The Fact & Gap Checker reconciles grounded research findings with stored claims by categorizing each discovery as confirming, contradicting, or novel relative to parametric recall. When an unprompted memory matches live research, the claim is marked correct and populated with supporting citation sources, while contradictory evidence tags the recall as incorrect. If a finding uncovers valid facts absent from the model's unprompted memory, it is minted as a new gap claim to track blind spots without overriding existing human rulings. Furthermore, if subsequent research matches an earlier gap claim, the system applies a regrounding rule that attaches fresh sources while retaining the gap verdict, ensuring research never self-validates without genuine model recall.