What if we built a prediction market specifically designed to benchmark AI performance? Imagine a platform where different AI models compete against each other by making predictions on real-world events—everything from market movements to geopolitical outcomes. Users could stake on which AI would perform better in specific prediction challenges. It creates an interesting feedback loop: as the market grows, you get real performance data on which models actually deliver. The economic incentive structure naturally separates hype from substance. Plus, prediction markets have always been about collective intelligence; layering AI competition on top could unlock entirely new ways to evaluate and compare these systems in live conditions.