> ## Content Index > Fetch the complete content index at: https://www.theglobalvoice.com/llms.txt > Use this file to discover other available public pages before exploring further. # OpenAI Says AI Has Solved One of Mathematics’ Greatest Unsolved Problems - URL: https://www.theglobalvoice.com/openai-says-ai-has-solved-one-of-mathematics-greatest-unsolved-problems/ - Published: 2026-09-09T06:30:47.000Z - Updated: 2026-09-09T06:31:57.000Z - Description: OpenAI says 10,000 AI agents have solved the 90-year-old Navier-Stokes problem in just 88 hours, producing a formal mathematical proof that, if independently validated, could mark a turning point in AI-driven scientific discovery. - Author: TGV Desk - Tags: The Day So Far, Tech, Top Stories OpenAI says an experimental artificial intelligence system has solved one of mathematics' legendary Millennium Prize Problems, potentially marking one of the most consequential demonstrations yet of AI performing original scientific research. The problem is known as the Navier-Stokes existence and smoothness problem. For roughly 90 years, mathematicians have struggled with a deceptively simple question buried inside equations describing how fluids move: can perfectly smooth fluid motion eventually develop a mathematical singularity a point at which quantities such as velocity become unbounded? OpenAI says the answer is yes. Its newly published proof describes an initially smooth fluid that develops a singularity in finite time, effectively demonstrating that the Navier-Stokes equations can “blow up” under the conditions covered by the Millennium Prize formulation. The equations themselves are anything but obscure. Developed from work by Claude-Louis Navier and George Gabriel Stokes, Navier–Stokes equations underpin modern understanding of fluid motion. They are relevant to everything from aircraft design and weather forecasting to ocean currents and blood flowing through the human body. In 2000, the Clay Mathematics Institute selected the Navier–Stokes question as one of seven Millennium Prize Problems, offering $1 million for a valid solution. Only one of those seven, the Poincaré conjecture had previously been solved. What makes OpenAI's announcement extraordinary is not merely the claimed solution, but how it was produced. ![](https://storage.ghost.io/c/9b/c5/9bc51e60-58a6-4a20-a6fd-f93758107131/content/images/2026/09/GPT-Astra-6-and-Internal-Model-Comparision--1.png) Performance of GPT-6 Astra and Open Ai's Internal Model on a curated set of open math problems (Photo credit openai.com) The company says it used an unreleased internal model that is “significantly more capable” than GPT-6 Astra. Researchers initially directed AI agents towards all six remaining Millennium problems. After the system began making promising progress on Navier–Stokes, OpenAI concentrated its computing resources there. Eventually, as many as 10,000 AI agents were working simultaneously. According to OpenAI researchers, the system exchanged approximately 2.7 million messages, generated around 130 billion output tokens and completed the work in roughly 88 hours. The computational effort reportedly cost millions of dollars. The result is not merely an answer generated in ordinary language. OpenAI released a lengthy analytical proof alongside a formalisation written in Lean, a theorem-proving system that allows mathematical arguments to be checked mechanically. The company's public repository contains formal certificates corresponding to the claimed finite-time blowup results. But there is an important reason not to declare the 90-year-old problem definitively closed just yet. OpenAI is claiming a solution. The wider mathematical community must now examine the argument. Clay Mathematics Institute president Martin Bridson called the announcement a major moment for mathematics, but formal recognition of a Millennium Prize solution involves independent scrutiny. OpenAI says it does not intend to claim the $1 million prize. There is also controversy surrounding how the breakthrough emerged. Mathematician Tristan Buckmaster of New York University and Levent Alpöge, an Anthropic researcher, had independently been pursuing closely related fluid-dynamics research using AI systems from both Anthropic and OpenAI. Buckmaster has questioned whether information from their use of OpenAI's systems could have influenced OpenAI's internal model. OpenAI denies that its researchers or agents accessed their unpublished work while developing the proof, although the company says it cannot completely rule out the possibility that de-identified usage data contributed indirectly to model improvements. That dispute may eventually become almost as consequential as the mathematics. Because if OpenAI's proof survives scrutiny, the story will not simply be that AI helped a mathematician solve an extraordinarily difficult problem. It will be that thousands of AI agents, coordinated at enormous computational scale, produced a solution to a problem generations of human mathematicians could not solve. And that raises a much bigger question. If 10,000 AI agents can spend 88 hours attacking one of mathematics' most formidable problems, what happens when similar systems are turned towards thousands of unsolved problems in mathematics, physics, chemistry and biology simultaneously? The Navier–Stokes proof may ultimately be remembered not simply for the problem it solved, but for what it revealed about the kind of scientific work AI is beginning to do.