All dates are real calendar dates (2026). Numbers are invented but internally consistent, and consistent with the worked examples in 06 — Training Experience and 05 — Food Analysis. One framing note: these weeks (like every dated example in docs 02–09) depict the Phase-2 steady state of the roadmap — every feature live, autonomy dials graduated — rendered on 2026 dates for readability, not the pilot build calendar.
1. Marta — "I don't have time to think about this"
1.1 Profile card
| Field | Value |
|---|---|
| Age / life | 34 · Kyiv · marketing lead · son Marko, 9 months (she is 8 months postpartum at program start) |
| Goal | First strict pull-up by week 12; stable daily energy; feel like an athlete again |
| Program | Return to Strength — 12 weeks, 3×/week full-body (Sessions A/B/C; currently week 8 = mesocycle 2, loading week 4) |
| Wearable | Whoop 4.0 (recovery %, sleep, HRV — via Health Sync) |
| Nutrition targets | 1,900 kcal · 110 g protein · 215 g carbs · 65 g fat · 28 g fiber (Nata's food program; kcal + protein are hard targets, carbs/fat are bands) |
| Constraints | Physio-cleared at 6 months postpartum; no breath-hold max efforts in mesocycle 1; cranky wrists → goblet over front-rack; trains only when Marko naps or her partner is home; quiet hours 21:00–07:00 |
| Tone preference | Warm, encouraging, zero guilt, celebrates small wins; emoji fine (💛 is Nata's signature); never frame a trimmed day as failure |
| Risk (from brief) | Overreaching on bad-sleep days — Marko's nights land directly on her Whoop recovery |
1.2 Marta's Wiki Brain
Per the Karpathy method (08 — Wiki Brain): raw/ is immutable, wiki/ is LLM-maintained, index.md + log.md + SCHEMA.md govern it. Marta's tree after 8 weeks:
marta/ (S3 bucket wiki-marta, versioned, git-style history)
├── index.md — catalog of all 12 pages with one-line summaries; rebuilt on every ingest
├── log.md — append-only: "## [2026-07-30] ingest | red morning 31%, Session B trimmed, accepted in 1 tap"
├── SCHEMA.md — page conventions, naming, cross-ref rules; the governing template from [08 §2]
├── raw/
│ ├── intake-2026-06-05.md — onboarding interview: goals, history, contraindications (physio clearance noted)
│ ├── program-return-to-strength-v3.yaml — Nata's authored program, current version (v3 = meso-2 loads)
│ ├── events/2026-07-30.jsonl — daily mirror of Postgres events (sets, meals, sleep, taps)
│ └── conversations/2026-W31.md — weekly chat transcript export
└── wiki/
├── profile.md — who she is: schedule, gym (dumbbells in 2 kg steps), family context
├── goals.md — first pull-up by W12; energy > aesthetics; how she defines "a good week"
├── training-history.md — session-by-session narrative distilled from events; links every PR
├── strength-progression.md — per-lift ladders (goblet squat 10→12 kg; pull-up: band → negatives, best 6 s)
├── recovery-patterns.md — ★ rendered below: the Marko-sleep correlation and what to do about it
├── food-priors.md — her portion biases from Meal Lens corrections (see 05 §3)
├── nutrition-patterns.md— protein dips on daycare-pickup days; skyr-before-bed fix works
├── preferences.md — what lands ("waiting for a green day") and what doesn't (any guilt framing)
├── constraints-injuries.md — postpartum notes, wrist workarounds; Nata-edited, ground truth
├── coaching-playbook.md — what works for HER: tone, motivators, red buttons (full page in [08 §7])
├── schedule-context.md — trainable windows: Tue/Thu ~18:00 or nap-time WFH; Sat morning
└── milestones.md — restart day, first 12 kg set, first 3 negatives, chin-over-bar hold 12 s(Core pages per 08's SCHEMA; strength-progression.md, food-priors.md, schedule-context.md, and milestones.md are per-user topic pages the SCHEMA allows.)
