Who he is
Levent Alpöge is a mathematician whose research includes number theory and arithmetic statistics. The supplied biographical excerpt says he studied mathematics at Harvard, completed further mathematical study at Cambridge, held a postdoctoral fellowship affiliated with Columbia and Harvard, and was a Junior Fellow in the Harvard Society of Fellows from 2021 to 2025. It says he later joined Anthropic as a member of its technical staff. [2]
His earlier research includes work with Manjul Bhargava and Ari Shnidman on how often integers can be expressed as sums of two rational cubes. The background excerpt reports that their 2022 work established an upper bound of about five-sixths, later improved by other researchers. It also describes a 2025 result with collaborators concerning Hilbert’s tenth problem over rings of integers of algebraic number fields. [2]
Why he is in the news
A report dated November 20, 2026 says Alpöge announced a possible proof of the Riemann hypothesis, the longstanding conjecture about the distribution of prime numbers. He said the argument was developed while working with Anthropic’s Claude and formalized in Lean, a programming language used to encode proofs for machine checking. The report says mathematicians, including MIT’s Andrew Sutherland, confirmed that the Lean proof compiles. That is meaningful evidence that the formalized argument is internally checkable, but it is not the same as the mathematical community having fully understood and accepted the proof. The report says that work of human interpretation and assessment remains ahead, and Alpöge committed to pursuing it. [1]
The same report notes that Alpöge and NYU mathematician Tristan Buckmaster had also released work related to the Navier–Stokes equations. A separate biographical excerpt says Buckmaster alleged that OpenAI accessed his private work after citing their AI-assisted research in connection with its own Navier–Stokes result. The supplied material does not include OpenAI’s response or an outcome, and the allegation concerns OpenAI—not wrongdoing by Alpöge. [1][2]
These accounts are excerpts and summaries, not the full reports or the proof itself. They support describing the Riemann result as a proposed proof whose Lean formalization has reportedly compiled, not as a settled resolution of the hypothesis. [1]
Why this matters for mathematics
Alpöge’s announcement sits at the intersection of two distinct questions: whether a proof can be mechanically checked, and whether mathematicians can understand what it establishes and why. Lean can help verify that a formal argument follows its encoded rules. But, as the report’s account of the ongoing discussion makes clear, successful compilation alone does not settle whether the argument is correct in the broader mathematical sense, nor does it provide a clear explanation that other researchers can readily evaluate. [1]
The episode also highlights a changing working relationship between mathematicians and AI tools. Alpöge described working through the argument with Claude, while expressing some concern about releasing a result without a human-intelligible explanation. His announcement therefore raises practical questions about how to communicate, review, and credit work produced with AI assistance. The supplied excerpts establish that these questions are being raised; they do not establish that AI itself has independently proved the conjecture. [1]
My take
The most responsible way to treat Alpöge’s announcement is as an important claim under examination—not as a solved problem. Independent confirmation that the Lean code compiles is a substantial check, but the report itself emphasizes that human understanding and scrutiny are still needed. [1]
That distinction should not diminish the achievement of making a proposed argument available in a form that can be checked. It does, however, set a high bar for headlines and public claims: until specialists can explain and validate the reasoning, uncertainty is part of the story. Alpöge’s stated intention to continue working toward an intelligible account is therefore central, not incidental. [1]