GPT-6 Astra cracks 217-year-old Napoleonic code in six hours

tomshardware.com

22 points·ortusdux··11 comments

https://carter.church/writeups/the-letter-to-marmont/

The original blog post of the person who orchestrated the attempt:

https://carter.church/writeups/the-letter-to-marmont/

Link to X thread: https://x.com/CarterWChurch/status/2105167567706816690?ref_s...

"What makes this impressive isn't actually the codebreaking, but that Astra completed the entire multi-modal workflow in ~6 hours from a single image and goal."

https://xxcancel.com/CarterWChurch/status/210516756770681669...

It’s an image of ink-on-paper cursive handwriting, with what looks like markings added during the original decryption to keep track of where they were in the process.

So image to text, then a brute force attack on a Napoleonic era cypher.

That link just gave me cancer from simply looking at it.

At least cut everything from the ? onwards.

amazing. can it also not butcher my medium size code base every time i give it a task?

Wonder when Astra is going to definitively "solve" the Voynich Manuscript.

Did it search for existing research and documented attempts at the problem before starting on the solution?

No, I will not un-block html-load (dot) com.

Glad the original blog post was linked in the comments here.

I mean is this good? How do we measure how "unsolvable" these codes are vs just unsolved because they are esoteric, uninteresting, and useless?

Yeah -- if you want to do something better; set it on the Beale Ciphers.

Reply on news.ycombinator.com