1.3 Rendered wiki page: wiki/recovery-patterns.md
The page below is what the nightly distillation actually maintains — synthesized claims with event citations, a Nata ground-truth edit, and a flagged contradiction, exactly per the wiki schema.
# Recovery patterns — Marta
Updated: 2026-08-02 by distillation · Cross-refs: [training-history](training-history.md),
[schedule-context](schedule-context.md), [preferences](preferences.md)
## Baseline (8 weeks of Whoop data, n=56 mornings)
- Median recovery 68% · median sleep 6 h 50 m · resting HR 58 bpm, drifting down (61 → 58).
- Band distribution: green 27, yellow 18, red 11 mornings.
## The Marko correlation (high confidence)
- 9 of 11 red mornings (≤33%) followed a night with ≥3 baby wake events
(wake events inferred from Whoop sleep-stage fragmentation; confirmed by Marta in chat 5×).
Mean recovery after such nights: 34% vs baseline 68% (−34 pts). [events: health.sleep 2026-06-17,
06-24, 07-02, 07-08, 07-21, 07-30 …]
- Teething bursts cluster: red mornings arrive in pairs (Jul 21+22, Jun 24+25).
A single red morning predicts ~45% chance the next morning is red or low-yellow.
- Her subjective report lags the score: on red mornings she says "I'm fine, just tired";
RPE data says the trim is still correct (red-day trimmed sets rated 🟢 easy 6/6 times).
## What works (validated)
- **Trim, don't skip.** 6/6 red-day trimmed-session proposals accepted with one tap;
all 6 completed. Framing that lands: "the 12s are waiting for a green day." [msg 2026-07-21]
- Sessions float on red days: no fixed pre-workout nudge; she trains in a nap window (13:00–15:00).
- Skyr before bed on daycare-pickup days closes the protein gap without a conversation.
## What doesn't
- Moving a red-day session to the evening: tried once (Jun 25), abandoned — evening is Marko time.
- Any message that compares this week to a "normal" week. See [preferences](preferences.md).
> **Nata (ground truth, 2026-08-02):** If two red mornings land in the same week, swap the
> third session to mobility + a walk. Protect the pull-up work; drop leg volume first.
## Contradictions / open questions
- [2026-07-30] Whoop measured 4 h 58 m; Marta felt "maybe 4 hours, honestly less."
Perceived-vs-measured gap stored as a note per 04 §5 — treat as fatigue signal, not data error.
- Does the Thursday-red pattern shift when teething ends? Re-check at week 12 lint.1.4 A week in the life — Mon Jul 27 → Sun Aug 2 (program week 8)
Whoop mornings that week: 72 🟢 · 78 🟢 · 65 🟡 · 31 🔴 · 58 🟡 · 74 🟢 · 69 🟢.
| Day | What the system did proactively | What Marta did | Nata min |
|---|---|---|---|
| Mon 27 | Morning Brief (green 72, rest day) + optional 10-min mobility; evening day-close | 3 meal photos, 3 ✓ taps, did the mobility | 1 |
| Tue 28 | Brief (green 78); Session A proposed with all weights (transcript in 06 §7); 17:30 nudge; close: 34 min, 2,742 kg, PR goblet 12 kg × 8 | Trained 18:03–18:40: nine ✓ taps, one adjust, three RPE taps | 2 |
| Wed 29 | Brief (yellow 65 — targets held, no progression today); 20:30 protein nudge (day short 18 g) | 2 meal photos; skyr before bed | 1 |
| Thu 30 | Red 31% → trimmed Session B proposed (signature transcript below); no fixed nudge, session floats | One tap ✓; trained 24 min in the 13:40 nap window | 4 |
| Fri 31 | Brief (yellow 58, rest day); deliberately quiet — no meal nag when breakfast photo never came | 2 meal photos (skipped one; nobody nagged) | 1 |
| Sat 1 | Brief (green 74); Session C proposed incl. pull-up work: 3×3 negatives; close celebrates a 6 s negative | Trained 41 min; filmed one negative → Form Check | 3 |
| Sun 2 | Weekly Review 18:00 (Mini App charts); weigh-in proposed from trend | ✓ 63.8 kg; read review; replied "good week 🙌" | 5 |
| Total | 19 proactive messages, 1 auto-adjustment, 0 questions she had to type answers to | 3 sessions, 11 photos, ~26 taps, 0 manual logs | 17 min |
Nata's 17 minutes: skimming the auto-send audit feed (meal feedback moved to auto-send with audit in week 6), approving 3 drafts, one 30-second voice note on Thursday, the Sunday digest, and one wiki edit. Under 30 minutes for a client who received 19 coach touches.
