Democratizing Abandonware - Evangelos “GeopJr” PaterakisI am not part of Flathub, I do not represent Flathub or anyone but myself.

On May 29, 2026, Bart announced that Flathub would no longer accept LLM-generated submissions and codebases. Chaos ensued. "Is this the end of Flathub?", "That's it I'm moving to snap!", "Remember Appimages? We won't ever modernize them, join us!".#History<br>Well, not exactly. This wasn't surprising to anyone even slightly involved. You see, reviewing Flathub submissions is a very tiring and thankless volunteer responsibility. Every manifest needs to be the most optimal it can be before getting merged, that means: minimal permissions, building from source, building for aarch64 alongside x86-64, proper cleanup, upstreaming and validating metadata, pinning everything to commits, validating app ids, domains, following build and translation best practices and so much more... Need to open a hole in the sandbox? You better have a good enough reason on why you are too good for the portals and then make a PR requesting an exception before the submission can proceed.It didn't start like this, but over the years the requirements got stricter to match Flathub's increasing amount of users and submissions. I can assure you that many of my early-day submissions would NEVER get accepted nowadays. The rules around submitting also got stricter, now submitters have to provide screen recordings of their flatpaked apps, verify that they built it successfully locally etc.#The AI Slop tag<br>The vibecoding trend and OpenClaw (Moltbot? Clawdbot?) turned Flathub submissions into a checkbox. Submitters started using these to automatically submit their apps and deal with the review process. Flathub reviewers were met with new PRs that didn't follow the PR template or instructions and when attempting to communicate with the PR authors, they got back huge word salads and a meaningless force push without addressing the actual comments.Around January 2026, the "AI Slop" tag was created. It wasn't meant to block a submission from getting merged, but to signal to the other reviewers that the submitter is using a chatbot to communicate with them or the manifest was completely vibecoded and they shouldn't spend time explaining things or be lenient.#What's the actual problem, though?<br>Flathub team's (or some people's in the team) issue is not actually ideological or based on "outdated information" on what LLMs produce. I wish it was, I wish we, as FL OSS developers actually took issue with the fact that our licenses mean nothing anymore and our contributions are being chewed up and sold back to us, I wish we stopped the whole "ignoring the political, social and economical issues" but alas, that's not the point of this.Flathub's issue (from my understanding based on Bart's post, read the disclaimer again) is that the volunteer-based community-maintained reviewing infrastructure can't handle the amount of spam coming in. There's like 3 reviewers in total dealing with people who not only demand their software gets in without any back-n-fourths but also refuse to even talk to the volunteer reviewers themselves, who spend their free time reading code they themselves didn't, and ask their agent to do it instead.The other problem with the lack of effort, is that the submitter doesn't learn why things are the way they are, why they should use the portals and how to be ready for the next time they submit an application or even become reviewers themselves. That's something GNOME Circle taught me. The review process doesn't only exist as a checklist, but it also trains you on the HIG and design principles.I talked about the AI spam on Flathub briefly on my LAS 2026 talk, before that decision was made, which Brage Fuglseth has cut into an individual video (thank you!):<br>Digging Through the App Cemetery: Sustaining a Fork | Evangelos Paterakis @ LAS 2026(does the subtitle of this post make more sense now?)For the record, I do agree with Bart that it's inevitable and despite being critical of LLMs, I did voice my concerns on this being overreaching. Flathub allows proprietary apps that are definitely filled with LLM written, LLM reviewed and LLM tested code. This is punishing those that open source their LLM generated code but not those that don't. I suspect this decision will be reversed eventually. And I'm still in favor of AI self-disclosure tags.#Present<br>Why am I even writing a blog post about something I am not involved in? Trust me, I barely write about the things I am actively involved in, so no, I'm not just shouting at clouds.Recently, there was a thread on fedi about anti-AI policies pushing people away from GNOME and Flathub (prompted by, I assume, a recent AI ban on GNOME Discourse). As someone who cares about those communities, I wanted to see some stats. Did these AI policies push valuable contributions and community members away? What are the stats?#PLAN.md<br>I wanted to see how many of the "AI Slop" tagged applications...