I do wonder how this influences projects like CoMaps which can currently accept code predominantly written by LLMs/agents. I've heard that the contributor must understands the code and is able to explain it, though I cannot find where this is a real rule. They do disallow using such tools in autonomous-looking ways (they say ‘agent mode’ or ‘vibe coding’).
Despite these restrictions, it seems Codeberg's new rules could put pressure on repos like CoMaps if the proportion of contributions is largely LLM-written.
See: https://codeberg.org/comaps/Governance/src/branch/main/AI_US...
Despite these restrictions, it seems Codeberg's new rules could put pressure on repos like CoMaps if the proportion of contributions is largely LLM-written.