Multi-agent system that reasons, monitors, critiques, and updates trial success predictions in real time

Instead of one model making one guess, several AI agents work together like a team: one watches for new evidence, one checks the reasoning, and another updates the prediction as things change.