OpenAI Navier-Stokes Proof: 10,000 AI Agents, 88 Hours, Credit Fight (2026)

OpenAI Navier-Stokes claim sparks credit fight
▲ OpenAI Navier-Stokes claim sparks credit fight

The Navier-Stokes equations are the math that describes how fluids such as water and air move. Whether their solutions can blow up, with speed shooting to infinity, is one of math's biggest open questions. On September 8, 2026, OpenAI said its AI answered it. Now the harder question is who deserves the credit.

What Is the Navier-Stokes Millennium Prize Problem?

It is one of 7 Millennium Prize Problems, each worth $1 million from the Clay Mathematics Institute, and only 1 has ever been solved. The core question is whether a smooth flow can develop a blow-up (a point where speed becomes infinite) in finite time. Real fluids never do that, so a yes would expose a flaw in the equations.




10,000 AI agents, 88 hours, one proof
▲ 10,000 AI agents, 88 hours, one proof

How OpenAI's Navier-Stokes Proof Came Together

OpenAI says an unreleased model more capable than GPT-6 Astra ran about 10,000 AI agents (autonomous AI programs) at the same time. They reached a solution 88 hours after launch and exchanged 2.7 million messages. Verifying the proof in Lean (software that checks math proofs step by step) took another 17 hours. The result assumes an outside force acting on the fluid, which the official problem allows, but experts disagree on whether it fully counts.




AI labs join the research race
▲ AI labs join the research race

The Navier-Stokes Credit Dispute

NYU mathematician Tristan Buckmaster and Anthropic researcher Levent Alpöge reached blow-up results on related equations on August 15. OpenAI began on September 1 after hearing rumors that rivals were close. Buckmaster says OpenAI researcher Sébastien Bubeck pushed during September 6 calls to remove Alpöge from authorship, which Bubeck denies. Both mathematicians also used OpenAI's Codex, and OpenAI says it cannot rule out that data from their usage helped improve its models.

What It Means for the US AI Race

The rivalry between US labs OpenAI and Anthropic now reaches into open math problems, and AI labs sell research tools while competing with the researchers who use them. Terence Tao warned that strip-mining open problems could destroy the ecosystem that trains future mathematicians. The Clay Institute still lists the problem as unsolved, and OpenAI says it will not claim the prize.

Key Takeaways

① 10,000-agent proof - OpenAI says its agents found a Navier-Stokes blow-up proof in 88 hours and verified it in Lean.

② Credit dispute - Mathematicians with related August results clash with OpenAI over authorship and data use.

③ No official ruling - The Clay Institute lists the problem as unsolved, and OpenAI won't claim the prize.

AI is solving hard problems faster than expected. The next open problem is whose name goes on the answer.

👉 OpenAI Model Solves 80-Year Geometry Conjecture - AI Math Milestone 2026 - also worth a read.


📌 Sources: OpenAI, Quanta Magazine, Fortune, Axios, TechCrunch, Science News, Scientific American (2026)

댓글

이 블로그의 인기 게시물

SK Hynix US ADR Listing 2026 — $10.5B SEC Filing to Close Micron Valuation Gap

$2.5B Nvidia Chip Smuggling: Thailand Route to Alibaba Exposed (2026)

Nvidia $3.2B Corning Bet: AI Data Center Optical Fiber Megadeal (2026)