ResearchNews
Claude surfaces 663 theorems in Gödel's notebooks
Helsinki logician Jan von Plato used Claude to work through Kurt Gödel's 1,009 pages of shorthand notebooks in a few days, after his team had managed about a page a week. According to the Coq proof checker, roughly 400 of the 663 theorem candidates held as stated.
· 3 min read

What happened
Since 2017 Jan von Plato, a logician at the University of Helsinki, has led a project on 16 notebooks Kurt Gödel wrote in the early 1940s (The Guardian). Most of the 1,009 pages are in Gabelsberger shorthand, already obsolete in Gödel's day; von Plato says only six people in the world can read it. Understanding the mathematics went at roughly a page a week.
In September von Plato gave Claude his transcriptions, six of his own books on Gödel and other 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 told the Guardian.
An almost complete proof on the axiom of choice
The first question was the independence of the axiom of choice. Paul Cohen won the Fields Medal for it in 1966, and Gödel then claimed he had proved it in the 1940s but could not reconstruct the proof. Claude's answer: yes, almost. Von Plato says Gödel was one step away, with one definition wrong, although he had used the right one two years earlier. The reconstructed proof passed Coq.
663 theorems, checked in Coq
Von Plato then asked Claude to list every theorem for which Gödel had a proof, a sketch or an idea for one. It found 663. According to von Plato, the Coq checks came out as follows:
- about 400 held exactly as stated;
- about 200 needed repair, such as a more precise statement or an added assumption;
- about 50 were false. Gödel himself sometimes wrote "falsch!" in the margins.
Von Plato says the notebooks anticipate dozens of results not known until the 1970s and 1980s, and contain 14 themes in logic nobody has pursued since.
How solid is it?
Caution is warranted. Every figure comes from one interview, and the 650-page digest, the theorem list and the Coq files had not been published as of November 14 (MACCUS). The three rounded counts add up to 650, not 663. It is also unclear who wrote the formal statements. That matters, because a proof checker only confirms that a proof matches its formal statement, not that the statement faithfully captures Gödel's idea.
Joel David Hamkins, a logic professor at Notre Dame, said that if it stands up it is remarkable and historically interesting, but how much is Claude and how much is Gödel is a job for historians. Resembling Cohen's proof, he added, does not mean Gödel had it.
Why it matters
The pattern matches DeepMind's Signal verification work (AlphaProof Code): a model produces a large set of candidates fast, and a formal checker filters them one by one. The bottleneck moves from reading to precise statements, checking against the source and deciding credit. The next step would be releasing the Coq files and theorem list; von Plato plans to move on to Gödel's physics and cosmology notebooks.
Key takeaways
- 1,009 pages of obsolete shorthand, a 650-page AI digest in a few days.
- 663 theorem candidates: about 400 held, 200 repaired, 50 false in Coq.
- The details are not public yet, and there is no independent check so far.
Produced with the help of AI, reviewed by an editor. How we make our articles
Reviewed and edited by the Neural Daily desk. About


