Frontier AI can now reverse-engineer software binaries. Here is what that means
 

https://www.popularai.org/p/ai-reverse-engineering-software-binaries

An interesting read, I hope it is not generated with AI :D

RIP decomp.dev. AI is claiming scalps.

The website decomp.dev said Thursday that it is shutting down because “the community aspect of decompilation has been broken” by AI, and that projects that once required intense collaboration between people with different skills can now be done trivially with AI. The administrator of the website said that it has become impossible to keep track of which projects are primarily human-led and which ones are done by AI.

Communities like ours dealing with non trivial protections will be the last man standing. AI wont trivially conduct creation of vital tools to deal with advanced anti debugger techniques or virtualization or packed and encrypted binaries etc. Vibe kiddies wont know to step back and do a focused software effort to unravel a binary layer by layer. But for sure all but advanced reversing is automated to an absurd level of even source recovery that is near byte identical when recompiled. Lots os signs that we are adapting and honestly I think ot is exciting. Reversing as a mindset and talent and hobby etc is not going away. But it is transitioning like msny fields are.

i.e : Get a new job within 1 to 2 years :D

As someone who has been in the reverse engineering game for a long time now, I can tell you that its not as simple as handing AI (even best-in-the-world AI) a binary and saying "give me source code for this". Things like "AI-generated Photoshop clone" or "game xyz ported to a whole bunch of new platforms thanks to AI" are absolutely not what they appear to be.

If you want an example of a AI-assisted reverse engineering project that's doing things the right way, check out open-bfme (project to reverse engineer the Battle for Middle Earth series)

I tested. Yes, they are not complete. They do not have all the functions of the original apps.

This generally follows the 80/20 rule: AI can easily knock out 80% of the feature set in no time, yet you’ll burn 80% of your time implementing the remaining 20% and making the whole thing rock-solid.

The new development is the crowd sourcing element as a lot of people have subscriptions they are not fully using. I am very much hoping this doesnt turn into an unmaintainable slop pile but rather a legit replacement for these important tools. Since this is a simple clone goal my guess it is achievable as the requirements are quite precise on that, and i doubt there will be real design issues. And most average users are testers who can fire up their agents to fix a bug or add a feature even if they dont know how to code and as long as the project maintenance and review process is solid enough, I think crowd sourcing the expensive 20% is just such a great way to go. AI is producing less slop and higher quality as of late even if it is still not perfect.