b2d9b46c30
7 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
jing
|
a99f733452 |
[lane A] MODELBEAST dresses the molecules: 5 matcaps, and the canal picks space-filling
The right axis wasn't "generate or don't" — it was WHICH HALF. Geometry stays procedural because a molecule's shape is already known exactly. **Material is exactly a generation problem**, and it's the whole difference between a textbook diagram and a game object. So MODELBEAST made the materials. 5 greyscale matcaps, flux_local, 69 seconds, $0.00, 68 KB total: mol_glass (O,N) — the big win: oxygen is a red glass marble with internal glow mol_matte (C) — graphite soot; carbon reads as the scaffold it is mol_chrome (Co) — B12's cobalt is now a chrome bearing in a cage mol_molten (P) — ATP's phosphate tail is incandescent AND pulses (uTime, free) mol_pearl (H) — satin white; hydrogens stop shouting The trick is Lane D's own law applied to spheres: author the LUMINANCE, tint in-shader. derive_maps already greyscales every matcap, so one glass ball serves oxygen AND nitrogen at their own CPK hues — 5 balls dress the whole periodic table we care about. They pack into one strip atlas indexed per-vertex by aMat, so **a molecule is still exactly 1 draw call** however many materials it's made of (7 molecules = 7 draws, unchanged). ?matcap=0 falls back to the built-in fake-lit path: the assets-optional law holds. SPACE-FILLING WINS FOR PICKUPS, and the canal decided it, not me. New `repr: 'space-filling'` draws every atom at its real van der Waals radius with no sticks. A/B'd at real pickup size against the real L2 wall, same camera: ball-and-stick goes spindly and dissolves; the solid blob holds its silhouette — and it's CHEAPER (21k vs 26k tris, no bond cylinders). Recommendation to B: space-filling in flight, ball-stick for hero/UI/collect close-ups where the chemistry is worth reading. Evidence: round2_molecule_materials.png (3-way). Two defects found by rendering it, both mine, both fixed: - Tint x luminance darkens TWICE: CPK carbon is 0.4 and a matte ball averages 0.5, so soot x soot vanished — the first pass sank glucose, caffeine and the cobalt into the background. Floored the matcap at 0.22, the same floor the fallback's key light always had. - The molten matcap was my own bad prompt: I asked for an "incandescent white hot CORE" and got exactly that — a black ball with a hot spot, so phosphorus lost its orange. Re-rolled for the whole sphere to glow (attempt 2 of the <=2 PIPELINE allows). Lesson for the kit, next to the organ/tissue one: a matcap is a LUMINANCE LOOKUP, so a dark-dominant ball makes a dark-dominant object. Also corrected the doc's own framing — v1 said "I'm authorised to burn GPU and I'm not going to", which was the wrong axis, not just the wrong tone. qa GREEN; 16/16 texture provenance verified (synced the FIXED batch json up first this time, so the box couldn't clobber it on the way back). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
jing
|
9f42a2e9ab |
[lane A] Molecules as pickups: real gut chemistry, procedural, 1 draw each
PROTOTYPE + proposal, not wired into boot. Pickups are B's systems, their economy is C's, object art is D's — this is a renderer and an argument, offered to all three. Bench: web/dev/laneA_molecules.html · docs/MOLECULES.md THE PITCH: the pickups already in L2 ARE molecules. C authored `nutrient`, `mucin`, `B12` and `antacid ammo` before any of this existed — those are the real biochemistry of a digestive tract, and glucose/cobalamin/bicarbonate are real compounds with known shapes. We don't need to invent a pickup language; we need to stop hiding the one the fiction handed us. GUTS is already a science game (an eosinophil is a real white blood cell, candida a real yeast, the hiatus a real choke point) — it just doesn't LOOK like one, because everything on screen is tinted tissue. THE RULE THAT MAKES IT WORK: CPK is the scanner. Everything identified is drawn in the standard element palette (O red, N blue, P orange, metals pink); everything else stays monochrome tissue. Against six biomes of tinted wall a CPK molecule is the ONLY saturated foreign colour on screen — it reads as artificial, valuable and targetable with NO HUD marker. Proof: round2_molecules_in_canal.png. That's the ART_BIBLE's synthetic-scanner fiction paying rent, and it hands Lane E a free HUD palette. WHY PROCEDURAL AND NOT MODELBEAST — and I'm authorised to burn the GPU: a molecule's shape is already known exactly. Glucose is a hexagonal ring because it IS one. FLUX+TRELLIS gives a plausible blob in 3-8 min that a chemist clocks as wrong and that can never be re-derived. Ball-and-stick from an atom list is exact, rebuilds in ms, and is ONE DRAW CALL per molecule (measured: 7 molecules = 7 draws / 26.5k tris). The split is a principle: small molecules = procedural (exact geometry known); proteins/enzymes = MODELBEAST blobs (real ones are 3000-atom blobs nobody reads — and PIPELINE says blobby organics are TRELLIS's sweet spot); cells/creatures = MODELBEAST, D's lane, unaffected. BUILT: water (chaff) · bicarbonate (antacid ammo — HCO3- + HCl is the real neutralisation, it genuinely fizzes