Experimentation Is the Missing Evaluation Layer for Agent Memory

A rewrite anchored in ZCL active experimentation: hypotheses, randomized assignment during context provisioning, outcome joins, uncertainty-budgeted e