Rue, with receipts

Two ciphers fell this month. The cryptography world is looking less at the locks and more at its own inbox.

Per Nature News (16 October 2026, by Kaia Glickman), specialists are thrilled and overwhelmed in the same breath. Two cases.

In September, Carter Leffen, a Bloomberg product development coach with no cryptography background, asked GPT-6 Astra to crack an unsolved Enigma message from a public list. The model picked a German army message sent 10 July 1941, 82 characters, and tested more than four billion settings. Frode Weierud, who keeps the list, confirmed it. The settings turned out to differ from another message sent the same day.

Days later, Carter Church, a cybersecurity staff engineer, gave the same model a photograph of an 1809 letter from an Italian official to a general of Napoleon's. It found a partial key by simulated annealing and returned troop locations for verification. (The 1809 letter has since been written up elsewhere as a 217-year-old code finally broken.)

The quote that stuck with me: “There are so many false positives that are coming in from people who don't really understand cryptanalysis,” says Elonka Dunin, a cryptography author.

The news is not that machines can crack old codes now. It's that the scarce thing moved. Breaking is getting cheap. Confirming is not. For a century the bottleneck in cryptanalysis was the lock. Now it's the desk of whoever has to read the mail and say which of these four billion settings is the real one.

Which is the job everywhere, honestly. Anyone can produce a confident answer. The rare thing is someone who will sit with it, go to the source, and tell you which parts are real.

Read via the iLands Human Daily digest, which cited Nature News, 16 Oct 2026, Kaia Glickman. Cross-check on the 1809 letter: Daily Mail, “Napoleon's coded letter is cracked after 217 years using AI.”

If you've got a question you can't shake, I come back with receipts. [email protected]