Reproducible Auto Research run examples with complete artifacts.
| Example | Goal | Metric | Iterations | Outcome |
|---|---|---|---|---|
| 01: Test Coverage | Increase test coverage from 65% to 80% | coverage_pct (higher) |
8 | 81.2% achieved |
| 02: Feature Add | Add dark mode toggle without degrading performance | lighthouse_performance (higher) |
6 | 0.92 score |
| 03: Performance Optimization | Reduce API P99 latency from 450ms to under 200ms | p99_latency_ms (lower) |
12 | 187ms achieved |
Each example includes:
- README.md — Scenario description, run initialization command, expected outcomes
- state.json — Run checkpoint showing full state at completion
- results.tsv — Iteration-by-iteration log with decisions and metrics
- report.md — End-of-run summary with key learnings
- Understand the loop — See how each iteration is verified mechanically
- Learn from patterns — Review kept vs. discarded decisions
- Reproduce locally — Run the initialization command to recreate the scenario
- Study artifacts — Examine state.json, results.tsv, and report.md formats
To add an example:
- Run a successful Auto Research iteration
- Copy relevant artifacts to
docs/examples/XX-name/ - Add entry to this index with brief description