This discussion has been going on for over a year. Copious evidence has been produced for the software freedom risks of using generative models, in a range of ways.
I’m now noticing the main F-Droid repo starting to fill up with low-effort, vibe-coded slop. Since slop is much quicker and easier to generate than quality software, the latter will soon be lost in the flood. Installs will reduce, and human developers will be discouraged from continuing their work.
So what it comes down to is, which of these outcomes would the F-Droid team prefer?
a) The F-Droid library app gets some kind of filtering, enabling us to avoid apps that use generative models. Ideally with flags for the different kinds of issues (vibe-coded, auto-generated code committed, generative models as a remote SaaSS, on-device Model-as-a-Software-Substitute, etc), so people can avoid the ones they’re concerned about.
b) A group of us forks the project to create a slop-free distribution of F-Droid, under its own branding, and educates the software freedom community about the Great Flood of Slop that F-Droid will inevitably become.
I’d much prefer a), and I’d be happy to put some time into helping with the UX design of the flag set. But if that’s not going to start by the end of 2026 (or already underway), then I’ll be looking for a core group to start working on the fork at the start of next year. The same group could potentially also do the same things with a Mobile GNU app library.