Phase 10: variety, twelve scenes, and tooling to write the next one

Watching several finished tracks side by side turned up the problem neither
Phase 8 (too few cuts) nor Phase 9 (no through-line) addressed: the same
scene cast in two different videos looked like the same footage twice.
Section bias is nearly identical between two tracks' drops, so both sampled
their parameters around the same centre, and the library's own averageness
did the rest.

Three answers, none of them a new scene:

  Temperament — a per-track hand on every parameter dial: intensity, pace,
  detail, and an extremity that decides how far toward the ends of a range
  the track is willing to sample. Bias comes from the section and is shared
  between tracks; temperament comes from the track and is not.

  Overlays — sometimes a second full scene composited over the shot, from a
  different family, in a blend that preserves what is underneath and never
  above 0.6 opacity. Not always: a stack that always doubled up would read
  as permanently cluttered rather than as occasionally layered.

  A wider palette — hue now derives from SPECTRAL TILT, the log ratio of
  treble to body. The centroid is a number most masters sit in the middle
  of, and the plain body/(body+treble) fraction is worse: low frequencies
  carry most of the energy in all music, so it read 0.98-1.00 for
  everything and four different battery tracks came out within 0.02 of
  each other. The ratio is multiplicative, so its logarithm is what
  spreads — the same four measure -9.3, -5.0, -4.1, -3.8. Also both ways
  round the wheel (violet, magenta and pink were unreachable by
  construction), four new schemes, and seeded chroma profile and lightness
  curve. Closest battery pair went from 0.005 to 0.113.

Twelve scenes take the library to 36, six per family: Aurora Veil, Vortex
Drift, Tide Rings, Ink Bleed, Dust Chamber, Salt Flat, Cargo Belt, Gate
Corridor, Circuit Bloom, Truchet Fold, Signal Decay, Storm Rift. Weighted
toward the 'space' and 'shape' traits, which were thinnest and so the
signatures most likely to run a track out of cast — the Phase 9 casting
rule means the pool a track draws from is smaller than the library.

Also fixes a real one in shots.js: heavy LRU weighting was not enough to
make a section reach its whole roster, and a five-shot section still came
out 0,2,0,2,0 about a fifth of the time. An unseen companion now wins
outright; which one is still free, so only the coverage is guaranteed.

Block Mosh declared the camera trait, assigned sigCamera(p) to a p it then
never read, and passed the lint's evidence grep. The Phase 9 render gate
measured its response to the camera at exactly zero.

--- tooling ---

Adding a scene was mostly boilerplate and round-trips, which is expensive
in both senses. The irreducible cost is the shader body; everything around
it is now mechanical:

  npm run new:scene -- "Name" --family=... --traits=...

writes the module, registers it, and leaves a skeleton that already passes
every gate, with name-derived constants so two skeletons are not twins.

The lint grew the rules that previously needed a GPU to catch: the dead
camera above, prev() with no base image, and large loops with no early
break (with a `// lint: fixed-cost` opt-out for a genuinely fixed-cost
sampling loop). checks.html?scene=Name runs the per-scene acceptance
battery for one scene — ten lines and a verdict instead of rendering the
whole library to find out whether one shader is alive. The same procedure
is a repo skill under .claude/skills/build-visualizer/.

--- checks changed, with the measurements ---

P5 determinism compared two WebGL CONTEXTS, which is not what it is for.
Measured: one context is bit-exact over 40 frames with feedback at 0.6;
two contexts disagree by up to 2/255 whether feedback is on or off. It now
asserts generation is byte-identical (hard) and rendering within 2/255,
since feedback compounds single-level variance.

