Notes on verification.

How we think about calibration, provenance, and reading a document claim by claim.

Provenance over volume: why coverage is not confirmation

Eight outlets running the same wire story are one piece of evidence. Here is how we weight provenance so coverage never reads as agreement.

A confidence score is only useful if it is right about as often as it claims. We hold every reading to that bar.

Reconciling the narrative against the filed statements, figure by figure, before a number reaches the wire.

What open scientific claim verification measures, how the evaluation was run, and where the hard cases remain.