Write and audit the report

The report is the final explanation of the investigation. It should make each claim easy to trace and keep the boundary of the evidence honest.

Goal

Write and audit a concise report that explains the question, implementation, checks, selected candidate, visible numerical and visual evidence, one limitation, and what Mathematics must still establish.

Inputs

Use output/experiment_config.json, output/generated_data.csv, output/search_scores.csv, output/selected_result.json, output/fit_and_differences.png, output/run_manifest.json, and the replay agreement or mismatch record. Read the saved files rather than relying on console output or memory.

Deliverables

Complete report.md with these sections:

1. Question and setup: state the supplied numerical question, input construction, configuration, and candidate grid.
2. Implementation and checks: explain the prediction, score, data generation, complete search, tie rule, and checks used.
3. Selected candidate: state its parameters, score, grid position, and one reconstructed prediction and signed difference.
4. Evidence: point to the generated data, score table, selected result, replay record, and the figure's separate source, candidate, and difference views.
5. Limit: state one specific limit, such as the candidate grid or bounded generator, and explain what Mathematics still needs to establish.

Audit every numerical statement against a saved artifact. Say “the candidate with the smallest score in this supplied grid” rather than “the best line” without that qualification. Do not introduce regression, probability, generalization, or causal claims that the project does not establish.

Checks

Check that the report names a real configuration and selected grid position, that the score agrees with selected_result.json, and that the stated replay conclusion agrees with the replay record. Follow every referenced path and confirm that the figure uses the saved arrays.

Check that the report mentions the tie rule, one limitation, and the difference between generation evidence and candidate-selection evidence. Remove any claim that extends beyond the supplied grid and recorded runtime.

Workspace

Keep the report at the project-level report.md, the stable path used by every milestone. Do not duplicate the full arrays in prose. The report explains the durable files; it is not a replacement for them.

Hints

Review

Read the report as a new reader. Can they tell what was supplied, what you implemented, which candidate was selected, how the result was checked, and what remains outside this project? Replace decorative wording with evidence or remove it.

How to check your work

Checks compare the report structure and evidence links with the supplied fixture. The supplied fixture is an exemplar of traceable writing, not a conclusion to copy without checking your own run.