Beelix — control the light, solve the boids into a helix

A hive at the top is a pooled particle emitter: it spawns bees from a free-list, and the death plane at the bottom recycles any bee that crosses it straight back into the pool — so the population holds a steady state (≈ emit-rate × fall-time). Same kernel shape as the swarm bench: struct-of-arrays, a pure fixed-1/60 s step(), reproducible from (seed, #steps).

The honest version (and what changed)

An earlier cut of this bench cheated: it drove each bee with analytic servos — hold radius r→R, hold tangential speed →ω·R, hold descent. Those three are the parametric equation of a helix, so the shape was authored, not emergent. This version throws that out. The bees now run real Reynolds boids — separation, alignment, cohesion — as self-propelled, constant-speed particles. The only things authored are the light pipe's controls.

Why pure boids can't do it alone — and what the light is for

Boids never spontaneously make a helix. Their emergent states are flocks, mills (a spontaneously rotating ring — a fish bait-ball), and swarms. A helix is a mill + descent. Two ingredients make that reachable here:

• Self-propulsion (cruise speed). A mill is agents circling at a steady speed; ordinary drag kills circulation and you just get a column. So bees hold a cruise speed (active matter), which lets a rotation sustain.
• A symmetry-breaker. Confined to the pipe by phototaxis (a pull toward the glowing axis — sets no radius) and pulled down by flow, the bees still need something to pick a rotation direction. Left to pure boids (Light swirl = 0) the handedness only locks about half the time — otherwise it's a wandering column or competing CW/CCW domains. That's the genuine, measured behaviour, and you can watch it: the HUD reads the swarm's angular momentum L.

So the light pipe carries one faint rotational input — Light swirl (the pulses spiral by a small amount) — and boids alignment amplifies it into one coherent, repeatable mill that the descent shears into a helix. The swirl's sign sets the handedness (try Counter-spin). Everything else — the radius, the pitch, the rope thickness — stays emergent. Crank the swirl and the light does more of the work; drop it to 0 and you're back to raw, flaky self-organisation. The selftest measures all of this: with a swirl, sign(L) tracks the swirl on every seed; at 0 it just runs.

The pulses

Bright rings launch from the hive and travel down at pulseSpeed. A bee within pulseBand of a ring gets a downward kick and an emissive glow (an iBright per-instance shader term), so a bright wave ripples down the column in lockstep with the light.

Files

Kernel + selftest: tjs/beelix/beelix.js, beelix.selftest.mjs (18 node checks — steady-state population & exact pool accounting, determinism, the handedness-follows-swirl control property, tube containment, raw-emergence stability, pulse travel). Bee mesh: tjs/beelix/beemesh.js. Reuses the swarm bench's curl-noise via its _internal export.