On September 8, 2026, OpenAI announced a breakthrough: its internal model had found a solution to the problem of existence and smoothness of the Navier–Stokes equations — one of the seven “Millennium Problems,” each carrying a $1 million prize. The Wall Street Journal called the discovery the “Holy Grail of mathematics.” The news stirred the scientific community: some see it as a triumph of technology, others as a reason for serious debate.

What the Problem Is All About

The Navier–Stokes equations describe the motion of liquids and gases — from water flowing through a pipe to air currents around an aircraft wing. They are used in aerospace engineering, meteorology, and medicine. Yet mathematically, a key question remains unresolved: do solutions to these equations in three‑dimensional space always exist and remain smooth (i.e., continuous and predictable) for any reasonable initial conditions? Or is it possible that a smoothly starting flow “blows up” in finite time — with velocity growing to infinity? This very question was formulated as one of the Millennium Problems.

How AI Reached the Answer

Work began on September 1, 2026. OpenAI didn’t rely on a single model but deployed a “team” of roughly 10,000 autonomous AI agents. Some searched for a proof, while others tried to find a counterexample — a refutation. An important stepping stone was solving a simpler problem: the regularity issue for the Euler equations (a special case of the Navier–Stokes equations that ignores viscosity). This success provided the necessary conceptual foundation.

The agents arrived at a solution to the Navier–Stokes equations on September 5 — about 88 hours after the work began. Next came 17 hours of formal verification using the Lean language, which allows mathematicians to rigorously check every step of a mathematical argument. During this process, the agents exchanged around 2.7 million messages and consumed roughly 130 billion output tokens — a volume comparable to a million books.

OpenAI released both a description of the proof and its formalization in Lean. At the same time, the company stated it does not plan to claim the monetary prize. According to the rules of the Clay Mathematics Institute, the solution must be published in a peer‑reviewed journal and withstand two years of scrutiny by the mathematics community before a committee will consider awarding the prize.

The Controversy Surrounding the Discovery

Shortly before OpenAI’s announcement, mathematician Tristan Buckmaster from New York University, together with Levent Alpöge from Anthropic, presented preliminary results on a closely related problem. Buckmaster argued that OpenAI might have drawn on the ideas from their work. OpenAI denied this, saying its team had not seen the research before its publication and that the proofs themselves differ significantly.

Thus, the story is not just about a technological breakthrough but also about complex issues of priority and scientific ethics. Right now, independent expert review is the key: it will determine whether the solution is officially recognized.

What This Means for Science

The solution mathematically demonstrates that smooth solutions to the Navier–Stokes equations can break down in finite time. This is a fundamental result, though it doesn’t mean water can literally “explode” in the real world — such a scenario is considered highly unlikely from a physics standpoint.

The breakthrough shows how AI can tackle problems that have stumped the best mathematicians for decades. At the same time, concerns have been raised: for example, renowned mathematician Terence Tao worries that ready‑made answers from AI might weaken deep human understanding of the subject, since it is often the laborious journey toward a solution that builds that understanding.

This case is a vivid example of how artificial intelligence is becoming a full‑fledged tool for fundamental science. But with new capabilities come new challenges: how to verify results, how to establish authorship, and how to maintain a balance between computational power and the depth of human knowledge.

Powered by Froala Editor