6. Four Capitals, Three Laws of Staying Alive

The action L = H + C + K names costs. A parallel question: what stocks does a community accumulate to lower those costs and raise achievable value?

Four capitals from three cost terms

Read capitals off the costs:

Cost term	What lowers it	Capital name
H (being wrong)	Accurate model of the value landscape	Needs two non-substituting inputs: material capital (data + physical substrate) and intelligence capital (patterns/models)
K (changing)	Shared channels that let updates propagate	Network capital
C (fragility / complexity under a brittle reference)	Spread of strategies and references	Diversity capital

Why four capitals and not three? Accurate prediction takes data and patterns as distinct inputs. A warehouse of uninterpreted facts without patterns is blind. Patterns without grounding are speculation. Neither substitutes for the other—so H takes two stocks where C and K take one.

V is not a fifth capital. V is the end you pursue, not a pile you store. You do not “accumulate an objective”; you pursue it with the means the structure provides.

Achievable value needs all four. Zero substrate, zero patterns, zero connection, or zero diversity each kills the whole in a different way—complementarity, not a nice-to-have portfolio slogan.

Three laws: one viability condition, three axes

Capitals are what you hold. Laws are what must stay true for the dynamics to remain alive at all.

The dynamics has a single attractor the population can share only while free energy stays finite and bounded below. It can run away in three principal modes—each a failure of one cost term along a different axis. The book names them flow, openness, and resilience. They are not moral stickers imported from outside. They are viability read three ways.

Flow. The value term must be conserved and circulated. Consume capital without accounting and your model diverges from reality—H blows up. Conserve value but stop circulating it and you stop generating the data that tests predictions—the model goes stale. Economically: keep value moving through the population rather than pooling out of the dynamics. Material and intelligence capital serve flow.

Resilience. The complexity term must stay bounded across states of the world. That requires diversity. A monoculture minimises C in the expected case and pays without bound when the unmodelled state arrives. A mixture of references pays at worst a finite log-weight price when a rare component matches the shock. Diversity capital buys resilience.

Openness. The distribution must stay one connected object. Connectedness fails two ways:

1. Exclusion — the community’s reference assigns some members’ configurations probability zero. The stationary target only lives on the support of µ; the dynamics cannot create mass where the reference put none. Exclusion is absorbing. Neglect (tiny positive weight) still leaves a path back; hard zero does not.
2. Detachment — a group’s realised life leaves bounded KL of the shared reference; free energy diverges; they fall out of the common attractor.

A society that drives people below viability or lets them exit the shared world has not only committed an external “injustice label.” In this framework it has fragmented its distribution and lost the ergodicity that let one dynamics carry everyone. Openness is network capital’s deepest job: connection fights entropy.

Fairness as operations, not only ethics

Under the doxa reading, these laws recover distributional and capability concerns as viability requirements:

• Flow ↔ dignity (meet needs without falling below viable support)
• Openness ↔ capability to move toward higher value while still connected
• Resilience ↔ survival of shocks the model did not rehearse

A society can be scored by asking whether free energy stays finite along all three axes. Moral traditions often condemn exclusion first; here exclusion is also a dynamical failure mode.

Next: how doxa is layered into existence—and why foundation-model training is hierarchical doxa engineering under other names.