OpenAI announced that its internal system has found a finite-time singularity in the three-dimensional Navier-Stokes equations, potentially resolving one of the seven Millennium Prize Problems in mathematics. The result describes a fluid that begins from a smooth, quiescent state and then a vortex collapses inward. Its core continues to stretch and accelerate until the velocity becomes unbounded, while the system still maintains finite total energy.

This combination makes the solution striking. The equations must spontaneously generate a breakdown in the presence of viscosity, which normally smooths out violent changes in fluid motion. Imagine a column of rotating fluid being pulled tighter and longer as it spins. The central region keeps contracting, concentrating motion into an ever-smaller area. Eventually, the velocity grows explosively in finite time. OpenAI says the surrounding mathematics remains sufficiently controlled that the fluid's energy stays finite.

The difficulty lies in making this behavior emerge naturally from the equations. Researchers cannot simply introduce an infinite external force to create a singularity. Near the collapse, acceleration, pressure gradients, momentum transfer, and viscosity all grow sharply, and they cancel one another with enough precision to leave a smooth external force. This balance gives the result its physical character: the singularity comes from the fluid's own dynamics.

Such a proof of the Navier-Stokes equations has eluded generations of mathematicians. Jean Leray proved the existence of generalized solutions in 1934, but mathematicians have never determined whether smooth solutions must remain smooth.

OpenAI did not arrive at a lucky answer through a single model output. The company assembled a large network of agents, giving different teams different versions of the problem. The work began on September 1, after researchers heard rumors that a major mathematical problem had been solved. Teams pursued competing results, including possible collapse solutions. At its peak, the Navier-Stokes project involved about 10,000 concurrent agents, which could run code in isolated environments and read cached internet materials.

An earlier result helped direct the search toward fluid singularities. Nearly 100 agents spent about 50 hours solving the regularity problem for the force-free version of the Euler equations. Researchers fed that result into the Navier-Stokes project, and Codex helped consolidate useful ideas across agent teams. The agents reached their conclusion on September 5, about 88 hours after the project began. GPT-6 Astra then spent another 17 hours formalizing and checking the proof in Lean. The Navier-Stokes work involved 2.7 million messages and about 130 billion output tokens.

OpenAI did not claim the $1 million Millennium Prize. It released the proof and its Lean formalization as evidence of progress in mathematical capability. Mathematicians still need to examine the argument and determine whether it meets all the requirements of the official problem. OpenAI also disclosed independent work by Levent Alpöge and New York University mathematician Tristan Buckmaster on the Euler equations with an external force.

If the proof passes expert review, it would show that machine-led mathematical search can discover a mechanism that has evaded researchers for 90 years: a vortex that tightens, stretches, and accelerates until the equations can no longer keep its velocity finite. The release, however, comes as 25 Fields Medalists have just signed a joint warning that AI mathematical proofs could disrupt the traditional processes of verification, attribution, and knowledge transmission. How OpenAI's Navier-Stokes proof withstands scrutiny by the mathematical community will test both the tension between AI mathematical capability and academic norms.