A mathematician says OpenAI raced his Navier-Stokes proof

A million-dollar math problem now comes with a credit fight.

What Buckmaster describes

Princeton mathematician Tristan Buckmaster says he and Anthropic researcher Levent Alpoge spent close to a year quietly working toward a blow-up proof for the forced Navier-Stokes equations, one of the seven Millennium Prize problems. By his account, OpenAI then learned about the work, ran an internal model down the same rarely used route, and came back with a roughly 100-page proof.

The offer he describes had two options: publish side by side, or write the paper himself while crediting OpenAI’s model. He also says OpenAI’s Sebastien Bubeck twice argued that Alpoge should be dropped as an author because he works at Anthropic.

The other side

Bubeck disputes the account and calls the allegations false and inflammatory. Terence Tao, commenting on the mathematics itself rather than the dispute, sees the approach as promising but far from settled.

Why it matters

The disagreement is about credit, not correctness, and that is what makes it new. Frontier models are now fast enough to compress a year of specialized human work into a sprint, which means priority disputes that used to play out between rival research groups can now play out between a mathematician and a lab with a large compute budget. The open question is what authorship and priority mean when a model finishes a race a human started, and the field has no settled norms for it yet.