My primary reason for leaving GitHub was not about a single feature or a single outage, but about the “enshittification” of the platform under Microsoft’s ownership. The web interface got rewritten into a sluggish pile of JavaScript that either broke things which used to just work, or made them so horribly slow that using them became a PITA. Beyond the technical decay GitHub had turned into de facto “public infrastructure” in much the same way that WhatsApp has, hosting the source code of a very large share of the world’s software and, through that, giving Microsoft a degree of leverage and surveillance over everyone’s projects, and by extension everyone’s digital lives, that no single company should hold. On top of that, stories about legitimate developers losing their accounts due to arbitrary bans by Microsoft only reinforced the feeling that it would be a good idea to at least have a backup somewhere else.

Codeberg looked like a viable alternative. It offered free and open-source projects a reputable home and, more importantly, an equally free one, run by a non-profit association rather than a subsidiary of the largest software vendor on the planet. Unfortunately, the latest update to its terms of service seems to mark a first step in changing one part I moved there for, namely the “freedom” part.

Human stupidity

Every project I’ve published so far was built with 100% human stupidity rather than “artificial intelligence”, or, more accurately, LLMs. I don’t hold particularly strong feelings about Codeberg banning projects that are predominantly LLM-driven, at least not feelings as strong as the ones I hold about the simultaneous ban of legitimate cryptocurrency projects, which reads as though it got lumped in for no reason other than that most people still remember the villain-du-jour that crypto was in the years before LLMs took that title. The two clauses landed within days of each other, the LLM prohibition on the 29th of June and the cryptocurrency prohibition on the 2nd of July, both as Assembly 2026 proposals, and the terms now file the latter under, of all things, “content that harms the reputation of Codeberg”, which sounds like legalese for “we don’t have a solid reason or an actual number of bad precedents to categorically ban it”.

The announcement blog post, however, reads very poorly, and the section titled “The development team of none” is the worst of it. It states:

Using LLMs to work with your code gives you a kick of adrenaline. You can develop at a rapid pace, build things as if you had a large team. Only that you have none. In fact, you are (often) alone, working with a statistical machine that turns energy into code.

And, a little further down, it says:

It seems like many ‘vibe coders’ don’t realize that they don’t actually have a community around them.

This is out of touch with how most free software gets made. The majority of FOSS developers are one-man-shows, and the only cOmMuNiTy they have around them are the users requesting features or reporting bugs while most of the time not contributing in any form whatsoever. I’ve been publishing silly little tools for decades, predating this website and even GitHub itself (remember when SourceForge was the hot sh.t?), and not one of them has ever had an actual “community” around it, at least not in the romanticized sense that Codeberg paints in that post. I’m a lone wolf who codes everything by hand and spends an absurd amount of time doing exactly that, and the notion that an LLM is the thing separating a real project with a real community from a fake one does not hold up once you look at how the average useful little tool on any forge comes to exist in the first place.

It’s frankly a bit snotty of Codeberg to make this argument at all, considering that the platform effectively lives inside the Forgejo bubble, and Forgejo mutinied inherited its community of active contributors from Gitea, who had spent the better part of six years building that community before Forgejo even existed. A project that acquired its own community by hard-forking someone else’s, then turned around to lecture solo developers about not having one, is a difficult position to argue from with a straight face.

In addition, Codeberg conflates “having a community” with “being legitimate software worth hosting”, when the bar for a personal project has always been a working build, ideally a license, and maybe a README, and not a channel full of contributors. A good deal of what makes the small, single-author tool ecosystem worth having is precisely that it doesn’t need a community to justify its existence, and a forge whose entire selling point is hosting the code of individuals is an odd place to argue the opposite.

Censorship

The part that bothers me isn’t the specific ban on LLM projects, or the specific ban on cryptocurrency projects. It’s that a hub built around “free software” is now telling its users which kinds of software are deemed good and which are not, and that is closer to censorship than it might seem. Once a platform writes into its terms that an entire category “harms its reputation” and can be removed on that basis, the deciding factor stops being whether the code is legal, or functional, or useful, and becomes whether it aligns with a position the platform has taken. I would argue that a significant share of the projects caught by a blanket ban of that kind are legitimate software rather than vibe-coded slop or sh.tcoin implementations.

