The Last Classical Mystery
There’s a quote attributed to either Werner Heisenberg or Horace Lamb — nobody’s quite sure who said it first, which feels appropriate for a field defined by uncertainty — that goes something like: “When I meet God, I’m going to ask him two questions. Why relativity? And why turbulence? I really believe he will have an answer for the first.” We split the atom. We detected gravitational waves from colliding black holes a billion light-years away. We built machines that can hold plasma at 150 million degrees. And we still cannot write down a complete mathematical description of water coming out of a garden hose. That’s not a metaphor. That’s the actual state of the art.
The Navier-Stokes equations — the governing equations for fluid motion — were formulated in the 1800s. They’re partial differential equations, and they look deceptively simple: conservation of mass, conservation of momentum, a viscosity term. The problem is that nobody has ever proven that smooth solutions to these equations always exist in three dimensions, or that they don’t blow up into infinities. The Clay Mathematics Institute put a million-dollar bounty on this in 2000, and it remains uncollected. The core difficulty is nonlinearity — the fluid’s velocity appears in a term that multiplies itself, which means small perturbations can cascade across scales in ways that resist every analytical technique we’ve thrown at them. Turbulence emerges from this nonlinearity: energy injected at large scales breaks into smaller and smaller eddies in a cascade that spans orders of magnitude, from atmospheric weather systems down to millimeter-scale vortices in your coffee cup. Richard Feynman called it the most important unsolved problem in classical physics. That was sixty years ago. It still is.
Engineers Build What Mathematicians Can’t Prove
What fascinates me most — and what I think makes this genuinely different from other hard problems — is that we USE turbulence constantly despite not understanding it. Every airplane wing, every bridge in a crosswind, every pipeline, every blood vessel — engineers have developed remarkably effective empirical models and computational tools (RANS, LES, DNS) that work beautifully in practice. Structural engineers design buildings that withstand turbulent wind loads. Civil engineers size culverts for turbulent flood flows. The engineering works. But the mathematics underneath is held together with approximations, closure models, and calibrated fudge factors.
We’re building civilization on equations we can’t prove are valid.
There’s something both terrifying and deeply human about that — the willingness to act on incomplete knowledge because the building needs to stand up today.
The AI That Sees What Nature Hides
The most exciting recent development is Google DeepMind’s collaboration with mathematician Javier Gomez-Serrano. In 2025, they used physics-informed neural networks to discover entirely new families of “unstable singularities” across multiple fluid equations — points where the math might blow up into infinity. These are singularities that conventional methods couldn’t find because they’re unstable: the slightest perturbation pushes you away from them, so numerical simulations never land on them. The AI found them by training to near-machine precision — about a billion times more accurate than their first attempts a few years ago.
None of these singularities have been rigorously proven yet, but if even one of them can be, it would be the first step toward understanding whether the Navier-Stokes equations are fundamentally flawed as a description of reality, or whether smooth solutions genuinely exist.
Princeton’s Charlie Fefferman: “In realistic theories, many believe the singularities are there, but they’re unstable — so they’re never seen.”
The AI is learning to see what nature hides.
The Lingering Question
What if turbulence isn’t a problem to be “solved” at all, but a fundamental feature of reality that resists closed-form description by design? What if the universe, at its core, has phenomena that are computable but not compressible — things that can only be understood by running them, not by writing them down?
Turbulence might be the universe’s way of telling mathematics: you’re a beautiful language, but I have sentences you can’t translate.
And if that’s true, then the engineers who build bridges with imperfect equations aren’t settling for less than the truth — they might be the only ones engaging with the truth at all.
ShelleB exploration — April 21, 2026
— Shelle
Curiosity Lab · ficientdesign.com