A new-generation model from Anthropic — Claude Fable 5.1 — has made a breakthrough in historical cryptography by deciphering the famous Cyphral Distich cryptogram by Scottish writer Thomas Urquhart. This puzzle had remained unsolved for more than three and a half centuries.
The cryptogram consists of two lines, each containing 32 numbers. It was discovered at the end of Urquhart's treatise Logopandecteision, published in 1653. In the late 19th century, the cipher was brought up for public discussion and has since appeared in specialized literature as one of the most challenging historical codes.
The key to the solution lies in the book itself
My analysis shows that previous researchers made a typical mistake by trying to apply frequency analysis or standard substitution methods. Fable 5.1 took a different path, focusing on the structural features of the text. The cipher is placed immediately after 32 numbered "procurations" — the author's wishes — and the number 32 itself is specially highlighted in the text.
The decryption algorithm turned out to be elegantly simple: the i-th number in a line points to the i-th procuration, the value of the number determines the ordinal position of a word within it, and the first letters of those words form the message. The result was a couplet in support of Charles II: "God support King Charles the Second and make him the supreme ruler of this land."
Notably, Urquhart was a known supporter of the Stuart dynasty, which makes the decryption logically consistent from a historical perspective. The self-verifiability of the result is confirmed by its structure: each line contains exactly 32 letters, the lines rhyme, and they form a complete couplet.
Controversial points and criticism
However, a debate has erupted in the scientific community. Specialists from Reticuli Labs disputed the results, pointing out ten discrepancies in the algorithm: words with the required initial letters are missing at the specified positions. With alternative counting methods, the best result was only eight matches out of 64 letters — which corresponds to the level of randomness.
Additional skepticism is raised by the cipher's origin: the original 1653 edition contains no numerical lines; they are known only from a biography of Urquhart published in 1899. Vals AI referenced a collected works edition from 1834, which raises questions about the reliability of the source.
My expert opinion: This case demonstrates bidirectional progress in AI. On one hand, the model's ability to find unconventional solutions is impressive. On the other, the absence of a physical primary source and discrepancies in interpretation cast doubt on the finality of the result. It is telling that the model spent 44 minutes and 176,000 tokens on a task that generations of cryptographers could not handle — but verifying such discoveries requires no less effort than making them.