Cryptid

cryptid is an open-source toolkit that tests whether a hash function is well-behaved. It was developed by RST using Anthropic Opus 4.6 over ~5 weeks starting in April 2026, and then cross-checked with Fable 5.1 in late Sept and OpenAI 6-Luna in early Nov 2026:

The tool started with a pretty silly idea: what if I could use AI robot magic to find some flaw in a cryptographic hashing algorithm that let me produce plaintext candidates from a hash?!?

Well, the robot let me know that original idea was probably not going to work, but I wanted to do something in the space and after kicking ideas around for a while we settled on a path that led to this tool and release that might be useful to someone someday :)

The tool uses statistical, differential, and structural black-box testing to attempt to determine if there are detectable weaknesses in hashing algorithms, and supports checking implementations of standard hashing algos against a reference to determine if they were implemented properly against the spec or if they diverge from expected behavior. In theory this let’s you test your HSM or hash library or other things to see if the claims about the hashing algo they use are actually accurate.

That “in theory” bit is in there because I am in no way anything approaching a cryptographer. In fact, I’m pretty bad with math in general ;) But that leads into the ‘real’ purpose that this project morphed into. If you’ve read posts on our site, you know I’ve been interested in exploring what AI will let people and teams do, and in this case the idea was can we find ways to demonstrate that AI can be leveraged to build something well outside of your comfort zone and/or expertise, and end up with something that is not AI slop.

We’ve been deep in dev land, building many fun and interesting things, but came back to this project to touch it up (found bugs and incorrect claims with better models, and fixed them) because we believe it actually qualifies as “not slop.”

If you know things about math and crypto and can tell us where we’ve gotten it wrong, we would love to hear from you, but for now we believe this project serves as validation that there are specific methodologies that can be leveraged to let normal people use AI to produce high-value outputs that are beyond what they could do on their own.

Cryptid is released as Open Source Software via MIT with no warranties and is available on our github.

If you want to ask questions or tell us how we got stuff wrong with cryptid, please let us know via the Contact page