Skip to content
Technology
ResearchAnalysis

Mathematicians Confirm Alpöge's Claude-Assisted Riemann Proof

Lean's formal checkers confirm Levent Alpöge's Claude-assisted proof of the Riemann Hypothesis is correct.

Edited by Kelly Dennis4 min read · 997 words · 7 sources
Andrew Sutherland of MIT, who reran Levent Alpöge's Riemann Hypothesis proof through independent checking software
“Seventeen million lines of code just answered a question that waited 167 years. Does a proof still unread by any single mathematician settle a hypothesis that outlasted two world wars?”

Levent Alpöge said Friday, Nov. 20, that he and Claude had proved the Riemann Hypothesis. The question about prime numbers has stood open since 18594.

One day later, the people who build Lean said they believe him. Lean is software that checks a math proof line by line, the way a calculator checks a sum1.

Four Checkers, One Answer

Andrew Sutherland teaches math at MIT. He pulled Alpöge's code out of Anthropic's own systems and ran it on his own machines2.

A kernel is the part of Lean that does the actual checking. Sutherland fed Alpöge's proof to nanoda, a kernel built outside Anthropic in the programming language Rust. It finished in about 28 minutes5.

Three more checks, built by other people in other code, read the same proof. Con-leche worked through all 1.47 million of its declarations. A tool called lean4lean replayed every one of its 175,187 modules, and Lean's own kernel did the same, with zero rejections5.

Sutherland called the result clean. "The Lean proof has been compiled and independently rechecked by the Lean kernel with only the three standard axioms," he said. "Honestly the chance that anything is fishy here is approximately zero"2.

atulit@atulit_gaur

SITUATION UPDATED: The Riemann Hypothesis proof has passed the independent kernel check. MIT's Andrew Sutherland exported Alpöge and Claude's full proof and ran it through separately written kernels. nanoda (Rust) accepted it in about 28 minutes. con-leche accepted all 1.47… https://t.co/4NfyhYHtuT

Open on X (opens a new tab)

The post that first reported the kernel check by name: nanoda, con-leche and lean4lean each rechecked Alpöge's proof overnight and agreed.Post by atulit (@atulit_gaur) on X, Nov. 21, 2026

The Lean FRO and the Mathlib Initiative, the two groups that steward the software, signed a joint statement Saturday, Nov. 21. They wrote plainly that Alpöge's proof holds1.

A Clean Compile Still Needs a Reader

Kevin Buzzard teaches math at Imperial College London. He started the project that taught Lean most of the number theory this proof now leans on.

A clean kernel check tells you the logic follows its own rules. It skips the reason the theorem is true, and Buzzard said so plainly. "At the minute, we don't even know why the theorem is true," he told Al Jazeera3.

Alpöge's proof also reaches past the question Riemann asked in 1859. It covers the Generalized Riemann Hypothesis, a wider version that applies the same idea to more number-counting functions than the original touched.

One rival claim complicates the picture further. A Stanford researcher named Alessandro Masullo posted his own attempt at the same hypothesis one day earlier4. He built it entirely by hand. Other mathematicians still have it under review, a separate story Ai Lately is tracking on its own clock.

The Huxley Building at Imperial College London, home to the department where Kevin Buzzard leads Lean's number theory formalization
The Huxley Building at Imperial College London houses the math department where Kevin Buzzard leads Lean's number theory project.Omar Shahpo / Wikimedia Commons, CC BY-SA 4.0

Seventeen million lines of code is the size Alpöge's proof reached1. Buzzard put that size into pages. "It's 17 million lines of Lean code, and this might correspond to tens of thousands of pages of mathematics," he said3.

Alpöge's own pages still await a first full reading. Buzzard described the current state bluntly. "At the minute, the proof is just millions of lines of gobbledygook written in a strange programming language," he said. He added that "there's people all over the world who are translating parts of this programming language back into English"3.

Four Checkers Confirmed Alpöge's 17-Million-Line Proof in a Day

Key moments in the Riemann Hypothesis proof's check, Nov. 20 to Nov. 21, 2026

Each pin is one dated moment: the blue pins are the claim and the open question around it, and the magenta pins are the independent checks that confirmed it.

Data through Nov. 21, 2026.

Sources: Lean community blog [1]; Al Jazeera [3]; X (@atulit_gaur) [5]. Chart by Kelly Dennis.

The numbers behind this chart
DateEvent
Nov. 20, 2026Alpöge posts his proof. Levent Alpöge releases a 17-million-line Lean proof built with Claude
Nov. 21, 2026A kernel confirms it. Nanoda, written in Rust, rechecks the proof in about 28 minutes
Nov. 21, 2026Lean signs off. The Lean FRO and the Mathlib Initiative post a joint statement of full confidence
Nov. 21, 2026Buzzard still unsure. Kevin Buzzard tells Al Jazeera mathematicians have yet to find the reason the theorem is true

Alpöge described the work itself as a blur. Claude itself, he said, "flew, and danced, and sparkled near that finish line." He called the whole experience a "stream-of-consciousness jaunt through the experience of finding all the pieces of the argument through incredible confusion with Claude"4.

A Prime Number Rule Guards Your Password

The Riemann Hypothesis pins down how evenly prime numbers spread out among all numbers.

Cryptographers leaned on that guess for decades, years before anyone built a check for it. Widely used encryption methods run faster because they assume the hypothesis is true. The hypothesis's assumed truth was a bet the field placed long before anyone could check it3.

