docs(evening): tomorrow is not a morning — the tent brief for J-4

PLN: "so tomorrow its not a MORNING.md note its an EVENING.md one :P" — he
travels at dawn, so the next time he and the rig are both awake it is evening, in
a tent, on headphones. MORNING.md's whole frame (the KRKs, the room, the last
speaker day) is spent.

Written for that context specifically: what needs ears vs what needs the room,
and §1 first because it is the only item that can destroy work — Pulsar is
probably still holding the broken desire.tidal buffer, and it already clobbered
this fix once tonight.

Carries forward the two things he has not heard yet (1-cycle xfade with the -6 dB
cut-group explanation, desire's repaired midiOn bass), the fact that NOTHING from
tonight was recorded and why that promotes #37 to a blocker, the unchanged ghost
list with the paste-at-top remedy, and the rig health numbers measured mid-set.

Also records the MIDI fork he diagnosed himself, because it is the kind of thing
that reads as impossible the second time it happens: Tidal hears the LCXL over
ALSA seq (24:0 -> SuperCollider in4), Ardour hears it over PipeWire's Midi-Bridge
into `ardour:MIDI Control In` via Midi Through. Break either and the other stays
perfect — which is exactly why "midi passes to midiOn/off but no fader moves"
was a real, coherent symptom and not a contradiction.

