Gödel and the logical limits of AI

A team at the University of Helsinki has achieved a breakthrough by having the AI model Claude analyse and summarise 16 of his “workbooks” -- which humans had been struggling through for years -- revealing that Gödel had proved a vast number of previously unknown theorems
If and when that becomes reality — the hype has been building up for a while — the question is how far it can go. Only Gödel may have the answer.
AI Narration
Listen
00:00
2 min readNov 17, 2026 06:15 AM IST First published on: Nov 17, 2026 at 06:15 AM IST
As of September 29, the world, or at least the United States, is in the era of super intelligence. “Artificial” intelligence no longer suffices as a descriptor; words like “super” hold out a promise of transcendence. But if that promise is redeemed, what will its limits be, if any? Can AI unravel every knotty problem preoccupying mathematicians — it has certainly made a start — or be a perfect guide for decision-making? Perhaps there is a limit: Gödel’s incompleteness theorems, among the most important results in mathematical logic, state that no consistent formal system capable of modelling basic arithmetic can prove all truths of arithmetic, or the system’s own consistency. Although LLMs are based not on formal logic but probability and statistics, the latter are themselves formal systems.
What is certainly within the powers of AI, however, is deciphering Gödel’s notebooks. The published work of Kurt Gödel (1906-1978), one of history’s greatest logicians, has been key to fields including the foundations of mathematics and computer science. Then there are his tens of thousands of pages of unpublished work, which includes difficult material written in an obscure shorthand. Now, a team at the University of Helsinki has achieved a breakthrough by having Claude analyse 16 of his “workbooks” — which humans had been struggling through for years — revealing that Gödel had proved a vast number of previously unknown theorems. The task took just a couple of days.
Gödel’s work on self-referential formulas has itself inspired efforts to theorise AI systems that examine and improve their own code autonomously. If and when that becomes reality — the hype has been building up for a while — the question is how far it can go. Only Gödel may have the answer.
KioskNews shows a cleaned-up reading view extracted from the publisher’s page — the original always lives on their site, not ours.