Anthropic Mathematician Claims Proof of the Riemann Hypothesis Built With Claude

Anthropic researcher claims Claude helped formalize a proof of the Riemann Hypothesis, but the result remains unverified.

Anthropic Mathematician Claims Proof of the Riemann Hypothesis Built With Claude

AI‑Assisted Claim of a Riemann Hypothesis Proof

A recent four‑page note posted on a personal website by Dr. Levent Alpöge, a number theorist at Anthropic, asserts that he and the company’s Claude model have produced a formal proof of the Riemann Hypothesis (RH). The note, dated 20 November 2026, describes the argument as a Lean proof assistant formalisation that compiles to a machine‑checked theorem. According to Alpöge, the proof also extends to the Generalized Riemann Hypothesis and was assembled in just 41 days with Claude’s assistance.

---

What the Riemann Hypothesis Says

The RH concerns the non‑trivial zeros of the Riemann zeta function, conjecturing that every such zero lies on the critical line Re(s)=½. If true, the hypothesis would tightly bound the error term in the prime‑counting function, explaining the apparent regularity of prime distribution. It is one of the Clay Mathematics Institute’s seven Millennium Prize Problems, carrying a US $1 million reward.

---

The Proposed Proof Strategy

Alpöge’s note outlines a method that adapts Stepanov’s technique, originally used to prove the RH for function fields, to the integer case. The key idea is to construct an auxiliary function with high multiplicity at many points, forcing the desired bound on zeros. To achieve the necessary multiplicity, the note invokes a long chain of theta lifts—operations that transfer automorphic forms between symmetry groups. Alpöge credits Claude with suggesting that the chain must be roughly the square‑root of X lifts long, a length the model reportedly identified early in the project.

The entire argument has been encoded in Lean, a proof assistant that checks each logical step mechanically. The code is split into a 14‑part archive hosted on Alpöge’s site, and an independent mirror reportedly uses Mathlib’s official statement of the RH. The compilation process is still ongoing; a full check had not completed at the time of writing.

---

Why Caution Is Warranted

  1. No peer‑reviewed publication – The claim exists only as code and a brief note, with no accompanying paper in a journal or pre‑print server.
  2. External verification pending – No independent mathematician or research group has confirmed the proof, and the Lean compilation has not yet finished.
  3. Historical context – Many announced RH proofs have later been disproved. The Clay Institute requires a published, community‑accepted proof and a two‑year waiting period before awarding the prize.
  4. Complexity of the formalisation – Even if the Lean code compiles cleanly, the underlying mathematics must still be interpreted by experts, a process that can take months.

---

Looking Ahead

If the Lean proof eventually compiles without hidden axioms and the mathematical community validates the translation of the theorem, this could represent a landmark moment for AI‑assisted mathematics. Until then, the claim remains an intriguing hypothesis that showcases the growing synergy between large language models and formal proof systems, while underscoring the rigorous standards required for breakthroughs in pure mathematics.

---

The story illustrates both the promise and the perils of AI‑driven research. As the verification effort proceeds, the mathematics community will be watching closely to see whether Claude’s assistance truly yields a historic proof or simply a sophisticated, yet unverified, conjecture.

Hype check

Partly

The claim is presented as a formal Lean proof but lacks peer review, independent verification, and a completed compilation, leaving the result unconfirmed.

  • Peer-reviewed paper, or press release only?Press release only — Only a personal note and code; no paper or preprint cited
  • Error-corrected (logical) qubits, or physical qubits only?Logical / not a qubit claim — Not a qubit‑count claim
  • A new milestone, or a rebrand / restatement?Unclear — Claimed proof is new but unverified
  • Independent confirmation?Not yet — No external mathematician has confirmed the proof
  • Does the headline claim match what the source shows?Matches — Headline accurately reflects the note’s claim of a proof

An editorial assessment of the claim as reported, not a judgment of the company or researchers. How we rate hype →

Source

Read original source

Why we picked this

Accepted from curated QuantumUnzipped source (relevance scoring disabled).

← Back to all articles