AI Deciphers Kurt Gödel's Notebooks
An AI model helped decode 16 long-inaccessible notebooks written by mathematician Kurt Gödel, recovering a nearly complete proof related to the axiom of choice.
Updated on Nov. 21, 2026 in Artificial Intelligence

Live Poll
Do you think AI should be used to speed up scientific research?
Researchers at the University of Helsinki used an AI model to decipher 16 notebooks written by Austrian mathematician and logician Kurt Gödel, which had remained largely inaccessible since the 1940s. The model produced a 650-page analysis within two days and identified an almost complete proof related to the axiom of choice.
Why it matters
Gödel's manuscripts were written in Gabelsberger shorthand, a stenography system now largely out of use, which made them exceptionally hard to study. The project suggests AI tools may open up historical scientific archives that would otherwise take years to read.
The AI extracted 663 mathematical statements from the notebooks, with about 400 transferred unmodified into the Coq proof verification system, while roughly 200 need clarification and 50 were incorrect. One step in the recovered axiom of choice proof is still missing because of an error in the definitions.
The players
Kurt Gödel
Kurt Gödel was an Austrian mathematician and logician whose work on incompleteness and set theory shaped modern mathematical logic.
Jan von Plato
Jan von Plato is a researcher at the University of Helsinki who studies the history and foundations of mathematics and led the effort to decipher the notebooks.
The details
Jan von Plato of the University of Helsinki submitted his transcriptions of the notebooks to an AI model in September, and the model returned its analysis within two days. The research team asked the model Claude to extract every mathematical theorem from the material, and von Plato then transferred the reconstructed axiom of choice proof into a proof verification system.
Timeline
1940s: Gödel's notebooks became largely inaccessible to researchers.
September 2026: Von Plato submitted transcriptions to an AI model.
- 2026-11-21
Findings were published by bgnes.com.
Big Picture
The recovered proof concerns the axiom of choice, the same question Paul Cohen resolved with his 1966 Fields Medal work on its independence. Gödel's earlier notes now add to that line of set theory research, which AI-assisted archive work is making newly accessible.
Researchers working with handwritten or archaic archives may find AI transcription and analysis far faster than manual reading. The same approach could make historical scientific records searchable without years of specialist effort.
The takeaway
AI tools can compress years of archival reading into days, though their output still needs expert checking before it is trusted. Verification steps like the Coq system remain essential for separating correct results from errors.
Further reading
For more on how these tools are being applied to research, see our Artificial Intelligence coverage.
Source note: This article includes information reported by BGNES: Breaking News, Latest News and Videos.
Live Poll
Do you think AI should be used to speed up scientific research?



