Reflection
Learn from what happened. Reflection closes the loop. It analyzes outcomes against original intent, captures learnings, and feeds insights forward into organizational memory and future elaboration cycles.
Key Artifacts
- Analysis report comparing results to intent
- Documented learnings — what worked, what failed, and why
- Recommendations for future initiatives
- Updated organizational memory
Activities
Analyze Outcomes
Compare results to success criteria and original intent. Did the initiative achieve its goal? Where did it exceed expectations? Where did it fall short? What unexpected outcomes emerged?
Capture Learnings
Document what worked, what failed, and why. Learnings should be specific and actionable — not abstract observations but concrete patterns.
| Category | Example |
|---|---|
| Process | "Autonomous mode worked well for data migration but needed supervised mode for schema design" |
| Quality | "Brand review gate caught 3 issues that would have required post-launch correction" |
| Estimation | "Research units consistently needed 2x the expected bolt count for data collection" |
| Collaboration | "Stakeholder involvement during elaboration reduced revision cycles by 60%" |
Feed Forward
Update organizational memory, refine processes, and inform the next iteration. Feed-forward is not a passive archive — it actively shapes future work by updating quality gate configurations, refining mode selection heuristics, improving success criteria templates, and enriching domain models.
The Compounding Effect
Reflection is what transforms HAIKU from a one-time framework into a compounding advantage:
- Success criteria get sharper as teams learn to define them
- Quality gates accumulate, creating progressively robust enforcement
- Mode selection becomes intuitive with practice
- Organizational memory expands, giving AI richer context over time
- Bolt success rates improve as patterns are established
Ad-hoc approaches don't compound — each session starts fresh, each prompt is one-off, no organizational learning occurs. The team with HAIKU gets better at working with AI; the team without starts over every time.
The Loop Closes
Reflection feeds directly back into Elaboration. The next initiative begins with richer context, better-calibrated success criteria, and refined quality gates. This is the HAIKU lifecycle in action: a continuous loop of definition, execution, management, and learning.