→ mean(μ): population gap-trait mean (0 → 1)
↑ var(μ): gap-trait variance (0 → 0.25)
What this take shows. A two-dimensional sweep of the population's gap-trait distribution. The X axis is mean(μ_i) — the average reflective bandwidth in the population. The Y axis is var(μ_i) — how dispersed it is. For each pixel we run a small simulation: 22 agents, one injected trigger, simulate forward and record the final cascade size. Color encodes outcome: cream = self-extinguishing, sage → earth → terracotta = increasingly cascading.
Click any pixel to re-run that exact configuration in detail. The moss contour line traces the critical boundary where the system flips. The interesting prediction: high variance with low mean μ may be more stable than uniform medium μ, because high-μ agents act as keystone firewalls.