The construction schedule
Toyota takes two to four years from groundbreaking to first vehicle, phases never skipped. Compressed here to quarters, with hard exit gates — no calendar dates, because dated plans rot and gated plans don’t:- Phase 0 — Stop the line. Known red andons outrank all new work: stale credentials retired, the export pipeline fenced until it carries real context. Exit gate: no known red andon.
- Phase 1 — Build the organization. The first compiled gates land — work-order schema, skill-gate enforcement, germline protection, handoff validation. Exit gate: a nonconforming work order physically cannot enter the queue; a germline file physically cannot change without review and changelog.
- Phase 2 — Trial production. Five pilot units, hand-built through all seven stations — cardboard engineering with real cardboard — the standard work revised after each. Exit gate: one unit of each family holds a valid certificate; the standard work’s changelog has five or more entries; first-pass yield is measured, whatever it is.
- Phase 3 — Run at rate. Sustained takt — 3 to 5 certified units a week — with WIP limits enforced and the silence andon live. Exit gate: one full cycle at takt with zero silent stalls — every gap in flow produced an alarm, and every alarm produced a response.
- SOP — Start of production. The line runs; kaizen begins metering its 10% budget; the pilot team hands daily quality to the workers it built; the operator’s attention retreats to the two takt-bound touches plus the obeya.
Growth changes the flow, not the plant
Expect scope to grow roughly tenfold during construction — Toyota’s greenfield plants did, without the timeline slipping, because the layout carries high ceilings and empty sockets. Concretely: the certificate schema carries fields not yet filled, the station interfaces are family-agnostic, the deploy adapter is one file. Growth changes what flows through the plant, not the plant. What deliberately does not change at any scale: the gates, the independence of verification, and the operator at the head of the plant bearing residual risk — someone must, and that is why a human principal is structural, not transitional.What the factory deliberately does not build
An overproduction list, enforced by its own absence from the queue: a multi-tenant external product, a second runtime, a public benchmark suite, bespoke memory infrastructure, further doctrine research. The corpus is complete. The factory is behind the corpus. Every unit of effort that widens the gap is waste.Self-audit
| Claim | Source |
|---|---|
| The one-paragraph definition of Zookooree | docs/factory-design.md:266 |
| Phases 0–3 and SOP with exit gates; dated plans rot, gated plans don’t | docs/factory-design.md:261–275 |
| Five pilot units; cardboard engineering; retro per unit | docs/factory-design.md:237 |
| ~10x scope growth; high ceilings and empty sockets; growth changes the flow, not the plant | docs/factory-design.md:277 |
| Residual risk: a human principal is structural, not transitional | docs/factory-design.md:181 |
| The deliberate do-not-build list | docs/factory-design.md:279 |