Levent Alpöge, a mathematician at Anthropic, has posted a claimed proof of the Generalized Riemann Hypothesis written in Lean formal verification language, stating he worked with Claude to produce the 250,000-page argument. Experts say independent verification is required before the claim can be accepted.
An Anthropic Researcher Claims to Have Proved the Riemann Hypothesis — With Claude's Help
A mathematician at Anthropic has posted a claimed proof of the Generalized Riemann Hypothesis, one of the seven Millennium Prize Problems and arguably the most famous unsolved problem in all of mathematics. If the proof holds up, it would resolve a 167-year-old question and earn its author a $1 million prize — and it would mark the first time an AI system played a material role in a breakthrough at that level.
The claim has not been independently verified. Experts are urging caution. But the mathematics community is paying close attention.
What Happened
Levent Alpöge, a researcher at Anthropic with a background in analytic number theory, quietly posted the proof to his personal website this week. The document is written in Lean, a formal verification language that lets mathematicians encode proofs in a form a computer can check for logical consistency. At over 250,000 pages in its expanded form, it is one of the largest formal proofs ever attempted.
Alpöge stated that he worked collaboratively with Claude — Anthropic's own AI model — to produce the proof. He did not elaborate on exactly how Claude contributed, whether as a code-writing assistant, a proof-search partner, or something else.
Background: Why the Riemann Hypothesis Matters
The Riemann Hypothesis was first stated by Bernhard Riemann in 1859. It concerns the behavior of the Riemann zeta function — a mathematical object that encodes deep information about how prime numbers are distributed among all integers. The hypothesis predicts that all non-trivial "zeros" of this function lie on a specific line in the complex plane, called the critical line.
This matters because the distribution of prime numbers underpins a large portion of modern number theory, and in turn, modern cryptography. Much of internet encryption — including the RSA algorithm that secures financial transactions — relies on the difficulty of factoring large numbers, a problem whose hardness is closely tied to prime distribution.
Mathematicians have verified the hypothesis computationally for the first 10 trillion zeros. No counterexample has ever been found. But verification is not proof: a claimed proof requires showing that all zeros, infinitely many of them, satisfy the condition.
The Clay Mathematics Institute included the Riemann Hypothesis on its list of seven Millennium Prize Problems in 2000, offering $1 million for a correct proof or disproof. Six of the seven remain unsolved. The only one resolved — the Poincaré Conjecture — was proved by Grigori Perelman in 2003, who famously declined the prize.
The Specific Claim: Generalized Riemann Hypothesis
Alpöge's proof targets the Generalized Riemann Hypothesis (GRH), a broader statement that extends Riemann's original conjecture to a larger family of related functions called Dirichlet L-functions. A proof of GRH would automatically imply a proof of the original Riemann Hypothesis.
Get this in your inbox.
Daily AI intelligence. Free. No spam.
The use of Lean is significant. Unlike traditional mathematical proofs, which are written in natural language and checked informally by human reviewers, a Lean proof is machine-verifiable: the Lean proof checker will either confirm or reject each logical step. This eliminates certain categories of human error — a mathematician cannot accidentally skip a step and have reviewers miss it.
However, a Lean proof is only as good as its formalization. Experts note that encoding a complex mathematical argument in Lean requires extremely careful translation from the underlying mathematical ideas into formal code. Errors can be subtle — a lemma might be stated slightly differently than intended, or a dependency might be miscategorized. Independent verification of a 250,000-page Lean proof is itself a significant undertaking.
What Experts Are Saying
According to Scientific American, mathematicians contacted about the claim are treating it as plausible enough to merit serious attention, but are explicitly not confirming it. The standard threshold in mathematics is independent verification by qualified experts who have studied the full argument — a process that for a proof of this complexity could take months or years.
No major mathematics institution has issued a statement. The Clay Mathematics Institute has not commented publicly.
Claude's Role: What We Know and Don't
Alpöge's description of Claude as a collaborator is notable but short on specifics. AI tools are increasingly used in mathematics research — primarily for:
- Writing and debugging formal proof code (Lean, Coq, Isabelle)
- Searching for applicable lemmas across large mathematical libraries
- Generating candidate proof strategies for a human mathematician to evaluate
It is not yet clear which of these roles, if any, Claude played in this specific work. Anthropic has not issued a statement. The distinction matters: an AI that helped write Lean syntax is a different kind of contribution than an AI that identified the core mathematical insight.
What is clear is that Alpöge chose to credit Claude explicitly — a decision that itself signals something about how he views the collaboration.
Why This Is Different From Past Claimed Proofs
The Riemann Hypothesis has attracted dozens of claimed proofs over the decades. The vast majority have been dismissed quickly by experts who identified errors in the argument. What makes Alpöge's claim different:
- The claimant is a credentialed mathematician at a serious research institution with published work in analytic number theory.
- The proof is formalized in Lean, which means it can be checked mechanically — it is not a sketched argument.
- The scale of the formalization (250,000+ pages) suggests substantial effort, not a casual submission.
None of these factors prove the proof is correct. They do mean it deserves serious scrutiny rather than immediate dismissal.
Impact: What a Confirmed Proof Would Mean
A verified proof of GRH would be one of the most significant events in the history of mathematics. The practical effects would take time to materialize:
- Cryptography: Many cryptographic systems are designed assuming GRH is true (they use it as a working hypothesis). A proof would not break these systems — it would actually strengthen their theoretical foundations by confirming the assumption. A disproof, by contrast, could have serious implications for certain cryptographic algorithms.
- Number theory: GRH is a dependency in thousands of conditional proofs — theorems that have been shown to be true if GRH holds. A proof would instantly promote all of those conditional results to unconditional theorems.
- AI and mathematics: A confirmed AI-assisted proof of a Millennium Prize Problem would accelerate the already-rapid development of AI tools for formal mathematics, drawing significant research investment to the space.
What to Watch
The immediate question is whether Lean experts and number theorists will confirm that Alpöge's formalization is correct and complete. Watch for formal responses from the mathematics community over the coming weeks — specifically from researchers specializing in analytic number theory and formal proof verification. A Clay Mathematics Institute review would be the definitive step toward prize consideration, but that process is not expected to begin until after broad peer scrutiny.
Hector Herrera covers AI developments at NexChron. Follow the story as it develops.
Did this help you understand AI better?
Your feedback helps us write more useful content.
Get tomorrow's AI briefing
Join readers who start their day with NexChron. Free, daily, no spam.