CO2) · glucose (nutrient/score) · caffeine (overdrive — really gut-absorbed, needs no tutorial) · ATP (boost — the orange triphosphate tail IS the charge, a pickup that's a diagram of its own mechanic) · capsaicin (burn hazard — long greasy tail reads wrong on sight) · B12 (rare treasure — a cobalt in a corrin cage, the only metal vitamin; C authored it as a pickup already). Size is a free rarity signal: `unit` mode keeps TRUE relative sizes, so water is a speck and B12 a chandelier. Found by rendering it, which is why the bench exists: fuseRing returns vertices starting adjacent to B, and bonding them from A instead draws a chord across the ring — still a valid 5-cycle so nothing errors, it just looked like a squashed pentagon. Fixed in caffeine + ATP's adenine, adjacency documented. Stated plainly: geometry is idealised (correct connectivity/rings/bond orders, believable angles, authored flat) NOT crystallographic — real glucose is a chair, not a hexagon. The B12 is a simplified corrin core. Untested above ~7 on screen; fifty would want instancing. qa GREEN. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
jing
|
111ade7d12 |
[lane A] The swept room + the acid sea, built — and it corrected me four times
The shape read in LANE_A_NOTES was written from reading code, so I built it to
find out if it was true. Flyable: web/dev/laneA_world.html?dbg=1&assets=1&lvl=swept
Nothing here is imposed on anyone: every addition is additive and a level that
opts out behaves exactly as today. C can bin the whole proposal and lose nothing.
WHAT THE PROTOTYPE DISPROVED (my own cost estimates, all four):
- "radial resolution must scale with radius" — wrong. 64 segments is a 0.05u
circle error at r=70. The arena needs it because it fbm-displaces; the tube's
wave is smooth in theta. No change made.
- "cap geometry at the span ends" — a thinko. The canal is continuous; a room is
a fat part of it. Only the fundus dome caps, and it stays an icosphere.
- cost — cheaper than feared: 9 draws / 98k tris worst-frame for a radius-70
cathedral vs 8 / 74k for the L2 corridor. No leak over a full sweep.
- "one schema flag" — held. `segments[].mode: "open"` is the entire ask.
THE TWO REAL COSTS, NEITHER PREDICTED:
1. D's `tile[0]` is a COUNT, not a density — only a texel size if radius never
changes. Measured 6.1:1 smear at r=70. The wall now derives the count from
each ring's own arc length (aRadius attr + uThetaSpan); D's authored numbers
keep their exact meaning at their reference radius, no re-authoring. That fix
then laid a seam down the canal (a fractional count can't wrap) -> rounded to
an integer, seam gone. Not a no-op on corridors: L2's hiatus takes 3 repeats
where it took 4, which is the texel size correctly holding still as the pipe
narrows. Eyeballed.
2. MY OWN PINCH GUARD WAS WRONG and would have blocked C. The fixture scored
1.30 and "failed" the >1.5 law; it was never bent. stats() divided the
tightest turn ANYWHERE by the widest radius ANYWHERE — 996 units apart, a
bend in the radius-13 pylorus against the radius of the sea. Locally its
worst point is 6.4. pinchRatio is now min over s of turnRadius(s)/radius(s).
Provably one-directional (global <= local by construction) so nothing that
passed can fail; real levels GAINED headroom: L2 3.32->4.16, L3 1.91->6.99,
L1 2.11->6.06. This mattered — the guard is what backs the round-1 promise
that C never has to think about curvature.
THE ACID SEA (world/acid.js, `level.acid {from,to,height,biome}`):
Flat emissive #c8ff3a, level and NOT tube-following — that's the mechanic. The
best find of the session: `height` is ONE NUMBER that tunes the level. Measured:
-18 -> 0% of the sea floor is dry shallow, -55 -> 14%, -62 -> 48%. So C's "position
IS the resource" falls out of the radius wobble they already author — the mucus
shallows are just "where the wall rises above the waterline", nothing to place —
and rising acid literally drowns them: free escalation, same scalar their event
pump drives. Also disproved my own claim: the centreline stays level (+/-1u over
700u), so shallows come from radius wobble, NOT the canal's bends. C must author
vertical bends deliberately if they want depth by anatomy.
Stated plainly, not papered over: the waterline is prototype quality (the plane
meets faceted rings and the shoreline steps). Acid does not drain the coat —
depthAt(pos) is the hook and the cost is Lane B's call. Nothing drives height but
DBG until C's events do.
qa GREEN incl. the new provenance gate; spline selfcheck green; L2 corridor
re-eyeballed for regression. Rulings requested from F in NOTES §-> Lane F.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
||
|
jing
|
3a6dbc30a3 |
[lane A] Complete the biome texture set (oral + colon) + a provenance gate
Generated on MODELBEAST flux_local, on-device, $0.00. Lane D's pipeline used
unchanged; **D owns these — ratify, retune or bin them.**
Why A did D's job: the six-biome tint contact sheet could only judge 3 of 6
biomes for real (oral/large_intestine/appendix had no pack and fell back to
procedural), and you cannot ratify a tint against a stand-in. All six are now
textured — the set ART_BIBLE §Biome palettes promises.
