Astra and Opus just passed Turing’s other test
techcrunch ·
Two independent cryptanalysts have successfully decoded previously unsolved Enigma messages using large language models from OpenAI and Anthropic. Carter Leffen utilized OpenAI's Astra model to break a message that had baffled researchers since 2005 by conducting archival research and simulating the machine. Retired engineer Frode Weirerud validated this solution, praising the AI for performing like a professional researcher in just two days. Separately, Jack Willis used Anthropic’s Claude Opus 5 to decode another message with specific guidance regarding an officer's name signature. These achievements highlight the growing capability of AI agents to perform complex historical cryptographic tasks. Weirerud noted that only seven unbroken Enigma messages remain, suggesting these barriers may soon fall. The events demonstrate how modern LLMs can replicate and accelerate human-level analytical work in specialized fields.
techcrunch