You mean llm, large language models? Would be fascinating if a compile backend could generate follow-up questions though :)