OpenAI says it found a blowup in the Navier-Stokes equations -- and a rushed rival result exposed how messy AI-assisted math has gotten
Roughly 10,000 concurrent agents ran for 88 hours to produce OpenAI's proof, formally verified in Lean, addressing a 90-year-old unsolved problem. Hours earlier, two other mathematicians published a related result after -- one says -- word of OpenAI's progress leaked and pressured them to publish early.
OpenAI announced on September 8, 2026 that an internal model, orchestrating roughly 10,000 concurrent AI agents over 88 hours, produced a proof addressing the three-dimensional Navier-Stokes existence-and-smoothness problem -- one of the Clay Mathematics Institute's seven Millennium Prize Problems, unsolved for about 90 years. OpenAI describes the result as identifying a "singularity" (a finite-time blowup configuration) in the equations; the proof has since been formalized and verified in the Lean proof assistant.
A priority dispute broke out within hours
Hours before OpenAI's announcement, NYU professor Tristan Buckmaster and Anthropic researcher Levent Alpöge published their own result -- on the related but distinct forced Euler equations, not Navier-Stokes itself. Buckmaster says word of OpenAI's progress had leaked to the team, pressuring them to publish before they were ready. He was equally unsparing about the AI-generated material involved along the way, calling one early draft something that "can only be described as AI slop. I am sorry for this."
What's actually settled, and what isn't
Per Quanta's reporting, OpenAI itself cedes priority on the 3D Euler result to Buckmaster and Alpöge while claiming the Navier-Stokes blowup result as its own -- the two teams solved genuinely different, if closely related, problems, not the same one twice. The capability milestone is real and independently checkable in Lean; the credit dispute around it, and a working mathematician publicly disowning an AI-drafted proof as slop rather than standing behind it, is a live preview of the authorship and quality-control friction AI-accelerated research is going to keep creating as it moves faster than the norms built around it.