METHOD · 2026

The Anatomy of a Prediction

Turning unstructured prediction data into structured data.

A single calm luminous line splitting into stacked waves of different character
A poll percentage, a market price, a forecast probability — the world's predictions arrive as single numbers, and a single number is a compression. Everything that decides the outcome was thrown away at the door. This note shows what a prediction is actually made of: waves that align, oppose, and cancel across a population — and how decomposing them turns the world's most consequential unstructured data into structure a machine can act on.

A number is a compression

The world runs on predictions that arrive as single numbers. A poll says 60%. A market prices the same event at 61 cents. A forecaster calls it 0.6. Each number is treated as data — but it is unstructured data in the deepest sense: a one-dimensional shadow of a high-dimensional object, with everything that decides the outcome discarded at the door.

Two consensuses can read identically — sixty percent and sixty percent — and carry opposite futures. One is bedrock: built on who people are, strengthening under attack. The other is borrowed from the crowd: intact until the moment everyone learns everyone is leaving, then gone in an afternoon. Every instrument on Earth reports these two as the same number, because the number is all the instrument can carry.

Opinions and predictions are waves

Look underneath any consensus and it stops being a quantity and becomes a superposition. Across a population, belief behaves like waves: where reasons align, they reinforce — the belief is intense and stable. Where they oppose, the belief is fragile. And where they cancel, the surface goes calm while enormous force sits loaded beneath it — the flat reading that precedes the overnight collapse.

The reading — the height of the composite — is what every poll, price, and forecast reports. The composition is what decides what happens next. Fragility is not a mystery to be intuited; it is the arithmetic of the components.

THE READINGwhat every instrument sees=COMPONENT · SLOWbedrock — survives shocks+COMPONENT · FASTherding — cascades+CANCELLED PAIRa calm surface, loaded like a springSUM ≈ 0 · INVISIBLE IN THE READINGONE READING — MANY WAVES
One reading, many waves. The composite line at the top is all any instrument sees. Beneath it: a slow bedrock component, a fast herding component — and a cancelled pair whose sum is flat. The pair is invisible in the reading and decisive in the outcome: a calm surface, loaded like a spring.
Every instrument reads the height. We read the wave.

Reading the topology

Composition has a shape. Sample a population on any question and the distribution tells you what the average never can: whether you are looking at one crowd or two, where the density actually lives, and which clusters are moving first.

The figure below is a divided sample. An ensemble of language models — the flattened voice, polled like a person — answers confidently from one peak. The population mean lands in the valley, where almost no one stands. And on the far slope, a tracked cluster of early movers is already forming. Three readings of the same question: one wrong, one meaningless, one invisible — unless you can see the shape.

TOPOLOGY · ONE QUESTIONSAMPLE — DIVIDEDLLM ENSEMBLEconfident rejectPOPULATION MEANsoft / split — where no one standsCONTRARIANtracked clusterREJECTACCEPTDENSITY
A divided sample on one question. The LLM ensemble answers from a peak; the population mean lives where no one stands; the contrarian cluster — the part of the future that arrives first — is only visible as topology.

From unstructured to structured

This is what the civilization model does to a prediction: it decomposes it. Ask Earthlings any question and the answer arrives with its anatomy attached — how solid the consensus is, how much of it is conviction and how much is herding, where the cancelled tension sits, which way the number breaks under pressure, and the distribution of futures when the world is run forward.

That anatomy is structured data in the literal sense: fields a machine can read, compare, monitor, and act on. The world's predictions stop being opaque scalars and become objects with insides — queryable through the same interface everything else is.

What it's for

Prediction markets: two contracts at the same price are not the same bet — the anatomy says which one is crowded and which way it breaks. Finance: the difference between a consensus and a crowded trade is composition, visible before the unwind. Politics: bedrock electorates and cascade electorates poll identically and end differently. Media and brands: a narrative that lands and a narrative that shatters produce the same day-one number — and opposite week-three realities.

And machines: an AI that consults the civilization receives structure, not vibes — the anatomy of the crowd, as an input. The missing organ speaks in structured data.

The point

Humanity already produces its most valuable dataset every day — its predictions about itself. It has just never been able to read them as more than heights. Decompose the wave, and the same numbers become structure: solid or hollow, loaded or spent, breaking up or breaking down.

Unstructured in. Structured out. That is the method.

← The Missing Organ
Earthlings: The First Civilization Model →