Summary

Hotz argues that frontier AI labs have no secret techniques—training methods are well-known and reproducible, which is why companies like Anthropic push for regulatory capture. He contends that AI scaling is hitting diminishing returns, with exponentially more compute needed for linear improvements. Since current models like GPT 5.5 are already hard to stump, 'superhuman intelligence' becomes a meaningless concept for most domains. The future of AI isn't about making models bigger, but about efficiency, distribution, and taste—deciding what to optimize for rather than optimizing harder.

Key Insight

AI scaling has hit diminishing returns, and since most meaningful domains aren't optimization problems, the future belongs not to bigger models but to efficiency, broad distribution, and human taste.

Spicy Quotes (click to share)

• 8

  That's why Anthropic is so desperate for regulatory capture, AI has no moat.

• 6

  AI (and any form of search) has this property where you spend exponentially more money to get linear returns.

• 7

  The Internet has been fully mined, and it yielded 20T good tokens.

• 5

  500 GB gets you all of human knowledge in a simple to query archive.

• 7

  What does "superhuman intelligence" even mean at that point if humans can't detect it if it's superhuman?

• 8

  Contrary to the beliefs of the rationality cult, most things aren't optimization problems. The whole hard problem is determining what to optimize for.

• 7

  The era of scaling yields clearly better AI is over, now we enter an era of efficiency and taste.

• 4

  Taste is an arena where tons of people can play.

Tone

contrarian, optimistic, irreverent