OpenAI and Anthropic models cracked two unsolved WWII Enigma messages. Seven remain.
Developer Carter Leffen gave OpenAI's GPT-6 Astra an Enigma message that had "baffled researchers since 2005", TechCrunch reported on 25 September.
The model dug through archive records and built its own Enigma simulator. It then recovered the German text.
Cybersecurity executive Jack Willis used Anthropic's Claude Opus 5 on a second unsolved message. That solution was validated on 21 September.
Frode Weierud, a retired engineer who keeps the Crypto Cellar list of unsolved Enigma messages, checked both. He had himself spent several weeks researching files from the German federal archives.
He said Astra did in two days what would take "a human researcher weeks or even months".
One more message has known text but a code nobody has broken.
Astra did the archive work and built the tool, and an expert checked both results. Which weeks-long investigation in your company would you run that way?
Sources
Our file on OpenAI
- 7 Oct
Analysts said Claude subscribers use 42% of Anthropic's computing. They pay 10% of sales.
- 4 Oct
AI chipmaker Cerebras debuted at $95 billion. It is now worth $39 billion.
- 1 Oct
Google's new AI guessed wrong on 15% of questions it couldn't answer. OpenAI's 51%.
- 30 Sept
OpenAI faced Florida's 49-page bid to halt model development without outside safeguards.
- 30 Sept
OpenAI sought $30 billion, valuing it at $1.4 trillion, Bloomberg reported.
Every story here is open to read. The ERP LEADERS brief goes one step further.
The week in enterprise software, in one email: the stories that mattered, the failures with their figures, one case taken apart. Read a sample or sign up for the brief.
Welcome back. · Issue 01
