Claude Fable 5.1 claims it solved a 370-year-old unsolved cipher in 44 minutes. The method is testable — but no one has independently verified it yet.
Vals AI says Claude Fable 5.1 cracked the Cyphral Distich, a 64-number cryptogram that cryptography researcher Klaus Schmeh had flagged as unresolved as recently as 2014. The model reportedly treated each number as a word-index into the 32 "Proquiritations" preceding the cipher in Urquhart's original text — decoding a royalist message calling for God to preserve King Charles II.
It's a bold claim, and unusually checkable. Unlike vague "the model reasoned well" claims, this one has a defined method: a specific source text, a specific numbering rule, and a specific output. That means anyone can, in theory, verify it.
And that's exactly what hasn't happened yet. Neither Vals AI nor The Decoder — which independently reported the claim — has published the full 64-number-to-word mapping or a reproducible breakdown of the decoding process. No historians or cryptographers have reviewed it. No alternative indexing schemes have been ruled out. The royalist message sounds historically plausible for 1653 — but plausibility isn't proof.
Bottom line: This could be a genuinely significant AI-assisted breakthrough in historical cryptanalysis — or a compelling-sounding pattern match that hasn't survived scrutiny yet. Until the full mapping is published and independently reproduced, the cipher's "solved" status remains a claim, not a confirmed result.
Vals AI says Claude Fable 5.1 cracked the Cyphral Distich, a 64 number cryptogram that cryptography researcher Klaus Schmeh had flagged as unresolved as recently as 2014.
The model reportedly treated each number as a word index into the 32 "Proquiritations" preceding the cipher in Urquhart's original text — decoding a royalist message calling for God to preserve King Charles II.
Unlike vague "the model reasoned well" claims, this one has a defined method: a specific source text, a specific numbering rule, and a specific output.
Continue reading