The Clay Mathematics Institute, which offers a $1 million prize for resolving the Navier-Stokes problem, said on 11 September 2026 that the problem "has apparently been settled." The four-paragraph statement names no solver, no company and no proof, and it does not start the prize process, which under Clay's rules cannot begin until a solution has been published in a qualifying outlet and scrutinised for at least two years.
Key facts
- Statement date: 11 September 2026, three days after OpenAI said an internal model had resolved the problem.
- Not named: OpenAI, the NYU mathematician Tristan Buckmaster, the Anthropic researcher Levent Alpöge, and the Lean formal proof all go unmentioned.
- Two years minimum: Clay's rules require a proposed solution to "survive rigorous examination by the global mathematics community for a minimum of two (2) years."
- Primary sources: Clay's announcement and its Millennium Prize rules.
What Clay said, and what it left out
The Navier-Stokes equations describe how fluids move, and Clay's official problem statement asks whether their solutions in three dimensions always stay smooth or can break down. It has been one of the seven Millennium Prize Problems since 2000.
On 8 September OpenAI announced that an internal model had resolved it, sharing "both a writeup of the proof and a formalization in Lean," and Ground Truth covered the claim that day. Clay's response is carefully worded. Most of it is about why the prizes exist. The operative sentence is one line: "Today, CMI shares in the excitement of the global mathematical community as we contemplate the announcement that the Navier-Stokes problem has apparently been settled."
It ends by pointing to its rules: "The process is deliberately unhurried, but we will provide updates." Clay's page for the problem still carries the status "Active."
On the Hacker News thread, which passed 200 comments, readers noticed the silence. "The word 'OpenAI' doesn't appear at all," one wrote. Another: "That 'apparently' feels load-bearing."
How a Millennium Prize actually gets decided
Clay's 2018 rules are a slow, deliberate procedure, and that is on purpose. Think of it less like a referee's whistle and more like a patent office crossed with a jury that must wait two years to convene.
A solution qualifies only after it "has been published by a Qualifying Outlet," normally a refereed mathematics journal "of worldwide repute." Clay will not accept solutions sent to it directly. After publication, at least two years must pass, the result must achieve "general acceptance in the global mathematics community," and Clay must convene "a special advisory committee" of at least three members, two of them experts in the problem's field. Clay's president, Martin Bridson, told AFP the day OpenAI announced its result: "The process of evaluation is deliberately unhurried, and we shall ensure that it is absolutely rigorous."
One rule matters a great deal for this case. Clay says it "will pay special attention to the question of whether a Prize solution depends crucially on insights published prior to the solution under consideration," and it may recognise that earlier work or "recommend the inclusion of the author of prior work in the award." Priority is already disputed. Buckmaster has said OpenAI asked him to drop his Anthropic coauthor from a joint announcement, and OpenAI's own post says it "began on September 1st after hearing a rumor" about Alpöge and Buckmaster's related work.
What happens to the money
No source says. OpenAI's announcement is explicit that it is not seeking it: "We do not intend to claim the Millennium Prize for this result." Clay's rules give it three options: award the prize to one person, divide it among several solvers, or award nothing. They add that if Clay "cannot come to a decision about the correctness of a solution to a Problem, its attribution, or the appropriateness of an award," it may conclude that no prize be awarded. A proof produced largely by an AI system, with contested human priority and a company that declines the money, is the kind of case that clause was written for.
Why it matters
Clay is the institution whose word turns a claimed solution into a recognised one, so its caution sets the pace for everyone else. The statement gives OpenAI's result a real acknowledgement without endorsing who deserves credit, and it lands a day after twenty-five Fields Medallists warned that rushed AI announcements raise "severe attribution and plagiarism questions." OpenAI's proof and Lean formalisation are public, so the two-year examination can begin once a qualifying publication exists.
The caveat
"Apparently settled" is not a verdict, and Clay has not said whether it considers the Lean formalisation or any publication to date to satisfy its rules. OpenAI updated its announcement on 10 September with the result of an internal investigation into whether Buckmaster's earlier chats could have influenced its system, and the credit questions remain open.
Originally published on Ground Truth, where every claim is checked against the primary source.
Top comments (0)