Over Labor Day weekend, NYU mathematician Tristan Buckmaster found himself at the center of a dispute between OpenAI and Anthropic over progress on the Navier-Stokes equations, one of the Clay Mathematics Institute's Millennium Prize problems offering $1 million. Buckmaster and collaborator Levent Alpöge, a mathematician at Anthropic, had made significant advances in August. However, OpenAI announced Tuesday it had completed a proof using AI agents, beating human mathematicians.

A firestorm erupted online, with suggestions OpenAI's system incorporated Buckmaster's recent work. OpenAI stated it was "categorically" impossible for its AI to have been influenced by Buckmaster's work from the past two months. Buckmaster said he cared more about contributing to mathematics than the prize money, calling himself a bystander in "petty games" between the rival AI companies.

OpenAI's Sébastien Bubeck offered Buckmaster first publication rights, lead authorship on an OpenAI paper, computing resources, and support for the prize money. The sticking point: Alpöge's affiliation with Anthropic. Buckmaster proposed the companies collaborate, but OpenAI resisted including Alpöge. The episode highlights distrust between competitive AI startups and the collaborative culture of mathematicians increasingly reliant on AI tools.