Anthropomorphizing also reaches it's limits. LLM's are not humans and have different strengths and weaknesses.

I believe agents can keep a check on each other. I have had better results running multiple agents