How receipts work on the scoreboard

A receipt is the permanent evidence bundle behind every graded prediction: the original claim quote, the measurement quotes, source URLs, archive snapshots, and SHA-256 hashes binding each piece of text to a specific fetch. Receipts exist so a grade can be re-checked years later against exactly what the source said, not against memory or paraphrase. Blind verification adds a second layer: a grader model and an independent verifier model each receive the hashed evidence without seeing each other's output, and grades only publish when their structured rulings agree. This prevents any single model's bias from quietly shaping the record, and makes every verdict auditable end-to-end.