LEARN CODING USA: REPRODUCIBLE TEACHING FIXTURE
All input records are synthetic, not Anthropic research data.
Requires Python 3.9 or newer; no dependencies. Also checked locally under the system Python version recorded in output/reproduction-verification.json.
From the extracted root, run:
python3 scripts/pipeline.py --input data/analysis-teaching-fixture.csv --output-dir output/run
python3 scripts/pipeline.py --input data/analysis-teaching-fixture.csv --output-dir output/tests --self-test
Expected cleaned values: A01=120,A02=120,A04=90,A05=90 seconds; mean 105, total 420.
Never change original input to make a check pass. Preserve baseline output before rerunning. Timing is not measured. One-machine two-interpreter reruns are not independent person or machine replication.
Original synthetic data: CC0. Original teaching code: MIT.
