finished doing all that — yeah, that brings up a lot of questions about where we are and what we’re running.

I don’t know what the right answer is, but I do know that I don’t think sticking it inside of DHS, which is already a lightning rod within the United States for all sorts of reasons, is the right answer.

Tony Stark: Final thought on this: there’s a bunch of regulations around AI that people tend to ignore because they want to go after the models. Until you have proper data privacy rights in this country, everything else — there’s a bunch of underlying things that I think drive more of the issues here. I’d also think there probably needs to be better regulations on AI marketing. Because it just feels extremely unethical to me that part of your marketing ploy can be, look at how evil our model turned out to be by accident, please give us more money to continue to build it.

Jordan Schneider: I don’t know, Tony. Forest for the trees here — what we’re worried about is this stuff getting out of control. Okay, data privacy. The world has been fine with no data privacy laws for the past twenty-five years. I think we can do another five. Marketing as well.

Tony Stark: But there’s a boy-who-cried-wolf factor, right? If you keep saying this and then people are like, they’ve said that five times before, that the model escaped containment or whatever — and then when it really does —

Jordan Schneider: I don’t know, Tony. Talk to me in two weeks. We’re here now. We’re living in the shit. Maybe it’ll take another three months where you have a Chinese model that’s open weight that’s Fable-y, or China gets something as good as Fable internally and can start hacking the world, or the Russians get something as good as Fable and can start hacking the world.

But I just think we’re much closer to the boy crying wolf happening than gives us the luxury to play around with cute data privacy stuff. Speaking of Congress not doing anything, the child stuff is the only legislation that might happen, and that’s been ten, fifteen years in the making on the social media side. I’m a little nervous.

Will Beijing Stay Open?

Tony Stark: I am curious here though. How do you think the PRC is gonna respond to this sort of thing? Because I know the CCP has a much different view of AI than our government does.

Jordan Schneider: I’ve given this take on like four other ChinaTalk podcasts over the past week. Open source is cute and good for the David Sacks and the JD Vances of the world until you’re staring down radical proliferation of cyber capabilities. And that only happened in the US in April of 2026 with Mythos. And that hasn’t happened quite yet in China, because the models aren’t good enough to really be able to do what Mythos and now what OpenAI have. So we’ll see. Maybe they’re talking a big game when it comes to open sourcing everything — Xi just gave this big speech — but that can change in a dramatic and surprising way, just like this White House changed their tune in a dramatic and surprising way.

Justin McIntosh: I would just say, do we really think they would be comfortable with a Tibetan separatist group all of a sudden having access to a Chinese open weight model that has Fable or Mythos capabilities and being able to turn it against the Great Firewall? I tend to agree with you, Jordan. The lockdown cometh for the PRC. They’re not gonna allow that to be turned internally. So I think that’s just a matter of time.

Tony Stark: The other half of this is we’re the government that couldn’t regulate TikTok when we literally wrote it into law. Put aside my arguments about privacy rights, which is just my stump speech — on being able to regulate anything that comes out of the box. Or is it just that there’s enough money that says actually what came out of the box is fine, it’s our friend now?

Jordan Schneider: If it gets obvious enough — there’s all this noise about cyber extortion, you know, ‘22, ‘23, ‘24. It’s kind of died down over the past year and a half. We’ve had a pretty strong defender advantage. But even then, the cyber extortion stuff was two hospital systems, two and a half hospital systems in America. Once it’s every single one all at the same time, I think that becomes a much more politically salient thing. Because it could happen really fast.

And the issue with TikTok was that it was all these latent worries and it was kind of a DC elite consensus. It wasn’t like the whole American public was clamoring for this to be shut off. It wasn’t real, it wasn’t affecting people’s daily lives. And if they time this wrong, or the technology goes in a weird way very quickly, this could be the most lucrative time for any criminal to ever do anything online. That’d be a really weird moment, and you’d have to expect there’d be some sort of response which has a little more teeth than what we saw from TikTok. Bryan?

Bryan Clark: I think the challenge ends up being, do we end up in a world where it’s just a free-for-all between our AI-enabled cyber defenses and our AI-enabled adversaries, and they’re going at it and we really can’t trust what we have on our computer networks, and you just gotta hope for the best? I don’t see how you avoid that through regulation or some kind of deterrence. It just seems like that’s sort of the inevitable outcome if you drive this out ten years. I’m sorry — I just haven’t seen anything that describes a more optimistic vision of where we end up.

Justin McIntosh: I think that’s Amodei’s entire argument: this is terrible. I do take Tony’s point — this is terrible, I need more money. Or it’s Sam Altman, this is terrible, I think the government should take a five percent stake of OpenAI. But in the end, that’s always my fear, that the economic incentives are going to outweigh the safety considerations until they don’t. And at that point it’s gonna be too late.

