AI solves Navier-Stokes — does it actually matter for the real world?

AI solves Navier-Stokes — does it actually matter for the real world?

AI labs are racing to solve the Millennium Prize math problems, with Navier-Stokes emerging as the latest target. Last year, both OpenAI and Google DeepMind achieved gold medals at the International Math Olympiad by scoring 35 out of 42, solving five of six questions despite disagreement over grading methodologies — OpenAI used external validation from former IMO gold medalists while Google applied its own internal rubric. The drama around attribution and methodology has become so prominent that observers now joke the hardest problem in advanced mathematics is determining authorship and who gets credit on the paper.

Yet solving Navier-Stokes mathematically may not actually move engineering forward. The equations are already known; engineers and physicists solve them numerically all the time for specific applications like airflow over wings, weather modeling, and water flow through pipes. Proving the equations formally is unlikely to produce measurable practical impact, a view shared by physicists who have weighed in on the topic. Better weather models and more efficient aircraft designs sound promising in theory, but proving abstract mathematical theorems doesn't translate directly into engineering breakthroughs.

The real value is signaling, not solving

These math victories function as benchmarks for AI progress rather than practical breakthroughs. They demonstrate raw capability acceleration in a way that resonates with researchers and technically literate audiences. But public perception lags far behind. Most people don't care about IMO gold or Millennium Prize proofs the way they do about generative tools like Sora or real-time 3D modeling from images — visceral, grounded applications where people can see immediate creative leverage.

The comparison is instructive. Sora generates video from text. Three-dimensional AI modeling tools let users walk around inside rendered spaces built from photos of things that don't yet exist. That feels like a breakthrough because it's tangible. An IMO proof, by contrast, reads as "cool calculator, bro" — another domain where computers outpace humans in ways that don't empower anyone's daily work.

The shift to biotech won't solve the reputation problem

Once AI labs finish with pure mathematics, the conversation will likely pivot to cancer treatment, an application messaged so heavily by AI companies that it's become the default next frontier. But curing cancer through AI faces a harder credibility problem. Pharmaceutical companies have been advancing cancer treatments for decades, yet they have the worst reputation of any business category. If an AI lab announces 20 percent progress on one cancer type, it reads as incremental science — the kind of thing happening in trials all the time — not as a breakthrough that shifts public sentiment.

GLP-1 drugs illustrate the problem. They've effectively addressed obesity and diabetes, meaningful interventions that improve health outcomes over decades. But because the benefit accrues gradually rather than as a dramatic rescue, they don't generate the cultural credit that breakthrough narratives demand. Incremental progress, however real, diffuses slowly and fails to move the needle on how people perceive technology broadly.

Math problems can be solved in a weekend with enough compute. Cancer requires years of testing, regulatory approval, and slow diffusion through healthcare systems. The labs will likely solve the math problems this year, possibly this week. The conversation will then move to biotech. The credibility gap, however, will remain.