Skip to content
InYourGeek
visiteur@inyourgeek — shell
↹ compléter↑↓ historique⏎ ouvrirhelp
FR

The Monday letter · the tech and security week in one email

Free, every Monday, via Brevo (EU), one-click unsubscribe · your data and rights

AI· 3 min read

Riemann Hypothesis: An Anthropic Researcher Says He Has Proved It

Levent Alpöge, a mathematician at Anthropic, claims to have proved the Riemann hypothesis with Claude's help. Nobody independent has checked the proof yet, and that is the one detail that matters.

Illustration of the Riemann hypothesis: prime numbers and the zeros of a function, alongside a proof still waiting for independent review

What is being claimed, and by whom

On 20 November 2026, The Verge reported that Levent Alpöge, a mathematician at Anthropic, says he has proved the Riemann hypothesis while working with Claude. The announcement sits on his personal website and nowhere else. Anthropic has neither commented on it nor put its weight behind it, and had not replied to the outlet by the time the story ran. So we have exactly one source, relaying the claim of exactly one author: read everything below with a large “allegedly” stapled to it.

Formulated in 1859, the hypothesis concerns the distribution of prime numbers and the zeros of a function. It is one of the seven Millennium Prize Problems, each worth a million dollars. Alpöge does not stop there: according to him, the method extends easily to the generalised Riemann hypothesis, and he expects, given more time, to settle its extended version as well. If you are going to announce one of the most eagerly awaited results in mathematics, you may as well order the full set.

Claude’s share of the work

The researcher does not describe the model as a glorified calculator. He writes that he discovered the proof while working with Claude, then formalised the arguments with it, with no additional assumptions. The proof is presented in Lean. The source says nothing more about who did what, human or machine.

The backstory gives the announcement some extra bite. In August, Anthropic said it was working on the Riemann hypothesis with Claude, while adding that it did not expect the techniques the model was using to lead to a proof. Three months later, one of its own mathematicians says the opposite on his personal page, and his employer has gone very quiet.

The timing deserves a look too. Alpöge says he texted Anthropic’s CEO, Dario Amodei, late on Sunday night. He hints that he held off publishing until the company’s IPO was out of the way, on Thursday 19 November. A problem that has been open since 1859 could presumably cope with four more days while the shares got listed.

Two years of scrutiny, at the very least

Nothing is settled. The solution has only been published on its author’s website, and the Clay Institute, which administers the Millennium Prize, has not publicly accepted it. Its rules are slow on purpose: a proof must be published, survive at least two years of scrutiny, and be generally accepted by the global mathematical community before it is even considered for the prize.

The track record is a good argument for keeping your champagne corked. In the quarter of a century the prize has existed, exactly one problem has been officially solved: the Poincaré conjecture.

Reviewers already drowning

The announcement lands in a crowded field. In September, OpenAI claimed to have solved another Millennium problem, Navier-Stokes, by throwing 10,000 agents and millions of dollars at it. According to The Verge, the result is widely considered credible, but it has received no formal recognition. It is also clouded by accusations about the research practices involved, which OpenAI firmly disputes.

Alpöge himself said in July that he had solved the Jacobian conjecture with Claude. Mathematicians, meanwhile, are struggling to keep up with the pile: proofs now arrive faster than the people qualified to check them.

So the million dollars will wait at least two years. The primes have been in the queue since 1859, and one more round of peer review is unlikely to be what finally breaks their patience.

Sources (1)

Written with AI assistance from the sources cited above, then reviewed and approved before publication by Sébastien Soulier.

The Monday letter

The tech and security week, every Monday in your inbox

What to patch first, data breaches, the articles with pictures. Free, no ads, one-click unsubscribe.

Free, every Monday, via Brevo (EU), one-click unsubscribe · your data and rights