Every platform I can think of that took this approach became divisive the moment it started enforcing an ideology on its users, whatever that ideology happened to be, and however justified it looked at the time. The mechanism is always the same, where a real problem shows up, an unpopular category becomes the obvious culprit, the platform bans the category instead of addressing the problem, and that ban then becomes the precedent for the next category, and the one after that. The category that is uncontroversial to ban today is the reason the mechanism exists tomorrow, and the users who applauded the first ban rarely get asked about the second one.

I do acknowledge that both categories aren’t free of problems. LLM-driven repositories do strain infrastructure, do generate unmanageable volumes of low-quality issues and pull requests, and do carry real questions about copyright and code provenance, all of which Codeberg names in its post. The cryptocurrency space, in turn, might have produced more outright scams than almost any other corner of software. However, a categoric ban on the villain-du-jour is not a solution to any of that.

We now even have people like Linus Torvalds making the fairly reasonable argument that an LLM is just a tool, and “clearly a useful one”, with a legitimate place in Linux kernel development when it’s used carefully and its output is held to the same standard as everything else. If the maintainer of the largest and most consequential open-source project on the planet can treat LLMs as a tool to be judged on its results rather than a category to be banned on sight, a backyard code forge can manage the same.

I, too, am worried about the impact of LLMs on tech, and on society in general, going forward, and I’d guess I’m about as worried as whoever wrote Codeberg’s policy. I just don’t believe that banning content, which is very much what this amounts to, is the way forward.

Reasonable solution

What I wish Codeberg had reached for is a solution that treats the actual problem, which by their own account in that same post is resource consumption and the infrastructure cost that comes with it, as an actual resource problem. A change to the terms of service could have required authors to tick a checkbox declaring that a repository contains LLM-generated code, or is cryptocurrency-related, and those repositories could then be segmented onto a separate tier of infrastructure that doesn’t get the same resources as everyone else. A tier that carries specific quotas, and that might require the author to pay for what they consume. Declaring the truth honestly would (at least at first) cost nothing, and failing to declare it, then getting caught, could be met with exactly the permanent, immediate ban that Codeberg is now applying to entire categories from the outset.

Similarly, projects that carry the LLM or Crypto label could carry automatically displayed disclaimers that explicitly state that Codeberg is in no way responsible for the quality or correctness of this specific repository. Heck, they might even go as far as to blatantly state that Codeberg does not approve of the use of LLMs or Cryptocurrencies in those warnings, to make extra-extra-extra sure that people get it and that there is no “reputational risk” for Codeberg.

An approach like this puts the cost of resource-hungry projects onto the people creating them, and it keeps the shared resources for the projects that were the reason the platform exists. All of that without Codeberg having to decide which categories of software are ideologically acceptable in the first place. The “we ban everything upfront that we don’t agree with” approach is the wrong signal to send, and it is a very slippery slope.

Despite not owning a single project that falls into either banned category, I’m now going to look into setting up my own public Git host, and I’ll move off Codeberg only a few months after moving there, because of this. Not because of the bans themselves, but because I don’t want to depend on a platform that rewrites its terms of service on a whim, without properly announcing that the change was even under consideration, and without giving its users a way to weigh in.

The decisions did go through Codeberg’s own Assembly 2026, which is more process than most platforms bother with, and yet as an ordinary user I found out about it the way probably most people else did, through a dark blue banner at the top of the site on the day it was already settled. While I appreciate the info about the ToS change, I wish I’d gotten a banner back when the platform was still deciding whether to go down this road, and I wish it had linked to a discussion thread, or at the very least a poll, so that I could have voiced the concern I have, which is about the freedom of the platform as a whole, rather than about any single category that ended up banned.