It would be far more efficient to have participants prompt LLM agents and have those agents debate one another. Who cares about what humans consider correct or logical in a world where LLMs will do most of the decision making and reality fabrication anyway? It's like asking a cow its opinion on the dairy industry.
That's an interesting idea for sure, a more relaxed mode. But then again, same could be said about video game championships, an aimbot would shred everybody. There's this kinda beauty in human's still trying to do stuff