type
Post
status
Published
date
Nov 21, 2026
slug
summary
tags
category
技术分享
icon
password
A 34-Year-Old Mathematician Claims a Proof of the Riemann Hypothesis with Claude, Publishing Lean Code
The mathematical world is abuzz with a bold claim from Levent Alpöge, a 34-year-old mathematician currently working at Anthropic. He has released a document stating that, in collaboration with the AI model Claude, he has found a proof for the Riemann Hypothesis. This is not just a conjecture; Alpöge has published the complete formalized proof in the Lean programming language, split into 14 compressed files for public download. If fully expanded, this code represents over 250,000 pages of detailed logical steps.
From a 4-Page Document to Formalized Code
Alpöge did not release a traditional, lengthy mathematical paper. Instead, he shared a concise 4-page PDF titled "Remark" that outlines the core methodology. He explained that he chose to publish the Lean code first because he had not yet fully grasped the extensive surrounding mathematical context required to write a complete, readable proof. He emphasized that the formalization was completed without relying on extra assumptions.
The Lean code serves as a machine-checkable record. Lean is an interactive theorem prover that allows mathematical derivations to be written as code and checked for logical consistency. By publishing the code, Alpöge invites the global community to verify if the proof compiles successfully and if the final theorem statement accurately reflects the Riemann Hypothesis.
The Role of Claude in the Proof
According to Alpöge, the breakthrough came at the final stage of the research when Claude "flew, danced, and sparkled near that finish line." The AI demonstrated technical capabilities that surpassed Alpöge's own understanding at the time. The proof borrows from the Stepnov method used for function fields, a technique that constructs auxiliary functions with high multiplicity zeros. However, applying this to the integer domain was difficult because integers lack a natural structure corresponding to function differentiation.
Claude proposed a critical new insight: using consecutive Theta Lifts within the framework of automorphic forms to accumulate the necessary multiplicity. Each lift in this chain adds one unit of multiplicity. Claude also identified a key parameter, suggesting that for a sum of length X in the von Mangoldt function, the Theta lift chain should reach a length of √X. The argument further incorporates modern number theory tools such as the relative trace formula, Iwasawa theory, and Bessel periods. Alpöge noted that he had prepared for this problem for decades, but the proof mobilized far more mathematical knowledge than he had anticipated.
Verifying the Riemann Hypothesis and Beyond
The Riemann Hypothesis, proposed in 1859, remains unsolved after 167 years. Alpöge claims that this method could also resolve the Generalized Riemann Hypothesis (GRH) and potentially extend to the Extended Riemann Hypothesis.
It is important to note that while Lean can verify the logical steps within a formalized system, it cannot guarantee that the input definitions are correct. Therefore, external researchers must check if the code uses valid axioms and if the final theorem is indeed equivalent to the original Riemann Hypothesis.
Levent Alpöge: Background and Previous Work
Levent Alpöge is a Turkish-American mathematician born in 1992. He studied at Harvard University, graduating with a perfect 4.0 GPA, and earned his PhD at Princeton University under Manjul Bhargava, a 2014 Fields Medalist. After postdoctoral research at Columbia University and Harvard, he joined Anthropic as a Member of Technical Staff, focusing on applying large models to frontier mathematics.
This claim follows a significant previous milestone. In August 2025, Anthropic released research where a non-mathematical employee, Jarred Sumner, used Claude to advance work on the Riemann Hypothesis. That work successfully increased the known lower bound of the proportion of non-trivial zeros on the critical line from 41.6% to 67.2%. Alpöge was one of the mathematicians who verified that earlier result.
The full PDF and the 14 Lean code files are available for public inspection, marking a potential turning point in how artificial intelligence contributes to solving century-old mathematical problems.
原文出处:量子位,链接:https://www.qbitai.com/2026/11/511727.html
- 作者:数字Baby
- 链接:myaigc.top/article/41fb7c2d-0e98-810c-8a29-f783d9129fe6
- 声明:本文采用 CC BY-NC-SA 4.0 许可协议,转载请注明出处。



