One prompt, every major AI model, side-by-side. The responses are then compiled then fed into a 'Master LLM', that then fact-checks, summarizes, and suggest improvements.