Did OpenAI Steal a Math Breakthrough?

A few weeks ago, I saw a headline that made me stop scrolling:

OpenAI had apparently solved the Navier-Stokes problem.

This is a big deal.

The Navier-Stokes equations describe how fluids move: water, air, blood, ocean currents, airflow around airplanes, and much more.

One of the great unsolved questions is whether these equations can ever produce a mathematical singularity, where a smooth flowing fluid suddenly develops an infinitely violent whirlpool in a finite amount of time.

It’s been an open problem for generations.

Then OpenAI apparently threw an absolutely ridiculous amount of compute at it.

According to OpenAI, its frontier model used thousands of agents, ran for roughly 88 hours, and consumed hundreds of billions of compute tokens while exploring enormous numbers of possible approaches.

Basically:

What happens if you throw an AI army at a problem humans have been unable to solve?

The answer, apparently, was controversial.

Things Got Weird