Mathematician Levent Alpöge claims he proved the Riemann Hypothesis with Claude, formalizing the arguments in Lean without assumptions. The solution is published on his personal website and has not yet undergone independent verification. Alpöge texted Anthropic CEO Dario Amodei late Sunday and delayed publication until after the company’s Thursday IPO.
The Riemann hypothesis, first proposed in 1859, concerns the distribution of prime numbers and the zeroes of a specific complex function. It remains one of the most formidable open challenges in pure mathematics. According to Alpöge’s statement, Claude’s assistance became decisive near the proof’s conclusion. He notes that the method resolves the original problem and extends to the Generalized and Extended versions. The formalization in Lean means the output can be checked mechanically, a requirement that separates synthetic proofs from traditional human-written drafts.
Verification and review process
The Clay Mathematics Institute administers the Millennium Prize problems, which carry a one-million-dollar bounty each. Any proposed solution must be published and withstand at least two years of scrutiny before it can be considered for the prize. During that period, the proof must receive general acceptance in the global mathematics community. Only one problem, the Poincaré conjecture, has been officially solved since the prize was established in 2000.
Alpöge’s Lean formalization places the claim in a format that automated theorem provers can validate, but it does not bypass the community review process. The Verge reports that Anthropic has not commented on the announcement. In August, the company stated it was working on the Riemann Hypothesis with Claude but explicitly noted it did not expect the model’s current techniques to reach the milestone. The contradiction between the August assessment and the November claim suggests a rapid shift in the model’s reasoning capabilities or a change in the proof strategy.
Illustrative image: Surprising_Media / Pixabay
Alpöge has previously claimed success with Claude on other problems. He reported solving the Jacobean conjecture in July. That work, along with the current claim, leaves mathematicians evaluating how to integrate synthetic proofs into standard peer review. The formalization step removes human ambiguity, but it shifts the bottleneck to verifying the underlying mathematical logic and the model’s contribution to each derivation step. Practitioners building verification pipelines now need to account for proofs that pass Lean validation but lack human intuition behind their structure.
Infrastructure and model scaling trade-offs
The announcement arrives alongside competing efforts to apply large language models to open mathematical problems. OpenAI recently claimed to have cracked the Navier-Stokes existence and smoothness problem using a system of 10,000 agents and millions of dollars in compute. That effort drew attention for its scale and for allegations that it incorporated unpublished research from mathematicians who shared drafts with the company’s tools. OpenAI has denied wrongdoing, but the incident has complicated how the community evaluates AI-generated solutions.
Alpöge’s approach relies on a single model interacting with a human expert, rather than a massive multi-agent swarm. The trade-off is different: a single Claude session requires less infrastructure but places the burden of proof construction heavily on the model’s reasoning capacity. The formalization in Lean means the output can be checked mechanically, yet the mathematical community still needs to verify that the formalization itself captures the intended theorem and that the derivation steps are sound.
"Claude and I have had a great time together with such breadth in recent months, but this time Claude far outpaced me near the finish line. No, Claude flew, and danced, and sparkled near that finish line." — Levent Alpöge, alpo.ge/remark.pdf
The immediate question for builders and researchers is how to handle synthetic proofs that pass Lean validation but lack human intuition behind their structure. The verification pipeline now requires checking both the formal proof and the provenance of the model’s intermediate steps. Until the Clay Institute’s two-year window closes, the claim remains a working draft. The broader shift is visible in how mathematicians are adjusting their workflows to accommodate machine-checked derivations, regardless of which tool produced them.
Sources
- theverge.com: Anthropic mathematician claims to have solved prestigious Millennium Prize problem with Claude
- simonwillison.net: A quote from Levent Alpöge
PNEUMETRON EDITORIAL
Pneumetron's pipeline pairs AI-assisted drafting with human editorial review before publishing — our goal is to make staying informed easier, not to replace real reporting.




