We have updated Flathub's LLM policy to explicitly disallow AI usage for both the submission process and applications being submitted.
https://github.com/flathub-infra/documentation/commit/992f57b30de98ddbd5e80959e9672998c83c8c97
I've had some reservations about it, so the wording before that commit was relatively milder. I know it's an unpopular opinion on the Fediverse, but I do think LLMs are inevitable, and the reality is that you can expect less organically grown code as time goes on. I believe it can be a useful tool in and outside FOSS; I hoped we will see a larger number of apps where authors made some effort beyond prompting an agent. Meanwhile, the number of unpleasant interactions I've had with entitled submitters acting as if they were bestowing their brilliant software upon us idiots who are rejecting it went through the roof in the last month. I'm tired.
As always, we are not applying this retroactively, so any vibecoded apps which were already published will remain available.
@barthalion I've been thinking quite a bit about this lately. Not that I am in the position to make those decisions, but I mean about what the best approach is towards LLM-generated code as a community.
Either way, I recently looked at the Flathub PR list, as I was considering submitting a new app soon. The amount of AI-slop being submitted is depressing and immediately makes me pity all volunteers that review Flathub submissions. Given that sheer volume, I think this may be for the better.
@barthalion
I think it's dangerous to not apply retroactively, because then, instead of prompting thought, you'll be prompting floods of BS ahead of changes.
@nexi@fosstodon.org@barthalion The problem of low-effort Flathub submissions is real, but a blanket ban on any AI-generated or AI-assisted code seems excessive.
Imagine if there was no exception for existing projects: Firefox, VSCode, Chromium would've been deleted.
It makes sense to introduce some sort of objective project maturity metric the software has to meet before submission to filter out low quality submissions instead.
@nexi if the project has community engagement, release cadence, CI, and seems to have been more than pure slop produced in 3 weeks, it would be likely exempted. That’s how the bar already was before that commit, and yet, here we are.
@nexi@fosstodon.org@barthalion This sounds quite reasonable. It would've been great to reflect that in policy for clarity (of course with a disclaimer that a project may be rejected anyways if the code quality is deemed too low, without any further explanations).
I found that that kind of elaboration inspires lawyering and takes even more of your time.
@eestileib @nexi Exactly. We tried to prevent slop submissions based on sufficient development history and overall "project health" and it didn't reduce the workload at all, only caused sloppers to dispute the rules.
@barthalion @nexi So there are absolutely no AI assisted submissions allowed, *unless* someone in a position to turn a blind eye likes you enough. Welcome to NepotismHub?
@barthalion @nexi How about we measure you by your actions? If you stand by "Applications containing AI-generated or AI-assisted code, documentation, or other content are not allowed." then Firefox, Chrome, VScode and maybe a dozen others have to go.
@Ember@blobfox.coffee@barthalion great that you're rejecting it, but it's just plain false to say it's inevitable, or that it can be a useful tool, it isn't and it cannot
@Ember It’s okay, the world is wide enough.
lumi-nhac 
@Ember @barthalion Agreed.
That said, great to see slop kept out of Flathub, and I hope you aren't pressured by any involved corporations to allow it.
@barthalion “The orphan crushing machine admittedly has its uses…”
There is no valid use for LLMs *anywhere*, given the code theft, pre-baked biases, and environmental impact. Please stop saying otherwise.
@Kroc There’s more nuance to this issue than good or bad. That’s not the place nor time for this discussion though.
@barthalion There is no “nuance” to code theft any more then there is “nuance” to someone hitting you over the head with a wrench
@barthalion @Kroc I think there is nuance in whether this technology can be useful in some ways, but there absolutely is no nuance regarding its (awful) ethics.
@engideer @barthalion @Kroc
...of the currently dominant practice.
@barthalion Hugs to you, and may you rest up well. Good on you for the boundaries.
