Nothing in this drawer ships. It exists so the structure can be judged separately from the decisions that are still open. The five red groups below are the five decisions the concept spec's section 16 hands to you by name. The prototype does not resolve any of them — it builds both answers and lets you look at each.
OPEN · Lance · spec s.16 item 1 — “this decides the concept”
Your 2026-08-21 guidance — “one at a time is easier than five at once” — is
still in warnUnfinished() in the shipped code. The constraint behind it is gone
(photos are stored, research runs on a Queue), but the preference may not be. With this on, Home
grows a second, quieter entry that shoots exactly one thing and opens its card as soon as it is
ready. Same screens, demoted emphasis.
OPEN · Lance · spec s.16 item 2, cost argued at s.15.1
On auto, the gate in s.5.1 starts research the moment the photos identify the thing well enough, and a wrong read costs about $0.67 that has to be spent again — Fabian's honest estimate is one wasted run in ten to fifteen. On ask, the summary screen lists every thing with what it thinks it is and you tap once for the lot. Watch the Done-for-now screen and the Floor; both change.
OPEN · Lance · spec s.16 item 3
Push costs a service worker, a VAPID signer, a subscriptions table and one more permission prompt for your parents to answer — and on iOS it only works from the home-screen install. With it off the concept still works; it is just quieter. The banner in the prototype is drawn by the prototype and says SIMULATED on it. A real push cannot be proven here.
OPEN · Lance · spec s.16 item 4
Pile order is stable and re-derivable, which is the relearnable rule. Price order gets the money first and puts the same thing in a different place each time. Switch it and open Ready to post.
OPEN · Lance · spec s.16 item 5, risk at s.17
The vision pass is already looking at every photograph. It can see wear; it cannot see whether a thing works, and the prompt has to say so. Every condition chip in Sort changes with this switch, and so does the sentence under it.
Prototype scaffolding — not a decision
A real pile of six is two waves: three runs start at once, three more ninety seconds later, each run up to five minutes. The prototype replays a fixture event log per item on its own timer, which is what makes the lanes genuinely independent rather than one bar in six pieces.
Prototype scaffolding
Offline is the back corner of the garage: chips go to “waiting for signal” and catch up. Failing a run puts that thing into the third hold and gives its card Try again.
Hard requirement, not a fork
The whole middle is choreography, which makes this the screen most at risk of needing movement to be legible. Under Reduce Motion every lane, every pulse and every card transition computes to nothing and the identical information is still on screen. Force it here and walk the same arc.
Verification scaffolding
The device frame only appears on a wide screen. On a phone the prototype is simply the app.
Prototype scaffolding
With markers on, every string or treatment that is still somebody's decision carries a dashed badge naming who holds it. Turn them off to read it as a product.
Prototype scaffolding
Fixture state. Nothing is persisted anywhere but this browser, no photographs leave this device, no model is called, no price is real, no notification is real. Built from BKM/Projects/Resale-Intake/2026-09-12_Concept-B-The-Pile-Spec.md. The wheels, Ray-Ban and mixer searches, comparables and confidences are copied from the app's own live runs of 2026-08-22 and 2026-08-31; the golf clubs, the bowl and the drill are written, not captured. See the header of fixtures.js, field by field.