journey title Marta week 8 of Return to Strength section Mon 27 Green brief and mobility offer: 4: Bot Meal photos with one tap each: 5: Marta section Tue 28 Session A proposed with weights: 5: Bot PR goblet squat 12 kg x8: 5: Marta section Thu 30 Red 31 detected after baby night: 2: Bot Trimmed Session B proposed: 4: Bot One tap accept and 24 min session: 5: Marta Voice note about trimmed days: 5: Nata section Sun 2 Weekly Review with negative hold chart: 5: Bot Reads review and feels coached: 5: Marta Digest and one wiki edit: 4: Nata
1.5 Signature transcript — the poor-recovery morning (Thu Jul 30)
── Thu Jul 30 · 07:12 ────────────────────────────────────────────
NataCoachBot:
Morning, Marta ☀️
Whoop: recovery 31% 🔴 · sleep 4 h 58 m, 4 wake-ups.
Marko had a night, huh. So today adapts to you, not the other way.
Session B, trimmed:
• DB bench 12 kg → 10 kg, 3 sets → 2
• Rows hold at 14 kg, core stays as is
• ~24 min instead of 40. Progression frozen — nothing is lost.
No fixed time today; ping ▶ whenever a window opens.
[ ✓ Do that ] [ Rather rest today ] [ Show session ]
07:13 Marta: [ ✓ Do that ]
NataCoachBot:
Loaded. The 12s are waiting for a green day. 💛
── 13:41 ─────────────────────────────────────────────────────────
Marta: ▶
NataCoachBot: Set 1 — DB bench 10 kg × 8. [ ✓ Done ] [ ± Adjust ]
Marta: [ ✓ Done ] (…six more one-tap set cards…)
── 14:05 ─────────────────────────────────────────────────────────
NataCoachBot:
Done — 24 min on four hours of sleep. That's the whole assignment.
── 19:50 ─────────────────────────────────────────────────────────
Nata 🎙 (voice, 0:31): "Marta, trimmed days like today are the ones
that actually build the pull-up. Proud of you. Sleep when he sleeps."One decision, one tap, before 07:15. The adjustment math came from the readiness rules in 06 §4; the framing came from preferences.md; the voice note is Nata, live — 31 seconds of her total.
