I've been working on a side project for some time: autonomous reverse-engineering powered by Gepetto.

I shared some thoughts here (https://blog.kwiatkowski.fr/gepetto-web), and am looking for beta testers to try out the platform.

DM me or reply below if you are interested!

Reverser friends, Gepetto has made huge leaps in the last weeks. I now consider it to be a decent IDA Pro agent.

With zero interaction (gpt-5), it solved a crackme all on its own. I opened IDA, typed in the prompt, and it did everything.

Prompt if you want to try it out: Analyze this crackme, starting from the main function, and using all the tools at your disposal, try to figure out what is the expected password. Rename everything you can in the process.

File: https://crackmes.one/crackme/68b6a36b8fac2855fe6fba66

The plan for Gepetto has always been to go further in the agentic direction. I've started exposing the context of the RE session to the LLM.

It may not look like much right now, but ~soon the model will create structures for you from the CLI and recast variables 🤯

Right now, the only tools available are getting the current EA and a function's decompiled code.

It's still bleeding edge, but try it out!
https://github.com/JusticeRage/Gepetto

One of the most unexpected and important things I learned working in threat intelligence is: governments will go nuts over maps.
Mislabel something or draw the wrong border and they'll hunt you down like a dog.

Whatever the map says goes in real life.

https://www.theguardian.com/technology/2025/jan/28/google-maps-will-rename-gulf-of-mexico-as-gulf-of-america-in-us