Which change earned the improvement?

Create a factorial comparison with identical cases across configurations. Report interactions and uncertainty across cases.

Hypothesis and criterion

Name the expected effect, metric and what would count against the hypothesis.

Method and reproducibility

Record fixtures, source, version, parameters, units, controls and how cases are split.

Paired evidence

Case and split Baseline result Changed result Interpretation

Counterexample and revised design

The benefit of filtering depends on whether retrieval introduces hostile content; component effects are not simply additive.

Record your new case and rerun the original cases after redesign.

Individual defence

Explain the mechanism, one exact result and what would overturn your conclusion.

Remaining uncertainty and handover

Case variation is not automatically independent sampling. Identify the population before interpreting uncertainty.

Next test: _ . Project filename: _ . Work that is mine and tools I used: ____ .

Keep your work

Use fictional data. Download a resumable project before changing devices. On a shared device, turn remembering off and clear your work when finished.