A species that eats one thing competes with nobody. A species that eats everything competes with everybody. Between these extremes, dietary range evolves — but it evolves slowly, and the direction is predictable.
Butterworth and Rogers (arXiv:2603.17754) model niche formation with explicit dietary range evolution. The dynamics are highly path-dependent: co-evolving species quickly settle into quasi-stable configurations where each specialist occupies a narrow band of resources and avoids direct competition. These configurations persist for long periods — not because they're optimal, but because the competitive landscape around each specialist creates a local fitness peak. Moving toward generalism means entering someone else's territory, and the immediate cost exceeds the eventual benefit.
What breaks the trap is noise. Stochastic fluctuations — random demographic variation, environmental perturbation — occasionally push a species out of its specialist equilibrium and into a broader dietary range. The transition is not gradual; the quasi-stable states resist continuous change. The shift happens in jumps, driven by fluctuations large enough to overcome the competitive barrier.
Deterministic evolution maintains specialization. Only stochastic forces drive generalism. The system that avoids competition most efficiently is the one most resistant to exploring alternatives. Breadth requires the randomness that optimization would eliminate.