diff --git a/Runs/2026-09-09.md b/Runs/2026-09-09.md
new file mode 100644
index 0000000..d41e212
--- /dev/null
+++ b/Runs/2026-09-09.md
@@ -0,0 +1,63 @@
+---
+date: 2026-09-09
+type: run
+tags: [run, late-morning, week2]
+---
+
+# Run — 2026-09-09
+
+**3.24 km in 19:21 — 5'58" pace at 157 bpm. Week-2 opener, ~11:30 start. Yesterday's "didn't go" (Tuesday — never a scheduled day, but it counted for him) got answered today: the opener happened, floor intact. Distance came in at 3.24 vs the 3.5 planned — close enough to call it a hit. The pace is the sharpest of the block (5'58" avg, 31 s/km faster than the Sep 2 controlled benchmark) and HR ran at the top of his real easy band again — 11 of 19 minutes above the aerobic ceiling. Unlike Sep 6 this isn't illness tax: the sleep base under it (HRV 127, score 88) was genuinely clean. This was pace choice on recovered legs — and the first run inside the smoking-craving peak window (Sep 9–13), where runs are the tool.**
+
+## Data
+
+| Metric | Value |
+|--------|-------|
+| Distance | 3.24 km |
+| Duration | 19:21 |
+| Avg pace | 5'58" /km |
+| Fastest pace | 5'30" /km (km 3) |
+| Avg HR | 157 bpm |
+| Max HR | 180 bpm |
+| Cadence | 168 spm (max 186) |
+
+| Metric | Value |
+|--------|-------|
+| HR recovery (2min) | 21 bpm (171→129) |
+| VO₂Max | 46 ml/kg/min (maintained, "average") |
+| Steps | 3,254 |
+
+**HR zones:** Extreme 1 min · Anaerobic **10 min** · Aerobic 6 min · Fat-burning 1 min · Warm-up <1 min — 11 of 19 minutes above the ~152 bpm aerobic ceiling. Avg 157 ≈ 83% of HRmax (~190). Splits: 6'06" → 6'11" → 5'30" + 0.24 km — even base, real pop on the last km.
+
+## Sleep — Night before
+
+**7h 37min** · 00:25→08:21 · REM: 18% · Deep: 30% · HRV: **127** · Resting HR: 50 · Score: 88/100
+
+> Sober late night (00:25 bed — the cost vs the 22:59 record night is the whole 145→127). Architecture fully repaired: REM back to 18%, deep 30% at 95 continuity. First run in a week starting from a clean base, not debt. See [[Sleep/2026-09-09]].
+
+## Comparison
+
+| Metric | Sep 2 (controlled) | Sep 4 (hot overshoot) | Sep 6 (long, taxed) | **Sep 9 (today)** |
+|--------|--------------------|-----------------------|---------------------|-------------------|
+| Distance | 3.01 km | 5.04 km | 6.07 km | **3.24 km** |
+| Duration | 19:32 | 36:31 | 41:59 | **19:21** |
+| Avg pace | 6'29" | 7'15" | 6'55" | **5'58"** |
+| Avg HR | 148 bpm | 159 bpm | 157 bpm | **157 bpm** |
+| Max HR | 163 bpm | 176 bpm | 179 bpm | **180 bpm** |
+| Cadence | 166 spm | 156 spm | 161 spm | **168 spm** |
+| VO₂Max | 46 | 46 | 46 | **46** |
+| Sleep HRV | 145ms | 111ms | 74ms | **127ms** |
+
+**Verdict:** The floor held and the dial drifted — both true, and today the drift is easier to live with than it was on Sunday. **The loop is the headline.** Yesterday's message was "didn't go"; today he went. Tuesday was never a scheduled run day (Wed/Fri/Sun is the contract), but it clearly mattered to him — and the never-miss-twice mechanic worked exactly as designed: miss the self-set Tuesday, bank the scheduled Wednesday. Week 2 is on the board: run 1 of 3, 3.24 km vs the 3.5 opener (0.26 short — not worth chasing; a route/loop thing, not a quit).
+
+**The dial:** 5'58" at 157 bpm is the sharpest average pace of the block and it wasn't an accident — the splits show a controlled build and a 5'30" last km with a genuine finish (HR recovery start 171, max 180). Compare the controlled benchmark five days before Sunday: Sep 2 ran 6'29" at 148 bpm when he meant easy. So the heat today is pace *choice* on fresh legs, not physiology — and that's the honest read, because unlike Sep 6 (HRV 74, infection signature) the base under this run was clean (HRV 127, score 88, resting 50). 11 anaerobic minutes in a 19-minute "easy" run is the same signature as Sep 6's 22-in-42 — scaled down, same finger. What's worth noting: his easy band has always been ~148–158, and 157 sits at its top edge; the plan's textbook 135–150 target has never been his reality. Friday (4 km, 48h before the 7 km Sunday) is the next negotiation: "Friday stays short and boring" — if any run this block is going to be genuinely easy, it's that one.
