Editorial recipe — not benchmarked — reviewed September 1, 2026
Turn a CSV or Parquet dataset into a reproducible decision brief
For analysts and operators who understand the source data and can validate the resulting metrics.
Run this recipe
Work through it here
Progress stays in this browser. The downloaded Markdown kit works in any notes app or repository.
0/5 complete
Procedure
- Freeze a read-only input snapshot, record its hash and data dictionary, and state the decision question plus expected checks before analysis.
- Use DuckDB to profile types, nulls, duplicates, ranges, and join cardinality; save every cleanup and metric in analysis.sql instead of editing source files.
- Ask Codex to create the smallest Jupyter notebook that executes analysis.sql, renders the necessary tables or charts, and records assumptions without hidden manual state.
- Restart the kernel, run all cells on the frozen input, export the notebook, and give the SQL, outputs, and data dictionary to Claude Code for a separate methodology and interpretation review.
- Have the analyst reproduce key totals directly from the input, resolve review findings, and label descriptive evidence, assumptions, missing data, and any causal claim separately.
Acceptance artifact
A frozen input manifest, executable SQL, restart-and-run notebook, exported decision brief, and resolved methodology review
Do not use it blindly
Do not upload restricted data, infer causation from correlation, or accept model-generated numbers that cannot be reproduced from the frozen input and saved queries.