Chapter
Loops of Agentic Engineering
The system you build around the model isn't a line — it's three nested loops (the value-factory model). A tight coding loop (red → green → refactor) sits inside a feature loop (specify → refine → verify), inside an impact loop (opportunities → hypothesis → impact). Each wider loop is slower, higher-stakes, and longer to close:
Each loop is built by specific tiers:
- Coding loop (red → green → refactor) — closed by Tier 4's tests; Tier 5 hardens that check so it runs every time.
- Feature loop (specify → refine → verify) — runs from the spec you shape in Tier 2 to that same Tier 4 check, which is why Tier 4 belongs to both.
- Impact loop (opportunities → hypothesis → impact) — Tiers 6–8 scale out to it, once one agent or one machine isn't enough.
Tiers 1 and 3 — prompting and context — are the groundwork under all three. Eight tiers, three loops: climbing the ladder is how you build the loops up.