Codeberg bans vibe coded projects

Original Article ↗ Hacker News Discussion ↗

Codeberg, a German non-profit code hosting platform, has updated its terms to prohibit projects that “mostly” consist of code generated by tools like Claude or OpenAI Codex, citing unclear copyright status and a higher risk of harmful or low-quality code. Commenters are sharply divided: supporters welcome a human‑written, AI‑free refuge and argue that Codeberg’s limited resources shouldn’t be spent storing disposable “vibe coded” repositories, while critics see the policy as ideologically driven, vague, and practically unenforceable. Many raise worries about how “mostly” will be defined or policed, and suggest that the rule will both incentivize hiding AI use and push AI‑assisted projects to other forges such as GitHub or self‑hosted alternatives.

Scope of the New Policy

  • Codeberg updated its ToS to ban projects that “mostly” consist of code written by generative AI tools (examples given: Claude, Codex).
  • Official justification in the text: unclear copyright status of AI-generated code and higher risk of harmful code.
  • Codeberg positions itself as a non-profit focused on free/open-source software, and some see this as consistent with hosting only clearly licensed, human-authored code.

Definition & Ambiguity

  • “Mostly” is undefined: is that >50% of lines, commits, files, or something else?
  • Unclear whether stepwise autocomplete or IDE-generated boilerplate counts as “AI-written.”
  • Several commenters note that serious projects may be heavily AI-assisted but still carefully designed and reviewed by humans, blurring the line.

Enforcement & Detection

  • No scalable, reliable way to detect AI-generated code is identified.
  • Proposed signals: “Co-authored-by” trailers to AI agents, AGENTS.MD/CLAUDE.MD files, characteristic verbose comments, or em-dashes in comments.
  • Many expect enforcement to be complaint-driven and selective, with incentives for users to hide AI usage.
  • Some argue the policy mainly gives Codeberg a ToS hook to remove problematic projects or accounts if needed.

Supportive Views

  • Supporters welcome a “human-written” refuge from “AI slop” and low-effort, agent-generated repos.
  • Some emphasize limited nonprofit resources: hosting vast volumes of disposable AI projects, backups, and scrapes is seen as unsustainable.
  • Others frame it as a principled stand for software freedom and clear licensing, and as a way to keep quality and maintainability higher.

Critical Views

  • Critics call the rule vague, ideologically driven, or “virtue signaling” against AI.
  • Concerns:
    • Rules based on how code was produced rather than demonstrated harm or infringement.
    • Punishing honest disclosure, while undisclosed AI use goes undetected.
    • Risk of driving away users as AI-assisted coding becomes widespread.
  • Some argue copyright concerns are overstated; AI output may be public domain or clearly attributable, and public-domain code is still “free.”

Broader Reflections

  • Debate over whether AI-written code will soon dominate, vs. continued niches of hand-written code.
  • Some see this as one community choosing a “human-first” space; others foresee migration to more AI-friendly forges or self-hosted instances.