/ce-compound
/ce-compound — “Close the loop by capturing learnings into docs/solutions/ so the next iteration starts smarter.” Step 6 of the loop and, per Every, “the money step… the whole point.” A council of research sub-agents (session-historian, pattern-recognition-specialist, security-sentinel, performance-oracle, best-practices-researcher, framework-docs-researcher, data-integrity-guardian) mines the just-completed work — its bugs, failed tests, and reusable solutions — and writes structured, schema-validated learnings to docs/solutions/ (artifact-solution-doc). Future ce-brainstorm, ce-plan, and ce-code-review runs read these notes as grounding (via the learnings-researcher persona), so “the next agent does not have to learn the same lesson from scratch.”
This is the capability that promoted stage-learn as a canonical stage. It gives learning-capture a second framework (alongside BMAD’s bmad-retrospective), clearing the ≥2-framework bar to split learning out of stage-release. It is the reference realization of pattern-knowledge-compounding — its output is consumed by the front of the next loop, making the lifecycle circular. See stage-learn for the full rationale.
Cross-framework equivalents
Learning-capture cluster: ce-compound ↔ bmad-retrospective — both harvest lessons at iteration close so the next starts ahead. Deliberately not an equivalent_to edge: the terminal acts differ (a maintained, machine-consumable solution corpus vs a one-off human retrospective). OpenSpec’s openspec-sync is a spec-level cousin (fold the change into the living spec) — noted on stage-learn.
See Also
- bmad-retrospective — the team-process flavor of the same stage-learn activity.
- ce-compound-refresh — keeps the corpus this produces healthy over time.
- pattern-knowledge-compounding — the signature technique; artifact-solution-doc — the output.
- stage-learn — the canonical stage this implements (and promoted).