P10's cross-track comparison measures distance RELATIVE to how much image
there is. Most scenes are mostly dark, so two genuinely different renders
— 25 bars against 53 — scored under 0.02 absolute purely because the black
background agrees with itself.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

flow-state/HOWTO-visualizers.md

@@ -1,5 +1,14 @@
 # How to build a visualizer (scene)
 
+> **Start here:** `npm run new:scene -- "My Scene" --family=glitch --traits=camera,style`
+>
+> That writes the module, registers it, and leaves a skeleton that already passes
+> every gate. Then write the `scene()` body, `npm run lint:scenes`, and open
+> `checks.html?scene=My%20Scene`. The rest of this file is the reference.
+>
+> There is also a repo skill — `.claude/skills/build-visualizer/` — which is the
+> same procedure in the form an agent will follow.
+
 A scene is **a fragment shader plus a params block** — nothing else. Uniform
 binding, the generated UI sliders, seeded per-track sampling, arc drift and the
 phase-gate checks are all derived from the schema. There is no per-scene wiring.
@@ -36,7 +45,10 @@ vec4 scene(vec2 uv, vec2 p) {
 };
 ```
 
-Register it in `src/scenes/registry.js` (import + push into `MODULES`). Done.
+The scaffolder writes all of that for you, including the registry import and
+entry, and bakes name-derived constants into the skeleton field so two freshly
+scaffolded scenes are not identical to each other. To do it by hand: write the
+file and add the import + `MODULES` entry in `src/scenes/registry.js`.
 
 ## The shader contract
 
@@ -128,19 +140,46 @@ The lint greps your shader and fails a declared trait with no evidence. Declarin
 ## Families
 
 Chosen by section kind in the arc driver — a breakdown never lands on a strobing
-glitch scene. Current counts (aim: 4-6 each): flow 2, organic 3, minimal 3,
-structural 2, geometric 3, **glitch 3**. The thin families are the best place to
-contribute next.
+glitch scene. The library is at **six per family** (36 scenes), which is the
+target. Check the current spread before adding another:
+
+```bash
+node -e "import('./src/scenes/registry.js').then(({scenes})=>{const b={};for(const m of scenes)(b[m.family]??=[]).push(m.name);console.log(b)})"
+```
+
+Depth matters more than it looks: the casting rule in `look/Personality.js`
+disqualifies scenes that do not honour the track's signature traits, so the pool
+a given track draws from is smaller than the library. Thin traits (`space`,
+`shape`) are worth more than thin families.
 
 ## Verify
 
+Three rungs, each about ten times cheaper than the next. Climb them in order.
+
 ```bash
-npm run lint:scenes   # schema/shader agreement both ways + determinism grep
-npm test              # audio pipeline (unaffected, but cheap)
+npm run lint:scenes   # ~1s, no browser
 ```
 
-Then the GPU gates in the browser at `http://localhost:5180/checks.html`:
-Phase 2 (param range sweep — no NaN/black/white), Phase 5 (flash-rate sweep),
-Phase 7 (per-scene acceptance: distinctness, liveness, animation, determinism,
-4K budget). These iterate the registry, so a new scene is covered automatically
-once registered.
+Static gates: schema and shader agreeing both ways, determinism grep, rate flags,
+a declared trait with no evidence in the source, a **dead camera**
+(`p = sigCamera(p)` and then nothing reads `p`), `prev()` with no base image, and
+loops with a large bound and no early break. A loop whose cost is genuinely fixed
+— sampling a curve at a set resolution — can say so with a `// lint: fixed-cost`
+comment just above it.
+
+```
+http://localhost:5180/checks.html?scene=My%20Scene
+```
+
+The per-scene acceptance battery for one scene: schema, renders, animates,
+deterministic, distinct from every other scene, param sweep, flash rate, and one
+line per declared trait proving the image actually responds to it. Ten lines and
+a verdict — this is the loop to stay in while writing.
+
+```
+http://localhost:5180/checks.html?slow=1
+```
+
+Everything. Phases 2, 5 and 7 iterate the registry so a new scene is covered
+automatically; Phases 8-10 cover how the look generator uses it. Run this once
+before committing.