13 Comments

I’m just glad you’re fired up

Expand full comment

Thanks for highlighting. The market is completely freaking out over Deepseek

Expand full comment

They shall learn to make their own high-end chips... since the rest of the world is *democratically "influenced"* by the Empire (of freedom and democracy, lol).

Expand full comment

Fascinating read! DeepSeek’s efficiency gains are impressive, but as you point out, they’re built on a foundation of substantial prior compute access—something that’s increasingly constrained by export controls. It’s also striking how much of the AI race now happens behind the scenes, with the most advanced models never seeing public release. This raises big questions about how we measure progress and maintain resilience in an era where capabilities are both diffusing and becoming more opaque. Great analysis!

Expand full comment

Yes, democracies should maintain a leg up in the AI competition, so they can more effectively finance the genocide, and regime change operations.

Expand full comment

How does Deepseek making its code open source affect AI industry?

Expand full comment

Continued antagonism between the 2 countries must be good for your substack business ?

Expand full comment

!

Expand full comment

Only took a week for the markets to digest. Deepseek just released an image model along with Qwen. The spigots are on.

Expand full comment

The release during Trump's first week is a good spot, but how are we 100% sure it isn't coincidental?

Expand full comment

That doesn't fit the 'narrative.'

Expand full comment

Really interesting. Thanks for writing.

This all assumes that China will not be able to create chips that are equivalent to the West’s highest end chips.

How likely is that to remain true? If they can produce their own chips how does that change your story?

Expand full comment

The book chip wars mentions that the time frame for reinventing current cutting edge chips would at least be 10+ years. Which is then 10 years behind the running cutting edge.

Expand full comment