EVENING.md

+# 🌙 Tuesday 2026-08-04 — J-4. The tent brief.
+
+Not a morning file. You travel at dawn, so the next time you and the rig are both
+awake it is **evening, in a tent, on headphones**. Supersedes `MORNING.md`.
+
+`master` clean, **everything pushed** (`70048d4`). The KRKs stayed home.
+
+**Read §1 before you open Pulsar.** It is the only thing here that can destroy work.
+
+---
+
+## 1. ⚠️ Reload Pulsar before you touch `desire.tidal`
+
+Tonight I fixed `desire.tidal` on disk, you quit and reopened Pulsar, and Pulsar
+**wrote its stale buffer back over the fix**. I had to redo it. The file on disk is
+now correct and committed — your editor may still be holding the broken version.
+
+    Window: Reload      (or just close the desire tab before opening it again)
+
+Then, and only then, evaluate it. This is #95 and it is not theoretical any more.
+
+---
+
+## 2. Two changes you have not heard yet 🎧
+
+### `xfade` is 1 cycle, not 4 — `15695de`
+
+Every `dN` now crossfades over **one** cycle. This is why "I change a `# gain` and
+the whole track swells down-and-up for four bars" is over.
+
+Why it dipped at all, since you asked the right question: the envelopes are
+equal-power (`g² + g² = 1`) but SuperDirt's law is `amp = gain⁴`, so at the
+midpoint each side lands on amp **0.500** exactly — and `overlay` fires the sample
+twice, so on any orbit with `# cut N` the second voice *kills* the first and you
+hear one voice at half amplitude: **−6 dB**. Without `cut`, incoherent sum: −3 dB.
+Shortening it makes the dip a blip, not a fade. It does not remove it.
+
+**Verbs for the tent** — all already defined, nothing to type twice:
+
+| want | type |
+|---|---|
+| instant, orbit-correct (adjusting a value) | `jumpTo 4 $ …` |
+| instant via xfade | `xfadeIn 4 0 $ …` |
+| the old long dissolve, one launch only | `xfadeIn 4 4 $ …` |
+
+`xfadeIn N 0` used to give you **silence**, not instant — Tidal guards
+`_slow 0 = silence` and `q \|* gain silence` empties the incoming pattern. Guarded
+now, so 0 is real.
+
+🎧 **Listen for:** one track→track seam at 1 cycle. It is faster than everything you
+rehearsed. If it feels abrupt for the set, `xfadeIn N 2` per launch, or flip the
+literal back in `BootTidal.hs` — it is one token.
+
+### `desire`'s bass compiles now — `70048d4`
+
+**The version you saved tonight did not compile.** The bass you liked existed only
+in the running ghci, in a take that was never recorded.
+
+`midiOn`/`midiOff` pass their argument straight to `someCyclesBy`, so it must be a
+**function** — that is why every working call in your corpus looks like
+`(# crush 6)`. Both note figures are now left sections: `((note (scale …)) #)`.
+`^44` ON = the moving arp, OFF = the grounded root. That was the intent; verify by
+ear that it is the *sound*.
+
+`"Claude's Weight"` is commented out — when its figure moved into Gesture's
+`midiOff` it was left with a `midiOn` and nothing to apply it to, which is what
+broke the file. The last `d4` wins, and that was never this block, so nothing
+musical is lost.
+
+🎧 **Test in the tent:** `^44` off → on → off on `desire`, headphones. Does the
+grounded↔moving flip land where you want it.
+
+---
+
+## 3. Nothing from tonight was recorded
+
+Measured, not assumed: 120 wavs in `Tidal Live`, newest **2026-08-02** (Take95),
+zero files today, no open write fds. The desire bass, the mafia d7 fix, the
+Sunshine→Vague work — code only.
+
+**#37 is now a gig blocker, not a nice-to-have.** Big red blinking dot when
+recording, big red triangle when not, notification on first eval with no rec.
+Watching *bytes landing on disk*, not the record button — the button, the saved
+session and the GUI have each lied to us this month.
+
+Not built yet. Until it is: **arm Ardour before you play anything you'd miss.**
+
+---
+
+## 4. Ghost list — the seams, unchanged
+
+`#77` is not fixed. An orbit the incoming track does not *declare* never gets
+evaluated, so it never transitions, so it keeps playing. Unrelated to the xfade
+work — opposite mechanism.
+
+    Mafia      -> Sunshine     GHOST  d7
+    Sunshine   -> Vague        GHOST  d9, d11     ← you hit this one tonight
+    Vague      -> Desire       clean ✓
+    Desire     -> REVOLUTION   GHOST  d6
+    REVOLUTION -> Hammer       GHOST  d7, d8, d12  ← worst seam in the set
+
+Paste at the **top** of the incoming file, own block. Now a 1-cycle fade:
+
+    d9 $ silence
+    d11 $ silence
+
+Under pressure: `hush`.
+
+---
+
+## 5. Tent-appropriate, and not
+
+**Do 🎧** — `^44` on desire; one seam at the new 1-cycle xfade; `jumpTo` as your
+adjusting verb; read `LAST_THIRD.md` cold; the three untested tracks (#127:
+desire / REVOLUTION / electric_hammer).
+
+**Don't** — gain staging or level work on headphones (#17, #91 need the room);
+loud seam checks; anything that needs the UMC if it is packed. And do not quit
+Pulsar with patterns running until #140 lands — that is why you had to reopen it
+tonight just to `hush`.
+
+---
+
+## 6. Rig state, verified tonight
+
+Measured mid-rehearsal, read-only:
+
+    pulsar renderer   5.0 / 5.5 / 5.0 %     (was ~82% idle before #7)
+    scsynth           1.0 / 0.5 / 1.0 %
+    load average      2.97                  (7.17 during Sunday's dropout)
+    UMC202HD          ~48k frames/s, 0 restarts, 0 xruns while Ardour ran
+
+The graph prune and #7 both held under a real recording load. This is the
+healthiest the rig has measured.
+
+**Two known-live oddities, neither urgent:**
+- Four orphaned `aseqdump` taps still alive (#129). Read-only subscribers, cost
+  nothing, but they are the thing to reap when you next restart the gear.
+- Ardour's control-surface input had come unwired from Midi Through — you found
+  it, and the fork is now understood: **Tidal hears the LCXL over ALSA seq
+  (`24:0 → SuperCollider`), Ardour hears it over PipeWire's Midi-Bridge.** Kill
+  either and the other stays flawless, which is why "MIDI works" and "the faders
+  are dead" were both true. `gig-up.sh:410` only ever checked the SuperCollider
+  half — a check for the Ardour half is the next thing I owe you.
+
+---
+
+## 7. Before the tent: unplug order
+
+1. `hush`
+2. **Save the Ardour session** — deliberately. Tonight's faders are correct, and
+   the *saved* session has shipped 9-of-12 faders at `-inf` three times (#124).
+3. Quit Ardour (it holds the UMC through PipeWire).
+4. Then MIDI → UMC → USB-powered.