ed.1 · Sep 1 · inputs (≤2, DECLARE): the owner's seven · the Sep 1 prototype census

Ship List

Your seven plus three you did not name — the fourth, defining shipped, turned out to be already ruled and became gate zero. Different list from the build queue: that one is features, this one is the conditions for a stranger to use this without you in the room. Where I can count the mess I count it rather than describing it.

{{ s.k }} {{ s.v }} {{ s.sub }}
Gate zero — shipped is already configured, in Backend Build

You are right, and I was wrong to write this as a new definition. The implementation plan already configures shipped: Backend Build holds B0..B5 with a gate on every phase — B1 walking skeleton, B2 active layer, B3 sync and operations, B4 content migration, B5 cutover — plus T0..T5 for the twin and DEC-01..20 as the decision register. Those gates are SE exit criteria already written; B5's is "a week of real work on production." Shipped is B5 passed. Nothing here needs inventing.

What the plan does not yet carry is the release-side vocabulary that turns those gates into something a stranger can check. Four rows to add to Backend Build rather than to this page — and the fifth is the one that matters most for us:

Missing SE artefactWhat B0..B5 does not currently carryHow a stranger checks it
{{ e.k }} {{ e.v }} {{ e.how }}

Two consequences. A baseline means the release stops being whatever is in the tab — items outside it are out of scope rather than late, which is the only reason a list like this can end. And this page should not hold the definition: the five rows above belong in Backend Build as an ed.4, with the Ship List citing them. Writing a parallel definition here is precisely the wiki-versus-architecture drift that item 03 is about, committed on a brand-new page.

{{ ph.k }}

{{ ph.v }} {{ ph.tag }}
{{ it.n }}

{{ it.k }}

{{ it.src }} {{ it.size }}

{{ it.v }}

{{ it.num }}

Done when — {{ it.done }}
Item 10 — the exam, as an acceptance test rather than a quiz

You asked me to test you at the end. Taken seriously it is the cheapest acceptance test here: where you cannot explain a mechanism, that is a design defect rather than a memory lapse. A system its own owner cannot narrate cannot be operated by a stranger, delegated to a deputy, or defended to a regulator.

{{ ex.k }} {{ ex.v }}

Ten questions, spoken, no notes. Any answer needing the wiki open becomes an item on the build queue — that traceability is the only reason the exam is worth taking.