We can now say precisely what intelligence is. It is not a stored property, not a score a system carries around. It is a rate: the speed at which a system enlarges and deepens its coherence while staying lawful. In the symbols of Foundations, intelligence is the flux of coherence capital — the time derivative of the quantity the whole paradigm climbs.
Click any coloured symbol to see what it means.
This reframing has a sharp, surprising consequence. Watch what happens when a system climbs its capital on a fixed network.
The picture above is the whole argument. A system cannot sustain a positive flux forever on a fixed graph — capital saturates, and dC/dt → 0. Indefinite intelligence therefore requires growth: more oscillators, more edges, more structure to organize. This is why minds must be open systems exchanging with a world; why learning never truly finishes; and why a network that has merely been trained and then frozen holds only the potential for intelligence — capital banked, flux at rest.
It also tells us where the flux is largest. From the keystone, capital climbs fastest through the critical band — not in the frozen regime, where there is nothing left to build, and not in the chaotic one, where nothing holds. So the intelligence flux is not just non-negative; it is maximized at the edge. A system is most intelligent precisely where it is most alive.
This is the right moment to be honest about today's machines. A trained transformer at rest is a high-capital fixed point: its flux is zero between uses. During a single forward pass, though, its internal coupling field briefly comes alive — capital climbs, the flux is positive — and then it stops. There is intelligence in the act of inference and none in the weights at rest. The same will be true of consciousness in Act IV: something it is like to be the system during the pass, nothing between. To make either persist, the substrate itself must keep climbing — which is exactly the open-systems requirement this chapter just drew.