chiprook
← AI
AISeptember 28, 2026, 10:46

OpenAI claims Navier-Stokes blow-up proof, mathematicians dispute it

On September 8, 2026, OpenAI said an internal model "significantly more capable than GPT-6 Astra" produced a proof and Lean formalization settling statements C and D of the Clay Navier-Stokes problem: smooth solutions can blow up. The run used about 10,000 concurrent agents over 88 hours, 2.7 million messages and ~130 billion output tokens (~$6.5M at Astra list prices). NYU professor Tristan Buckmaster says OpenAI started after learning of his work with Levent Alpöge and that he was pressed to drop a co-author.

OpenAI claims Navier-Stokes blow-up proof, mathematicians dispute it
#OpenAI#GPT-6Astra#Lean
Read next
AI

Mathematicians clash over AI ethics after OpenAI solves Navier-Stokes problem

AI

Anthropic: Claude wrote a Lean proof of Fermat's Last Theorem in 11 days

AI

Anthropic investor Lonsdale says AI firms stoke fear to sway policy

AI

OpenAI taps independent mathematicians to responsibly share math-related AI advances