A detail view of the Riemann zeta function, showing two of its zeros sitting on the critical line the hypothesis describes
A detail of the Riemann zeta function, with two of its zeros on the critical line the hypothesis says every zero sits on.Empetrisor / Wikimedia Commons, CC BY-SA 4.0

Buzzard framed what a verified answer buys the rest of us. "We can sleep slightly safer in our beds at night because it is true," he said3.

Abhishek Saha teaches math at Queen Mary University of London. He expected this result years away. "My own intuition was that we were still several years away from a proof of the Riemann hypothesis. This claimed proof, if correct, shatters my model of reality," he said. He added that Alpöge "should get a Fields Medal for this"3.

Loud Claim, Quiet Confirmation

A Hacker News thread on Alpöge's original claim drew a crowd on Nov. 206. The thread on Saturday's joint statement drew about 13 percent as much of one7.

The Confirmation Drew 13 Percent of the Claim's Hacker News Points

Points on each Hacker News thread, read Nov. 21, 2026

Each bar is one thread's points: the longer magenta bar is the thread on Alpöge's original claim, and the shorter blue bar is the thread on the confirmation.

Hacker News points keep moving while a thread stays active, so both counts are one Saturday reading.

Data through Nov. 21, 2026.

Sources: Hacker News [6]; Hacker News [7]. Chart by Kelly Dennis.

The numbers behind this chart
ItemValue
Original claim2,808
Confirmation362

Attention chased the announcement far more than the proof. The X post carrying Sutherland's kernel result still pulled 8,347 likes and 115 replies by Saturday afternoon5. Plenty of people watched the quiet part too.

What to do

Treat "checked" and "understood" as two separate facts about this proof, and track them separately. The checking finished this weekend. Understanding it, by Buzzard's own estimate, still needs mathematicians a week or two with the Lean code in front of them3.

Give that same week to the plain-language paper, once it lands, over the next viral post. It will tell working mathematicians what new idea actually cracked the problem.

By the numbers

  • 17 million: lines of Lean code in Alpöge's proof1.
  • 28 minutes: how long the independent kernel nanoda took to recheck it5.
  • 175,187: Lean modules that lean4lean rechecked with zero rejections5.
  • 1.47 million: declarations the checker con-leche worked through and accepted5.
  • 2,808: points the original claim earned on Hacker News6.
  • 362: the confirmation thread's own Hacker News total, about 13 percent as many7.
  • 8,347: likes on the X post reporting the kernel check5.
  • 115: replies to that same post5.

What to watch

Buzzard's promised week or two of translation work matters most. Turning Lean code back into ordinary mathematical English will show one thing clearly3. It will say whether the proof holds a genuinely new idea, or mostly automates known methods.

Cryptography deserves a watch too. Security researchers assumed the Riemann Hypothesis for decades on faith. Any method that quietly rested on that same assumption deserves a second look now that the assumption has a check behind it.

Sources

  1. Lean and Mathlib communities, "Lean verification of the Riemann Hypothesis," Lean community blog, Nov. 21, 2026, https://leanprover-community.github.io/blog/posts/riemann-hypothesis/
  2. Clara Moskowitz, "The AI Riemann hypothesis proof is true, independent mathematicians attest," Scientific American, Nov. 20, 2026, https://www.scientificamerican.com/article/the-ai-riemann-hypothesis-proof-is-almost-certainly-true-independent-mathematicians-attest/
  3. Heba Habib, "Riemann hypothesis: Has the world's most famous maths problem been cracked?," Al Jazeera, Nov. 21, 2026, https://www.aljazeera.com/news/2026/11/21/riemann-hypothesis-has-the-worlds-most-famous-maths-problem-been-cracked
  4. Tim Fernholz, "AI claimed a solution to the 167-year-old Riemann hypothesis — and a human did, too," TechCrunch, Nov. 20, 2026, https://techcrunch.com/2026/11/20/ai-claimed-a-solution-to-the-167-year-old-riemann-hypothesis-and-a-human-did-too/
  5. atulit, post on X (@atulit_gaur), Nov. 21, 2026, https://x.com/atulit_gaur/status/2124028966373916967
  6. "Levent Alpöge claims proof of the Generalized Riemann Hypothesis," Hacker News, Nov. 20, 2026, https://news.ycombinator.com/item?id=50574968
  7. "Lean Verification of the Riemann Hypothesis," Hacker News, Nov. 21, 2026, https://news.ycombinator.com/item?id=50589618

Header photo: Ken Richardson / Wikimedia Commons, CC BY-SA 4.0

Cite this piece

Kelly Dennis, "Mathematicians Confirm Alpöge's Claude-Assisted Riemann Proof," Ai Lately, Nov 21, 2026, https://ailately.com/articles/mathematicians-confirm-alpoge-riemann-proof

Tags: Riemann Hypothesis · Levent Alpöge · Claude · Lean · Mathlib

Related, lately

More

Keyboard

j k
Move through a list
Enter
Open the selected piece
/
Search, and browse the columns
g then r
Right Now
g then a
Articles
g then s
The Signal
g then o
Opinion
g then b
Analysis
g then p
People
?
This sheet
Esc
Close

AI Right Now is free with an account.The Citation Audit is free with an account.Skills are free with an account.Tutorials are free with an account.You have read your 3 free articles.

Every article and comic, AI Right Now, the Citation Audit, Skills and Tutorials.

Join Ai Lately for free

Join with Google or your email. Already a member? Sign in

Press anywhere else for the home page.