The true belief is behind the wall — it needs the sum you can never finish. So carry one you can move: a position on a line from 0 to 1. Then learn to tell a good position from a bad one, by three pressures you can actually feel — and prove, by your own hand, that no single one is enough.
Same five doors as before — the plain line runs straight through; take whichever fit: ▶ Picture it, ✎ Try it (commit before the reveal), ✓ Check it yourself, ▸ Go deeper, ⬡ See it in UNI. Runs locally — no account, no network — and adapts to light and dark.
The one fence that never moves. “Belief,” “fit,” “stretch,” “looseness” are functional quantities — a position a model holds and the forces on it. For you we use felt words, because you’re a person reading a sky; for UNI we use functional words — it holds a number and its dynamics move it. Neither is awareness, feeling, or sentience.
No scrolling back to Module 2. A cook pulls a tray from the oven — the edges are darker than usual, the smell is sharp, there’s a thin haze above it. The cook wants the true chance the batch is burnt — not a hunch, the honest answer. Reconstruct, cold: why can’t they just compute it?
From memory, in a new domain (no rain, coins, dice or sailor): say why the true number is out of reach, pick the reason, then check. Reconstruct it — no re-reading.
If that came back cold, you’re ready for the way around the wall.
You can’t reach the true belief about the hidden cause — that needs the forbidden sum. So don’t try. Carry a stand-in belief instead: a guess you’re allowed to pick, hold, and slide around until it’s as good as you can make it.
Back in your doorway’s two-cause world from Module 2 — front on or off. The truth about how much to believe “on,” given the sky, was 0.913 — but that came from dividing by the evidence 0.69, the sum you’re avoiding. So set it aside. Instead, just pick a number for how strongly you’ll believe the front is on — and get ready to move it.
So look at what a belief actually is now: not a fact, not a number handed down — a position. A spot on a line from 0 to 1.
(a) 0 = “certain the front is off.” (b) 0.5 = “genuinely torn — could be either.” (c) 1 = “certain the front is on.” Everything between is a shade of leaning. That one number is your whole belief about the hidden cause, and you can put it anywhere.
That’s the earned idea: a belief is a position b on [0, 1] you can move. Now — how do you tell a good position from a bad one, when you can’t peek at the true answer?
You can’t score a position against the truth (behind the wall). But you can feel three honest pressures on it, each judged from things in your hands. Meet them one at a time — agreeing case first.
Commit, for each pressure, which of three positions it prefers — then reveal where the three actually point. The lesson is the CONFLICT: no one position wins all three. (This shows the three pulls, not a scored curve — the single score comes in Module 4.)
You’ve made the first move around the wall. A belief is no longer a fact you can’t reach — it’s a position you can hold and slide, and you can feel three honest pressures on it: fit, stretch, looseness. And you proved, by your own ranking, that they pull in different directions — no position is best on all three at once.
So the next rung is forced: put the three pressures on one scale and add them into a single number — a height that’s low where a belief is good and high where it’s bad. That single number is the one this whole book is named for, and here’s the beautiful part waiting: it will contain no evidence term at all. The forbidden sum won’t appear. That’s how a mind scores its beliefs without ever walking through the wall. You assemble it in Module 4.