Generate a Deep Dive¶
/deep-dive produces a comprehensive technical reference document (DEEP_DIVE.md) after an investigation or experiment phase is complete.
When to use¶
Run /deep-dive after:
- A full
/ml-labinvestigation completes - A major experiment phase finishes
- You want a standalone technical reference covering implementation details
Generate the document¶
If experiment_path is omitted, /deep-dive looks for experiment artifacts in the current directory.
What it covers¶
The generated DEEP_DIVE.md includes:
- Data construction — how inputs were generated or collected
- Model architecture — design choices and parameters
- Scoring mechanics — rubrics, judge configurations, aggregation
- Statistical methods — tests used, multiple comparison corrections
- Per-test detail — individual test breakdowns with results
- Quality gates — validation checks and their outcomes
- Aggregation — how individual results compose into conclusions
- Key design decisions — sourced from journal entries and plan documents
How it finds information¶
/deep-dive surveys:
- Experiment scripts (
.pyfiles) - Results schemas (JSON/JSONL output files)
- Journal decisions (from
.project-log/journal.jsonl) - Existing documentation (REPORT.md, CONCLUSIONS.md, plan files)
It works with or without a project journal and degrades gracefully when artifacts are missing.
You have now...¶
Generated a comprehensive technical reference document that covers implementation details end-to-end, with sources cited from scripts, results, and journal entries.