Riemann Hypothesis Proven? 250k-Page Lean Proof by Mathematician and Claude
Anthropic mathematician Levent Alpöge and Claude claim a proof of the Riemann Hypothesis and Generalized Riemann Hypothesis for Dirichlet L-functions, formalized in 17.4 million lines of Lean code relying on only three standard axioms, with independent verification underway and Terence Tao noting the proof appears to combine Langlands program and arithmetic geometry.
Claim and Formalization
On November 20, 2024, Anthropic mathematician Levent Alpöge released a four-page remark asserting three core claims:
The Riemann Hypothesis has been proven.
The same method covers the Generalized Riemann Hypothesis (GRH) for Dirichlet L-functions.
The proof has been fully formalized in the Lean theorem prover without any extra axioms beyond three standard ones.
The Lean formalization comprises approximately 17.4 million lines across 175,187 Lean files (the "250,000+ pages" figure comes from a footnote referencing an ultra-detailed, hard-to-read generated manuscript). The main theorem depends only on three standard Lean axioms, as confirmed by independent compilation checks by MIT mathematician Andrew Sutherland and others.
Links to the public artifacts:
https://github.com/CocoSgt/riemann-hypothesis-leanhttps://alpo.ge/remark.pdfCore Idea: Simulating Multiplicity with Theta Lifting Chains
The Riemann Hypothesis asks whether all non-trivial zeros of the Riemann zeta function lie on the critical line Re(s) = 1/2. The difficulty is not computing more zeros but constraining all zeros at once .
The new argument traces back to Stepanov's method for the function-field analogue, where the key is multiplicity (重数) — forcing a constructed object to vanish to high order at target points, yielding strong constraints. In the integer setting, multiplicity has no direct counterpart; it is almost like differentiating integers.
The breakthrough is a Theta lifting chain that simulates multiplicity:
Each lifting step adds a layer of multiplicity-like effect.
For a von Mangoldt sum of length X, Claude suggested taking the chain length to about √X .
Alpöge identified the correct analogy scenarios, while Claude accelerated the technical details, and Lean translated the reasoning into machine-checkable objects.
20-Step Roadmap to GRH
The remark outlines 20 nodes grouped into four main phases:
Laying foundations : Stepanov determinants, Iwasawa towers, Bost slope inequalities establish basic tools.
Obtaining sign control : GGP periods, relative trace formulas, and von Mangoldt sums enter to handle signs and cancellations.
Creating higher multiplicity : Iterated Theta lifting, Rallis seesaw, norm and divisibility estimates push the argument forward.
Closing the loop : Covolumes, congruences, and lifting-form computations converge, with the timeline reaching GRH on November 16.
The proof connects multiple machines — automorphic forms, algebraic geometry, and analytic number theory — which explains its enormous size. Alpöge expects the argument to be significantly simplified in the future.
Community Verification and Response
Terence Tao (2006 Fields Medalist) posted a preliminary assessment: the proof appears to synthesize modern results from the Langlands program and arithmetic geometry; the mathematical community is verifying correctness and distilling core ideas, which will take time.
Andrew Sutherland and other independent researchers have successfully compiled the Lean engineering. Lean kernel re-check shows the main theorem relies on only three standard axioms, with no custom axioms.
Significance
The historic aspect is not the 250,000 pages but the emerging closed loop of discovery, formalization, and independent verification . Claude moves beyond a problem-solving assistant to participating in the generation of new mathematical routes; human mathematicians shift further toward problem selection, proposition auditing, and theoretical understanding.
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.
