Giménez-Romero et al. model an ecosystem with habitat heterogeneity — patches of wet, dry, cool, warm — and ask how much diversity this heterogeneity can support. Below a critical threshold of environmental variation, the answer is none. The dominant species wins everywhere. The heterogeneity exists but does nothing. Above the threshold, multiple species coexist stably. The transition is sharp: not a gradual increase in biodiversity as heterogeneity increases, but a switch from zero coexistence to stable coexistence at a specific critical value.
Sinet et al. study what happens when a system temporarily crosses a tipping point. The conventional model is binary: cross the line, the system tips irreversibly. Their analysis finds a safe overshoot region that follows an inverse-square law — the product of overshoot magnitude squared and duration squared must stay below a coupling-dependent constant. You can cross the line. You just can't cross it far, or for long, and the relationship between “far” and “long” has a precise geometry.
These are not the same system, but they share a structural feature that matters more than their differences. In both cases, the boundary between regimes is not a gradient but a geometric object. The habitat threshold isn't just “more is better” — it's a cliff below which more is nothing. The overshoot boundary isn't just “less is safer” — it's a curve that trades magnitude for duration along a specific mathematical contour.
Gaba makes this formal in a different context entirely: computational complexity classes (P, NP, EXPTIME) arise as stable sets under a scaling homeomorphism on a quasi-metric space. The boundaries between what is easy and what is hard are not artifacts of how we count operations. They are topological features of the space. The geometric structure preceded our measurement of it.
The practical implication cuts across all three domains. In ecology, modest conservation interventions — creating small patches, introducing minor variation — are predicted to fail unless they cross the heterogeneity threshold. The intervention must be large enough to reach the geometric boundary where coexistence becomes structurally possible. In climate policy, the question is not whether you will cross a tipping point, but whether you can stay within the inverse-square envelope that permits return. And in computation, the question of whether a problem is tractable is not a matter of degree but of which side of a topological boundary it falls on.
We are trained to think in gradients. A little more effort produces a little more result. But these systems have boundaries where the relationship between input and output is discontinuous. The boundary doesn't just separate two states. It defines the geometry of what is possible on each side.