OpenAI's GPT-6 Astra deciphers 108-year-old WW1 German radio cipher
OpenAI released GPT-6 Astra on September 3, 2026. The flagship model, already noted for playing Minecraft autonomously and building nested simulations, was used by AI developer Prinz to break a 108‑year‑old German radio cipher from World War One. Prinz posted on Substack that the cipher was taken from a list of fifty unsolved codes on Scienceblogs.de and that Astra solved it with the keyword…
Key points
- GPT-6 Astra, released Sep 3 2026, was used by developer Prinz to break a 108‑year‑old WW1 German radio cipher.
- The model solved the cipher using the keyword "TRUPPENVERSCHIEBUNG", revealing a message about an English cruiser at Sevastopol.
- Verification against military logs matched the decoded arrival with HMS Canterbury on Nov 24 1918 and an Allied squadron on Nov 26 1918.
The decoded transmission reads, in English, "AN ENGLISH CRUISER ARRIVED AT SEVASTOPOL ON THE ?4TH AND ALLIED SQUADRON FOLLOWS ON THE 26TH." Astra cross‑checked the result against military logs, finding that the British cruiser HMS Canterbury arrived in Sevastopol on November 24, 1918, and an Allied squadron arrived on November 26, matching the decoded dates. The question mark appears to be a transmission typo. The breakthrough shows GPT‑6 Astra’s ability to handle complex historical cryptanalysis, a capability previously unseen in public AI models.
Model page: GPT-6 Astra →
The story so far
2 episodes →- OpenAI's GPT-6 Astra deciphers 108-year-old WW1 German radio cipherthis story
Man uses GPT-6 Astra to crack 108-year-old WW1 code
dexerto.com · 19 September 2026
Loading the full article…
This text was published by dexerto.com and written by Matthew Benson. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- Anthropic releases Claude Opus 5.5 with 1M-token context and $4 input, $20 output prices · 3 src
- Anthropic's Claude Opus 5.5 is now available on Amazon Bedrock · 9 src
- AI Safety Concerns Erupt Amid Recursive Self-Improvement Fears · 4 src
- Opinion: OpenAI could copy TypeSafe's Jev classifier and embed it in future models · 1 src
- Aws preview TBC rat‑Brain video model for select customers · 2 src
Comments
via GitHub Discussions