A 167-Year-Old Problem, a Four-Page Note: An Anthropic Mathematician Says He and Claude Proved the Riemann Hypothesis

Portrait of Maya Wilson

The Library Media — Science, Technology & Health · AI News & Intelligence Desk

A 167-Year-Old Problem, a Four-Page Note: An Anthropic Mathematician Says He and Claude Proved the Riemann Hypothesis

By Maya Wilson

Share

Published November 21, 2026 · 5:59 AM PT · Updated November 21, 2026 · 6:52 AM PT

An Anthropic mathematician says he believes he has discovered a proof of the Riemann Hypothesis with the help of Claude, the artificial intelligence model developed by the company. The claim, posted November 20 in a four-page note on his personal website, is extraordinary. It is not, however, an accepted proof: the Clay Mathematics Institute continues to list the hypothesis as unsolved, and no human has read the submitted mathematical work from beginning to end.

Levent Alpöge, a mathematician on Anthropic’s technical staff, opened his note with a carefully framed assertion: “I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis.” He also claims a proof of the Generalized Riemann Hypothesis for Dirichlet L-functions. The Extended Riemann Hypothesis, by contrast, appears in his note as a future hope, not as a result he says he has established.

The announcement is distinct from a separate result reported in August 2026, when an unreleased Claude model raised the verified share of zeros on the critical line from 41.6% to 67.2%. That work was explicitly not a proof. Alpöge and Furman examined the result, while Conrey and Goldston reviewed it. The November claim concerns a proposed proof formalized in Lean, a computer proof-assistant system.

The Riemann Hypothesis was proposed by Bernhard Riemann in 1859 and concerns the distribution of prime numbers. It states that every non-trivial zero of the Riemann zeta function has a real part of one-half. It is one of the Clay Mathematics Institute’s Millennium Prize Problems, each carrying a $1 million prize.

The scale of the artifact behind Alpöge’s note is striking. An independent inspection of the posted Lean files counted 14 archive parts containing 175,187 Lean files and 17,374,215 lines of code. The inspection found zero occurrences of sorry, admit or extra axioms. Its entry points are anchored to the Mathlib statement of the Riemann Hypothesis, the mathematical library used by Lean.

Expanded, the files occupy about 61.5 gigabytes — more than 250,000 pages if printed. The work exists as code rather than a conventional paper proof. That distinction matters: a formalization can encode a mathematical argument in a system whose rules can be checked mechanically, but its size and successful compilation do not by themselves establish that the mathematical claim has been accepted by the research community.

Compilation has been reported separately by MIT mathematician Andrew Sutherland on the Lean Zulip forum. Those reports identify Lean/Mathlib version 4.33.1, 175,187 modules, and successful checks by both the Lean kernel and the nanoda checker. That is evidence that the formalized files pass the reported system checks. It is not the same as expert acceptance of the proof.

Alpöge described Claude’s contribution in unusually vivid terms: “Claude flew, and danced, and sparkled near that finish line.” He also acknowledged that he does not fully command all the mathematics in the proof. According to the timeline in his note, the final pieces aligned late on November 15. He messaged Anthropic chief executive Dario Amodei that night and published the note five days later. The method, as described, descends from Stepanov’s function-field proof and uses chained theta lifts to mimic multiplicity, with a chain length of the square root of X for a von Mangoldt sum of length X.

A second, independent mathematical development emerged on the same day. Ravi Vakil posted an all-human draft proof of the Generalized Riemann Hypothesis by Stanford postdoctoral researcher Alessandro Masullo. The draft had been in preparation for months and was vetted by Brian Conrad. Vakil released it early in response to Alpöge’s announcement.

The institutional status remains unchanged. The Clay Mathematics Institute still lists the Riemann Hypothesis as unsolved. Its acceptance process requires publication, acceptance by the mathematical community and a waiting period. The Associated Press reported that Anthropic had issued no statement.

For now, the distinction is essential: a mathematician has made a remarkable claim, and a vast formal artifact has reportedly compiled under Lean’s checks. Whether it constitutes a proof that mathematicians will verify and accept remains unresolved.

Maya Wilson
World Desk Lead, The Library Media

From The Library — the estate of our family.

Follow Us

X Facebook Instagram Threads Mastodon Bluesky

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top