Bryan Clark: Unless, Jordan, to your earlier point — something that’s been addressed in previous podcasts — unless you do like we do with nuclear weapons and you regulate the raw material you use to create these capabilities. So you’d have to control compute.

Justin McIntosh: But we’re unwilling to do that. We have David Sacks arguing that Nvidia should be able to sell whatever it wants to whoever it wants. We have Jensen Huang saying — and honestly, his argument seems more prescient now, because he’s saying they’re catching up anyways. Yes, did they catch up with Blackwells and distillation and whatever, all that aside. But we’re not enforcing even the ITAR restrictions and the export control regime that we currently have, because there’s a lot of money to be made.

Is There Enough Upside?

Jordan Schneider: It’ll be interesting though. Tony brought up the TikTok thing. There was a strong political economy on the other side of, you don’t want to be the politician to shut this down because a lot of people use it and it’s really popular.

Bryan Clark: Businesses depend on it.

Jordan Schneider: It’ll be interesting with AI. Is there enough societal benefit that lots of people are feeling positively about it — so that whatever ripples you get from cyber attacks is kind of washed out by, my God, my kids’ school experience is so much better, or my business is so much more productive, I can work fewer hours at my job? You’re gonna need to have something on the other side. So the answer is not just AI data center moratoriums, let’s shut this all down. What have these assholes done for me? All of a sudden I can’t even access my bank account.

Justin McIntosh: I saw some numbers come out, I think it was in The Argument the other day, that basically showed that since the launch two years ago of one of the ChatGPT models, the number of new businesses registered in the United States — LLCs and sole proprietorships and things like that — has increased by like five or 10%. And that’s just people who are now able to navigate that system and start a business strictly because they have something that can actually give them the answers as they’re working through it.

So again, 100%. I’m not saying the economic benefits are accruing just to millionaires and billionaires or just to politicians who are willing to take it. I’m saying the economic benefits globally, even down to individual people, get to the point where they outweigh the security concerns. The issue becomes, if they’re right and it’s a terminator, what do we do?

Bryan Clark: There’s a difference between the model that’s able to support your business development and the model that’s able to do cracking of the most capable crypto. And just like with nuclear weapons, even though it’s not a great example — the underlying technologies yield a lot of economic benefits. How do you control the access to the parts that are maybe the most useful for weapons and not that useful for the economic side of it?

That gets to Jordan’s point that there are maybe thoughtful regulations that try to capture those nuances, to allow for economic development while not fostering the kind of free-for-all that we have now, where everybody’s just racing to deploy the most capable models and the most capable compute as fast as possible.

Zelensky and the Gamification of Strike Packages

Justin McIntosh: Let’s be honest, democracy worked in Ukraine, I think, this week. That’s also an awesome thing to at least close on. A burgeoning democracy who’s actually answering the cause of the civilian population and making changes to their government. That’s the least Soviet thing that possibly could have happened in Ukraine, where they had protests and they made changes.

Jordan Schneider: But do we think Zelensky is losing it though? What was he thinking?

Justin McIntosh: I talked to Rob a little bit about this. There is a little bit where it’s slightly overstated, the capabilities that were lost by removing the defense minister. There is real consternation within the military, at even some of the tactical levels, over the way that the gamification of strike packages has occurred. What they’ve basically done, to conceptualize it, is they’ve taken a lot of strategic thought and they’ve pressed it down into tactical action. And that’s it. It’s individual commanders deciding to go after targets based off of a value structure that they put into this gamified system. This is worth 5,000 points, this is worth 10,000 points, this is worth five points.

What that is not directly tied to — and where I think there was some concern within middle and upper parts of the Ukrainian military, at least from what I’ve been able to gather from people I’ve talked to — is that none of those strikes are necessarily tied to achieving some type of end state or goal that the Ukrainian military is driving towards. It’s just fully baked on imposed cost. Maybe that’s the model. Maybe the model is just imposed cost. And maybe he’s right that that is the case.

But I think what we were actually watching is not Zelensky making a move by himself. I think we were watching him actually start to weigh in on the side of some of the older, more established military guys that are saying, hey, we need to start defining what we’re actually going after. And we need some changes that Fedorov just wasn’t willing to make from his position. I think that’s what we’re seeing.

Now, again, take that with a grain of salt. That’s an outside perspective, and that’s obviously a guy who’s over there that’s just watching and me just talking to people. Don’t take that as the truth. But that is at least the things I was hearing as I talked to Rob and other people this week and last week on why this decision would have gotten made, and also what would have been the catalyst for it. Because again, for all intents and purposes, they’re doing much better than they were.