Ravi Vakil Posts 940-Page Riemann Proof Draft as Reviewers Flag Problems

Hours after Anthropic mathematician Levent Alpöge claimed Claude proved the Riemann Hypothesis on Friday, November 20, Stanford math professor and American Mathematical Society president Ravi Vakil rushed out a competing 940-page all-human proof draft. Roughly eight hours later, Vakil updated his Math 216 blog to warn that mathematicians using AI models had already reported problems in the manuscript.

The competing manuscript, titled The Music of the Primes, was written by mathematician Alessandro Maria Masullo over about 12 years to tackle the Generalized Riemann Hypothesis (GRH), a stronger version of Bernhard Riemann’s 1859 conjecture about prime numbers. Vakil posted the draft just before an American Mathematical Society Executive Committee and Board of Trustees meeting, writing, “It is all human, with no AI.” Stanford number theorist Brian Conrad, who advised Masullo’s 2020 Stanford PhD thesis, had quietly reviewed the draft for months.

Vakil wrote that Stanford researchers originally planned for Masullo to spend a year working with Conrad to polish the draft before going public, but Alpöge’s AI claim pushed them to release the current version immediately. Eight hours later, Vakil appended a candid update: “Several people have contacted me to report problems in the proposed proof, identified with the assistance of different AI models. (Thanks to several mathematicians in particular who were first to let me know.)” He added, “As is usual, all issues need to be resolved to the satisfaction of the community before this proof would be accepted as true. (I make no claim as to the correctness of the proof strategy.)”

Where the two Riemann Hypothesis claims stand

Vakil’s hurried post came hours after Alpöge published 14 archives of Lean 4 code, a proof-checking language that acts like a digital referee for logical steps, in Friday’s Lean code release. Although some early reports described a full paper alongside the code, Alpöge’s website holds only a four-page note and 17.4 million lines of Lean code that he says would span more than 250,000 printed pages. “I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis,” Alpöge wrote, adding that the argument covers GRH and relies on chaining mathematical transformations of length square root of X for a sequence of length X.

Neither proof has passed peer review or full community verification. Independent checks of the nanoda kernel, theorem statement, and axioms on Alpöge’s Lean files were reported passed by Saturday morning, while MIT mathematician Andrew Sutherland continued running the main comparator audit on an MIT verification page last updated at 12:55 UTC on November 21.

DetailLevent Alpöge and ClaudeAlessandro Maria Masullo
InstitutionAnthropicStanford PhD, posted by Ravi Vakil
Claimed scopeRiemann Hypothesis and GRHGeneralized Riemann Hypothesis
Format and size14 Lean packages (17.4 million lines) and a four-page note940-page manuscript (The Music of the Primes)
AI involvementCo-developed with Anthropic’s ClaudeNone
Verification statusKernel, statement, and axiom checks reported passed; MIT comparator ongoingNot formalized in Lean; AI-assisted checks flagged problems after eight hours

How top mathematicians reacted to both releases

Conrad kept a cautious tone on his former PhD student’s manuscript, telling Scientific American that Masullo’s draft is undergoing major revision and declining to say the proof works. He was much sharper about Alpöge’s release in The New York Times, calling it “frankly completely outrageous” to publish a proof the author has not yet fully mastered.

Alpöge openly acknowledged that gap in his note, writing, “This puts us in the rather extraordinary situation this age has made possible, in which I am presenting the world with a complete argument before having mastered a massive neighbourhood around its mathematics.” Number theorist Andrew Granville told Scientific American that Alpöge’s four-page sketch was “garbled.” UCLA Fields Medalist Terence Tao also weighed in on his blog on Friday, calling Alpöge’s release an “(enormous) Lean-only proof” built with an internal AI model, noting Masullo’s manuscript will take time to digest, and declining to endorse either claim.

What is at stake and what happens next

Proposed by Bernhard Riemann in 1859, the Riemann Hypothesis predicts that all non-trivial zeros of the Riemann zeta function lie on a vertical line with real part one-half, a pattern tied directly to how prime numbers are distributed. Because GRH extends that claim to a wider family of equations, a valid GRH proof would automatically settle the original hypothesis and confirm hundreds of theorems that assume it holds. Even if one of the two proofs survives expert review, Clay Mathematics Institute rules require formal publication and multi-year community acceptance before awarding its $1 million Millennium Prize, putting any decision no earlier than 2028 or 2029.

Alpöge’s release landed one day after Anthropic’s Nasdaq IPO on Thursday, November 19, and on the same Friday he and NYU mathematician Tristan Buckmaster also posted a separate claim on the Navier-Stokes fluid equations, following Alpöge’s July 2026 Claude-assisted counterexample to the three-dimensional Jacobian conjecture. Alpöge said he made the Riemann breakthrough late Sunday night, requested clearance from Anthropic CEO Dario Amodei, called the post-IPO timing unfortunate, and rejected suggestions of a PR stunt. Neither Anthropic, Stanford University, nor the Clay Mathematics Institute had issued an official statement on the competing claims.

Sutherland’s MIT audit still needs to finish its comparator check, and mathematicians must verify that Alpöge translated the standard statement of GRH into Lean without hidden definition errors before anyone treats the Riemann Hypothesis as solved. Alpöge plans to simplify the code and write a human-readable paper over the coming months, while Vakil has urged researchers to inspect Masullo’s draft, translate it into Lean, and determine whether the newly reported problems are fixable gaps or fatal flaws.

Recent Articles

Related News

Leave A Reply

Please enter your comment!
Please enter your name here