I've been really inspired by it. Ever since that sort of joking project came out, "Malus," that would clean room reimplement GPL software under more permissive licensing, I've been chewing on doing the reverse for proprietary software. I've been calling it the Manfred Macx theory of disruption, for the character in Accelerando that would patent and then release to public domain profitable ideas before a megacorp could lock them down.

I've got my wedding coming up so haven't had the time I want to dedicate seriously to finding some proprietary things to disrupt, but I've had fun using this as an opportunity to play with subagent orchestration, open weight models, various harnesses, and local models, to see what happens if I let some LLMs churn on it. That resulted in a hodge podge researched list of potential targets: https://github.com/508-dev/genairosity/issues

One thing that stands out is a lot of software kinda does already have a FOSS replacement, it's just not really how people want it to be. GIMP being the representative example. Photoshop people just don't like it, I'm sure for not entirely invalid grievances. However the maintainers of these kinds of tools are often strongly opposed to LLM involved contributions, again , often for not entirely invalid reasons on these incredibly complex projects.

So the long and short of it is that as powerful as LLMs are, we aren't quite where some of the doomsayers are saying we are, insomuch as proprietary software is dead. Even with reverse engineering we aren't there. There's genuine labor, time, and expertise moats around most of these programs.

Personally I'm shifting to trying to find niche abandoned software with no export flow. Even if I can only help out a couple hundred people, it sounds like a decent use of my time.