Kurt Gödel’s 1,009 pages of secret notebooks, decoded with artificial intelligence, have revealed hundreds of unknown theorems and 14 unexplored themes in mathematics and logic. Finnish logician Jan von Plato of the University of Helsinki asked Anthropic’s Claude to read the workbooks Gödel filled in the early 1940s, and within days it returned a 650-page explanation. Among the findings is an almost complete argument for the independence of the axiom of choice. Von Plato says the workbooks also contain dozens of anticipations of results not known until the 1970s and 1980s, and that Gödel was far ahead of his time.
Why AI decoded Kurt Gödel’s 1,009 pages of unpublished notebooks matter to mathematics and modern computing
According to The Guardian, the Austrian logician Kurt Gödel was one of the most original thinkers of the 20th century. His two incompleteness theorems, published in 1931, transformed mathematics and paved the way for the digital age. They showed that in every mathematical system there are true statements that can be neither proved nor disproved, an insight that inspired Alan Turing’s 1936 paper, which began the theory of modern computation.Gödel fled Austria under the Nazis in 1940, settled in Princeton, New Jersey, and became a close friend of Albert Einstein at the Institute for Advanced Study. After his landmark results of the 1930s, he published only a few papers, but his unpublished work runs to tens of thousands of pages. Among it are 16 notebooks written in the early 1940s, known as the workbooks, which hold 1,009 pages of dense writing on mathematics and logic. Much of it is in Gabelsberger, a German shorthand that was already obsolete when Gödel used it.
Jan von Plato has been decoding Kurt Gödel’s mysterious notebooks since 2017
Since 2017, von Plato has led a project at the University of Helsinki to understand the workbooks. He secured a grant of €2.5 million (£2.1 million) from the European Research Council, taught himself the shorthand and assembled a small team. Even so, progress on the mathematics was painfully slow, at about a page a week.The turning point came in August, at a conference in France on artificial intelligence and theorem proving, where computer scientists suggested he enlist Claude. Back in Helsinki in September, he showed the model his transcriptions of the workbooks, six books he had written on Gödel and other scientific literature.
Claude read 1,009 pages of Gödel’s notebooks and produced a 650-page digest in days
Within a couple of days, Claude had produced a comprehensive digest of the entire material, which revealed that Gödel had proved hundreds of theorems that no one knew about. Von Plato told The Guardian that the explanation had to be checked page by page, but that it is basically a work impossible for human beings to produce. People read one page at a time and tend to forget things, he said, whereas the model can connect all the pieces across a thousand pages and see Gödel’s writings as one piece. He then asked Claude to extract every theorem for which Gödel had a proof, a sketched proof or an idea for a proof. It found 663.
Image: Wikipedia
Why researchers combined AI with a strict system to check Gödel’s work
Von Plato put those results into Coq, a program that verifies the validity of mathematical proofs. About 400 were verified as stated. About 200 were repaired, meaning the statement had to be made more precise or an assumption added. About 50 were false, which is unsurprising, because Gödel himself sometimes marked passages in the margins of his notes as incorrect or bad.The AI for Developing Countries Forum argued in its analysis that this combination is the real breakthrough. The model acted as a bridge that connected references across thousands of pages, while Coq enforced strict logical validity. The Forum described this hybrid of a language model and a formal proof checker as a template for recovering scientific knowledge without diluting mathematical standards.
Gödel came one step before proving the independence of the axiom of choice, before Paul Cohen
In 1966, the American mathematician Paul Cohen won the Fields Medal, the highest honour in mathematics, for showing the independence of two famous statements: the continuum hypothesis and the axiom of choice. Gödel reacted by saying he had already proved the result in the 1940s. Yet when his peers pushed him for a proof, he could not reconstruct one from his notes.The first task von Plato gave Claude was to check whether the workbooks held a proof of the independence of the axiom of choice. The answer was almost. In the same interview, von Plato said Gödel was “just one step away”. One definition was wrong, but Gödel had written the right one two years earlier, so he would probably have reached the correct version within months. When Claude presented the reconstructed proof, von Plato ran it through Coq, and it passed. He described the architecture of the proof as wonderful, because it makes all the set theory results follow a uniform pattern.In the same report, Joel David Hamkins, a professor of logic at the University of Notre Dame in Indiana, said he has not seen Claude’s reconstruction. If it stands up, he said, it is remarkable and historically interesting, though how much is Claude and how much is Gödel is a job for historians. Resemblance to Cohen’s proof does not mean Gödel had it, he cautioned. It shows he was on the path, but credit does not go to work you tell nobody about.
The 14 unexplored themes in logic and dozens of early anticipations
Von Plato said he was flabbergasted by what the digest revealed. The workbooks contain 14 themes in logic that have not been pursued, and Claude puts it as more than a dozen themes in logic and the foundations of mathematics that have gone unfound in the 80-plus years since they were written down.Gödel, von Plato said, starts from very clear ideas and then proceeds methodically, step after step, rather than playing with formulas and hoping to arrive somewhere. He is sure that continued study will reveal further treasures, and he plans to explore Gödel’s physics and cosmology notebooks, which he calls unknown territory, once his current logic projects are finished.
Why the Gödel decoding method could help recover neglected archives in developing countries
The AI for Developing Countries Forum argued that the workflow could matter well beyond mathematics. Across Latin America, Africa, the Middle East and South Asia, it noted, millions of pages of scientific, medical and legal manuscripts sit unstudied in national archives. Many are written in obsolete scripts, such as classical Arabic variants, Ottoman Turkish, Ge’ez, pre-modern Tamil or Mesoamerican scripts, and trained paleographers are scarce. If the approach can be adapted to low-cost, openly available models, the Forum said, developing nations would no longer depend only on well-funded Western universities to decipher their own records.It also stressed that the method rested on deep human expertise, since years of painstaking transcription came before the model could synthesise anything. It suggested watching for the peer-reviewed release of the full synthesis and the Coq files, for von Plato’s move to the physics and cosmology notebooks, and for open-source projects that pair language models with low-resource scripts.