Les frictions

Vigie — Phase 0 friction log (Lookout Arrival)

Date opened: 2026-09-16 Last verified: 2026-09-16 (instrument re-walk post Phases 1–6 + DESIGN) Owner: Bucky + le porteur

Phase 0 verdict (honest)

Phase 0 is incomplete. Shipping Arrival before resident walks was a deliberate instrument-first bet. It does not satisfy the canvas DoD (“5 resident walkthroughs recorded; pain ranked”). Do not mark Phase 0 done until the resident table below has five real rows.

Refuse: inventing resident quotes, personas, or fake pain scores.

Baseline — first paint (before Arrival redesign)

Method: structured co-founder walkthrough of then-live public/index.html (desktop + mobile CSS).

Kill list (binding — must not ship in hero / Arrival)

• For You / personalized feed
• Ranking weight changes (w_impact stays 0 until ranking.md gates)
• Bias / trust meters
• Hero stats strips, score worship in Arrival
• Second ranking brain for Approaches (store order unless declared FACETS.md opt-in)
• Invented news or crowning a voice for cleaner UI

Instrument re-walk — live Arrival (2026-09-16)

Evidence: public/index.html DOM order + CSS after beauty-without-fog polish. Not a resident substitute.

Remaining instrument pain (ranked — co-founder only)

Resident walkthrough protocol (required for Phase 0 close)

Goal: five Quebec City–life residents (or near-daily visitors), not co-founders. Task: open the live French brief cold; think aloud ~5–8 minutes; no coaching. Surface: the deployed front door — https://vigieqc.com (custom domain; the alias https://vigie-deemto.vercel.app also routes to production). Aligned 2026-09-18: this script replaces the old English workbench walk (Arrival/Approaches/Life facets/Stage); the questions now follow the live brief’s sections.

Script (same for R1–R5)

1. Open / on their phone or laptop (note which). No tab of the site already open.
2. Ask: “Sans faire défiler — qu’est-ce que c’est ?” (“Without scrolling — what is this?”)
3. Le point (#essentiel): “What does this tell you about your city today?”
4. Ce qui a changé (#changements): “Can you tell what moved since the last edition? Does the ‘Suivi depuis…’ line on a dossier mean anything to you?”
5. Travaux et entraves (#travaux): “You’re crossing town tomorrow — what here would help? What’s missing?”
6. Les dossiers (#dossiers): “Who spoke / who stayed silent — can you tell? Does the silence line change your trust?”
7. Repères utiles (#agir): “Would you tap any of these? Which? Why?”
8. Ask: “What would you tap first? What would you ignore?”
9. Stop before pitching Vision.

Sessions may be remote (screen share) or in person. Record only with consent; quotes land in the table below only with permission. No session, no row — the table stays empty rather than filled with invented residents (kill-list law).

Record row (fill only with real sessions)

Pain ranking rule (after R1–R5)

1. Count tag frequency across five walks.
2. Weight by median severity.
3. Publish a new table “Resident pain ranked” in this file — replace instrument rank where residents disagree.
4. Only then check [x] on Phase 0 DoD.

Definition of done — Phase 0

• [x] Friction table published (baseline F1–F8)
• [x] Kill list published and still binding
• [x] Instrument re-walk logged against live Arrival (2026-09-16)
• [x] Residual F8 closed on instrument (ops behind details)
• [ ] Five resident walkthroughs recorded (0 / 5)
• [ ] Resident pain ranked and published here

Phase 1 status (2026-09-16)

• [x] Arrival first viewport: brand · line · Approaches · CTA
• [x] Lookout field hidden until invitation / Approach tap
• [x] No For You; w_impact unchanged (0)
• [x] Tests in tests/test_arrival_lookout.py

Phase 2 status (2026-09-17) — Approach objects

Canvas DoD: Replace card lists with Approach (geo + unit + voice count); Unit/silence visible without opening Stage.

• [x] Approaches are interaction objects (buttons), not archive cards
• [x] Glance: nest (geo) · voices · silent count · official quiet · units or “no units yet”
• [x] Official Quiet name preview on Approach (≤2) — silence glanceable without Stage
• [x] Ambient morning digest shares same meta chips + silence preview
• [x] Store order unchanged; w_impact still 0
• [x] Tests: tests/test_phase2_approaches.py

Phase 2 DoD: MET (instrument).

