I Regret Migrating to Codeberg: Concerns Over New Terms Banning LLM and Crypto Projects

Overview

The author regrets migrating to Codeberg because the platform’s updated terms of service now ban projects that are predominantly LLM‑generated or cryptocurrency‑related, which they view as a move away from the freedom that motivated leaving GitHub.

Background: Leaving GitHub for Codeberg

The primary reason for leaving GitHub was not a single feature or outage but the perceived “enshittification” under Microsoft ownership: a sluggish, JavaScript‑heavy web interface, the platform’s role as de facto public infrastructure giving Microsoft leverage over developers, and stories of arbitrary account bans. Codeberg appeared as a viable alternative—a free, open‑source host run by a non‑profit association rather than a corporate subsidiary.

Codeberg’s New Terms and Bans

In June–July 2026 Codeberg adopted two Assembly 2026 proposals: a ban on projects that mostly consist of code written by generative AI tools (effective 29 June) and a ban on cryptocurrency‑related projects (effective 2 July). The terms now file the cryptocurrency ban under “content that harms the reputation of Codeberg”. The associated blog post argues that using LLMs to work with code gives a “kick of adrenaline” but leaves the developer alone, and claims many “vibe coders” lack a real community.

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.

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

Critique of the “Lone Wolf” Argument

The author points out that most free‑software developers are solo maintainers whose only community consists of users filing issues or requesting features, not contributors. They note that Codeberg itself inherited its community from a hard fork of Gitea (via Forgejo), making the lecture about lacking a community seem hypocritical. The bar for a personal project has always been a working build, a license, and a README—not a channel full of contributors.

Concerns About Censorship and Precedent

The author’s main worry is not the specific bans but that a platform built around “free software” is now deciding which kinds of software are welcome, turning the deciding factor from legality, functionality, or usefulness into alignment with a platform‑taken position. They warn that once a category is banned on the basis of “harms reputation”, the mechanism can be reused for future categories, creating a slippery slope of ideological censorship.

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.

They acknowledge that LLM‑driven repos can strain infrastructure and generate low‑quality issues, and that the crypto space has many scams, but argue that a categorical ban on the “villain‑du‑jour” does not solve those problems.

Proposed Reasonable Solution

Instead of outright bans, the author suggests treating the actual problem—resource consumption—as a resource issue. Repositories containing LLM‑generated or crypto‑related code could be required to declare that fact, then be placed on a separate infrastructure tier with specific quotas or payment requirements. Disclaimers could state that Codeberg is not responsible for the quality or correctness of such repositories, and could even note that Codeberg does not approve of LLMs or cryptocurrencies in those warnings. This would shift costs to the creators while preserving shared resources for other projects, without requiring the platform to decide which software categories are ideologically acceptable.

Community Process and Transparency

The changes went through Codeberg’s Assembly 2026, which included a meeting and an email ballot. However, as an ordinary user the author learned about the change only via a dark blue banner on the day it was already settled, wishing they had been notified earlier with a link to a discussion thread or poll so they could weigh in on the freedom implications.

A comment from @Aachen notes that the assembly meeting allowed clarifications but no discussion, and that many members likely only received the email “Wanna ban AI? Vote now!”. After the results were announced, questions were moved to a ticket that was then shut down, with users directed to a Matrix chat room.

Personal Decision to Self‑Host

Despite not owning any project that falls into the banned categories, the author is now looking into setting up their own public Git host and will move off Codeberg only a few months after migrating there, because they do not want to depend on a platform that rewrites its terms of service on a whim without proper announcement or user input.

Commentary from Hacker News

• Some users defend Codeberg’s right to set its own hosting rules, comparing it to private instances or community spaces (@tebruno99, @anticorporate).
• Others worry about arbitrary enforcement, noting the difficulty of identifying LLM‑generated code and the risk of bans based on vague evidence (@BadBadJellyBean, @rmast).
• Several commenters point out that Codeberg already limited non‑FOSS projects and that the new rules are an extension of its existing stance (@embedding-shape).
• A few mention practical issues such as site lagging, overlay warnings when JavaScript is enabled, and a “Collaborative” repository filter that may hint at further restrictions (@Yokohiii, @Rendello, @animesh).
• Many suggest self‑hosting Forgejo or Gitea as a way to retain full control (@periphery, @chlorion, @tehabe).
• One user highlights the irony of banning LLMs while Linus Torvalds has described them as a useful tool when used carefully (@vintagedave).

Conclusion

The author’s regret stems from seeing Codeberg adopt a policy that treats entire categories of software as unacceptable based on reputation concerns rather than concrete harms, which they view as a threat to the freedom that initially drew them to the platform. They advocate for a solution that addresses resource usage directly, preserves user autonomy, and maintains transparency in policy changes.