The owner's current P6-00–P6-05 instruction supersedes the seven-checkpoint proposal retained below. Execution started at fetched/synchronized 9d6d29dba4733778e478909d4ff74a25409b3115 on codex/ui-redesign-phase4, preserving seven modified source/test files. Execution plan and commit log track actual work.

Current checkpoint	Deliverable
P6-00	Baseline, all requested surfaces and prioritized design consistency register; pushed `7da083c`
P6-01	Shared semantic, contrast and Settings panel corrections
P6-02	Workflow/error/retry review and approved AI patterns
P6-03	Fresh locale/theme/responsive/keyboard evidence and axe triage
P6-04	Native performance and component reuse review
P6-05	Integrated acceptance, 57-group preservation and Phase 7 plan

Final decisions and today's fresh counts are in P6-05; inherited results below are attributed history. Production remains NO-GO, main unchanged. P5-13, P5-05/followMode, human reader/physical-device and production qualification remain deferred or unverified. No Phase 7/8 implementation or deployment is authorized by this review.

Historical transition/planning record

Review-planning readiness: CONDITIONAL GO. Phase 5's implemented local scope can enter bounded design/UX review; the originally planned full scope is NOT COMPLETE and production release remains NO-GO. This is the transition/planning checkpoint, not Phase 7 migration or a release authorization.

Verified handoff

Phase 5 final acceptance baseline: 9528504866054da10e20d5196e65de536c4c9499, development codex/ui-redesign-phase4, committed/pushed with remote=HEAD and a clean tree. Final application change: 22455bba0a016f0cefabde46c539df588f789667. Production local/remote main remains 27657f31131cd95d42f0e1290853ec9c9b3959a4; a main push triggers deployment and is prohibited while NO-GO.

Authority: original Implementation Plan, Phase 5 plan, final acceptance, 57 original-group reconciliation, 32 routes, pushed-unit ledger, remaining findings and production gate. Historical failures and incomplete acceptance are retained.

Inherited fresh gate	Result
Backend / frontend / ordinary browser	1,033 / 498 / 148 passing
Separate diagnostics / sidecar / focused AI / real HTTP	3 / 49 / 108 / 2 passing
Revision safety	5/5
Standard fidelity	16/16, 36 preview=36 PDF pages
Settings exports	8/8, 24 preview=24 PDF pages
Rate-limit / ingress / PDF admission / configuration	9 / 6 / 1 / 3 passing
TypeScript / lint / builds / secrets / links / whitespace	Passing; exact native commands/times/fingerprints in final manifest
PRES / routes	0PASS / 56PARTIAL / 0FAIL / 1DEFERRED =57; all32 route patterns remain PARTIAL for full acceptance

Counts are separate lanes/subsets, not summed into a fictional total. Final ordinary/native runs have zero skips, expected failures, flaky results or retries. P2-D01 is ordinary passing. This planning-only change does not claim another full runtime run; document/link/secret/whitespace validation is performed for the handoff.

Accepted implementation boundary

Shell/navigation and utilities, Jobs Table/Board, Dashboard/Analytics, Job header/Overview/deeper sections, CV frame/navigation, section editing/source clarity, preview chrome, template/document controls and scoped responsive CV workflow have representative implemented/native evidence. All original routes/capabilities remain; later visual migration is not represented as a removed workflow.

Five new audit barriers were repaired independently: keyboard-native file selection, hidden-initial preview measurement, drawer focus/return, CV-to-Career navigation and report target sizes. Two P1s and three P2s are no longer demonstrated failures on their tested paths. No PDF geometry, API status, ownership, source/override/history, submitted package or provider semantics were changed for appearance.

The transition is conditional because whole-workflow/human/device/deployed acceptance is incomplete. It authorizes review planning and finite inspection of implemented scope, not certification of every capability. A newly demonstrated critical security/data or severe accessibility failure requires containment and a separate bounded repair before dependent work; it must not be dismissed as design preference.

Deferred scope and unresolved acceptance

P5-13 remains DEFERRED / UNQUALIFIED / Production-withheld. Actual deployed model/runtime identity, reachable inference, unchanged366 calls, independent human semantic review, AI review UI and final integration remain outstanding. Zero configured-target calls/zero human reviews; frozen corpus and deterministic authoritative grounding/stale/tenant guards retained. Production service registration remains withheld; no automatic activation. Existing unrelated AI is retained.
Production hardware is owner-reported Ubuntu24.04/GTX1060 6GB/32GB RAM, distinct from Windows/RTX3080 development. qwen3.5:4b-q4_K_M is only a candidate. No actual driver/runtime/offload/throughput qualification or Docker recovery was performed. Separate approved feasibility/qualification work must precede target selection and activation.
Human reader acceptance: NOT COMPLETE / NOT VERIFIED. Owner postponed available-operator work; no actual reader/version/device/spoken results supplied. Physical touch, native dragging, software keyboard and mobile AT also remain unverified. M01–M13 operator packet remains mandatory.
Fresh312axe scans retain39all-enabled violation-node occurrences and2,367incomplete occurrences. Contrast/meaningful initials, heading/landmark structure, drawer AX names, panel associations and broader reflow require contextual disposition; no zero-violation compliance claim. Actual Chromium100/200/400 zoom and emulator input are finite evidence only.
Production Linux font/portrait/asset qualification remains open despite passing local16/8 corpora and four tiny synthetic photo exports. Do not widen tolerances or infer production metric equivalence.
Historical preference500/notification contention causes remain open: twelve native concurrent rounds pass but do not diagnose the old exceptions or clear swallowed language-save errors. CV Unsaved incident remains open; the800ms test boundary is qualified, historical trace absent. Other readiness rows retain their own status/owner.
Board200/1,000 exact membership passes locally (render1,040/1,612ms), but O(N) full-array capacity and physical dragging remain unqualified. No supported-load/memory/latency budget exists yet.
P5-05 quick preview and followMode reviewed compose UI remain explicitly deferred; full Job entry and compatibility query/state round-trip retained. No fake AI/compose/preview screen is included in acceptance.
F04, current complete backup/isolated restore, actual configuration/schema/secrets, rollback and deployed smoke/monitoring remain release blockers. No production network/service/data change was performed. Prepared operator commands are not verification.

