Perception · 2026-09-16 · 1:38

A 110-year-old "unsolved cipher" that may not be a message

Name: A 110-year-old "unsolved cipher" that may not be a message
Uploaded: 2026-09-16
Duration: 1 min 38 s
Description: the most useful result was an honest one. Some mysteries are not locked messages but demonstrations mistaken for them. Telling the two apart — the plausible from the true — is the whole game.

In 1901 two people traded coded newspaper ads — "Tissie" and "Jabber" — that resisted decryption for over a century. Set loose on unsolved ciphers, an AI reached a different conclusion: they may not be a secret message at all.

In favour

Every group uses the same fixed frame letters plus one variable, swept systematically — the ads read as a demonstration of a cipher table, not as sentences.
It would explain 110 years of failed decryption: there is (almost) nothing to decrypt.

Worth watching

It is a hypothesis, not a solve — one lone non-systematic group is the only candidate "message", and without the original key it cannot be verified.
The finding came from an AI told to attempt unsolved ciphers; it reported this openly and refused to fabricate a "reading".

Our takethe most useful result was an honest one. Some mysteries are not locked messages but demonstrations mistaken for them. Telling the two apart — the plausible from the true — is the whole game.

Source: Klaus Schmeh — τα cryptograms των Tissie & Jabber

← All Focus posts

AI-narrated · AI-generated content

Transcript

Before asking how to break a code, it is worth asking something simpler. Is there a message in it at all? In 1901, two people traded coded ads in a newspaper. The ads resisted decryption for over a century. Then an AI, set loose on unsolved ciphers, reached a different conclusion. The ads may not be a secret message at all. The clue is not in what they say, but in how they are built. Every group uses the same fixed frame letters, plus one variable, swept systematically. Read that way, the ads look like a demonstration of a cipher table, not like sentences. And it would explain 110 years of failed decryption. If that is what they are, there is almost nothing to decrypt. Two things to keep straight. This is a hypothesis, not a solve. One lone group breaks the pattern, and it is the only candidate message. Without the original key, it cannot be verified. And the finding came from an AI told to attempt unsolved ciphers. It reported this openly, and it refused to fabricate a reading. So the most useful result was an honest one. Some mysteries are not locked messages, but demonstrations mistaken for them. Telling the two apart, the plausible from the true, is the whole game.