Who Is Levent Alpöge, The Anthropic Mathematician Who Says He Proved The Riemann Hypothesis With Claude?

A Harvard prize-winner, a student of Fields medalist Manjul Bhargava, and the central figure in this year’s most heated AI-mathematics dispute, Levent Alpöge has become the face of AI-assisted mathematics in 2026.

Until this year, Levent Alpöge was known mostly to number theorists. That changed quickly in 2026. In July he used Claude to disprove an 87-year-old conjecture. In September he was at the centre of a public credit dispute with OpenAI over Navier-Stokes. And on November 20, a note on his website claimed a proof of the Riemann Hypothesis, the most famous unsolved problem in mathematics, found working with Anthropic’s Claude.

So who is the 34-year-old mathematician behind these claims?

Early life

Levent Alpöge was born on April 1, 1992, and grew up on Long Island, New York, where he attended Half Hollow Hills High School West in Dix Hills. He holds both American and Turkish citizenship.

His talent showed early, and not only in mathematics. In 2010 he was a finalist in the Intel Science Talent Search, one of the most prestigious science competitions for American high school students, with a computer science project on the algorithmic analysis of MRI scans of blood vessels. His first published paper, on a GPU-accelerated method for detecting blood vessels in medical images, dates from that period. He also has a main-belt asteroid named after him: 25898 Alpöge.

Education: Harvard, Cambridge and Princeton

Alpöge went to Harvard, where he compiled one of the most decorated undergraduate records in recent memory. He graduated in 2014 with a bachelor’s degree in mathematics and a master’s degree in physics, with a perfect 4.0 GPA, summa cum laude. He won the Sophia Freund Prize for the highest GPA at Harvard and the Captain Jonathan Fay Prize for the best senior thesis. He also shared the Mumford Prize for the most outstanding mathematics concentrator with Akhil Mathew, the “close friend akhil” he later thanked in his Jacobian conjecture announcement.

His senior thesis, supervised by Jacob Tsimerman, was titled “The average elliptic curve has few integral points.” It became a paper of the same name and won him the 2015 AMS-MAA-SIAM Morgan Prize, the top American award for research by an undergraduate. Tsimerman, a Fields medalist, later joined OpenAI, so Alpöge’s undergraduate advisor now works at the rival lab.

Alpöge then went to the University of Cambridge as a Churchill Scholar, completing Part III of the Mathematical Tripos in 2015. He moved on to Princeton, where he earned his PhD in 2020 under Manjul Bhargava, who won the Fields Medal in 2014. His dissertation was titled “Points on Curves.”

An academic career in number theory

Alpöge’s research sits in number theory and arithmetic statistics, a field that asks how often equations have solutions and what “typical” behaviour looks like across large families of equations. His papers show that, on average, elliptic curves have a bounded number of integral points, and that the average number of rational points on genus-two curves is bounded too.

His best-known academic result came with Bhargava and Ari Shnidman. In 2022 the trio proved that a positive proportion of whole numbers can be written as the sum of two rational cubes, and that a positive proportion cannot, settling a conjecture of Harold Davenport. Quanta Magazine covered the work, noting that the trio stepped up their collaboration while Alpöge and Shnidman both had Covid-19. In 2025, with Bhargava, Wei Ho and Shnidman, he also published a proof in Inventiones Mathematicae that Hilbert’s tenth problem has a negative answer over the ring of integers of every number field. It was an independent route to a result first established by Peter Koymans and Carlo Pagano.

After his PhD, Alpöge held an NSF Postdoctoral Fellowship affiliated with Columbia and Harvard from 2020, and was a Junior Fellow at Harvard’s Society of Fellows from 2021 to 2025, one of the most selective early-career posts in American academia.

When did Levent Alpöge join Anthropic?

Alpöge joined Anthropic as a member of technical staff after his Harvard fellowship, which ran until 2025. Neither he nor the company has publicly given an exact start date, but he was at Anthropic by the time of his Jacobian conjecture result in July 2026. He has written that the arrival of GPT-4 convinced him that AI was among the most interesting things humans had built, and that he wants to help steer the technology in a good direction.

At Anthropic he has been one of the company’s leading mathematical voices. He and fellow Anthropic mathematician Ralph Furman reviewed Claude’s improvement of a Riemann-related bound from 41% to 67% earlier this year. In October he shared Claude’s independent disproof of the Exponential Time Hypothesis, which the company said was found the same day as OpenAI’s.

