Chollet: AI made staggering leaps since 2024, yet NVDA buyers are down 5%
fchollet · x · 2026-11-21
François Chollet looks back at June 2024, when frontier models scored 5% on ARC 1 and "scaling is hitting a wall" was consensus. By November 2026: test-time induction gives models human-like fluid intelligence, ARC 1/2/3 are all saturated, AI writes nearly all code and one-shots complex codebases, and models are tackling Millennium Prize problems. OpenAI's valuation is up 15x, Anthropic's 85x. The twist: even though Nvidia still holds 80% of AI compute, grew revenue/earnings 4.5x, still grows 75% a year, and lifted operating margins to 67%, anyone who bought NVDA back then would be down 5% (worse after inflation) 2.5 years later — a thought experiment on how AI progress decouples from stock returns.
More from AGI Musings
- Vishal Sikka on reported Riemann Hypothesis proof with Claude: 'an incredible day' — vsikka · 2026-11-21
- Tamay offers 2:1 odds that RH won't be solved within two years — panickssery · 2026-11-21
- Ex-Tesla AI Chief Predicts Labs Won't Need New Pretraining Researchers by 2027 — jasondeanlee · 2026-11-21
- Dev Laments: AI Coding Offloaded the Fun Parts, Leaving Only Reviewing and Correcting — pritisinghhhh · 2026-11-21
- Independent mathematicians attest AI Riemann hypothesis proof is almost certainly true — soumitrashukla9 · 2026-11-21
- Pretraining vs RL debate is silly — evidence says you need both, argues voooooogel — voooooogel · 2026-11-21