BitcoinWorld AI Cracked a Nazi Code. Crypto Twitter Panicked for Nothing. Key Takeaways An AI just broke a German Enigma message from 1941 that nobody had solved in 83 years. It took about te
BitcoinWorld
AI Cracked a Nazi Code. Crypto Twitter Panicked for Nothing.
On 10 July 1941, a German soldier sent a short radio message. Eighty two characters. He wanted to know which road to take, said he was in a town called Rosenow, and asked for a fast reply.
Nobody could read it. For 83 years it sat on CryptoCellar, a public archive of unsolved wartime ciphers, marked as broken open by no one.
Then Carter Leffen, who works in product development at Bloomberg in New York, pointed OpenAI’s GPT-6 Astra at it. Ten hours later, he had the plaintext.
Astra spotted that the word “ROSENOW” showed up in another message sent the same day, one that had already been solved. It guessed the same word was buried in the unsolved text. From there it built its own Enigma simulator, wrote the attack code, ran keys in parallel, and checked its answer against the telegram header.
Roughly 14.8 million keys tested. One correct answer at the end.
This is where the takes went sideways. “AI breaks encryption” sounds terrifying if you hold coins.
Here is the honest comparison. Once you have a crib word like ROSENOW, Enigma collapses to something a laptop can chew through. Bitcoin uses a 256-bit key on the secp256k1 curve. Attacking it means around 2^128 operations. That number is bigger than the count of atoms you can see in the sky. Ten hours of compute does not dent it. Ten billion years does not dent it.
Enigma was 1930s hardware. Bitcoin is not.
Leffen is not a cryptographer. He set the goal. The model did the archive digging, the coding, the testing, the verification.
One detail sells it. The recovered German still had the operator’s typos in it, like BTTE instead of BITTE. A fake answer would have looked cleaner. Real messy beats fake tidy.
Not the math. Never the math.
The weak spots are old wallets built with bad random number generators, dead brainwallets, forgotten seed phrases, unaudited contracts. Those are search problems, not brute force problems. And search is exactly what these agents are good at now. Remember October 2025, when a broken PRNG left around 220,000 wallets exposed by design? That flaw was sitting in plain sight for years.
Astra did not break Bitcoin. It proved AI can finish jobs that beat human experts for generations. Use a reputable wallet, generate keys properly, and stop worrying about the elliptic curve. Worry about yourself.
This post AI Cracked a Nazi Code. Crypto Twitter Panicked for Nothing. first appeared on BitcoinWorld.