The Jacobian conjecture

Alpöge’s first big public moment came on July 20, 2026. He posted a single explicit polynomial map that disproved the Jacobian conjecture, a problem in algebraic geometry formally posed in 1939 and listed by Steve Smale among the most important mathematical problems of the century. He credited Anthropic’s Claude Fable 5 model with the search. The counterexample is a map in three variables whose Jacobian determinant is a nonzero constant but which has no polynomial inverse. Other mathematicians checked it within days.

He followed it with a run of AI-assisted results over the summer, including new Hadamard matrices, a counterexample to the Carathéodory conjecture and a claimed resolution of a problem of Heinz Hopf about the six-dimensional sphere.

The Navier-Stokes proof and the OpenAI controversy

The episode that made Alpöge a household name in AI circles began in September.

For about a year, Alpöge and NYU mathematician Tristan Buckmaster had been working, with heavy help from AI models, on whether fluid equations can “blow up” in finite time. This is the question at the heart of the Navier-Stokes Millennium Prize Problem. Buckmaster has said the collaboration was personal and had no institutional backing from either Anthropic or NYU. On August 15, they proved that the Euler equations, the frictionless cousin of Navier-Stokes, can blow up when a smooth external force is applied, and they did the same for two related equations. After checking the proofs in Lean, they published them in early September, and Terence Tao called the work “a remarkable achievement”.

Alongside the papers, Buckmaster published a statement that turned a mathematical result into a scandal. He accused OpenAI of trying to take credit. He alleged that OpenAI had begun its own Navier-Stokes push after rumours of the pair’s work reached the company, and that in calls OpenAI’s Sébastien Bubeck had floated arrangements Buckmaster rejected. On September 8, OpenAI announced its own Navier-Stokes proof, produced by thousands of AI agents and relying on an external force.

Bubeck called Buckmaster’s claims “false and inflammatory”. He later published text messages to Alpöge to argue that he had proposed a coordinated release in good faith. OpenAI said neither its researchers nor its agents had seen the pair’s work before it was public. The fight soon spread to researchers at both Anthropic and OpenAI, who traded barbs on X.

Despite the feud, Alpöge has been generous to his rivals in public. When OpenAI released more than 300 AI-generated proofs in October, he called it the “most significant moment in mathematical history”.

The story came full circle on November 20. Buckmaster posted a new paper with Alpöge claiming that Navier-Stokes can blow up with no external force at all, the version of the problem many mathematicians consider the real Millennium Prize question. Buckmaster wrote that “after last time” they could not risk waiting.

The Riemann Hypothesis claim

On the same day, a four-page note appeared on Alpöge’s website claiming a proof of the Riemann Hypothesis and the Generalized Riemann Hypothesis. The note is written in the first person but unsigned, and Alpöge has not shared it publicly. It credits Claude as the main technical collaborator and says the argument has been fully formalized in Lean. Only the Lean code has been released, with no human-readable paper. According to its timeline, the work took just 41 days, from October 6 to November 16.

The claim has not been verified. Lean experts are checking whether the code compiles and whether its final theorem really states the hypothesis, and mathematicians have reacted with a mix of excitement and caution. If it holds up, Alpöge will have played a central part in settling two Millennium Prize Problems in a single year. Either way, few mathematicians have done more to show what happens when a strong human mathematician works closely with a frontier AI model.

Levent Alpöge: quick facts

How old is Levent Alpöge? He was born on April 1, 1992, which makes him 34.

Where is Levent Alpöge from? He grew up on Long Island, New York, and holds American and Turkish citizenship.

Where did Levent Alpöge study? Harvard (AB in mathematics and AM in physics, 2014), Cambridge (MASt, 2015, as a Churchill Scholar) and Princeton (PhD, 2020, under Manjul Bhargava).

What is Levent Alpöge known for? In academia, for work on rational and integral points on curves, sums of two cubes and Hilbert’s tenth problem. In 2026, for AI-assisted results including a disproof of the Jacobian conjecture, the Navier-Stokes blow-up work with Tristan Buckmaster, and a claimed proof of the Riemann Hypothesis.

Where does Levent Alpöge work? He is a member of technical staff at Anthropic, the company behind Claude.