Carter Leffen announced that he used GPT-6 Astra to recover the text of a German military Enigma radio message dated July 10, 1941. According to Leffen, just two days ago the model cracked a previously unsolved message with the identifier MV, which had been sitting in the CryptoCellar archive since 2005.

Breaking Down the Claim

Leffen says GPT-6 Astra handled the entire process on its own: digging up historical materials, cross-referencing sketchy letter recognitions, and hunting for context clues. He also claims the system built its own Enigma simulator from scratch, wrote all the code needed for cryptoanalysis, and ran the computations in parallel. That’s what made it possible, he says, to finally recover the July 10, 1941 radio message.

What Archive, What Signal?

We’re talking about the MV-identified message, which Leffen says had gone unsolved in the CryptoCellar archive since 2005. Now, thanks to GPT-6 Astra, he claims the text has finally been restored using next-gen AI tools.

Why It Matters for AI and Cryptoanalysis

If this result checks out, it’s a big flex for modern AI systems in tackling real-world cryptoanalysis — from gathering raw data to building tools and running the numbers. The spotlight here is on how AI can connect context-finding with software emulation of cipher systems to recover long-lost historical communications.