OpenAI says it solved a $1M math problem. Rivals object

OpenAI says one of its own unreleased AI models cracked part of the Navier-Stokes problem. It's a fluid-dynamics puzzle mathematicians have chased for over 80 years. A $1 million prize is attached. On Sept. 8, OpenAI said about 10,000 AI agents worked 88 hours to build the proof. Another model then spent 17 more hours writing it up clearly.

That's not the whole story. Two mathematicians, NYU's Tristan Buckmaster and Anthropic's Levent Alpöge, say they'd already made unpublished progress on the same problem. They say an OpenAI researcher learned about their work shortly before OpenAI's announcement. They also say OpenAI offered Buckmaster credit only if he left his collaborator's name out. Buckmaster refused. OpenAI denies trying to take credit that wasn't earned.

Keep reading — what to watch

Did an AI genuinely solve this on its own? Or did OpenAI's model get a head start from watching two humans nearly finish it first? Nobody outside OpenAI has verified the full proof yet. An independent mathematician still has to check it line by line. Until then, this is a claim from the company that benefits most from people believing it.