1.6 What Nata saw in the Coach Console that week
| Console item | Detail |
|---|---|
| Flags raised | 1 informational red-recovery flag (Thu, auto-handled by the engine — shown, not blocking); 1 lint note: perceived-vs-measured sleep gap on Jul 30; 0 pain flags |
| Approvals made | 12 queue items: 9 auto-sent audits skimmed (~10 s each), 2 drafts approved unchanged, 1 draft edited — she softened one line of the Weekly Review coach note. Avg 21 s/item |
| Autonomy dial (Marta) | Morning Brief: autonomous · meal feedback: auto-send with audit (since week 6) · session summaries: auto-send · red-day adjustments: auto-send with audit · pain/injury: always escalate |
| Wiki edit performed | Sunday, 2 min: added the ground-truth rule to recovery-patterns.md — "two red mornings in one week → third session becomes mobility + walk; protect the pull-up work". The rule compiles into next week's adjustment logic immediately |
| Human touch | One voice note (Thu). Total: 17 min |
2. Denys — "Travel weeks are where this always dies"
2.1 Profile card
| Field | Value |
|---|---|
| Age / life | 41 · product manager · travels ~1 week/month (hotel gyms, client restaurants) |
| Goal | Fat loss 96 kg → 88 kg; keep strength; currently 91.8 kg at week 10 (−4.2 kg, ~0.45 kg/wk — matches the trend in 05 §7) |
| Program | Recomp Hybrid — 16 weeks, 2 strength + 2 zone-2/week (Strength A/B + two 40-min zone-2 blocks at 130–138 bpm; week 10 of 16) |
| Wearable | Garmin Forerunner (Body Battery, sleep, auto-imported runs — via Health Sync) |
| Nutrition targets | 2,100 kcal · 160 g protein · 205 g carbs · 68 g fat · 30 g fiber. Travel-day budget: up to 2,200 kcal with a 140 g protein floor (Nata's rule — trend over perfection on the road) |
| Constraints | 2019 disc scare, no pain since → hinge form matters (Form Check on RDLs), no loaded spinal flexion; knees cap runs at 45 min; client dinners mean wine |
| Tone preference | Blunt-but-kind: plan first, number second after hard days; no cheerleading; always end with a next step |
| Risk (from brief) | Travel weeks derail him; skips logging when eating out — so restaurant meals may be described, not photographed (02 §6.2) |
2.2 Denys's Wiki Brain
denys/ (S3 bucket wiki-denys, versioned)
├── index.md — catalog of all 14 pages; travel-playbook flagged "high-traffic"
├── log.md — "## [2026-08-05] ingest | Lisbon: hotel session done, menu assist followed (seabass)"
├── SCHEMA.md — the governing template from [08 §2]; adds a per-city note format Nata requested
├── raw/
│ ├── intake-2026-05-28.md — onboarding: goals, 2019 disc history, travel cadence
│ ├── program-recomp-hybrid-v2.yaml — Nata's program incl. pre-authored hotel/bodyweight variants
│ ├── events/2026-08-05.jsonl — daily event mirror (sets, meals, runs, taps)
│ └── conversations/2026-W32.md — weekly transcript export
└── wiki/
├── profile.md — who he is: schedule, home gym (full barbell), iPhone + Garmin
├── goals.md — 96→88 by week 16; "keep the bar numbers" is non-negotiable to him
├── training-history.md — sessions + auto-imported zone-2 runs, weekly volume trend
├── strength-progression.md — barbell ladders (RDL 80 kg 3×10) + frozen/resume state for travel
├── travel-playbook.md — what each travel week taught: Warsaw (Jun), Berlin (Jul), Lisbon (Aug)
├── hotel-gym-profiles.md— per-hotel equipment maps (Lisbon: DBs to 26 kg, cables, no barbell)
├── cardio-engine.md — zone-2 responds fast: resting HR 66 → 61 in 10 weeks
├── recovery-patterns.md — Body Battery dips on flight days; alcohol > 2 drinks costs next-day sleep score
├── food-priors.md — restaurant-portion bias +15%; corrections shrinking it (05 §3)
├── nutrition-patterns.md— ★ rendered below: the travel-week pattern and what fixed it
├── preferences.md — "straight talk" confirmed; never open with the overage after a hard workday
├── constraints-injuries.md — disc history, knee cap on runs; Nata-edited, ground truth
├── coaching-playbook.md — what works for HIM: plan-first framing, budget-not-rule language
└── milestones.md — first sub-92 kg, Berlin week survived at 74% adherence, 10k steps × 21 days(Same core pages as Marta's tree; travel-playbook.md, hotel-gym-profiles.md, cardio-engine.md, strength-progression.md, food-priors.md, and milestones.md are his per-user topic pages.)
