GPT-6 Astra Deciphers an Undeciphered 1918 German WWI Radio Message—And Verifies Its Own Work
GregKamradt · x · 2026-09-18
deredleritt3r reports that GPT-6 Astra deciphered a 1918 German radio transmission that, to his knowledge, had never been broken. The message reads: "An English cruiser arrived at Sevastopol on the ?4th; an allied squadron follows on the 26th."
More striking, the model double-checked its own work: it verified that HMS Canterbury reported arriving at Sevastopol on November 24, 1918, with an allied squadron following on the 26th. The message is one of 20 WWI German radio messages on a public cipher challenge list that had resisted decryption.
GregKamradt adds a takeaway: a side effect of more capable AI is a better understanding of which problems are actually interesting—many open puzzles may simply be waiting for a model to try.
More from Fun
- "There's a 5% chance AI kills us all — but almost all 5% chances don't happen" — NathanpmYoung · 2026-09-18
- Satire: claiming AI will destroy humanity is just a relevance play for replaced experts — bindureddy · 2026-09-18
- Grandpa uses Codex to remotely teach his 10-year-old grandson Mars landing math — burhop · 2026-09-18
- Which AI products count as 'genius'? One dev's list: GPT-1, ChatGPT, Claude Code, Midjourney — wordgrammer · 2026-09-18
- Tesla claims solar panels generate more energy in shade, netizens call 'chart crime' — Yamapama · 2026-09-18
- Hollywood fears AI? "Imagine machines recycling the same old stories instead of a room full of executives" — TinfoilTricorn · 2026-09-18