OpenAI announced that its internal artificial intelligence model produced a Lean-verified proof solving the Navier–Stokes existence and smoothness problem, a legendary mathematical challenge dating back nearly a century. According to OpenAI, the system demonstrated that the dynamics of the Navier-Stokes equations for fluid motion can develop a singularity in finite time. The Clay Mathematics Institute established the Navier–Stokes problem as one of seven Millennium Prize Problems in 2000, offering a $1 million bounty for its resolution. The equations themselves trace back to nineteenth-century work by Claude-Louis Navier and George Gabriel Stokes.
OpenAI Announces Breakthrough on Navier-Stokes Millennium Prize Problem
OpenAI reported that the resolution was achieved in approximately 88 hours, arriving on Saturday, September 5, after the first agents were launched. The company utilized an internal model significantly more capable than GPT-6 Astra, operating on the order of 10,000 concurrent agents subdivided into communicating groups with access to tools including code execution and a cached version of the internet. The resulting analytical proof and Lean formalization establish that an initially smooth fluid at rest can develop a singularity with finite energy.
Controversy and Allegations of Data Usage
The announcement was swiftly met with controversy following allegations from New York University mathematics professor Tristan Buckmaster. According to CNBC, Buckmaster stated that he had been working on related mathematics problems in a personal collaboration with Levent Alpöge, a researcher at rival firm Anthropic. Buckmaster claimed that information regarding their progress was passed to nypost.com, alleging that the tech giant’s route mirrored their own work and raising questions about whether OpenAI models had been trained on or had access to their sessions in OpenAI’s Codex.
Buckmaster also alleged that OpenAI scientist Sébastien Bubeck made threatening remarks during communications regarding joint announcements, though Bubeck later posted on X that his response was a poor choice of words
intended only as a warning about unfounded claims. Decrypt noted that OpenAI, Bubeck, and CEO Sam Altman denied Buckmaster’s characterization of events. While OpenAI insisted no specific user data was accessed to solve the problem, the company stated it cannot rule out that de-identified data derived from their usage of our products helped improve our models.
Historical Context of Millennium Prize Problems
The Millennium Prize Problems were conceived by the Clay Mathematics Institute to highlight deep, unsolved mathematical frontiers and recognize historical achievements. In the 26 years since the list was compiled, only one problem has been officially solved: the Poincaré conjecture, which was cracked between 2002 and 2003 by Russian mathematician Grigori Perelman, who famously declined the $1 million prize in 2010. Another high-profile attempt occurred in 2018, when mathematician Sir Michael Atiyah claimed a solution to the Riemann hypothesis, though it was later rejected due to fatal logical flaws.
Independent verification of OpenAI’s 100-page Navier-Stokes proof has not yet been conducted outside the company. Meanwhile, Buckmaster and Alpöge’s own advanced result remains unpublished pending a final Lean check, unfolding alongside concurrent AI-math announcements across the tech sector, including a computer-checked proof of Fermat’s Last Theorem attributed to Anthropic’s Claude.
Keep reading