AI unlocks Gödel's notebooks, said to contain hundreds of proved theorems
Points and comments are a snapshot, not live.
AI model Claude deciphered Gödel's notebooks, revealing hundreds of unproved theorems.
Jan von Plato at Helsinki University led a project to interpret 16 notebooks (1,009 pages) written by Kurt Gödel in the 1940s, mostly in Gabelsberger shorthand. Progress was about a page per week until September 2026, when von Plato used Anthropic's Claude AI model. Within days, Claude produced a 650-page digest, revealing Gödel proved hundreds of theorems that were unknown, including anticipations of results from the 1970s-80s and 14 unexplored themes in logic. Claude found 663 theorems, of which about 400 were verified as stated, about 200 repaired, and about 50 false. Claude reconstructed a proof that Gödel was one step away from proving the independence of the axiom of choice decades before Paul Cohen won the Fields medal for it in 1966.
What commenters are saying
Commenters were excited by the discovery, calling Gödel the most important logician in history and noting that von Plato is the world's leading expert on Gödel's shorthand. A link to von Plato's slides and the research group's site was shared. Some commenters debated whether this is an example of someone making themselves obsolete with AI, or keeping themselves relevant by using AI as a tool. One camp argued von Plato's value as an interpreter of the shorthand evaporates now that the workbooks are accessible, while others countered that his value was always his mathematical ability, not being a gatekeeper, and that he used AI to achieve a historic result.