I mean you can make any set of LLMs produce any set of output you want. The problem is that it isn't terribly efficient and you have to filter between the real

I mean you can make any set of LLMs produce any set of output you want. The problem is that it isn't terribly efficient and you have to filter between the real stuff and the hallucinations.