Every dot below just sits and buzzes in place — nothing self-propels. The only thing that moves the crowd is what happens when two dots touch. Turn the dials and watch a quiet canvas decide, on its own, whether to stay quiet or turn into a gas.
Each particle carries no motor and no preferred direction — left alone it only trembles with a small constant thermal jitter, which a constant drag pulls back down to almost nothing. The only way energy gets added is a collision: when two dots touch, the model resolves it as an elastic bounce and then scales the impulse by a restitution factor e. At e = 1 a collision is ordinary and energy-neutral. Push e above 1 (super-elastic — the model's stand-in for the paper's contact-charge discharge on impact) and every collision hands back more speed than it received. More speed means faster, harder collisions, which means more collisions per second, which means more injected energy — a feedback loop with nothing self-propelled anywhere in it.
The transition below is genuinely emergent: the physics is simulated directly from that one collision rule, not scripted from the sliders. Whether a given (density, e) pair settles or runs away is measured live from the particles' own kinetic energy, against the resting jitter floor the same rule predicts for an empty room. The exact threshold curve isn't derived from the paper's kinetic theory — explore the sliders to find where it flips.
"Mean kinetic energy" is measured straight from every particle's current speed, smoothed over about a second. "Quiet-room floor" is what that same measurement settles to with collisions switched off — the resting buzz from thermal jitter and drag alone, at whatever e is set. The state flag reads buzzing once the live energy sits well above that floor for a sustained stretch, and quiet otherwise. Nothing about that call is hand-set per density — it's the same ratio test at every setting.