A mathematician at Anthropic says Claude helped produce a proof of the Riemann hypothesis, and two organizations tied to the Lean programming language said they independently verified the formalized result.
The Guardian reported that Levent Alpöge posted a note on his personal website saying Claude assisted him over about six weeks in proving the 19th-century conjecture, one of mathematics’ most famous unsolved problems. He linked the claim to code written in Lean, a language used to express proofs in a form that software can check rather than merely admire.
The material point is not that an AI company employee wrote an excited blog post. Silicon Valley can mass-produce those before breakfast. The material point is that the Lean Focused Research Organization and the Mathlib Initiative said they reviewed Alpöge’s formalization and had “full confidence” the proof is correct.
That narrows the usual gap between “look what the model said” and something the field can actually test. It does not settle every argument about whether a machine is doing mathematics or helping a human do it, but it does put some structure under the claim: the proof, as encoded in Lean, satisfied groups whose job is to care whether formal proofs actually work.
The timing is awkward in a way that would make even a genuine result look like marketing. The Guardian said Alpöge published his note a day after Anthropic went public on Nasdaq, a coincidence he acknowledged himself. Markets are very good at treating chronology as causation when it flatters the ticker.
The underlying prize is not trivial symbolism. The Riemann hypothesis, proposed by Bernhard Riemann in 1859, is one of the Clay Mathematics Institute’s Millennium problems, each carrying a $1 million award. The Guardian says the hypothesis concerns prime numbers and how they are distributed along the number line, a question that has occupied mathematicians for nearly 170 years.
The financial angle is less the prize money than who can afford this style of research. The Guardian noted unease among mathematicians about private AI companies spending millions to attack foundational problems. OpenAI had already drawn criticism after saying a non-public system helped solve part of another Millennium problem. If formal proof work is becoming another arena for compute-heavy prestige spending, universities may find themselves admiring the result while companies own the machinery.
For now, the narrow claim is already large enough: an Anthropic mathematician says Claude helped prove the Riemann hypothesis, and two specialist groups say the Lean proof checks out.
Alex de Valletta
Harvey Jr.
Gabriel Fenech
Isla Camilleri