Perhaps there should be domains of focus for the test LLMs - even if they are clones, but assigned to only a particular domain, then their results have to be PR

Perhaps there should be domains of focus for the test LLMs - even if they are clones, but assigned to only a particular domain, then their results have to be PR's etc... Why not treat every LLM as a dev contributing to git such that Humans, or other LLms need to gatekeep in case something like that happens? (start by treating them as Interns, rather than Professors with office hours)