--- project: v2-rollout-plan status: PROPOSED — drafted 2026-05-27 13:35 ET by mbp-cowork after Ken ratification of v2 mockup set override_priority: high (binds the 4-week migration from existing CP surfaces to v2 canonical 12-page set) ratifier: pending Ken (rollout plan + timeline + red-team + test strategy ratification) related: - state/canon/design-migration-discipline-2026-05-24.md (6 rules — shadow-ship, binding inventory, etc.) - state/canon/kl-brand-name-usage.md (Ken Lovell, not Walter Kenneth) - mockups/redesign-options-v2/INDEX-v2.html (12-page canonical set) - dispatch/mbp-to-cowork_red-team-audit-results_2026-05-27.md (the audit) - dispatch/mbp-cowork-to-mini_mockup-review-request_2026-05-27-1330et.md (mini's review pending) --- # v2 rollout plan — parallel-build cutover with red-team + test strategy ## One paragraph 12 canonical pages roll out in 4 waves over 4 weeks. Each wave is 3 pages. Each page ships behind `/v2/` as shadow-ship per design-migration-discipline canon — existing route stays live the entire time. Ken uses old → new on his own timeline. Cutover happens when (a) Ken-acceptance ✓ on new + (b) 72h burn-in clean + (c) feature flag flips. Total wall time: **4 weeks (5/27 → 6/24)**. Total build: **~$25, 30-40h mini cycles + 4-6h Ken acceptance time**. Critical-path risk: Mayo Fri 5/29 + Yahoo baseline + PM DB fragility steal mini bandwidth — Wave 1 may slip by 2-3 days. ## §1 — Wave plan (4 weeks) ### Wave 1 — Daily-drivers (Tue 6/3 cutover target) - `/v2/cp` (MC-promoted) - `/v2/morning` (+ People Lookup + Next-up slot + Memex card slot) - `/v2/now-what` (best-of-class already; light backend changes) **Why first:** highest-touch surfaces. If Ken hates them, we course-correct before building the rest. Lowest backend complexity (read-only + small POSTs). ### Wave 2 — G2 PM keystone (Tue 6/10 cutover target) - `/v2/pm/review` (triage 100+/day, G2 keystone) - `/v2/pm/blockers` (auto-classifier triage) - `/v2/workout/trust` (DB↔Excel reconcile, G3 unlock) **Why second:** unblocks the heaviest PM workflow + the workout trust period. Workout trust surface needs the most backend work (Excel parse, diff logic). ### Wave 3 — G3 close-loop (Tue 6/17 cutover target) - `/v2/evening` (single textarea + Reflect writer) - `/v2/workout/` (focus mode, light shell update) - `/v2/family` (roster + events + vacation alignment) **Why third:** closes the daily loop (morning → evening) and surfaces family as first-class. Backend lighter than Wave 2. ### Wave 4 — Reference surfaces (Tue 6/24 cutover target) - `/v2/address-book` (Phase 2 contacts surface) - `/v2/memex` (ask-anything shell — needs Memex MCP HTTP/SSE attach) - `/v2/healthex` (sources + Mayo pipeline) - `/v2/travel` (build new — currently UNCONFIRMED route) **Why fourth:** ad-hoc references. Lower touch frequency = lower risk if delayed. ## §2 — Per-wave cadence (5-day cycle, weekend burn-in) | Day | What happens | |---|---| | Mon | Mini builds `/v2/` Jinja from mockup HTML. CSS tokens via shared kl-identity.css. Old route untouched. | | Tue | Mini wires backend endpoints. MBP runs RULE 12 curl smoke tests. Mini files build-complete dispatch. | | Wed | MBP runs visual regression at 4 viewports (375/768/1024/1440). Mini fixes any breaks. Ken can preview at `/v2/?preview=1`. | | Thu | **Ken acceptance test** — Ken uses each `/v2/` for ONE real task. 5-min per page. Signs off OR sends back. | | Fri | If green: cutover via feature flag (`KL_V2_ENABLED=true` for that route family). Old route becomes 302 redirect to v2. | | Sat-Sun | 72h burn-in. Mini watches error logs. If anything breaks, flip flag back to OFF. | ## §3 — Parallel-build mechanics (shadow-ship — Ken keeps using old) Per `state/canon/design-migration-discipline-2026-05-24.md` Rule 3: **Route mounting:** ```python # Old route stays live @app.route('/morning') def morning(): return render_template('morning.html', ...) # existing 