alt Hacker NewsI know this is satire, but I have an adjacent problem I could use help with. In my company, we have some legacy apps that run, but we no longer have the source, any everyone that worked on them has probably left the planet.
We need to replatform them at some point, and ideally I'd like to let some agents "use" the apps as a means to copy them / rebuild. Most of these are desktop apps, but some have browser interfaces. Has anyone tried something like this or can recommend a service that's worked for them?
Replies
I've done a little bit of this and Claude is pretty great. Take the app and let Claude run wild with it. It does require you to be relatively familiar with the app as you may need to guide it in the right direction.
I was able to get it to rebuild and hack together a .NET application that we don't have source for. This was done in a Linux VM and it gave me a version that I could build and run on Windows.
We're past the point of legacy blackbox apps being a mystery. Happy to talk more, my e-mail is available on my profile.
Interested to keep updated on this point. As a consultant, I've worked on transformation of legacy applications so this would help me greatly as well. We've worked on pretty archaic systems where no one knows how the system works even if we have the source code.
Well, what kind of desktop apps?
Unless obfuscated C# desktop apps are pretty friendly to decompile.
I have actually very convincingly recreated a moderately complex 70s-era mainframe app by having an LLM reimplement it based on existing documentation and by accessing the textual user interface.
The biggest trick is that you need to spend 75% of your time designing and building very good verification tools (which you can do with help from the LLM), and having the LLM carefully trace as many paths as possible through the original application. This will be considerably harder for desktop apps unless you have access to something like an accessibility API that can faithfully capture and operate a GUI.
But in general, LLM performance is limited by how good your validation suite is, and whether you have scalable ways to convince yourself the software is correct.