2.3 Rendered wiki page: wiki/nutrition-patterns.md
This is the full page whose excerpt appears in 05 §7 — same claims, plus the travel mechanics and the Nata edit from the Lisbon week.
# Nutrition patterns — Denys
Updated: 2026-08-09 by distillation · Cross-refs: [travel-playbook](travel-playbook.md),
[goals](goals.md), [food-priors](food-priors.md), [preferences](preferences.md)
## Reliable patterns
- Home weeks: 85–95% adherence; protein target (160 g) hit 5–6 days/7.
- Travel weeks: adherence historically drops to ~60%; **dinner is the failure point**
(restaurant + 1–2 beers). Breakfast/lunch stay clean even when traveling.
[weekly rollups W25 Warsaw 58%, W28 Berlin 63%]
- Responds well to pre-ordering menu assists: 5/6 recommendations followed
(was 4/5 before Lisbon). [meal events 07-07, 07-09, 08-05 …]
- Describes meals honestly when photos are awkward (client dinners): text estimates
vs later corrections differ by only ~8% — trust his descriptions.
## Travel rules currently in force (Nata's food program, v2)
- Travel-day budget 2,200 kcal, protein floor 140 g (home: 2,100 / 160).
- Buffet breakfast = protein-first: eggs + skyr before anything else. Followed 9/10 mornings.
- Menu-forward move at ~19:00, before the restaurant: ONE recommendation, not three options.
- Two-drink cap framed as "budget", never as a rule. Framing matters: cap held 4/5 dinners
since reframing; 1/4 before.
- Skyr backstop: a corner-shop skyr on the way back closes a protein gap without a debate.
## What worked in Lisbon (W32) — new
- Menu-photo flow beat text descriptions: he photographed the menu, got one pick
("the seabass"), ordered it, then photographed the plate. Dinner adherence 3/4 nights
vs 1/5 in Warsaw. [meal.scored 2026-08-05T20:52]
- Week landed 78% adherence — best travel week on record (prior best: Berlin 63%).
## Coaching implications
- Never open with the overage number after a hard workday — plan first, number second.
- Weight trend 96 → 91.8 kg over 9 weeks (~0.45 kg/wk): estimates are calibrated
well enough; no target change needed. Skipped travel weigh-ins are expected —
interpolate, don't nag. [weight.skipped 2026-08-07]
> **Nata (ground truth, 2026-08-09):** He actually *likes* fish — lean on it in every
> port city. Keep the two-drink cap framed as budget. Next travel week: propose the
> restaurant *before* the client picks one; he said he'd forward the invite.2.4 A week in the life — Mon Aug 3 → Sun Aug 9 (program week 10, Lisbon Tue–Fri)
Garmin Body Battery on waking: 68 · 55 (flight) · 71 · 64 · 58 · 76 · 80.
