The loop samples claims seeded with honest priors, selects testing candidates via an acquisition policy, generates crafted edits to test those hypotheses, and updates Beta-distribution posteriors based on observed rank outcomes until rank 1 or convergence is achieved.