@barthalion Excellent reason to take an action like this. I think people don’t really think about it, but the culture around LLM generated code and the people who use LLMs is often extreme and not compatible with the kind and collaborative approach it takes to sustain open source communities. An even better reason than anything technical, in my view. Bravo.
@barthalion hey Bart, I'd love to help reviewing submissions, is there any room/chat or resource to get started?
@barthalion wouldn't that add fragmentation into the flatpak and yet-another-way-of-distribution on linux which is the very thing flatpak tries to solve? How's that going to play with proprietary apps? I.e. Slack and Spotify are available on flathub and imo it's a good thing to have them sandboxed. Or closed code will have an advantage because nobody knows the way it was written?
@dmantis The new policy states an exemption for mature and well-maintained project.
While we can scratch "well-maintained" from a lot of the proprietary applications, this is not a ban on such applications. Neither is this a ban on established and proven FOSS applications that have used an LLM somewhere.
And yeah, if it's a brand-new proprietary application from an unknown developer, it might be a good idea to not immediately publish it on Flathub anyway. Even without this policy.
@sstendahl I understand the reasoning and the slop problem, but the complete ban on even AI documentation sounds like a mistake to me.
Realistically, flatpak and linux desktop ecosystem is not in the position of power to dictate to businesses the way they should behave. I recently visited the store and was positively surprised to see some apps I had only as AppImages for years to be supported now officially as flatpaks. I'm afraid after these policies most of them will just avoid proceeding in this way. I'd rather see fee charge for AI-based submissions to pay for reviewers work in apple style, rather then put complete ban.
It's basically saying: before you have N users you should go distribute elsewhere. But if a person of a well-maintained app with, let's say, 10% of tests/docs automation by LLM, setups a distribution in another place and it works well for him, then there is no motivation to switch. That can cause network effect.
@barthalion I feel like we're all going to have to adopt a 'cooldown' philosophy to new apps, anything that hasn't existed for a number of releases and for long enough for a second human contributor is likely going to be too risky
@barthalion Thank you for all your work and time to keep the lights on and flathub safe from slop 
@zeenix@toot.cat@barthalion I'm baffled as to why so many people are so aggressive about their opinion on LLMs. One person was banned from the embedded Rust matrix channel recently after they continued attacking multiple people, who agreed with them about LLMs themselves but disagreed on banning their usage for code contributions.
@zeenix While we should be extremely careful about witch-hunts, which is definitely happening.
I think part of the reason why people are so aggressive are because the push behind LLM's from big-tech is incredibly aggressive. They're really pushing it down our throats whether we want it or not. Some employers mandate people to use LLM's in their work, with the threat of being fired if not complied to. My entire feed is being filled with slop. 1/2
@zeenix Even simple things like search are just ruined forever now
Nobody wants it everywhere. "Agentic future" is incredibly niche, people overall hate this shit. Yet everything is turning into homeopathic residue of something that almost resembles human work.
Having said that, in isolation it's a cool technology and the witch-hunt that exists against people using it is not good. But the aggressive push from big-tech and everyone in power on this is really getting on my nerves. 2/2
@sstendahl @zeenix so you're saying that we literally know there are evil witches here, many paid, but hunting them is bad?
@davidgerard
All I'm saying that it doesn't help for the community, to burn everyone that played around with brewing a potion at one point. Even though you and I actually do hold mostly similar opinions on witchcraft in this metaphor.
If you want to interpret that as defending AI boosters or slop in general, be my guest. But then you'd have to be misinterpreting me on purpose.
@abucci @sstendahl it's a bit odd
it feels like the performance of centrism
if people choose to use the bot, they're people who chose to use the bot
@sstendahl @zeenix at this point, search, chatting with a bot etc have nothing to do with it. Using llms for code is a standalone thing, things are proofable with code, trade offs are clear. It needs its own category to be judged.
@zeenix@toot.cat@sstendahl I have seen people who it is definitely not being forced on, being aggressive against people use the tech so I think that's not a good explanation and even if it was, it's no excuse for being aggressive to others who didn't force it on you.