| Day | What the system did proactively | What Denys did | Nata min |
|---|---|---|---|
| Mon 3 | Weigh-in proposed pre-flight (✓ 91.8 kg); timezone jump detected at landing → travel mode offered until Fri; step target 10k for flight day | ✓ travel mode; 11,400 steps through airports; 2 meals by text | 3 |
| Tue 4 | Brief with hotel-gym check ("send me a 10-s pan of the gym") → equipment profile built: DBs to 26 kg, cables, no barbell; zone-2 #1 proposed: 40 min treadmill @ 130–138 bpm | Filmed the gym (9 s); ran 40 min — Garmin auto-imported, zero logging; buffet photo ✓ | 2 |
| Wed 5 | Strength B re-resolved to hotel gym (signature transcript below); 19:00 menu-forward nudge; seabass recommendation from menu photo | ✓ Load it; 38-min session; filmed one DB RDL set → Form Check cue; ordered the seabass; skyr on walk back | 6 |
| Thu 6 | Zone-2 #2 with a route: Tagus waterfront out-and-back, 40 min; evening asks for a description, not a photo (client dinner) | Ran 42 min (auto-imported); described dinner in one line, ✓ estimate | 2 |
| Fri 7 | Weigh-in proposed with an explicit [ Skip — hotel scale ] (per 04 §7); travel mode off at landing | Tapped Skip; flew home | 1 |
| Sat 8 | Strength A proposed at home gym — barbell numbers resume exactly where frozen (RDL 80 kg 3×10) | Trained 52 min; nine ✓ taps | 2 |
| Sun 9 | Weekly Review: travel week at 78% adherence vs 60% historical; weight trend interpolated, no gap-shaming; next week previewed | Read it; tapped ✓ on next week's plan; replied "ok that was the least-bad travel week ever" | 8 |
| Total | 21 proactive messages, 2 session re-resolutions, 1 menu assist, 0 nags | 2 strength + 2 zone-2 done, 9 meals logged (5 photos, 4 text lines), 0 manual entries | 24 min |
Nata's 24 minutes: the Monday travel re-plan approval, the Wednesday menu draft (edited + approved in 38 s) plus a voice note, daily audit skims, and Sunday's digest + a program tweak for September travel + the wiki edit. A historically catastrophic week type cost her under half an hour.
journey title Denys travel week in Lisbon section Mon 3 Timezone jump detected and travel mode on: 4: Bot One tap accepts the travel plan: 4: Denys section Wed 5 Strength B re-resolved to hotel gym: 5: Bot Session done in 38 min: 4: Denys Menu photo answered with one pick: 5: Bot Orders the seabass: 5: Denys Approves draft in 38 seconds: 4: Nata section Fri 7 Weigh-in offered with skip option: 4: Bot Skips without guilt: 4: Denys section Sun 9 Weekly Review shows 78 percent adherence: 5: Bot Best travel week on record: 5: Denys Digest and travel playbook edit: 4: Nata
2.5 Signature transcript — Wednesday in Lisbon (Aug 5)
── Wed Aug 5 · 07:20 (Lisbon) ────────────────────────────────────
NataCoachBot:
Morning. Body Battery 71 — decent for a conference week.
Today: Strength B, re-built for your hotel gym (DBs to 26 kg, no bar):
• Barbell RDL 80 kg 3×10 → DB RDL 2×24 kg, 3×10 (form cue: stop at mid-shin)
• Lat pulldown 55 kg → Single-arm cable row 30 kg, 3×10/side
• Bench 70 kg 3×8 → DB bench 2×26 kg, 3×8
Your barbell numbers are frozen, not lost — they resume Saturday at home.
[ ✓ Load it ] [ Show all 6 ] [ Swap something ]
07:21 Denys: [ ✓ Load it ]
── 19:02 ─────────────────────────────────────────────────────────
NataCoachBot:
Dinner reservation tonight? Send me a menu photo when you sit down —
I'll give you one pick, not a lecture.
So far today: 1,250 kcal · 87 g protein. Budget left: ~850 kcal · 73 g.
── 20:04 ─────────────────────────────────────────────────────────
Denys: <photo: menu, Cervejaria O Tejo>
NataCoachBot 20:07:
Order the seabass. Grilled robalo with roast potatoes + greens —
about 640 kcal, 49 g protein, and it's the best thing on that menu
anyway. One glass of vinho verde fits the budget (~120 kcal).
Skip the bread basket; the francesinha is a 1,100-kcal decision
for a different kind of week.
[ 👍 Ordering it ] [ Show 2nd option ]
20:09 Denys: [ 👍 Ordering it ] done. seabass it is
── 20:50 ─────────────────────────────────────────────────────────
Denys: <photo: grilled seabass, potatoes, greens, glass of wine>
NataCoachBot:
Estimate: 640 kcal · 49 g protein · matches the plan. ✓ to log.
