>compose LLMs in hierarchical agents with effective guard rails Could you elaborate? Is this related to the "teams of specialized LLMs" concept I saw last year

>compose LLMs in hierarchical agents with effective guard rails Could you elaborate? Is this related to the "teams of specialized LLMs" concept I saw last year when Auto-GPT was getting a lot of hype?