A second Riemann hypothesis proof claim lands hours after the first
Two claims on math's most famous open problem arrived within hours, and only slow, brutal verification will decide which one, if either, survives.
Within hours of a blockbuster announcement that an AI-assisted effort had solved the Riemann hypothesis, a second claim emerged, leaving two unverified proofs of math's most famous open problem on the table. For executives watching AI's push into scientific discovery, the episode is a live test of whether machine-generated breakthroughs can survive the referees who decide what is real.
Within hours of a blockbuster announcement that an AI-assisted effort had solved the Riemann hypothesis, the most famous open problem in mathematics, a second claim landed. That is the full public record so far: two claims, hours apart, on a problem that has beaten back more than 160 years of attack. No referee has signed off. No proof has been verified. And that gap, between the announcement and the verification, is where this story actually lives.
Here is the problem in plain English. In 1859, Bernhard Riemann noticed that the way prime numbers thin out as you count higher is governed by a hidden rhythm, encoded in a function whose zeros sit on a specific line in the complex plane. His hypothesis: every one of those nontrivial zeros sits exactly on that line. If it is true, the primes are as evenly distributed as they can possibly be, and a huge amount of number theory snaps into place. If it is false, a lot of what mathematicians believe about primes is built on sand. The prize money is real, too. The Clay Mathematics Institute named it one of seven Millennium Prize Problems in 2000, each worth $1 million. Exactly one has fallen: the Poincare conjecture, proved by Grigori Perelman in work he posted in the early 2000s. He declined the $1 million in 2010. The other six, Riemann included, are still standing.
In mathematics, a claim is cheap and a proof is expensive. There is no lab, no experiment, no prototype that either works or does not. A proof is a social object until it is checked, and checking is done by humans who are slow, careful, and frequently unconvinced. Announced proofs of the Riemann hypothesis have come and gone for decades. The typical failure is not fraud; it is a single step that looks obvious and is not, buried on page 40. So the field's default posture is polite skepticism, and the fact that two claims arrived hours apart does not double the odds of either. It raises harder questions. Are they independent? Is one a variant of the other? Is one a correction of the other? Is this a priority dispute in the making? None of that is answerable yet.
The AI-assisted label is doing a lot of work in that first announcement, and it deserves scrutiny. AI systems have been pointed at hard mathematics with increasing seriousness, and the interesting shift is not that a model can generate a plausible argument. It is that proofs can now be checked mechanically. Formal verification tools let a computer confirm every logical step, which makes a proof far harder to fake and far easier to trust. That is the real change in the economics of mathematical discovery. But AI-assisted covers a wide range. It can mean a model suggested a promising direction that a human then developed. It can mean a model produced a full argument that a human cleaned up. It can mean a model found a counterexample to a lemma everyone assumed. The label alone settles nothing. What settles it is whether the argument can be formalized and machine-checked, and whether humans who know the terrain can follow it.
What happens next is a gauntlet, not a coronation. Expect the work to be posted publicly, where specialists will pick at it line by line. Expect referees at a serious journal to spend months, possibly years. Expect independent groups to try to reproduce the argument or break it. For a result of this size, the failure modes are well known: a gap that cannot be closed, a lemma that is assumed rather than proved, an argument that quietly assumes what it is trying to show, or a computational step that works at small scale and collapses at large scale.
The second-order stakes are bigger than the math. For AI labs, a verified proof of the Riemann hypothesis would be the single strongest advertisement imaginable for AI as a discovery engine, worth more than any benchmark. For the field, it would be a generational event that reshapes number theory and the cryptography that leans on it. And for the broader AI-does-science narrative, a failed claim is just as informative as a successful one. It tests whether the hype cycle can survive contact with referees who have no incentive to be impressed. There is a credit dimension too. If two claims are live at once, the question of who got there first, and what counts as getting there, becomes a fight over priority that can outlast the mathematics itself.
Watch for specific signals, not vibes. A formal verification in a proof assistant is the strongest one. A named author with a track record in analytic number theory is another. Independent confirmation from mathematicians with no stake in the outcome is a third. Journal acceptance is the slowest and most definitive. Until several of those land, both claims should be treated as unverified, and the honest headline is that we do not know yet. The strategic read for anyone watching this space: as AI systems get better at generating plausible mathematics, the scarce resource shifts from generation to verification. The labs, journals, and institutions that build real capacity to check machine-produced arguments will hold the bottleneck, and the bottleneck is where the credibility lives.
This story's Key Insights and Take-aways are locked.
Create a free account to unlock Executive Actions for one credit.
Register to UnlockAlways free for Executives Club members. Join the Club
More in Science
Anthropic's I.P.O. Made Rank-and-File Researchers Worth Tens of Millions
Researchers who joined just a few years ago are now worth tens of millions, and the windfall resets the pay benchmark for every private A.I. lab still paying in paper.
Sea Lion Culling Strategy Fails: New Model Says Killing Them Won't Save Salmon
The latest research from AAAS overturn a long-held conservation assumption, forcing policymakers to rethink critical resource allocation in marine ecosystems.
Microplastics do more than pile up: they react with water to damage cells
Scientists have found that plastic particles can generate highly reactive molecules in water, a mechanism that could explain links to heart attacks, strokes, and neurological disorders.


