OpenAI says it solved a $1M math problem, NYU mathematician says he did first

OpenAI said its internal AI model proved that the Navier-Stokes equations can blow up. The equations describe how fluids move. A blow-up is a point where the mathematics predicts infinite speed. The problem is one of seven Millennium Prize Problems worth $1 million each from the Clay Mathematics Institute. OpenAI said the result took 88 hours. OpenAI said roughly 10,000 coordinating AI agents worked on the result. The agents were copies of its models working in parallel. Lean checked the proof step by step. NYU mathematician Tristan Buckmaster said he and Levent Alpöge spent nearly a year using AI models to pursue fluid blow-up proofs. Buckmaster said they had a solution by August 22. Buckmaster said he told a mathematician at OpenAI about the project on September 3. He said he described the project as personal and unaffiliated with either company. Three days later, Sébastien Bubeck told Buckmaster that an internal model had produced a 100-page proof for forced Navier-Stokes. Buckmaster said the approach was the same narrow approach that almost nobody else was pursuing. Buckmaster said Bubeck offered to let OpenAI publish the day after his team. Buckmaster said Bubeck also offered to let him write the paper alone without Alpöge because Alpöge worked at a rival lab. Buckmaster said Bubeck asked, "Why would you ruin your career?" Buckmaster later published research papers with his findings. The papers credited Alpöge. OpenAI did not credit either researcher in its announcement. A quote tweet acknowledged that OpenAI could not rule out de-identified data from product usage helping improve its models. Bubeck, OpenAI and Altman denied Buckmaster's characterization. Bubeck shared texts proposing a coordinated release with Alpöge. Bubeck also offered OpenAI's prompts. Bubeck said he acted in good faith. Sam Altman said Bubeck acted with integrity and generosity throughout. Altman said OpenAI offered Buckmaster's team the option to publish first after realizing the team had solved only Euler rather than full Navier-Stokes. OpenAI said it never saw the pair's work before publication. OpenAI said it accessed no specific user data. OpenAI did not explicitly rule out that de-identified data from product usage helped improve its models. Fields Medalist Terence Tao called Buckmaster and Alpöge's underlying mathematics a remarkable achievement. Their most advanced result remains unpublished pending a final Lean check. Nobody outside OpenAI has independently reviewed the 100-page proof.