Configure your models in the sidebar, then start chatting. AI 1 responds, AI 2 reviews, AI 1 synthesizes the final answer.