Compound engineering
A development philosophy: each unit of engineering work should make subsequent units easier, not harder — the inverse of accumulating technical debt. Named/operationalized by Every Inc; introduced here via compound-engineering-plugin (the skillpack now lives in the sibling agentic-tooling-wiki, linked cross-wiki). Kept here as a bridge concept: it is Engelbart’s bootstrapping (cluster A) applied to agent-assisted software.
The idea
Traditional development compounds negatively (every feature adds complexity, every fix buries local knowledge). Compound engineering inverts this by front-loading planning and review (~80%) over execution (~20%) and, crucially, codifying learnings so they are reusable — “a good compound note means the next agent does not have to learn the same lesson from scratch.” Each cycle leaves better context for the next.
Where it sits in this wiki
- It’s douglas-engelbart‘s bootstrapping (“use the tools to improve the tools,” regeneratively) applied to software engineering with AI agents.
- It’s the engineering sibling of the compounding-knowledge thesis behind llm-wiki (a “compounding artifact”) and gbrain (“compound interest” on notes): same dynamic — knowledge/leverage accumulates instead of dissipating.
- At organizational scale it’s the enterprise-context-layer‘s core claim — “the tenth agent is dramatically smarter than the first” on accumulated, governed context — i.e. compound engineering as enterprise infrastructure (with explicit versioning/governance against decay).
- Mechanically it leans on spec-driven-development (plan before code) and agent-skills (the loop ships as reusable skills).
- Reflexive: this wiki embodies the same principle — every ingest makes the next question cheaper to answer; the synthesis and links are the “compound notes.”
Related
compound-engineering-plugin · douglas-engelbart · llm-wiki · gbrain · spec-driven-development · agent-skills