Hah, glad I wasn't the only one that had this idea. Been chewing on it ever since malus.sh or whatever that joke clean room reimplementation of FOSS software project was.

I noticed that others are taking a shot at the big boys like Photoshop, but I don't think that makes sense since 1. Jesus Christ imagine trying to rewrite Photoshop, with or without Astra and reverse engineering, let's be real! And 2. GIMP already exists and people don't use it for various reasons an ultra nerd that doesn't have osx or windows on any device in his house (me) could understand. Instead I'm looking for small niche markets that are locked in by the software equivalent of a petty village despot. Decent examples so far include quilting software and funeral home SASS. I'm also looking for anyone stuck on old dead software with no migration path because of no export functionality or nothing supporting whatever given file format anymore. Apparently QuickBooks desktop falls into this category, but the space feels full enough, with e.g. GnuCash, that I don't think I would be much help there.

Investigations so far: https://github.com/508-dev/genairosity/issues

Basically leveraging subsidized tokens to liberate proprietary software. A bit of a petulant table flip maybe, but it kinda feels like the only thing left to us. Either we can kill ourselves scrambling around doing 996 for startups to try to avoid being a permanent underclass... Or we can light it all on fire.

Also inspired by open weight models tanking possible profit margins on frontier LLMs. One form of anti capitalist activism could be to just make it impossible to monetize something. The Manfred Macx form of activism.

Many other people in this thread with similar projects seem to be following the same workflow:

1. LLM agents sent out to find locked in proprietary software without a FOSS alternative

2. Targets selected by a human for deeper research on feasibility and alternatives

3. Clean room plan established

4. Human uses the software and takes notes

5. Feed notes into a token subsidized frontier model for orchestration. Architecture and plan written, beginning normal agentic development flow.

I'm slowed down significantly by step 4. Perhaps I will look into contributing my efforts to a larger project such as the one linked instead.

OccupySaaS