Phase 3 status (2026-09-17) — Fight theater

Canvas DoD: Stage as calm confrontation; institution voices; silence arc; who spoke / who didn't in <10s.

• [x] Stage = fight-theater: Spoke + Did not speak arc before voice panels
• [x] Institution names (not feed-desk archaeology) on arc + voice panels
• [x] Dense status/items/sources meta soup removed; glance chips only
• [x] Faces demoted under links (name + claims first)
• [x] Side zones soft-dim when field open; center Stage emphasized
• [x] No bias meter / For You; w_impact still 0
• [x] Tests: tests/test_phase3_fight_theater.py

Phase 3 DoD: MET (instrument).

Phase 1 verification — Arrival Lookout (2026-09-17)

Canvas DoD: No dashboard in first paint; Lighthouse mobile pass.

Check (live public/index.html)

• Composition: brand → line → Approaches → CTA; facets/ops after CTA (collapsed).
• Field: #field-shell ships hidden + aria-hidden="true"; openField() reveals; CSS display:none until body.field-open.
• Site footer deferred until field open (not in first-paint surface).
• Skip link, viewport, meta description, favicon.svg, async font load.
• Kill list holds on Arrival; W_IMPACT=0.
• Tests: tests/test_phase1_arrival.py + suite green.

Lighthouse mobile (Edge headless → http://127.0.0.1:8767/)

Recorded in data/lighthouse_mobile_p1.json:

Phase 1 DoD: MET.

Act

• Phase 0 resident walks remain deferred until live (separate).
• Side-rail density after assent soft-dimmed in Phase 3; deeper rail redesign deferred.

Phase 4 status (2026-09-16) — Approach polish + Since you left

• [x] Glance chips: nest label · voices · silent · official quiet · units
• [x] Store fingerprint data-fp + issue_id (no reorder)
• [x] #since-left strip from localStorage vs #vigie-pulse
• [x] Badges new/changed in place only

Phase 4 verification (2026-09-17) — store-timestamp delta

Canvas DoD: Local delta strip from last visit (store timestamp); returning visit shows delta approaches / fights.

• [x] Strip shows store clock (clustered_at compact UTC)
• [x] Same clustered_at ⇒ “same Approaches” (no fingerprint false delta)
• [x] New store pulse ⇒ new/changed badges in place + “left Approaches” count
• [x] Fingerprint v2: voices|silent|remix|unit_count
• [x] Pure contract since_left_visit mirrored by Arrival JS
• [x] Tests: tests/test_phase4_since_left.py
• [x] No For You / no reorder; w_impact still 0

Phase 4 DoD: MET (instrument).

Phase 5 status (2026-09-16) — Ambient morning pulse

• [x] Digest from same store builders as Arrival (ambient_pulse.py)
• [x] Identity gate refuses write if Approaches diverge
• [x] public/morning.html + data/pulse/latest_morning.{json,txt}
• [x] Arrival CTA “Morning pulse”; ?approach= deep-link
• [x] No second ranking / no LLM / no For You; w_impact unchanged
• [ ] ~~Optional alert delivery channel (RENT v1)~~ — withdrawn 2026-10-06: Vigie is permanently non-commercial and has no paid alerts (see RENT.md); the store twin stays a payload, not a product

Phase 5 verification (2026-09-17) — same pulse as Stage

Canvas DoD: Morning digest + optional widget copy from same store; one pulse = same method as Stage; no second ranking priest.

• [x] Method ambient-pulse-v0.2 same-store-as-arrival-and-stage
• [x] Emit gates: Approaches identity · Arrival #vigie-pulse twin · Stage fight issue_id set
• [x] Optional widget: data/pulse/latest_morning.widget.txt (paste twin, no LLM)
• [x] TXT carries silence + unit honesty (no units yet)
• [x] Tests: tests/test_phase5_ambient.py + ambient suite
• [x] Alert delivery not built (the paid RENT v1 candidates were withdrawn 2026-10-06); widget/json/txt are the payload

Phase 5 DoD: MET (instrument). No delivery channel: the paid RENT v1 candidates were withdrawn on 2026-10-06 (see RENT.md).

