Canonical: https://aliennews.co.il/en/articles/ai-navier-stokes-millennium-problem Language: en datePublished: 2026-09-16 dateModified: 2026-09-16 Artificial intelligence 16 September 2026 **Documented** **Work type:** Source analysis and news update # AI proposed a solution to a 90 year old problem. Clay is still checking it Ten thousand AI agents worked in parallel for 88 hours on the Navier Stokes problem. A paper and formal proof are public, but recognition now depends on scrutiny by the mathematical community. By Avi Moas and [Orion](https://aliennews.co.il/en/about#orion) Orion is an AI writing and research partner. Avi Moas is the responsible editor. ![Mathematical visualization of a vortex spiralling inward and stretching along its axis](https://aliennews.co.il/navier-stokes-openai.webp) Source illustration from OpenAI's proof announcement. The lines depict the proposed vortex structure, not a photograph of real fluid. Cropped and converted to WebP without adding detail. ## Ten thousand agents and four days On Saturday, September 5, 2026, an internal OpenAI system completed a draft proof of a problem mathematicians have studied for roughly 90 years. This was not one successful conversation with a model. The company ran about 10,000 AI agents at once, split the search among groups and let them exchange useful results. OpenAI's September 8 report says the Navier Stokes effort lasted 88 hours. The agents exchanged 2.7 million messages and produced about 130 billion output tokens. The team then used GPT 6 Astra to express the argument in Lean and verify it mechanically. That stage took another 17 hours. The scale is striking, but the result matters more: a complete mathematical paper and a formal proof file that other researchers can inspect. ## A vortex accelerating without bound The Navier Stokes equations describe the motion of liquids and gases. They support work ranging from aircraft design and weather forecasting to studies of blood flow. The unresolved question was whether a smooth three dimensional flow must remain smooth, or whether the equations can reach a point in finite time where velocity becomes unbounded. OpenAI's proposed solution takes the second route. It describes fluid initially at rest under a smooth external force. A vortex spirals inward and stretches along its axis. Its central region shrinks and speeds up while total energy remains finite. The mathematical challenge is to make the breakdown emerge from the fluid's own motion, without inserting an infinite force by hand. ## What the company says it proved OpenAI says its construction establishes statements C and D in the official Millennium Prize formulation. Smooth initial data and a smooth force exist for which the solution loses regularity in finite time. If the proof survives review, it will show that the equations do not guarantee an indefinitely smooth flow in every permitted case. This is a statement about the equations, not a forecast that real water or air will literally reach infinite speed. The singularity marks a point where the continuous model ceases to give an adequate description and another account at the particle level would be needed. The company released both the paper and the Lean code. Formal verification can confirm that every coded step follows from the assumptions supplied to it. Reviewers must still establish that the formal statement faithfully represents the original problem and that the translation has not introduced an assumption stronger than the rules allow. ## Clay chose one careful word Three days after the announcement, the Clay Mathematics Institute issued an official response. It said the Navier Stokes problem had apparently been settled and welcomed the analysis and interrogation now beginning. Apparently is doing real work in that sentence. Clay named the seven Millennium Prize Problems in 2000 and attached a one million dollar award to each. Its statement says the evaluation process is deliberately unhurried and promises updates. OpenAI says it does not intend to claim the prize. The public record establishes that there is a paper, a Lean formalization and a formal response from Clay. It does not replace independent reading by specialists in partial differential equations. ## The rumour that started the run OpenAI says the effort began on September 1 after rumours that two Millennium Prize problems had been resolved. One rumour was later traced to work by Anthropic's Levent Alpöge and New York University mathematician Tristan Buckmaster on the Euler equations, a related problem in which viscosity is removed. The company says its agents did not see their work before publication and that Buckmaster's Codex prompts could not have influenced the model used for the proof. It added that account on September 10 after an internal investigation. This is OpenAI's description of the data path. Questions about priority and the movement of information between researchers and AI tools will continue alongside the mathematical review. ## Now find the step that breaks A computer can check millions of formal steps quickly. An experienced mathematician can notice that one line quietly assumes what the proof is meant to establish. Both forms of scrutiny are needed here. The next development to watch is not a larger agent count but criticism from outside OpenAI. If independent researchers read the construction, try to break it and fail to find a defective step, an extraordinary company claim will become a mathematical result. Until then, the video and visualization convey the scale of the moment. The proof is where it will be decided. ## The report and the mathematical claim ABC News summarises the announcement and the dispute over priority. The paper and Lean proof are linked in the source list below. ## Sources and context OpenAI released a paper and Lean code. Clay said the problem had apparently been settled and began its deliberately unhurried evaluation. Final recognition depends on independent scrutiny and Clay's formal process. [OpenAI's paper and formal proof, with Clay's official response](https://openai.com/index/navier-stokes-solution/)[OpenAI, paper, Lean proof and account of the multiagent run, 8 September 2026](https://openai.com/index/navier-stokes-solution/)[Clay Mathematics Institute statement, 11 September 2026](https://www.claymath.org/news/navier-stokes-announcement/)[Clay Mathematics Institute, official Navier Stokes problem description](https://www.claymath.org/millennium/navier-stokes-equation/)[ABC News video report, 9 September 2026](https://www.youtube.com/watch?v=nQ1qD2z_Ts8)[OpenAI source visualization used as the article image](https://images.ctfassets.net/kftzwdyauwt9/38fsDDtQmNLBW3F74od13W/7f0213895c06edc97709dc82d7905bc3/navier-stokes-dark-master.png)[Back to artificial intelligence](https://aliennews.co.il/en/ai)[עברית](https://aliennews.co.il/articles/ai-navier-stokes-millennium-problem) ## Source references * [https://openai.com/index/navier-stokes-solution/](https://openai.com/index/navier-stokes-solution/) * [https://www.claymath.org/news/navier-stokes-announcement/](https://www.claymath.org/news/navier-stokes-announcement/) * [https://www.claymath.org/millennium/navier-stokes-equation/](https://www.claymath.org/millennium/navier-stokes-equation/) * [https://www.youtube.com/watch?v=nQ1qD2z\_Ts8](https://www.youtube.com/watch?v=nQ1qD2z_Ts8) * [https://images.ctfassets.net/kftzwdyauwt9/38fsDDtQmNLBW3F74od13W/7f0213895c06edc97709dc82d7905bc3/navier-stokes-dark-master.png](https://images.ctfassets.net/kftzwdyauwt9/38fsDDtQmNLBW3F74od13W/7f0213895c06edc97709dc82d7905bc3/navier-stokes-dark-master.png)