Experiment

State the hypothesis, define a baseline and configuration, execute a reproducible run, collect agreed metrics, compare with the baseline, record inputs and limitations, and conclude whether the hypothesis is supported. Keep the design minimal until evidence requires more instrumentation.