AI Deciphers 1,009 Pages of Kurt Gödel's Secret Notes, Hundreds of Theorems Revealed
Baca dalam 60 detik
- Model AI Claude berhasil mengurai buku catatan Kurt Gödel yang ditulis dengan stenografi usang, mengungkap ratusan teorema yang belum pernah diketahui.
- Penemuan ini membuktikan Gödel telah membuktikan independensi aksioma pilihan puluhan tahun sebelum Paul Cohen meraih Fields Medal.
- Kolaborasi AI dan sejarah matematika membuka jalan bagi eksplorasi naskah-naskah kuno lainnya, termasuk catatan fisika dan kosmologi Gödel.

Kurt Gödel, the Austrian logician who shook the foundations of mathematics in 1931, left behind a legacy of 1,009 pages of notes written in the archaic German shorthand Gabelsberger. For nearly eight decades, the contents of those notes were almost unreadable—until the artificial intelligence model Claude helped reveal hundreds of theorems never before known to the public. The findings not only change the understanding of the history of mathematics but also show how AI can serve as a surgical tool for scientific manuscripts long considered impossible to decipher.
The ambitious project was led by Jan von Plato, a logician at the University of Helsinki, who since 2017 had sought to understand 16 of Gödel's notebooks from the early 1940s. Progress had been very slow: only about one page per week. Von Plato, one of only six people in the world able to read Gabelsberger shorthand, said he was astonished when Claude produced a comprehensive 650-page summary within days. “I was stunned,” he said. “The notebooks contain not only dozens of anticipations of results only recognized in the '70s and '80s, but also 14 themes of logic that had never been traced. Gödel was far ahead of his time.”
Gödel is known for two incompleteness theorems proving that true statements exist that cannot be proved or disproved within any mathematical system. That work inspired Alan Turing and laid the foundations of modern computing. Yet after the 1930s, Gödel published almost nothing. His private notes—now held at the Institute for Advanced Study, Princeton—became a hidden treasure. Von Plato, facing mandatory retirement, decided to learn the shorthand and received a €2.5 million grant from the European Research Council.
The collaboration with Claude began at a conference on AI and theorem proving in France last August. Von Plato fed transcriptions of the notes and six of his own books into the model. As a result, Claude was able to connect the threads across a thousand pages that are difficult for humans to hold in memory. “AI can see all of Gödel's writing as a single whole,” von Plato said. He then verified the reconstructed proof of the independence of the axiom of choice using Coq, a mathematical proof assistant, and the result passed the test.
Even so, not everyone is fully convinced. Joel David Hamkins, a professor of logic at the University of Notre Dame, Indiana, called the findings “amazing and historically interesting” if proven valid. But he cautioned, “how much is Claude's contribution and how much is Gödel's is a task for historians.” Hamkins added that similarity to Cohen's proof does not mean Gödel already had a complete proof—at least he was already on the right track.
For the academic world, this case opens a new chapter in the use of AI for archival research. Von Plato plans to explore Gödel's untouched physics and cosmology notes. Meanwhile, in Indonesia, similar methods could potentially be applied to ancient Nusantara manuscripts held in libraries and museums, from lontar to Arabic-Malay manuscripts, which have so far been constrained by a shortage of paleographers. If AI can read a dead shorthand, it is not impossible that it could also accelerate the digitization and understanding of local intellectual heritage.
The question now is: how far can AI be trusted to reconstruct the thinking of a genius, and who has the right to claim authority over findings produced by a machine? The history of mathematics may need to rewrite some of its chapters—with a footnote about the role of algorithms.



