Methodological framework to build AI agents that investigate first, focus on contradictions, and persist until work is done