> Nobody wants it everywhere. "Agentic future" is incredibly niche, people overall hate this shit.
This is part of the aggression I was talking about. I didn't ask for your completely unfounded personal claims. I've heard it all a thousand times and sick & tired of it.
@zeenix it is forced upon us all though, I gave up on Google because of this. Every other major tech-product is full of AI, and in many cases you can’t even opt-out.
Regarding everyone hating it, it has literally worse approval ratings than ICE and Donald Trump. Despite that, it’s in every other product witht leadership bullying us into thinking there’s no alternative.
Having said that. Be kind to people be ruthless to systems. I don’t intend any anger to users that leave me be.
@sstendahl @zeenix "Be kind to people, be ruthless to systems". It's something we tend to forget a lot.
@endorama systems don't build themselves 
@FinFangFoomed surely but either you assume that everyone in a system is complicit, or you need to give some slack. Societal systems are generally too complex to easily point fingers, and very few people, if any, would pass a purity test. We still all live in the same place.
That's not to say there should be no consequences (the bigger your influence the bigger your responsibility in my opinion) for behaviours or lack of it. And caveats apply (Nazi regime as one example, or the French revolution).
But if anyone in a system becomes an enemy the only route to deal with that is violence. There is no room for dialogue.
@endorama you're the only one talking about violence here.
You can't hold systems responsible because they're not people. Whether we're talking governments, a board, an application, or whatever, those are man made systems that have no concept of mortality, ethics, or responsibility. Only the people operating and reinforcing those systems are able to be held to those standards.
So yes, I will hold the people responsible because they are the ones who designed or maintain the system doing harm
@barubary@infosec.exchangeI have seen people who [AI] is definitely not being forced on
Are those people in the room with us right now?
@zeenix @sstendahl You seriously posted an unprompted public reply to someone with, as you call them, "aggressive and unfounded personal claims", and then got up in arms and started insulting others when they replied with their personal claims to yours. You should reflect on your behavior, since you are projecting.
@engideer @zeenix @sstendahl How can a "reply to someone" be "unprompted"? Doesn't it have to be a reply to something?
@zeenix @barthalion dunning-kruger syndrome
hacker @strk@floss.social@zeenix I don't find any aggressive tone in the linked commit, do you agree ?
Whoever hosts a service has the right to set the usage policy. More flatpak repositories are welcome, centralization is bad 
.
@zeenix @barthalion is it really surprising that people have strong feelings about it?
@zeenix@toot.cat@benjamineskola @barthalion no, but so strong that you get yourself banned from a channel full of anti-LLM and LLM-sceptic folks because you attack anyone who disagrees with you on specific actions, is just a level that's indeed surprising to me.
@zeenix @barthalion okay, if they’re also attacking other anti-LLM people for not being anti-LLM *enough* (or going about it in the wrong way or whatever), that’s counterproductive at best. Fair point.
@zeenix I wish I could say that anti-LLM crowd behaves any better, but yes, people addicted to LLM usage are sure special.
@zeenix@toot.cat@barthalion I think you misunderstood. I was talking about anti-LLM crowd. I've yet to see pro-LLM crowd to be aggressive to others. Passionate perhaps and being too vocal about it but not aggressive.
@zeenix Ohh, yeah, I have misunderstood. Sadly both crowds are like that, vibecoders throw a tantrum when called out and play victimized. The entire discourse seems to have been taken over by polarized people but I guess that’s not surprising given everything else happening around the world.
@zeenix@toot.cat@barthalion I have not seen that but I can imagine. AFAIK most folks are not vibe-coders though but just people who see the value of these tools and use them. Also, I'd happily take a cry-baby over naked aggression, any day.
@zeenix@toot.cat@eschwartz @barthalion that's not an example of a pro-LLM person being aggressive to others but rather some vibe-coder letting their LLM loose