Phase 6 status (2026-09-16) — Life facets (opt-in)

• [x] Published FACETS.md facet→weight method
• [x] Catalog: renter · transit · energy · civic · health
• [x] Arrival opt-in UI; localStorage; Clear restores store order
• [x] Reorder Approaches only — same issue_id set; data-i / store_index preserved
• [x] Does not mutate latest_ranked / score_item; w_impact still 0
• [x] Morning digest remains store-order twin

Phase 6 verification (2026-09-17) — published weights + gate

Canvas DoD: Declared facets reorder Approaches only; published facet→weight method; w_impact still gated.

• [x] catalog_matches_facets_md lockstep (FACETS.md table ↔ FACET_CATALOG)
• [x] score_item free of facet imports; W_IMPACT=0
• [x] data-i / data-store-index preserved after reorder; Stage index intact
• [x] data-unit-kinds on Approach + JS fallback when signals thin
• [x] Ambient morning stays store order (facets are Arrival lens only)
• [x] Status copy names w_impact gated; tests: tests/test_phase6_facets.py

Phase 6 DoD: MET (instrument).

Beauty without fog (2026-09-16) — design law

• [x] Published DESIGN.md (Do / Do not)
• [x] Arrival: brand hero · Approaches · CTA; facets after CTA (opt-in)
• [x] Nest depth inset on Approaches; fleuve/stone palette (no purple/cream)
• [x] Nest wash only — no invented QC photo; no multi-layer radial glow
• [x] No score / bias / trust chrome in hero; w_impact still 0
• [x] Clock & method collapsed (#arrival-ops) — residual F8 instrument close

Beauty without fog verification (2026-09-17)

Canvas / co-founder DoD: One composition; brand hero; expressive type; nest depth; no purple/cream cliches; no invented place glow; cards only as Approach.

• [x] DOM order: brand → line → sub → Approaches → CTA → facets → ops
• [x] Brand clamp floor (3.6rem) > line (1.25rem); Newsreader + Figtree stacks
• [x] No radial-gradient / cream hex / purple indigo / lavender silence chrome
• [x] public/place/ empty until named in DESIGN.md
• [x] Arrival has no decorative .card, no score/bias/For You
• [x] Tests: tests/test_beauty_without_fog.py (+ test_design_law.py)

Beauty-without-fog DoD: MET (instrument). Full-bleed QC place image remains backlog.

Resident walkthrough — ready-to-run script (2026-09-19)

Phase 0 still shows 0/5. Everything we have shipped is instrument evidence until this table holds five real rows. This replaces the older English-workbench walk; use the live French brief at https://vigieqc.com.

Who: five Quebec City–life residents (or near-daily visitors). Not co-founders. Setup: their own phone or laptop, cold (no introduction), 5–8 minutes. You: say only the task line below. Do not teach, point, or defend a choice. Record what they do, not what they say they would like.

Task (read verbatim):

« Ouvrez vigieqc.com. Imaginez que vous voulez savoir ce qui a bougé à Québec cette semaine. Prenez cinq minutes. Dites à voix haute ce que vous cherchez, ce que vous comprenez, et ce qui vous bloque. Dites-moi quand vous avez fini. »

Then ask, in this order (no leading questions):

1. Qu’est-ce qui a attiré votre œil en premier ?
2. Qu’est-ce que cette page vous dit de Québec aujourd’hui ?
3. Avez-vous trouvé quelque chose que vous pouviez vérifier vous-même ? Comment ?
4. À quel moment avez-vous hésité ou perdu le fil ?
5. Qu’avez-vous cherché sans le trouver ?
6. Reviendriez-vous demain ? Pourquoi, ou pourquoi pas ?

Capture sheet — one row per person, raw notes, no paraphrase:

Rules: no invented quotes, no personas, no fake severity. A pain ranks only from a real row. After five rows: rank the pains here and write the Phase 0 close.