Key takeaways

  • Levent Alpöge earned a PhD in mathematics at Princeton in 2020, advised by Manjul Bhargava, after degrees from Harvard and Cambridge.
  • Before Anthropic he was an NSF postdoctoral fellow at Columbia and Harvard and a junior fellow of Harvard's Society of Fellows from 2021 to 2025.
  • On November 20, 2026 he released a Lean proof of the Riemann hypothesis made with Claude, which the Lean and Mathlib communities say is correct, though no human-readable paper exists yet.

Levent Alpöge is a number theorist who works at Anthropic. He earned his PhD in mathematics at Princeton in 2020 and was then a junior fellow at Harvard's Society of Fellows until 2025. On November 20, 2026, he said that he and Claude had found a proof of the Riemann hypothesis, long the most famous unsolved problem in mathematics, and released it as Lean code.

His published papers are mostly about counting solutions to equations on curves. For the claim itself, the formal check and the math behind it, see our articles on the Claude proof claim, the Lean verification and what the Riemann hypothesis is.

Levent Alpöge's education

His CV, linked from his website, lists this path:

Year Degree Institution
2014 AB in mathematics, summa cum laude Harvard University
2014 AM in physics Harvard University
2015 MASt in mathematics University of Cambridge
2016 MA in mathematics Princeton University
2020 PhD in mathematics Princeton University

His Harvard senior thesis, advised by Jacob Tsimerman, was titled "The average elliptic curve has few integral points." His Princeton thesis, "Points on Curves," was advised by Manjul Bhargava.

The CV lists a long run of student awards. Among them: the AMS-MAA-SIAM Morgan Prize; Harvard's Captain Jonathan Fay Prize, which the CV notes is for the best senior thesis; a Churchill Scholarship; and a Goldwater Scholarship. An asteroid, 25898 Alpöge, carries his name.

Academic positions before Anthropic

After the PhD, the CV lists two posts:

  • NSF Postdoctoral Fellow at Columbia and Harvard, 2020 to 2022.
  • Junior Fellow of the Society of Fellows at Harvard, 2021 to 2025.

In 2019 he taught a minicourse on arithmetic statistics at the Institut Henri Poincaré in Paris. With Dennis Gaitsgory and Gurbir Dhillon he wrote course notes for a two-semester algebra course, published as AMS Open Math Notes.

What his research is about

Most of his papers sit in number theory, and most of those ask how many whole-number or fraction solutions an equation has. Elliptic curves are the main object: equations like y² = x³ + ax + b, which are central to modern number theory.

His arXiv listing shows the thread:

  • The average number of integral points on elliptic curves is bounded (2014), from his senior thesis. "Integral points" are solutions in whole numbers.
  • With Bhargava and Ari Shnidman, papers showing that a positive share of cubic and quartic number fields are "not monogenic" (2020 and 2021), and a paper on which integers are the sum of two rational cubes (2022).
  • "Modularity and effective Mordell I" (2021), on making a famous finiteness theorem about rational points on curves effective, meaning it gives an actual bound you can compute.
  • "Conditional algorithmic Mordell" with Brian Lawrence (2024).

His early papers include two on zeros of L-functions attached to Maass forms, written with Steven J. Miller and others in 2013. L-functions are the family the Riemann zeta function belongs to, so those zeros are the same kind of object the Riemann hypothesis is about.

His role at Anthropic

His website lists Anthropic as his employer.

The first public trace of his work there is Anthropic's August 2026 research post on a smaller result. An unreleased research version of Claude raised the known share of zeta zeros that provably sit on the critical line from 41.6% to 67.2%. The post names Alpöge and Ralph Furman as the Anthropic mathematicians who checked Claude's paper and wrote a short note stating the proof.

Outside Anthropic's posts, he has worked with NYU mathematician Tristan Buckmaster on fluid equations. Terence Tao's November 20 blog post describes a proof by Alpöge and Buckmaster that solutions of the 3D Navier-Stokes equations can blow up in finite time, and a similar proof for the Euler equations with Buckmaster and Coiculescu. Tao says both were written with heavy AI help and come with Lean formalizations. Scientific American also reports that Alpöge posted a Claude-generated solution to the Jacobian conjecture in July.