[ ✓ Looks right ] [ ± Adjust ]
Denys: [ ✓ Looks right ]
NataCoachBot:
Day so far: ~2,010 kcal · 136 g protein. Grab a skyr on the walk
back and you land at ~2,100 · 152 g — over your 140 g travel floor,
bang on budget. Straight talk: this is what a good travel Wednesday
looks like.Invisible to Denys: the 20:07 recommendation was a draft — his restaurant recommendations still sit at draft → approve on the autonomy dial. Nata approved it from the queue at 20:06 in 38 seconds, adding one line ("skip the bread basket"). Arithmetic check: 1,250 + 640 + 120 + 90 (skyr) = 2,100 kcal; 87 + 49 + 16 = 152 g protein.
2.6 What Nata saw in the Coach Console that week
| Console item | Detail |
|---|---|
| Flags raised | Travel-mode banner Mon–Fri (context, not action); 1 blocking draft — Wednesday's restaurant recommendation (approved in 38 s with one edit); weight.skipped Friday (informational, expected pattern); Form Check auto-sent Wed with a mid-shin depth cue (audit only, confidence 0.82 per 07) |
| Approvals made | 16 queue items: 13 audits skimmed, 2 approved unchanged, 1 edited-then-approved (the seabass draft). Avg 24 s/item |
| Autonomy dial (Denys) | Morning Brief: autonomous · training re-resolutions: auto-send with audit · meal feedback: auto-send with audit · restaurant recommendations: draft → approve (kept manual — highest-leverage moment of his week) · pain/injury: always escalate |
| Wiki edit performed | Sunday, 3 min: appended the ground-truth note to nutrition-patterns.md — fish-forward picks in port cities, keep the two-drink "budget" framing, ask for the restaurant invite in advance |
| Human touch | One voice note after Wednesday's session ("that hotel session was the week, right there"). Total: 24 min |
3. What these two cases prove
| Claim | Evidence above |
|---|---|
| Zero-logging adherence. Marta hates logging and logged nothing — yet her week produced ~140 structured events (sets, meals, sleep, RPE) because every datum arrived as a photo, a wearable sync, or a one-tap confirmation. | Marta §1.4: 3 sessions + 11 photos + 26 taps, 0 manual entries; the red Thursday cost her exactly one decision before 07:15 |
| Travel resilience. The week type that historically killed Denys's progress (60% adherence) became his best on record (78%) — not by demanding more discipline but by re-resolving the program around the hotel he has and the menu in front of him. Progression state froze and resumed; nothing was "lost". | Denys §2.4–2.5: hotel-gym re-resolution, the seabass moment, skip-without-guilt weigh-in, Saturday's exact barbell resume |
| Trust via propose-confirm. Every number either of them saw carried provenance and a one-tap accept; editing stayed the exception (Marta adjusted 1 of ~26 proposals; Denys 0 of ~20). Acceptance is the trust metric — 6/6 red-day trims accepted, 5/6 menu assists followed — and it feeds back into the Wiki Brain, which makes the next proposal better. | Marta §1.5 one-tap trim; Denys §2.5 one-tap re-resolution and "seabass it is" |
| Coach-in-the-loop is cheap where it's safe and human where it counts. Nata spent 17 + 24 = 41 minutes on two clients for a combined 40 proactive touches, and both weeks contained a real voice note and a real wiki edit — her judgment now compiled into next week's automation. | §1.6, §2.6: dial states, sub-30-min totals, one ground-truth edit each |
The same machinery served a postpartum strength client on a red-sleep Thursday and a traveling fat-loss client in a Lisbon restaurant — different programs, different tones, different failure modes, one system. That is the scaling thesis of 01 — Product Vision made concrete: the user does less, the coach touches only what matters, and the Wiki Brain remembers why.
Siblings: 02 — User Experience (all flows these weeks ran through) · 06 — Training Experience (the weight math behind every proposal here) · 05 — Food Analysis (Meal Lens mechanics) · 08 — Wiki Brain (how the rendered pages are built and linted) · 09 — Admin Analytics (the Console views in §1.6/§2.6).