The complete readiness register retains other provider, domain concurrency, company-dialog error, worker, deletion/billing and cutover obligations. Phase 6 does not erase them or start B07.

Objectives and reusable foundations

Use the original Phase 6 scope: consistency, usability, information density, light/dark quality, responsive behavior, Norwegian localisation, existing AI patterns, component reuse and performance before broad rollout. Review actual implemented workflows and representative empty/loading/error/conflict states, not idealized screenshots.

Reuse Phase 4 foundations: semantic theme/focus/status colors, typography/spacing, named modal controls, deliberate preview entry/exit, PageLayout/section primitives and Comfortable/Compact table density. Existing MUI controls, shell, Jobs query state, Job workspace sections and CV chrome are the starting point. Add no second saving/undo/navigation/renderer system and no new framework for stylistic preference.

The rendered document is a separate physical contract from application chrome. Any review-driven change touching CV preview/template/layout must keep both unchanged fidelity suites and revision/override/frozen-package evidence green; ordinary UI reflow must not alter printed geometry.

Proposed bounded checkpoints

Checkpoint	Scope and entry	Required evidence / exit	Dependencies / push boundary
P6-00 — Handoff and review programme	This record; verified final Phase5 baseline and scope rulings	Accurate baseline, complete deferred/manual/release obligations and review order	Planning-only documentation commit/push; no release/migration
P6-01 — System consistency and workflow review	Next execution checkpoint: implemented shell, Jobs, Dashboard/Analytics, Job and CV surfaces	Representative route/state/locale/theme comparison; hierarchy/density/control/action consistency, duplicate/dead affordances; prioritized findings tied to PRES and existing tokens	Review first; any correction gets reproduction, focused regression and independent small commit/push. No broad redesign
P6-02 — Usability and information density	P6-01 findings and existing product capabilities	Complete task journeys, progressive disclosure, navigation/orientation, long-data and source-scope clarity; no feature removal or persistence change	Product/design owner; record actual user/manual evidence separately from agent review
P6-03 — Locale/theme/accessibility closure	Fresh triage and M01–M13 packet	EN/Bokmål labels, meaningful contrast, settled drawer names, headings/panels, enabled-action reflow; real reader/device/software-keyboard outcomes before complete acceptance	Qualified human/device operators; unavailable checks stay NOT VERIFIED. Bounded severe-barrier fixes before dependent rollout
P6-04 — Reuse, capacity and rendering environment	Existing components, finite Board and16/8 rendering baselines	Agreed performance/account/device envelope, measured CPU/memory/query/render cost; production-equivalent font/portrait tests; avoid duplicate components	Platform/CV/operator. Any virtualization/renderer change needs its own compatible design and membership/fidelity regressions; no speculative migration
P6-05 — Existing AI interaction review	Currently approved/operational AI only	Scope/provenance/review/failure conventions and minimum-context privacy; retain account/provider/Production-withholding guards	Deferred P5-13 is visibly excluded, not completed. No model calls, invented UI or activation from a consistency review
P6-06 — Review acceptance and Phase7 handoff	Prior findings/remediation and required evidence reconciled	Fresh affected/full regressions,57groups/32routes, three separate scope/full/production decisions, unresolved work and migration order	Phase7 starts only after a separate reviewed handoff; production GO is an independent operator/release decision

Manual/operator work may run alongside independent review; unavailable evidence does not repeatedly halt unrelated inspection. It cannot be impersonated by an agent or silently converted into PASS. Preserve independently reviewable commits and immediate remote synchronization; never force-push or touch main without a separately verified authorized release decision.

What Phase 7 inherits

Phase7 will inherit the verified representative design language, original preservation/route inventory, compatible shared foundations, measured document renderer, test fixtures and unresolved gates. The original full-migration list still includes remaining authentication/onboarding/Career/library/Documents/application/mail/Notifications/Settings/Admin/error/dialog states. Already migrated representative Analytics/Settings portions are retained; remaining coverage must be reconciled explicitly rather than rebuilt or assumed complete.

Before that migration: resolve system-level findings with bounded proof, complete required reader/device/rendering acceptance or keep accurately scoped conditional decisions, and retain all production prerequisites. No Phase7 implementation, B07, P5-13 return work or deployment begins in this planning checkpoint.

README.md Unescape Escape

Phase 6 — Design and UX Review

Current execution — 2026-10-11