- wall_oral_a/b, wall_colon_a/b + normals. Seams 6.35/9.30/14.06/6.37 -> 1.01/
1.08/0.74/1.36 (D's "indistinguishable" band; the shipped pack's weakest is
2.39). Means normalised to ~0.50 by D's derive_maps, untouched.
- appendix wears the colon pack at its own gold tint (TEXTURE_SHARE in
world/index.js) — anatomically it IS colon tissue, and it's ART_BIBLE's
"one texture, two biomes" law for zero bytes. Distinct from the temporary
slug map; keep it when that dies.
- oral_a took three subjects. v1 named an ORGAN ("tongue") and FLUX drew the
organ's silhouette — D's "one urchin of villi" in a field's clothing. v2 named
the tissue but papillae are 3D projections, so perspective gave it a vanishing
point that tiles into starbursts. v3 asks for squamous cell pavement: flat by
construction. Rule for the kit: prompt the SURFACE, not the FEATURE.
Evidence: docs/shots/laneA/round2_oral_prompt_evolution.png
PROVENANCE DEFECT FOUND AND FIXED (-> Lane D):
batch_textures.json recorded, for 4 of 6 walls, the prompt from the framing
experiment D tried and rejected — not the prompt that made the shipped pack.
record() only runs on generation, so reverting a prompt while keeping its image
(exactly what D correctly did: "four winners kept untouched rather than churned")
drifts the record silently. Proved by regenerating esophagus_a at seed 1101 under
both prompts: the code's prompt reproduces the shipped wall, the recorded one
gives the bland wood-grain D described rejecting. Evidence:
docs/shots/laneA/round2_provenance_probe.png. Records re-pointed at the code's
prompt (images untouched — they were never wrong).
New: gen_textures.py --verify-provenance + qa.sh gate 5b, so it cannot rot again.
qa GREEN incl. the new gate; 11/11 provenance verified; all six biomes eyeballed
in engine (docs/shots/laneA/round2_tint_*.png).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
||
|
jing
|
ecf43940ec |
[lane A] Round 2 close-out: colorspace re-eyeball, stomach shape read, ruling #5
Finishes the lane A round-2 list that
|
||
|
jing
|
f0982e70c3 |
[lane A] Round 2 WIP (session cut off): crest-speed law, taper blending, schema-v2 wave, TBN walls
Landed before the cut: crestSpeed(s) + CREST_FACTOR 1.6 (ruling #1) with selfcheck asserts; radius blend widened +/-12 -> +/-25 (C's #4 dependency) + no-cliff selfcheck; per-segment wave.amp override as per-vertex aWaveA (ruling #8); colorspace law in the wall shader (ruling #2); TBN normal maps + matcap + dual detail layers (D's perturb(), trap documented in-shader); sample(s, out) v1.2; slug map shrunk to the two real mismatches. Evidence: docs/shots/laneA/round2_L2_*.png. Cut off before: NOTES/progress, stomach-arena shape read for C (task #7). Committed by F to protect the shared tree; spline + qa selfchecks GREEN at commit time. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
jing
|
689c478a7f |
[lane A] Round 1: the canal v0 — spline, streaming tube, scanner wall, arenas
Implements THE WORLD CONTRACT (TECH.md) in web/js/world/; drop-in for the stub. boot.js now runs this on C's L2_esophagus (3600u, hash cefc4f83), zero console errors. - spline.js: centreline + arclength LUTs + parallel-transport frames + radius law + peristalsis phase + project() + hash(). No three.js import, so qa can run its 14-assertion selfcheck headless. - curviness is solved from a curvature budget, not authored as an amplitude: the pinch guard caught the tube folding through itself on a wide+curvy join (turn radius 9.9 vs tube radius 17.2). C can no longer author a pinch. - peristalsis: global OMEGA, k(s) = OMEGA/flow(s), phase K(s) = integral k ds. A crest therefore travels at exactly the local flow speed, so "surf the crest" == "ride the current". Esophagus wave amp 0.9 -> 1.4 (measured): moves wallRho for Lane B. - tube.js: chunked streaming, seams aligned to biome joins, arenas own their s-span. - arena.js: displaced icosphere shells v0 (three's polyhedron detail is 20*(detail+1)^2, not 20*4^detail — 720 tris was far too coarse); fog sized to the room, since biome fog tuned for 100u corridors renders C's 360u acid sea as a black rectangle. - index.js: prefers assets.texture() over get() (get returns raw JSON a shader can't eat), plus an explicit slug map — D's wall_smallint_a vs biome id small_intestine would miss silently forever under the assets-optional law. Measured (C's real L2 + D's real pack, 1920x1080, renderer.info): 8 draws worst frame (budget 150), 74k tris (500k), no geometry leak across a full sweep, <120ms load, pinch ratio 3.32. fps not claimed — the screenshot pane runs tabs hidden, pausing rAF. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> |