What Alpöge claimed on November 20

In a statement on his website, he wrote: "I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis." He says the same method proves the Generalized Riemann Hypothesis, a version that covers a whole family of L-functions, and that he expects it to reach the Extended Riemann Hypothesis with more time.

He says he steered the work toward combining two ideas from a simpler setting, and that Claude then outpaced him, especially on automorphic forms on higher-rank groups, an area where he says he has little mastery.

He released only Lean code, in 14 archive files, and no paper. He says the argument is so long that a written version would be unreadable, and that he expects the proof to get much shorter and intends to write it up properly.

How the proof was checked

Lean is a programming language in which a computer checks every step of a proof. On November 21, the Lean and Mathlib communities published a joint statement, endorsed by the Lean FRO and the Mathlib Initiative. It says Alpöge's library is about 17 million lines of code, that the theorem it proves matches Mathlib's vetted statement of the Riemann hypothesis, and that the proof passed Lean's kernel and several independent checkers. Their conclusion: "We have full confidence that Alpöge's proof of the Riemann Hypothesis is correct."

The statement doesn't mention Claude. It vouches that the proof is valid, not that anyone understands it yet.

What other mathematicians say

  • Terence Tao calls it an "(enormous) Lean-only proof," the least digested of the three results in his post. From a first look, he says it seems to use the full power of the Langlands program and arithmetic geometry. He expects understanding it to take time, and notes that there is a human author to talk to. He does not say he has checked it himself.
  • Andrew Sutherland of MIT told Scientific American that he compiled the Lean proof himself and called the day historic.
  • Hector Pasten and Carlo Pagano told Scientific American that, if the proof is correct and understood, it could tie together branches of number theory and open new lines of work.

The Clay prize is not automatic

The Clay Mathematics Institute offers $1 million for a proof of the Riemann hypothesis, as Scientific American notes. Its rules set three conditions before it will even consider a solution: it must be published in a qualifying outlet, at least two years must pass after publication, and it must have general acceptance in the mathematics community.

No paper on the proof has been published yet, so the two-year clock has not started.

What is still not known

  • How the proof works, in human terms. Nobody has published a readable account, and Tao expects digesting it to take time.
  • How much Claude did. Alpöge credits Claude with much of the work near the end, but his statement does not say which Claude model he used or how the work was run.
  • When a paper will appear. Alpöge says he intends to write one. He gave no date.

What is settled is the formal check, which the Lean community says is complete. The open question is the explanation, and Alpöge says that is his next job.

Frequently asked questions

Does Levent Alpöge work at Anthropic?

Yes. His website lists Anthropic as his employer, and Anthropic's August 2026 research post names him as one of its mathematicians.

Did Claude prove the Riemann hypothesis?

Alpöge says he found the proof working with Claude, and that Claude did much of the work near the end. The proof is a Lean library that the Lean and Mathlib communities say is correct, but no written explanation has been published yet.

What is Levent Alpöge known for in mathematics?

Before 2026 his best-known work was in number theory, on counting integral and rational points on curves, including a proof that the average number of integral points on elliptic curves is bounded.

Will Levent Alpöge get the $1 million Clay prize?

Not yet. Clay's rules require publication in a qualifying outlet, two years after publication and general acceptance by mathematicians before a solution is considered.

Sources

  1. Levent Alpöge, personal websiteLevent Alpöge
  2. Statement on the Riemann hypothesisLevent Alpöge
  3. Curriculum vitaeLevent Alpöge
  4. Articles by Levent AlpögearXiv
  5. Claude has improved on a longstanding lower bound for the fraction of zeros of the Riemann zeta function that satisfy the Riemann hypothesisAnthropic
  6. Lean verification of the Riemann HypothesisLean and Mathlib communities
  7. Unforced Euler singularity, unforced Navier-Stokes singularity, and the generalized Riemann hypothesisTerence Tao
  8. Millennium Prize rulesClay Mathematics Institute
  9. Anthropic mathematician proves Riemann hypothesis, most famous problem in all of mathScientific American