THE ESSENTIALS IN 30 SECONDS

  • University of Helsinki logician Jan von Plato used Claude on 16 Gödel workbooks from the early 1940s, about 1,009 pages largely in Gabelsberger shorthand.
  • He says Claude extracted 663 theorems; checked in Coq, about 400 held as stated, about 200 needed repair and about 50 were false.
  • Von Plato says Claude also reconstructed a near-complete proof linked to the independence of the axiom of choice; no paper or peer review has been reported.

Kurt Gödel published little after the 1930s but left thousands of pages of private notes. According to a report by Alex Bellos in The Guardian on 13 November, logician Jan von Plato used Anthropic's Claude to work through 16 workbooks Gödel filled in the early 1940s, about 1,009 pages written largely in Gabelsberger, a German shorthand that was already out of use in his day.

From a page a week to a full digest

Von Plato has led the project since 2017 with a €2.5 million European Research Council grant. He told the newspaper that he is one of only six people able to read the shorthand, and that his transcription had advanced at roughly a page a week. After computer scientists met at a conference in France suggested trying Claude, he supplied the model with his transcriptions, six of his own books on Gödel and other secondary literature. Within a couple of days, he says, it produced a digest of about 650 pages, which he then had to check page by page.

Von Plato says Claude identified 663 theorems accompanied by a proof, a sketch or a proof idea. He ran them through the Coq proof assistant: roughly 400 held as stated, about 200 needed a more precise statement or an added assumption, and around 50 turned out to be false.

A claimed lost proof

The most striking claim concerns the axiom of choice. Paul Cohen received the 1966 Fields Medal for proving that the axiom of choice and the continuum hypothesis are independent of the standard axioms of set theory. Gödel said he had found such a result in the 1940s but could not reconstruct it when asked. Von Plato says Claude rebuilt a nearly complete argument from the notebooks, held back by one faulty definition that Gödel had stated correctly two years earlier, and that the reconstruction passed Coq.

What remains unverified

These are von Plato's own accounts. The Guardian does not report a paper, preprint or data release, so other logicians cannot yet examine the extraction, and the Claude version used was not given. Joel David Hamkins of the University of Notre Dame, who had not seen the reconstruction, called it potentially remarkable but said separating Claude's contribution from Gödel's is a job for historians, adding that "you don't get credit for work you don't tell anyone about." A Coq check confirms that formal statements are valid; it does not establish what Gödel actually knew. The Helsinki project lists exploiting Gödel's unpublished discoveries, such as his proposed independence proof for the continuum hypothesis, among its stated goals.