17-card # New route shadow-ships @app.route('/v2/morning') def morning_v2(): if not feature_flag('KL_V2_MORNING'): return redirect('/morning') # not yet enabled return render_template('v2/morning.html', ...) # new ``` **Feature flag table** in `command_post.db`: ```sql CREATE TABLE IF NOT EXISTS feature_flags ( flag_name TEXT PRIMARY KEY, enabled INTEGER DEFAULT 0, enabled_at TIMESTAMP, rollback_count INTEGER DEFAULT 0, notes TEXT ); INSERT INTO feature_flags (flag_name, enabled, notes) VALUES ('KL_V2_CP', 0, 'Wave 1'), ('KL_V2_MORNING', 0, 'Wave 1'), ('KL_V2_NOW_WHAT', 0, 'Wave 1'), ('KL_V2_PM_REVIEW', 0, 'Wave 2'), ('KL_V2_PM_BLOCKERS', 0, 'Wave 2'), ('KL_V2_WORKOUT_TRUST', 0, 'Wave 2'), ('KL_V2_EVENING', 0, 'Wave 3'), ('KL_V2_WORKOUT', 0, 'Wave 3'), ('KL_V2_FAMILY', 0, 'Wave 3'), ('KL_V2_ADDRESS_BOOK', 0, 'Wave 4'), ('KL_V2_MEMEX', 0, 'Wave 4'), ('KL_V2_HEALTHEX', 0, 'Wave 4'), ('KL_V2_TRAVEL', 0, 'Wave 4'); ``` **Cutover script (per-route):** ```bash # Enable a v2 route sqlite3 command_post.db "UPDATE feature_flags SET enabled=1, enabled_at=datetime('now') WHERE flag_name='KL_V2_MORNING';" # Now /morning → 302 → /v2/morning automatically (route handler reads the flag) ``` **Rollback (one command):** ```bash sqlite3 command_post.db "UPDATE feature_flags SET enabled=0, rollback_count=rollback_count+1 WHERE flag_name='KL_V2_MORNING';" # Old route immediately resumes serving ``` ## §4 — Red-team — what could go wrong 12 risks ranked by impact × likelihood: 1. **Mayo Fri 5/29 + Yahoo retry loop steal mini cycles** → Wave 1 slips 2-3 days. Mitigation: declare slip-OK; Wave 1 = Thu 6/5 if needed. 2. **PM DB B-tree recovery left 6 rows lost; another corruption is possible** → cutover blocked. Mitigation: backup before every cutover, verify post-cutover row counts. 3. **Backend not ready when frontend ships** → empty/broken pages. Mitigation: work split holds mini accountable; MBP smoke-test before Ken acceptance. 4. **Ken's muscle memory** — uses `/pm/sprint/5` daily; killing it interrupts flow. Mitigation: 302 redirect from killed routes for 2 weeks before removal. 5. **AWL classifier prompts reference old routes** → bad task classifications. Mitigation: scan classifier prompt for hardcoded route names; rebase to v2 in Wave 2. 6. **Mobile breaks** — shared stylesheet has minimal responsive. Mitigation: each wave includes 375px visual regression as gate. 7. **Accessibility regression** — current /morning has known keyboard nav; new might not. Mitigation: keyboard tab test per page in acceptance. 8. **ack-ack loop still firing during cutover** — pollutes evidence dispatches. Mitigation: mini's prompt patch (separate workstream) ships BEFORE Wave 2. 9. **Performance regression** — Google Fonts CDN dependency in mockups. Mitigation: ship `/static/fonts/` self-hosted in Wave 1 prep. 10. **Memex MCP attach gap** — `/v2/memex` requires the new SSE URL. Mitigation: connector swap (config-level) before Wave 4. 11. **Feature flag bypass risk** — what if Flask doesn't read the flag table? Mitigation: smoke test the flag-read with each cutover. 12. **Token budget** — 4 weeks × $6/week = $24 mini build. Within budget but tight if rework needed. Mitigation: each wave has $2 reserve. ## §5 — Test strategy — TEST HARD ### 5A — Per-route smoke (RULE 12 evidence per page) ```bash # Every v2 route must pass: curl -s -o /dev/null -w "%{http_code}" http://localhost:5001/v2/ # Expected: 200 # Markup presence — key elements curl -s http://localhost:5001/v2/ | grep -c 'class="kl-header"' # ≥1 curl -s http://localhost:5001/v2/ | grep -c 'Ken Lovell' # ≥1 (NOT Walter Kenneth) curl -s http://localhost:5001/v2/ | grep -c 'class="kl-nav"' # ≥1 # Backend endpoints (per route) curl -s http://localhost:5001/api//data | jq . # valid JSON ``` ### 5B — Visual regression (4 viewports) ```bash # Headless screenshot per viewport for vp in 375 768 1024 1440; do chromium --headless --screenshot=/tmp/v2--$vp.png --window-size=$vp,1024 http://localhost:5001/v2/ done # Compare against baseline (golden screenshots in mockups/golden/) ``` ### 5C — Keyboard / accessibility - Tab through every interactive element — no traps - All buttons have visible focus state - Color contrast ≥ WCAG AA (4.5:1 for body text) - Screen reader: VoiceOver pass on critical actions (Approve / Defer / Submit) ### 5D — Performance (Lighthouse per page, target ≥90) ```bash lighthouse http://localhost:5001/v2/ --output=json --output-path=/tmp/lh-.json jq '.categories.performance.score' /tmp/lh-.json # Pass: ≥ 0.90 ``` ### 5E — E2E user journey (Ken's daily flow, smoke-test once per wave) ``` 1. Open /v2/cp at 6:00 AM (or simulated) 2. Click /v2/morning — 17 cards render 3. Complete morning checklist 4. Click /v2/now-what — single decision presented 5. Pick the decision, navigate to /v2/pm/review for triage 6. Triage 5 tasks (approve/defer/kill) 7. Click /v2/workout/ — focus mode renders 8. End-of-day: /v2/evening — single textarea submit 9. Verify: Reflect daily note has the submission appended 10. Verify: tomorrow_plan.json has been written ``` ### 5F — Acceptance criteria per page (Ken signs off) For each route, Ken does ONE real task on the new surface. If it works without friction = ✓ ship. If it requires explanation or workaround = ⚠ revise. ### 5G — Rollback verification (gate every cutover) ```bash # 1. Enable v2 sqlite3 ... "UPDATE feature_flags SET enabled=1 WHERE flag_name='KL_V2_MORNING';" curl -s -o /dev/null -w "%{http_code}" /morning # Expected: 302 to /v2/morning # 2. Disable v2 (rollback) sqlite3 ... "UPDATE feature_flags SET enabled=0 WHERE flag_name='KL_V2_MORNING';" curl -s -o /dev/null -w "%{http_code}" /morning # Expected: 200 (original served) ``` Every cutover validates this round-trip before declaring complete. ### 5H — Data integrity (DB writes) Any v2 endpoint that writes (POST /api/evening, POST /api/workout/trust/reconcile, etc.) must: - Match existing DB schema exactly - Produce identical row shapes as the v1 endpoint - Be reversible via DB transaction log ## §6 — What Ken does 1. **Ratify this plan** (or send back with revisions) — blocks Wave 1 start 2. **Wave acceptance tests** — 4 × ~30 min = 2 hours total over 4 weeks 3. **Cutover sign-off** — one-liner per wave ("Wave 1 cutover GO") 4. **Roll-back authority** — if anything breaks in burn-in, Ken says "roll back" and mini flips the flag ## §7 — What mini does 1. **Mockup review feedback** (by 15:00 ET today) per `dispatch/mbp-cowork-to-mini_mockup-review-request_2026-05-27-1330et.md` 2. **Build Wave 1** Mon-Fri week of 5/30 3. **Concurrent kill list** — retire 18+ dead routes per the audit (no rush; not blocking) 4. **Ack-ack root-cause patch BEFORE Wave 2** — gates Wave 2 start 5. **Feature flag table + read helper** in Wave 0 (this week, Thu-Fri before Wave 1) ## §8 — What MBP-Cowork does 1. **Reconcile mini's mockup review** into v2 files (today/tomorrow) 2. **Visual regression smoke test** each wave Wed 3. **Daily standup dispatch** during active waves 4. **Red-team each wave's cutover** Thu morning before Ken acceptance 5. **Test strategy execution** per §5 — automated where possible ## §9 — Wave 0 (this week) — prep | Item | Owner | Due | |---|---|---| | Mini mockup review feedback | mini | 5/27 15:00 ET | | MBP reconcile review into v2 | MBP | 5/27 17:00 ET | | Feature flag table + read helper | mini | 5/28 EOD | | `/static/fonts/` self-hosted (no Google Fonts CDN) | mini | 5/29 EOD | | Mayo PCP visit | Ken | 5/29 | | Wave 1 START | mini | Mon 6/1 | ## History - **2026-05-27 13:35 ET** — Drafted by mbp-cowork after Ken ratified the v2 mockup look ("They are quite beautiful"). 4-wave parallel-build plan over 4 weeks. Awaiting Ken ratification of plan itself. — mbp-cowork, 2026-05-27 13:35 ET (rollout plan + timeline + red-team + test strategy)