OpenAI AI System Produces First Finite-Time Blowup Proof for Navier–Stokes Equations

OpenAI announced on September 8 that an internal AI system had produced a formal proof showing that the three-dimensional incompressible Navier–Stokes equations can develop a singularity in finite time, resolving a core version of the Clay Millennium Prize problem. The proof, spanning 166 pages with a Lean formalization, was generated between September 1 and 5 by a swarm of roughly 10,000 concurrent AI agents that exchanged approximately 2.7 million messages and produced around 130 billion output tokens. The system was not OpenAI's publicly released GPT-6 Astra model but a more capable internal model still undergoing training, with Astra handling only the final Lean verification step in about 17 hours. Rather than directing agents to solve Navier–Stokes directly, OpenAI split the work across groups assigned to different formulations of the problem, with a related result on the unforced Euler equations solved first and then used to guide the main proof. The breakthrough comes roughly 90 years after mathematician Jean Leray first studied turbulence and regularity in fluid equations, and raises open questions about prize eligibility and the downstream impact on dependent mathematical theorems.