Five-condition comparisons of AI coding workflows against a fixed
specification. Each card is one immutable CSV in data/reports/
— open one to see the metric breakdown.
python scripts/run_pilot.py for a pilot,
or auditor experiment for a real five-condition run.