Did AI crack the Navier–Stokes problem? And why the quantum industry should pay more attention.
This year OpenAI published a 166-page paper claiming its internal AI system solved a version of the Navier–Stokes Millennium Prize Problem. I spent the weekend reading it, and I keep coming back to it.
A bit about where I'm coming from. My background spans in Machine Learning and leading RAG/LangChain projects from Microsoft. Plus with my graduate study in Quantum Computing, where my thesis framed the coordination of AI agent teams as a QUBO problem solved with quantum annealing. So this type of solution sits right where my work lives.
Reality Check
Navier–Stokes are the equations behind almost everything that flows: air over a wing, weather, blood. For about 90 years, mathematicians couldn't answer a simple-sounding question: if water or air starts out moving calmly, can the equations ever predict that it suddenly goes out of control? The paper says yes. It describes a whirlpool that keeps tightening and spinning faster, until the equations say its speed becomes infinite. Real water can't do that, so it shows a point where the equations stop matching reality.
According to the paper, around 10,000 AI agents worked in parallel and reached the proof in roughly 88 hours. It was then written in Lean, a program that checks every step of the logic automatically. So instead of asking "can we trust the AI?", we can ask "did the proof pass the check?"
The proof only works when an outside push is applied to the fluid. The case with no push, which many experts see as the more important one, is still unsolved. Other mathematicians haven't formally reviewed it, and OpenAI says it won't claim the prize. There's also a credit debate with Tristan Buckmaster and Levent Alpöge, who published closely related work just before.
Intersection between Quantum and AI
First, fluid simulation is one of the most promising future uses of quantum computers. This result shows the equations can behave in extreme ways, and any quantum solver has to be built with that in mind.
Second, our field runs on hard proofs, from error correction to post-quantum security. AI that can write and check that kind of math could certainly speed up quantum research.
Third, this took thousands of agents and billions of tokens. That scale of computing is exactly where efficient quantum-classical methods can help.
Where WISER fits in
At WISER, our Solutions Launchpad program brings quantum research to real industry problems. One of the projects we've worked on is for solving fluid dynamics equations, the same family of problems behind Navier–Stokes.
Whether this exact proof survives review or not, the bar has moved. The next step is making sure quantum and AI grow together, with humans still checking the results.
Find out more about WISER at thewiser.org.
Comments