# Claude Pulled 663 Theorems From Gödel's Notebooks; a Proof Checker Found About 50 False

Mohit Tyagi · 2026-11-15 · Research

[https://timesof.ai/2026/11/claude-godel-notebooks-663-theorem-candidates-coq-checks](https://timesof.ai/2026/11/claude-godel-notebooks-663-theorem-candidates-coq-checks)

> Jan von Plato, who leads a University of Helsinki project on 16 workbooks Kurt Gödel wrote in the early 1940s, told The Guardian that Claude produced a 650-page explanation of the 1,009 pages in a couple of days, after years of progress at about a page a week. Asked to extract every theorem Gödel had proved, sketched or had an idea for, Claude found 663; checked in the Coq proof assistant, about 400 were verified as stated, about 200 were repaired and about 50 were false, von Plato said. A Claude-reconstructed proof of the independence of the axiom of choice passed Coq, with one definition Gödel got wrong. An outside logician who has not seen it called it remarkable if it stands up.

The useful part of this story is the ratio, not the headline. A logician says Claude read the unpublished notebooks of Kurt Gödel faster than any human team could and surfaced hundreds of theorems nobody knew he had proved. He also says a proof checker rejected about one in thirteen of them. The model made the archive legible; the checker decided what was true.

## What von Plato says Claude did

Since 2017, **Jan von Plato** has led a University of Helsinki project on **16** workbooks Gödel wrote in the early 1940s: **1,009** pages on mathematics and logic, much of it in Gabelsberger, a German shorthand that was obsolete even then, The Guardian reported on **November 13**. Von Plato told the paper he is "one of only six people who can read it in the world," and that progress had run at about a page a week. In September, after computer scientists at a conference suggested it, he gave Claude his transcriptions, six of his own books on Gödel and other secondary literature.

"In a couple of days, I got to see a 650-page explanation of the workbooks. It had to be controlled page for page, but it's basically a work that is impossible to do for human beings," he said.

He then asked Claude to extract every theorem for which Gödel had a proof, a sketch of one or an idea for one. Claude found **663**. When those were put into **Coq**, a proof assistant that checks each step of a proof: "About 400 were verified as stated, about 200 were repaired, meaning that perhaps the statement of the theorem had to be made more precise, or some assumption had to be added. About 50 were false." Von Plato noted that Gödel himself sometimes wrote "falsch!" in his margins.

## The axiom of choice

The first question von Plato put to Claude was historical. In 1966, after Paul Cohen won the Fields medal for showing the independence of the continuum hypothesis and the axiom of choice, Gödel said he had proved the result in the 1940s but could not reconstruct the proof. Claude's answer, according to von Plato, was "yes, almost": the workbooks contain the architecture of a proof with one definition Gödel got wrong, having had the right one two years earlier. Von Plato said the reconstructed proof, presented by Claude, passed Coq.

**Joel David Hamkins**, a logician at the University of Notre Dame who has not seen the reconstruction, told The Guardian that if it stands up "it is remarkable and historically interesting. Although how much is Claude and how much is Gödel is a job for the historians." He added: "You don't get credit for work you don't tell anyone about."

The workflow is the transferable part, and it matches what other teams report: [DeepMind's machine-checked Signal proofs](https://timesof.ai/2026/11/deepmind-alphaproof-code-formally-verifies-signal-protocol-lean) also put a formal checker between model output and any claim of correctness. Here the checker sorted hundreds of candidates at a pace no page-a-week reading could match.

## Limits

- Every figure is von Plato's account to The Guardian. We found no published paper or data release to check the 663, 400, 200 or 50 counts against.
- The Guardian does not say who translated the theorem statements into Coq or how. A Coq check verifies the formal statement, so a mistranslation from Gödel's notes would not be caught by it.
- The reconstructed axiom-of-choice proof uses a corrected definition, so it is not Gödel's proof as written. The one outside expert quoted had not seen it.
- Claude's model version is not given.

## Sources

- [The Guardian: 'Ahead of his time': AI helps unlock discoveries hidden in notebooks of man who transformed maths (November 13, 2026)](https://www.theguardian.com/science/2026/nov/13/kurt-godel-notebooks-ai-helps-unlock-hidden-discoveries-man-who-transformed-maths)
