Is resistance futile?

From: Drew DeVault <drew@ddevault.org>
To: ~sircmpwn/sr.ht-discuss@lists.sr.ht
Subject: Proposal to prohibit vibe coded projects from being hosted on SourceHut

Codeberg recently passed a proposal which limits the use of LLM tools for projects hosted there:

https://blog.codeberg.org/protecting-our-floss-commons-from-llms.html

SourceHut staff are impressed with their reasoning and their approach, and we share their sentiments. We have contemplated doing something similar on SourceHut for some time now.

We face many of the same practical problems as Codeberg with the additional burden vibe coded projects tend to impose on our infrastructure. For example, a few weeks ago we suffered a builds.sr.ht outage when a vibe coded project submitted several hundred builds at once. Of course, it is also well-known that SourceHut infrastructure suffers from large scale scraping operations which are intended to populate large language models with our user's data, and we spend a lot of our time and energy mitigating this, leaving us with an unfavorable view of the means by which these models are produced. We feel some solidarity with Codeberg in this respect.

We also agree broadly with all of the political and ethical objections raised by Codeberg. LLMs are harmful to the free software ecosystem as a whole, as their main use-case in software is to appropriate free software regardless of license and regurgitate it new new projects without any respect for the copyright, provenance, or obligations of the source material. Moreover, these tools enable harmful techniques like mass rewrites of software for the explicit purpose of circumventing a copyleft license, such as in the case of the chardet project.

The externalities of LLMs are also extreme, and we are not comfortable with tacitly supporting the effects they have on the environment and the political sphere. They are an environmental catastrophe, expand the already-severe gap between the wealthy and the poor, reduce the leverage and security of the working class, manipulate and seduce vulnerable users into episodes of psychosis or worse, and they inflame political tensions and exacerbate conflicts, including violent conflicts, at the expense of the vulnerable and innocent.

This is our view, but I now turn it over to SourceHut users to discuss further. Please take extra care to approach this conversation cautiously and respectfully of the viewpoints of others. We can discuss the moral and political matters that shape each of our views on the subject, but be careful not to start or exacerbate a flamewar.

	Are you in favor of a AI/LLM/Vibe Coding ban on SourceHut?
	+1 to the idea, for all of the reasons listed above, including technical, ethical, ecological, etc. … repositories that feed back into the larger AI ecosystem are just kind of unfit for SourceHut, and tarnish its image.
	+1 I'd be in favour of an even stronger policy than merely banning vibe coded projects.
	Please, do. Go ahead. If free software (as in freedom) means something, its meaning is opposed to what LLM represents and implies.
	As a paying user and contributor, I fully support this proposal. What I want to fund and contribute to is a platform that supports human work rather than LLM output. If anything, I think Codeberg's policy doesn't go far enough.
	+1 if people want to use llms to write (or "assist") they can use github, gitlab, or self-host something i pay for sourcehut because it's opinionated and i resonate with the opinions expressed through the software, platform, and by the team i don't care about splitting hairs about whether technically a human could conduct similar abuses. the llm industry is a scourge and i want to be part of projects that refuse it, and specifically don't want to be part of projects that accept it
	I fully support a blanket ban. The people addicted to these tools are already heavily subsidized by venture capital and will be just fine if SourceHut forbears to contribute its own resources to the pyre.
	I use Sourcehut in part because of its strong moral principles, and a prohibition on vibe-coded projects feels perfectly inline with its existing stances on other topics.
	+1 from another paying customer. The ethical and practical reasoning align on this one, so as much as a ban would upset some people, I think we're already living in a world where a project's mere presence on Codeberg is a better trust signal than any Github star count, and that's the kind of model SourceHut could follow and benefit from as well.
	As paying users, we fully support this blanket ban, the whole rest of the internet will welcome these projects with open arms, let them go elsewhere.
	As a platform for the free software community, by humans for humans, it seems naturally to follow that this is not a platform for software not produced by humans, and I see no issue with expressing this conviction in an anti-ai policy.
	Big +1 on prohibiting vibe coded projects on SourceHut. LLMs have been incredibly harmful to any sort of creative community, as well as ones were facts are importants, and it so happens that software is a combination of both.
	I'm in full support of full and indefinite ban on anything-LLM at sr.ht. Thank you for writing this up.
	Another paying customer here, and I also fully endorse and support the exclusion of LLM use on ethical and epistemic grounds.
	I'm in favor. I wouldn't want Sourcehut staff to spend an inordinate amount of time trying to police it, or for users to go on witch hunts, but I think it would be good to have some wording in place to point people to. At the very least it would communicate "these are our values, if you don't agree with them you might be happier elsewhere".
	+1 the reason why i started paying and migrated to sourcehut was to be part of a platform that shared the same values as i do, lets make this the place to harbour software that we can trust and rely on to be developed in a sane and moral way
	+1 As Rek, Dev, and dozens of others have said, LLM tools intrinsically cannot be used in a good-faith way, and we should let projects that don't care move on to other platforms.
	I support this initiative to limit or ban AI generated code.
	+1 As a paying user I strongly support taking an uncompromising stance against LLMs / so-called AI* on the platform
	+1, I'm tired of regurgitative AI.
	+1 am even in favour of an even stronger policy, and just ban them all.
	+1 sounds like a good start.
	+1 Happily paying member, fully in support of a blanket ban.
	as a paying user I fully support a ban on hosting LLM-generated projects on Sourcehut.