Claude Fable 5.1 cracks a 370-year-old cipher that stumped scholars since 1653

Anthropic's Claude Fable 5.1 has cracked the Cyphral Distich, a 370-year-old cryptogram buried at the back of Sir Thomas Urquhart's 1653 work Logopandecteision. The cipher had appeared on Klaus Schmeh's list of the Top 50 unsolved historical encrypted messages and had resisted decryption for centuries, according to a writeup by Geby Jaff at Vals AI.

The model solved it in 44 minutes using 176,000 tokens with no human intervention, producing the decoded message: "O GOD UPHOLD KING CHARLS THE SECOND AND MAKE HIM THE SUPREME RULER OF THIS LAND" — a Royalist prayer entirely consistent with Urquhart's known political loyalties in the 1650s.

The key was the book itself

What made the Cyphral Distich particularly resistant to classical cryptanalysis is that previous attempts searched for external cipher alphabets — standard tools for historical ciphers. Fable 5.1 instead noticed a structural clue in the text: Urquhart repeatedly emphasized the number 32, and the cryptogram followed exactly 32 Proquiritations (numbered sections in the book).

The model's method: for each number at position i in a cipher line, locate the i-th Proquiritation, use the encoded number as a word index within that section, and extract the first letter of that word. Applied across both 32-number lines, the plaintext message appeared without ambiguity. Using the same approach, Fable 5.1 also decoded a second cipher, the Cyphral Octastich, embedded in Urquhart's 1652 work The Jewel.

Not cryptanalysis — persistence

Researcher Geby Jaff emphasized that the achievement wasn't about cryptographic virtuosity. The model "kept looking" and rejected dead ends that had trapped human researchers. The bottleneck for solving historical ciphers is often not the reasoning required but the sustained attention across obscure source materials — precisely the kind of task where AI has no fatigue limit.

"Budget for the check, not just the solve," Jaff wrote, noting that confirming the solution required cross-referencing historical dictionaries, bibliographic records, and physical verification of page counts in 17th-century books. The Octastich solution has 231 confirmed exact matches out of 275 readable positions; the remaining 44 differ by 1–3 words attributed to transcription artifacts across centuries.

Dispute from researchers

A group of researchers published a rebuttal the day after, arguing that the cipher isn't placed where claimed in the 1653 book and that Fable's proposed method doesn't hold up under close scrutiny. The dispute is unresolved. Whatever the final verdict on the Cyphral Distich specifically, the episode illustrates AI's growing capability in archival and historical research tasks that require sustained attention and pattern matching across large bodies of obscure text.

As reported by Vals AI, this is among the first documented cases of an AI solving a historical cipher from Schmeh's unsolved list.