docs(armada): log 037 — the surface follows the track

The documentary entry for the night the driver stopped being a one-track
instrument: the follow bus that already existed and nobody had read, PLN's
carry policy in his own words, the tempo knob that became the clock, and the
transition move the screen now serves. Plus the three findings that cost only
reading (an innocent v2 service, MIDIIn.connectAll, a disproven off-by-one) and
the two traps worth retelling (the detent that welded itself to the first notch
twice, and zsh eating the backticks out of a commit message).

armada/tasks/037-the-surface-follows.md

+# 037 — The surface follows the track (2026-08-30)
+
+**Goal:** stop the driver from being a one-track instrument. PLN, mid-set, three
+ways in one hour: "i see still ROSE_ROUGE in title", "why dont i see the lcxl
+colors evolve as i map buttons on the new track?", "im starting a new track,
+probably this will be an opp to confirm you read current file well". One bug,
+three faces — a driver that parsed its `.tidal` once, at launch, and never
+looked again. It played sept1 correctly the whole time, because the v3→v2
+translation is track-agnostic, and it labelled everything wrongly, because the
+map was rose_rouge's. The sound was right and every light lied.
+
+**What shipped** (`claude/lcxl3-surface`, `7c297a6..ac86478`):
+
+- **Track-follow.** The bus already existed and nobody had read it: the Pulsar
+  HUD publishes the loaded track to `~/.cache/parvagues/current-track` and
+  appends every ctrl+enter to `eval-events.jsonl`. A different path is a track
+  change (remap, relabel 32 targets, repaint, reset the lap clock); a changed
+  mtime on the same path is a re-read, so mapping a control and saving lights
+  the board in under a second. Verified live twice in the log while PLN worked.
+- **Carry policy**, his spec: DJFs carry (they ARE the transition — "so i can do
+  djf-and-fader-crossfaded-transitions"), effect knobs go to 0, faders are
+  physical. Two he did not name carry too, because the sound would otherwise
+  have taught us: the tempo knob (reset = the range floor = a 60→120 transition
+  slamming back to 60 as the new track lands) and every button (emitting to a
+  gate is the mutebomb).
+- **Live tempo.** `# cps ((range 60 180 "^29")/60/4)` makes row-B knob 1 the
+  clock. BPM now comes from the knob, and bar phase is integrated rather than
+  read off the wall clock — the wall-clock form re-scales all of history the
+  moment the tempo moves, which would make the countdown jump exactly
+  mid-transition.
+- **The move, served.** PLN: "start holding kick mute, move kick to the 120bpm
+  position (i used the center notch, now ill rely on OLED), drop kick mute at
+  end of a bar." So the screen became the notch — BPM not wire units, an arrow
+  to the nearest landmark that vanishes when he is on it, and while a mute is
+  held the bar countdown ("DROP kick / ~120 BPM >2") with the ring breathing at
+  the beat to count it in. Plus software detents, since the encoder has no
+  ridge.
+- **The mute overlay** finally names the held SET ("MUTE x2 / kick+percs",
+  "all") instead of the last finger down.
+- **A probe instead of a fourth guess** (`lcxl3-probe-display.py`, Opus worker):
+  the frozen lap clock and the 1-second DJF Hz share one unproven rule, the
+  protocol has no display readback, so the instrument is PLN's eyes asked
+  precise questions in order.
+
+**Three findings that cost nothing but reading.** The v2 `lcxl-leds-watch`
+service has been logging "no LCXL sequencer port" every 30 s since the v3 board
+arrived — its name regex cannot match "LCXL3 1", so my prime suspect for the
+paint-reset mystery never touched the surface at all, and that mystery is
+reopened with no suspect. The Midi-Through→SuperCollider re-add is
+`MIDIIn.connectAll`, re-running on sample reload, exactly as
+`midi-autoconnect.sh`'s own header predicted. And the A1 label off-by-one that
+sat on the board was never real: "d9 level [ard]" is correct — the grid puts
+d9-d12 on row A and d1-d8 on the faders.
+
+**Two traps worth the retelling.** The detent welded the knob to the first notch
+it met, twice: the closest STEP to a landmark sits just below it, so stepping up
+re-crosses the same landmark and snaps back — 24 consecutive clicks stuck at
+114.8 BPM before the fix tracked *which* landmark holds you by identity. And a
+rich commit message passed inline under zsh lost every backticked code
+reference to command substitution, silently; the only hint was a stray "command
+not found: resetCycles". Rich messages go through `-F` from now on.
+
+**Answered at the desk, unbuilt on purpose:** PLN asked why rapid ctrl+enter
+retriggers the kick "as if it was a once, not a d1". Because `resetCycles` shares
+a blank-line block with the patterns, so every eval restarts cycle 0 — and
+separately, every `dN` is a 1-cycle crossfade launch by definition
+(`d1 = xfade 1 . (|< orbit 0)`). Both diagnosed, neither touched: it is his file
+and his taste. The compose/perform mode that would make it a switch instead of a
+shadowed `let` is now C-1 on the register.