AI Deciphers 370-Year-Old Cipher in Minutes

Claude Fable 5.1 solved a 370-year-old cipher in forty-four minutes.

Claude Fable 5.1's accomplishment

The single, concrete report at the center of this item is succinct: Claude Fable 5.1 cracked a cipher that has survived for roughly 370 years, and it did so in forty-four minutes. That timing and outcome are the raw facts the author offers; they are presented as a discrete event rather than part of a broader procession of similar wins.

The 370-year-old cipher

The puzzle in question is described only by its age — approximately 370 years — and by the fact that it had resisted solution until Claude Fable 5.1’s effort. The source does not identify the cipher by name, origin, or language; it presents age and persistence as the primary signals of the object’s historical significance.

Forty-four minutes: what the timing tells us

The elapsed time — forty-four minutes — is given as the measurable performance metric. Standing alone, that duration emphasizes two facts the source implies: first, that the exercise completed quickly when measured against centuries of silence; and second, that speed was a notable attribute of the outcome, worth reporting alongside the fact of solution.

AIs doing mathematics: search and testing

The author links this success to an argument previously made: “It’s good at things that involve lots of searching and testing.” That sentence frames the win as an instance of a broader behavioural profile for contemporary artificial intelligences. The source thus offers both the isolated accomplishment and an interpretation: solving long-standing, search‑heavy mathematical or cryptanalytic problems aligns with how some AIs operate.

What this means for cryptographers, historians, and the public

• Cryptographers and technical analysts will likely note the concrete datum — a complex, centuries-old cipher solved in forty-four minutes — and treat it as evidence of a capability pattern the author has already discussed: efficacy on tasks dominated by search and testing.
• Historians and custodians of archival puzzles will see that an archival problem described only by age and persistence has been resolved by an AI tool; the source’s emphasis on age frames the solution as culturally and temporally significant even though no further contextual details are provided.
• The public and nontechnical readers are offered a compact narrative: a modern AI, identified by product name and version, delivered an outcome measured in minutes for a problem measured in centuries. The contrast is the reporting’s central rhetorical device.

The original item is brief, timestamped at September 9, 2026 at 7:08 AM, and bears a modest reader engagement marker of “4 Comments.” Beyond the headline fact — Claude Fable 5.1 plus forty-four minutes plus a 370‑year span — the source leaves its readers with a tightly drawn pair of claims: the solved cipher itself and the author’s broader observation that AIs excel at search‑and‑test problems.

Readers can follow the single, specific thread offered here: a named model and version achieved a named result in a named interval of time, and the author interprets that result as consistent with earlier statements about AI capability profiles. That compact factual lattice is the story’s entirety and its strongest claim.

Is this a solitary demonstration or an exemplar of a repeatable pattern? The item does not say. It does, however, make two uncontested points available for scrutiny: Claude Fable 5.1 solved a 370‑year‑old cipher, and it did so in forty‑four minutes — a pairing that invites further, concrete reporting and verification.

https://www.schneier.com/blog/archives/2026/09/claude-fable-solves-a-historical-cipher.html