Good, bad, and half-true — every belief gets the same test. Each question puts its competing claims on trial, then settles them with a data-backed ruling graded by how strong the evidence is. And half-true doesn’t get to shrug: any claim holding a real share of the truth wears its staked percentage right on the stamp, and the stake ledger measures every position — the ruling included, which never claims 100 — on the same scale. Synthesized, argued, and revised when someone brings better data. No numbers at all means the claims didn’t earn any: the ruling is the ground truth.