Healthcare AI benchmarks are saturated, and the evaluation that answers “does this help?” looks like product analytics — in private. One argument, in three acts and an appendix.

Anatomy of a visit: the question stream, the scope, the same patient twice, and the question that never gets asked.

The four eras of measuring medical AI, and the one chart with two questions.

Medicine runs on tickets. The ticket nobody closes. No one grades the outcome.

Every card, sparkline, and source behind the chart.