Good ideas we have deliberately removed, deferred, or left dormant — kept here so a short-term scope call never loses a long-term idea. Every card says where the idea lands when it returns, why it's parked, and what triggers re-entry.
Side-by-side comparison of tray selections before attaching them to a job — a triage surface between Explorer results and the grid.
A $/acre contour layer on Explorer derived from the firm's Ring-1 verified sales — the defensible successor to the removed CAD-assessed heat grid (assessed values are a questionable basis for appraisal work; 1-d-1 ag-use divergence).
Per-firm assistant-usage visibility. Dropped from the tenant Billing panel because a visible turn meter reads as metered-billing-in-waiting beside "assistant included on every seat — never an upsell." The durable usage ledger still records it ops-side.
The removed inline candidate row (C6) had a kernel worth keeping: surfacing "matches your #1042 — merge or skip" where the appraiser is already looking, not only in the candidate lane.
Listed here for the record; each is marked in place.
Seats, pack line-items, usage tiles — good IA, out of launch scope by the domain-model ruling. The committed hook is the durable usage ledger.
v1 evidence kinds are ruled: sales · leases · cap-rate observations. List/ask prices remain candidate-lane metadata meanwhile.
Ring-2 catalog members not in the ruled TX launch layer set (StratMap parcels, FEMA NFHL, SSURGO, NAIP refs, Protomaps).
Photo capture on site, parcel-at-point from the truck, candidate confirmation on the go — a designed mobile experience that slots into the existing IA (Subject · Candidates · Home) without new navigation.
Ruled: stripped from v1. The 🔒 layer row is removed from the Explorer panel; listing photos surface only on the record/media pane, where the embed gate (the four-rights vector) governs them.
Ruled: Ring 1 + the firm's own staging lane only. Public-record discovery arrives via connectors → candidates on their sync cadence — never live federated search. The results dock now reads "from your verified database + your candidate lane," and candidate cards are the firm's own staging rows badged by feeder.
Ruled: bounded out of v1. The Ctrl K palette searches jobs, comps, and actions. Parcels live in county-sharded artifacts, and a keystroke-latency surface over them needs a search-index sidecar that hasn't been ruled or costed — Explorer remains the parcel-lookup surface. Topbar placeholder copy reconciled across all pages.
Cross-tenant comp sharing, opt-in and contribution-based. Parked for v1 by ruling, but the schema ships ready day one: per-record sharing-consent (default OFF), provenance, contribution lineage — so the co-op is a feature later, not a migration. Acres.com selling confirmed TX comps re-based the moat language; the whitespace is no longer uncontested.
The full pipeline feature behind the trainee role: supervisor linkage, an hours-by- property-type experience log, and supervisor sign-off in the shape state boards require for licensure upgrade. No incumbent treats the trainee pipeline as a first-class feature, and the industry's aging demographics make it a real wedge for growing firms. v1 ships only the non-signing assist participant + certification disclosure (gap-analysis C1, option C); this program builds on top of those participant rows plus the audit trail.
Tracker board, status chips, drag-to-move, activity log — deliberately deferred at kickoff; returns as an add-on pack. The event log and job graph are its hooks; Home's needs-attention feed carries the daily load meanwhile.
"Connect your own AI tools to your comp database" — machinery proven, parked post-launch; needs bridging to Firebase identities when it returns.
Full Argus-style DCF stays a partner-import boundary until commercial pack v2 (Wave 4); external-schedule import keeps DCF reports producible meanwhile. The Dallas GSA exemplar pins the rendering ceiling long before the engine exists.
Launch is single-active-editor with a visible, stealable turn lease + presence. True simultaneous co-editing is wanted later; the provider design must not foreclose it.