Astra and Opus crack unsolved Enigma codes with LLMs
two llms, openai's astra and anthropic's opus, have solved long-unsolved enigma messages, showcasing their cryptanalysis capabilities. this breakthrough has significant implications for cryptography and historical research.
openai's astra and anthropic's opus, two large language models, have successfully cracked two long-unsolved enigma messages. astra, tested by developer carter leffen, searched a database of enigma messages, found context clues, and built a simulator of the enigma machine to recover the plaintext of a message that had baffled researchers since 2005. anthropic's opus 5 model, guided by cryptanalyst jack willis, used a known signature of a particular officer's name to break a different unsolved message. these breakthroughs have significant implications for cryptography and historical research, and may aid in the decryption of the seven remaining unsolved enigma messages.