Tristan Buckmaster alleges OpenAI pressured him to drop Anthropic co-author

The dispute centers on the Navier-Stokes equations, one of the seven Clay Millennium Problems carrying a million-dollar prize. Buckmaster and Alpöge used AI models like Anthropic's Claude and OpenAI's Codex over several months, making breakthroughs by mid-August. They also uploaded all their drafts into OpenAI Codex sessions, which forms the crux of the data-use question.

OpenAI claims its own proof was produced by about 10,000 coordinated AI agents in 88 hours using an internal model it says is "significantly more capable than GPT-6 Astra." Sébastien Bubeck says the company began training its model on August 28 and shifted resources after hearing rumors of Anthropic's progress. Buckmaster says he was told the model did not look up user data but received no answer about training data.

The stakes involve academic trust in AI platforms and fair collaboration across rival firms. OpenAI's Bubeck claims the company's solution "differs fundamentally" from Buckmaster and Alpöge's approach, but Buckmaster says the mention of forced Navier-Stokes suggests the same unusual route. Whether both proofs are validated, and whether any training-data overlap emerges, will likely determine who gets credit.