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’).

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.