OpenAI Says It Solved a $1M Math Problem. A Rival Mathematician Says He Did It First
OpenAI announced that its internal AI system had found a computer-checked proof showing a blow-up scenario for the Navier-Stokes equations, a Millennium Prize Problem. The company said the result took 88 hours and about 10,000 coordinated agents verified in Lean. The claim was immediately disputed by NYU mathematician Tristan Buckmaster, who said he and Anthropic researcher Levent Alpöge had already developed a related proof by August 22 after months of using AI tools. Buckmaster says he contacted OpenAI, then was told OpenAI had a 100-page proof along a similar forced Navier-Stokes approach and was pressured over publication timing and authorship. OpenAI, Sébastien Bubeck, and Sam Altman deny wrongdoing, saying they acted in good faith and did not see the pair’s work before publication. Buckmaster and Alpöge’s strongest result is still unpublished pending final Lean verification, and no outside group has independently reviewed OpenAI’s proof.
