Reviewer / PilotLocal gate

Market-Ready UX Gate

The current falsifiable gate for NamicGreen pilot readiness: what is proven locally, what remains human-tested, and what reviewers must try to break.

Current verdict

The project is not yet declared market-ready in the full sense. The discipline layer is strong, Slice 1 is locally hardened, and browser smokes are green. The remaining gate is the complete human browser walkthrough plus the production Flow-A happy-path reproduction.

What is already proven

Slice 1 route/surface grounding

The workspace map, story reader, work item, panels, tools, aliases, invalid-story recovery, impact route, and debug-route scope are enumerated in the local route inventory.

Map entry no longer has an unbounded Core wait

The map route has a 1.2s SSR Core-data budget. Slow story data is classified as delayed, not invalid, and exposes retry plus Back-to-map recovery.

Mapbox lifecycle hardened

Layer events bind only after Mapbox load; cleanup unregisters only bound handlers. Map markers use GPU layers, larger hit targets, hover state, and localized status guidance.

Story exits are explicit

Reader and work-item surfaces expose Close case / Back to map. Exits strip stale story, panel, tool, focus, and creation URL state.

Trust File readiness remains Core-bound

Readiness UI explains blockers, but it does not replace Core/governance truth. Human gates, verifier behavior, and Flow A invariants remain preserved.

Map/List/Reader route state is canonicalized

List and reader transitions now write canonical URL state from one helper, preserving unrelated query parameters while clearing stale story-scoped state.

Reader readiness no longer overclaims publish readiness

Reader copy now says case text is ready, not that the Trust File is ready. Trust File readiness remains a separate Core/gate-driven state.

Generate actions are protected against duplicate runs

A local in-flight guard and disabled generating state prevent secondary Create/Regenerate entry points from presenting a running text creation as ready.

Human review decisions are explicit

The active review step now states that the next click records a human review decision in Core for later Trust File use.

Trust File proof freshness is localized and fail-closed

The Trust File drawer no longer shows stale local publication facts as current. Proof loading/unavailable states are localized in EN/DE/ES.

Local browser smokes passed

Headless Chromium proved dummy-login to map, Mapbox canvas visibility, map drag, Map -> List -> Story -> Map-tab exit, and Map -> List -> Story -> Close-case exit.

Still open before external claim

Human reviewer walkthrough

A non-technical user must still complete the workflow in a real browser on desktop and mobile: login, map, list, story, review, generate/read, Trust File, verify/share, and exit.

Production happy-path reproduction

Flow A is proven locally/hybrid and verifier policy is proven in production. Before an external pilot claim, reproduce the happy path against production with a real L2 bundle.

Production generate freshness

Production must run the current robust generate stack: no silent baseline, fail-closed degradation, real synthesis, hedge/lead fixes, and honest provenance.

Verifier distribution

For independent verification, NamicGreen needs a pinned/auditable verifier version plus the pinned trust anchor and out-of-band fingerprint.

Reviewer matrix

Review as a non-technical user, not as an implementer.

Each row is Pass / Fail / Concern with the exact route, screen, and one-line reason. A fail in navigation, review, Trust File, or API consistency is a pilot blocker until reproduced and resolved.

Navigation and exits: no dead ends; every route has a visible way back.

Map and case selection: markers/list/read/work item are understandable and stable.

Evidence and statements: a non-technical reviewer understands source, statement, status, and missing work.

Review/HITL: human decisions are deliberate and visible; the UI never implies that AI fulfilled a human gate.

Generate: progress is visible; output reads like a briefing and does not claim more than the sources support.

Trust File: blocked, ready, signed/L2, L1, and verifier states are truthful and clear.

Errors: no raw technical dead ends; every blocked state gives cause and next action.

Performance: map first paint and interactions feel usable; slow Core data does not trap the user.

Mobile: the same path works on a narrow viewport without hidden exits.

API/Flow A: UI states align with the external Trust File API story; no second truth.

Reviewer start

Start from the app URL and the instruction: "Please check whether this case can be shared as a Trust File." Do not explain AIJIM terms.

Browser proof

Desktop and mobile screenshots are part of the gate. Targeted tests do not replace the real user walkthrough.

Trust boundary

The UI may guide, but Core remains the trust authority. Any mismatch between UI label and Core truth is a blocker.

Reviewer package

What to hand over

Market-ready UX roadmap

documentation/MARKET_READY_UX_ROADMAP.md

Slice 1 route/surface inventory

documentation/SLICE_1_ROUTE_SURFACE_INVENTORY.md

AIJIM 1.0 tester walkthrough

/handbook/testing-1-0

Open AIJIM 1.0 tester walkthrough

Current state

/current-state

Open Current state

Scope boundary

Pilot-ready means workflow, trust states, and Flow A discipline. It does not claim empirical validation of all scientific axes. Analytical validity remains explicitly scoped as empirical-validation pending.

External signal rule

Do not say "market-ready" without the human pass.

The accurate signal today is: "The core trust and navigation substrate plus Slices 2-6 code gates are locally hardened and ready for reviewer falsification." After the human walkthrough passes with zero P0/P1, the signal can move to "NamicGreen pilot-ready for Flow A."