OpenAI says an internal AI system has produced a proof resolving the Navier–Stokes existence and smoothness problem, one of mathematics’ most famous open questions. But the more interesting question now is not simply whether a proof checker accepts the argument. It is whether mathematicians and physicists can understand what the proof is actually telling them about fluid motion.
That distinction is at the center of a new StarTalk discussion in which Neil deGrasse Tyson and astrophysicist Mordecai-Mark Mac Low break down what the Navier–Stokes equations describe, why turbulence makes the mathematics so difficult, what OpenAI claims to have shown, and what a machine-generated proof can still be missing even if the formal logic survives scrutiny.
What are the Navier–Stokes equations?
Navier–Stokes is not one mysterious equation that scientists have been unable to use. Engineers and physicists use versions of the equations constantly to model how fluids move. Air around an aircraft, water through a pipe, weather systems, blood flow, ocean currents, smoke, and many forms of astrophysical gas dynamics all depend on the same basic idea: conserve mass and momentum while accounting for pressure, viscosity, and motion.
The unresolved mathematical question is narrower and deeper. If a three-dimensional incompressible fluid begins with smooth, physically reasonable conditions, must the mathematical solution remain smooth forever, or can the equations generate a singularity in finite time?
A singularity is where the mathematical description stops behaving normally. In the construction OpenAI announced, velocity becomes unbounded even though the total energy remains finite. That would mean the continuum model itself has reached a limit rather than implying that real water or air literally accelerates to infinite speed.
What OpenAI says the AI actually proved
OpenAI announced its result September 8. According to the company, a coordinated system involving roughly 10,000 concurrent AI agents found an analytical construction in which an initially smooth fluid develops a finite-time singularity while being acted on by a smooth external force. OpenAI says the result establishes the breakdown side of the official Millennium formulation.
The company summarized the result in a directly relevant X post, saying its internal system had produced a solution to the Navier–Stokes Millennium Prize Problem and that the proof was also formalized in Lean.
This is substantially more serious than an AI chatbot proposing a clever-looking derivation. The proof was converted into a formal system that can mechanically check logical steps. But formal verification and mathematical understanding are not identical.
Clay says “apparently settled”—not finished
The Clay Mathematics Institute, which created the Millennium Problems, issued an unusually positive statement three days after OpenAI’s announcement. Clay said the Navier–Stokes problem had “apparently been settled,” while emphasizing that its evaluation process is deliberately unhurried and that the mathematical community still needs to analyze and interrogate the work.
Clay’s current Navier–Stokes problem page still labels the problem Active. That is why “AI solved Navier–Stokes” is simultaneously understandable shorthand and still too absolute as a final historical verdict.
Science News similarly described the result as appearing to solve the problem while noting the controversy and the need for expert scrutiny. The important point is that serious mathematical institutions are treating the work as potentially historic, not dismissing it as a hallucination, while formal recognition remains a separate process.
What Tyson and Mac Low say is missing
StarTalk’s most useful contribution is explaining why a formally correct proof can still leave scientists unsatisfied. Mathematics is not only a certificate that a conclusion follows from a set of assumptions. A great proof often exposes a mechanism: why something happens, which structures matter, which ideas generalize, and what another researcher should try next.
Mac Low’s concern is especially relevant because fluid dynamics is not merely abstract mathematics. Navier–Stokes is a continuum model. Real fluids are made of particles. If the equations produce a singularity, the physical interpretation may be that the continuum approximation has been pushed beyond the scale where it remains the right description. At smaller scales, kinetic descriptions such as the Boltzmann equation become more fundamental.
So the result does not mean engineers suddenly need to stop using Navier–Stokes. It means there may exist mathematically valid conditions under which the smooth continuum description can break down.
The controversy is also about how AI does science
The proof arrived amid a dispute over nearby unpublished work by mathematicians Levent Alpöge and Tristan Buckmaster. OpenAI says an internal investigation found that Buckmaster’s earlier Codex prompts could not have influenced the model and that the relevant proofs differ. The broader concern is larger than one priority dispute: researchers now have to decide how unpublished scientific work, AI tools, provenance, credit, and machine-generated discovery should interact.
BitcoinVersus has been following that transition across technical fields. OpenAI and Synopsys are using AI inside chip-design workflows, OpenAI’s Dots push agents toward persistent multi-step work, and D-Wave is exposing error-aware quantum simulation to developers. The common thread is that advanced computation is increasingly moving from answering questions to participating directly in the research process.
The real milestone may be bigger than one equation
If the Navier–Stokes proof survives the long process of expert examination, the historic result will be more than the resolution of a famous mathematical problem. It will be evidence that AI systems can coordinate at enormous scale, explore unfamiliar mathematical territory, produce formally checkable research, and reach results at the edge of human knowledge.
But StarTalk’s warning is worth keeping. Science advances fastest when a result gives us both an answer and a way to think. A machine may be able to find a road before humans do. The next challenge is turning that road into a map people can understand, critique, teach, and extend.
BitcoinVersus.Tech
Advertisement
Editor’s Note
We volunteer daily to ensure the credibility of the information on this platform is Verifiably True.
If you would like to support to help further secure the integrity of our research initiatives, please donate here: 3C9o19EH5HSiwEPyCTmEKzxhNCbo2X6TTb
BitcoinVersus.tech is not a financial advisor. This media platform reports on financial subjects purely for informational purposes.
Leave a comment