2. Why Prediction Wins

Before the math of choice, a fair question: why start from “agents compare by value, with limited capacity, over time”? Why not some other triple?

The answer is not a free taste in psychology. It is a claim about what survives.

Decay is the default

The second law of thermodynamics, said without equations: left alone, ordered structure tends to erode. The ways a system can fall apart vastly outnumber the ways it stays put. Persistence is what needs explaining. It is not the free default.

A structure that lasts in a changing world is doing work against that erosion. Among systems competing for finite resources, those that anticipate the environment spend less to stay alive than those that only react after every shock. Predict what is coming, position early, pay less. Fail to predict, pay full repair costs forever. Over many rounds, the system with an accurate, cheap model of its surroundings is the one still present later.

This is not a pure mathematical theorem. It is an observation joined to a selection argument—and the paper marks it as the one load-bearing empirical premise: persistence under the second law selects for systems that hold accurate, cheap predictive models of their environment. The companion book The Last Economy argues the case at length under the name Intelligence Theory. Here it is taken as the starting fact about the world.

Three properties fall out

From that premise, the three properties the choice-chain needs are almost descriptions, not inventions:

1. Valued comparison. A system selected for prediction acts as if it ranks configurations—preferring those its model marks as good for persistence over those it marks as threats. That ranking is what the framework calls comparison by value. The value function V is the formal trace of “what I’m trying to bring about.”

2. Bounded capacity. Prediction is not free. Every irreversible step of modelling has a physical cost floor (the paper later ties this to Landauer’s bound—the minimum energy cost of erasing information). So capacity is finite. The agent never resolves the single best configuration with infinite precision. Choice stays noisy and approximate.

3. Action over time. The whole story is about survival across time. Persistence is the frame. Yesterday’s choices condition tomorrow’s options.

Valued comparison, bounded capacity, and action over time are what a persistent predictive system looks like from outside. That is why they are the right premises for an economics: an economy is a population of such systems provisioning themselves against decay.

Sorter’s Law (name for later)

The companion book names a principle: a persistent system minimises the sum of predictive error, model complexity, and the cost of updating—the Intelligence Lagrangian or action L = H + C + K. The next chapters derive why those three costs appear and why their static and dynamic forms are forced. For now, hold the slogan:

Systems that last cheaply predict well; systems that predict well under limits choose in one forced pattern.

Given an agent that compares by value with bounded capacity and acts over time—plus consistency requirements stated as they are used—the structure of its choice is unique. No alternative survives the requirements.

Next: that unique shape, in plain language—what ρ, µ, V, and τ mean without drowning in symbols.