+
+**The other context:** this is the first run inside the smoking-craving peak window (Sep 9–13, per the plan). Runs are the craving tool. He used the tool. Day 3 smoke-free engine also gets its first real data point here — 5'58" at 157 with a 5'30" finish suggests the engine wasn't the constraint today. Worth watching whether Friday's HR at ~6'00–6'30 pace starts drifting down as the smoke clears the system; that's the earliest honest signal, not the pace.
+
+**Physical notes:** none reported — no achilles tightness, no foot numbness at this distance. The 3-4 km shoe-limit range was touched but nothing surfaced; logged as clean.
+
+## Connected
+
+- [[Sleep/2026-09-09]]
+- [[Runs/2026-09-06]] (previous run — the taxed 6.07 long run)
+- [[Runs/2026-09-02]] (the controlled benchmark this week should rhyme with)
+- [[Half-marathon-plan]]
diff --git a/Sleep/2026-09-09.md b/Sleep/2026-09-09.md
new file mode 100644
index 0000000..96dff87
--- /dev/null
+++ b/Sleep/2026-09-09.md
@@ -0,0 +1,44 @@
+---
+date: 2026-09-09
+type: sleep
+tags: [sleep, late-bedtime, hrv-high-normal, rem-recovered]
+---
+
+# Sleep — 2026-09-09
+
+**7h 37min** · 00:25→08:21 · REM: 18% · Deep: 30% · HRV: **127** · Resting HR: 50 · Woke: 1× · Score: 88/100
+
+> Score 88 — better than 99% of users. No narrative given.
+
+**Interpretation:** The architecture is fully repaired. Two nights after the 7%-REM debt-repayment night, the stages are textbook: **REM 18% (1h 21min) back in range, Deep 30% (2h 18min) with 95-point continuity**, light 52%, woke once, breathing quality 98. HRV **127ms** sits high-normal in range (77–137) — no alcohol signature (his drinking nights run 71–94ms), so this reads as a *sober, late* night rather than a clean-early one. The 00:25 bedtime is the one wobble: ~1.5h later than the 22:59 that produced Tuesday's 145ms record, and Huawei flagged it ("a little late… can weaken your immune system" — template advice, but the timing point is real). The 90-minute later onset plus slightly shorter sleep (7h37 vs 8h40) is the whole 145→127 story. Three-night shape: **32 → 145 → 127**: crash, clean rebound, settle into high-normal. This was a genuinely recovered base under Wednesday's run — the first one since Sep 2 that didn't start in debt.
+
+| Metric | Value | Reference |
+|--------|-------|-----------|
+| Night sleep | **7h 37min** | 6-10h (Normal) |
+| Bed → Rise | 00:25 → 08:21 | — |
+| Deep sleep | 30% (2h 18min) | 20-60% (Normal) |
+| Light sleep | 52% (3h 58min) | <55% (Normal) |
+| REM sleep | 18% (1h 21min) | 10-30% (Normal) |
+| Deep sleep continuity | 95 points | 70-100 (Normal) |
+| Times woke | 1 time | 0-1 (Normal) |
+| Breathing quality | 98 points | 70-100 (Normal) |
+| HRV | **127ms** | 77-137ms (Within range, upper) |
+| Resting HR | **50 bpm** | 48-58 bpm (Within range, low) |
+| SpO2 | 98% | 97-99% (Within range) |
+| Respiratory rate | 14 brpm | 12-16 brpm (Within range) |
+| Score | 88 / 100 | Better than 99% |
+
+### HRV progression
+
+| Date | HRV | Context |
+|------|-----|---------|
+| Sep 5 | 98ms | Sober, 23:15 bed, 10h17 debt-repayment sleep |
+| Sep 6 | 74ms | 0% beers ×4, home late 01:11, residual illness |
+| Sep 7 | **32ms** | Heavy night, ~4h sleep — near all-time low (chat-reported) |
+| Sep 8 | **145ms** 🚀 | Sober, bed 22:59, 8h40 — ties all-time high |
+| **Sep 9** | **127ms** | Sober, bed 00:25 (late), 7h37 — high-normal settle |
+
+## Connected
+
+- [[Sleep/2026-09-08]] (the 145ms record this settles from)
+- [[Runs/2026-09-09]] (the run this night powered)