Key Takeaways

  • AI research labs are targeting Millennium Prize problems, including the Navier-Stokes existence and smoothness equations, to test automated mathematical reasoning.
  • John Coogan points out that engineers and physicists already approximate Navier-Stokes equations numerically every day, meaning formal mathematical proofs will not alter real-world industrial output.
  • Jordi Hays argues that pure mathematical proofs are closer to fine art than engineering tools, serving as high-water marks for human knowledge.
  • Consumer attention and market sentiment track visual video generators like Sora rather than abstract mathematical benchmarks.

The Disagreement

AI labs love grand academic milestones. When researchers point machine intelligence toward the Navier-Stokes equations (one of the unsolved Millennium Prize math problems), the tech world splits into two camps: pure theory versus practical utility.

John Coogan argues that solving Navier-Stokes with an AI model is an impressive technical demo that changes nothing on the factory floor. “The interesting thing about Navier-Stokes is, does it matter, Tyler? You think it does? I'll debate you on this a little bit. It seems like it doesn't because the equations themselves are known. They're just not fully proven. Engineers and physicists already solve them numerically all the time for particular situations,” Coogan said. He added that physicists agree: “Proving these equations is unlikely to actually have any measurable impact... this won't actually move the discipline of engineering forward at all, but it is a cool demo.”

Jordi Hays defended the effort from an intellectual standpoint. He agreed that practical applications are distant, but argued that pure mathematics has intrinsic value: “I don't think you're going to feel it when you're on an airplane. You're not going to feel like less turbulence now because of this. But I think these are the most important open math problems. These are more close to a beautiful painting than some practical engineering breakthrough.” For Hays, the achievement is philosophical: “To me it just comes down to humans are able to use machines that humans built to further humanity's general knowledge.”

At the same time, Hays recognized the PR problem facing deep research benchmarks: “The thing is the average person is probably more impressed by Sora than by this.”

Who's Right (and When They're Wrong)

Coogan is right on enterprise utility. If you run a startup building computational fluid dynamics, aerodynamics simulations, or HVAC software, you already rely on numerical approximations. An existence proof from an AI theorem prover does not speed up your rendering pipeline or make your wind tunnel simulations cheaper next quarter. Building product around academic prestige benchmarks creates vanity software that customers will not pay to deploy.

Yet Coogan misses why frontier labs target these problems in the first place. AI labs do not tackle Millennium Prize problems to help Boeing reduce wing drag next month. They tackle them because theorem proving eliminates the hallucination problem inherent in natural language. Formal verification provides a ground-truth environment where a model either produces a valid mathematical proof or fails completely. If a system can navigate the logical search space of Navier-Stokes, the underlying reasoning architecture can transfer to chip design, compiler optimization, and autonomous software synthesis.

Hays is right that the public market does not judge AI progress by formal logic. A photorealistic video from Sora generates immediate global distribution, cultural panic, and investor frenzy. A rigorous mathematical proof published in a formal verification language barely ripples outside academic departments.

What to Do With This

Audit your product roadmap and strip out vanity benchmarks that sound impressive to researchers but do not solve a customer's immediate operational bottleneck. If you build software for technical industries, interview three lead engineers this week to identify which numerical approximations they already trust, then optimize for their calculation speed and compute cost instead of theoretical perfection.