When AI Solved a 370-Year-Old Cipher Nobody Could Crack

For 370 years, a short encrypted poem sat unsolved in the back of a 17th-century book. Cryptographers tried everything — frequency analysis, substitution ciphers, homophonic mappings. None of it worked. The puzzle became one of Klaus Schmeh's Top 50 unsolved encrypted messages. Then an AI spent 44 minutes on it and cracked it wide open.

The Cyphral Distich: A Royalist's Hidden Prayer

At the end of Sir Thomas Urquhart's Logopandecteision (1653) sits a cryptogram: two lines of 32 numbers each, called the Cyphral Distich. The numbers are:

5.3.27.38.32.14.21.8.66.8.70.39.5.9.12.18.2.3.56.5.1.7.3.2.13.19.3.25.9.3.16.6.

25.15.13.6.11.20.5.1.2.12.1.20.20.49.20.20.35.33.4.6.8.35.5.33.5.5.18.10.3.11.32.42.

The puzzle was posed as an open problem in Notes and Queries in 1899. It appeared repeatedly in 20th-century cryptography literature. Every attempt followed the same assumption: the key was external — some cipher alphabet or mapping that had to be reconstructed from outside the text.

That assumption was wrong.

Enter Fable 5.1

Claude Fable 5.1 was given an open task: solve the Cyphral Distich. No hints about the method. No hand-holding. Just the 64 numbers and a goal.

After 44 minutes and 176,000 tokens of reasoning, Fable 5.1 arrived at a verified solution. It tried several approaches before landing on two crucial realizations:

• The cryptogram appears immediately after Urquhart's 32 Proquiritations, and Urquhart himself emphasized the number 32 — even writing "there can no number like that of two and thirty be pitched upon."
• The accompanying poem promises that an honest reader will find "his own heart's wishes, and the Author's minde." The Proquiritations repeatedly conclude with phrases like "is the desire," "wish," or "hope of."

The key was not an external cipher alphabet. The key was the book itself.

The Solution Hiding in Plain Sight

The rule was elegant in its simplicity: for the i-th number in a cipher line, go to the i-th Proquiritation, use that number as a word index, and take the first letter of that word.

The result:

O GOD UPHOLD KING CHARLS THE SECOND AND

MAKE HIM THE SUPREME RULER OF THIS LAND

A royalist prayer for Charles II, hidden in a book by a committed Royalist during the interregnum. It's self-verifying: each line contains exactly 32 letters, they rhyme (end/land), and the message is historically consistent with Urquhart's politics.

But Wait — There's a Second Cipher

Urquhart left a second, larger cryptogram in The Jewel (1652) — the Cyphral Octastich, containing 285 numbers. Fable 5.1 cracked this one too, using the same principle but with pages instead of paragraphs: the k-th number keys to page k of the book, taking the first letter of the indexed word.

The plaintext is an ottava rima royalist prayer written in London, March 1652:

GREAT LORD, MANTAINE THAT REGAL FAMILIE / WHEREOF KING CHARLS THE SECOND IS THE HEAD, / AND GRANT THAT HE MAY BEARE THE SUPREME SWEIGH / WHERE ENGLISH, SCOTS AND IRISH ARE BORNE AND BRED...

The word "sweigh" is Scots for "sway, controlling power" — documented in the Dictionary of the Older Scottish Tongue with the exact idiom "to bear the swey" circa 1600. It rhymes perfectly with "familie" and "authoritie."

What This Means for AI and Cryptography

This isn't just a neat historical puzzle solved for fun. It's a demonstration of something deeper about how AI approaches problems differently than humans do.

• Human cryptographers spent centuries looking for external keys. The AI found the key was internal — the book itself was the cipher.
• The AI didn't just apply known cryptographic techniques. It read the surrounding text, identified structural clues (the emphasis on 32, the word "wishes"), and built a hypothesis from contextual reasoning.
• No other frontier model the researcher tried over several months produced a verified solve. Fable 5.1 was the first.
• The entire solve took 44 minutes and 176k tokens — less computing power than a single Netflix movie consumes to buffer.

The Caveats (Stated Plainly)

The solve isn't perfect. Nine letters in the Octastich remain uncertain — either due to Urquhart's own errors, misprinted numbers, or transcription issues. Line 5 contains a sequence that doesn't decode cleanly into English. And confirming the remaining positions requires a physical copy of the 1652 Jewel, since no free page images exist online.

But the core solution — the method, the plaintext, the historical context — is solid. This is a 370-year-old cipher, solved by an AI in under an hour, with a method that humans never tried because they were too busy looking outward for the key when it was right there in the text all along.

The Bigger Picture

What makes this remarkable isn't just that AI solved an old puzzle. It's that the AI approached the problem like a reader, not like a cryptanalyst. It read the book. It noticed the clues Urquhart left. It understood that the number 32 wasn't a coincidence — it was a signal. And it understood that "wishes" wasn't just a word — it was an instruction.

This is the kind of reasoning that humans are supposed to be good at — the holistic, contextual, literary kind. The kind that requires reading between the lines. And an AI just did it better than 370 years of human cryptographers.

As AI models continue to improve, we should expect more of these moments — not just in math and code, but in the humanities, in history, in the spaces where intuition and context matter. The Cyphral Distich is one of the oldest unsolved ciphers in the world. It won't be the last to fall.