Flows and the impossible temperatureSquinting has destinations — and one point in the dial that refuses to choose

The last section ended on the important observation: squinting doesn't just shrink a lattice, it changes it — usually into a cleaner or a messier one. So imagine squinting over and over, and tracking where the system goes. It turns out almost every starting point rushes to one of two dead ends. The exceptions — the starting points that refuse to commit — are rare, special, and they are where this entire library lives.

Three fates

Below are three copies of the same thermal lattice, prepared at three temperatures: cold (well-ordered, a little speckle), near the transition, and hot (nearly random). One button squints all three at once. Under them, a chart tracks each one's agreement a level by level — its trajectory under repeated squinting.

The faint dashed curves are reference trajectories (means of repeated offline runs); the bright dots are your three live lattices. Cold rushes upward and saturates: total order, one solid color — squinting a mostly-ordered thing distills its order. Hot crashes to the coin-flip line and stays: pure noise is a destination too. The middle one does neither — it descends slowly, still between the two dead ends at every level, structure appearing under every squint. level 0

The two dead ends deserve names. A lattice at a = 1 squints to itself forever: solid gold is solid gold at every scale. A lattice at a = 0.5 also squints to itself forever: the majority of four coin-flips is a fresh coin-flip. Both are fixed points of the squinting operation — states that look the same at every scale for the dullest possible reason: they have no scale-dependent structure left to lose. Order is one attractor; noise is the other; nearly every lattice you can prepare is secretly en route to one of them, and squinting just reveals which.

The point that refuses

Which makes the middle trajectory the interesting one. Between "flows up" and "flows down" there must sit a knife-edge — a temperature where the system, squinted, reproduces itself: not solid, not random, but the same statistical texture at every magnification. Domains inside domains inside domains. At that special point the system is scale-invariant — a third fixed point, nothing like the two dead ends, rich instead of empty. Physics calls the neighborhood of that point criticality.

Hunt for it yourself:

Pick a temperature, press run: the machine prepares a fresh lattice there and squints it twice, drawing its trajectory over the three reference fates. Far below the transition it commits to order almost instantly; far above, to noise. The closer you dial toward the transition — from either side — the longer the trajectory lingers between the dead ends before committing. Exactly at the critical point, in the infinite ideal, it would linger forever. verdict = —

Now connect this to everything you have read on this site. The essays keep insisting that minds, and living structure generally, sit at the edge of a phase transition — the critical band of the keystone chapter, the sharp threshold you first heard as beats stopping in primer 00. This section is what that placement means in scale language: the critical point is the one place on the dial where the system is not en route to a dead end. Order has no news left in it; noise never had any; the scale-invariant point is the only regime where every level of magnification contains real structure — where zooming out never exhausts the thing. If you want a system whose coarse description is as alive as its fine one — a scale ladder worth climbing — there is exactly one place to stand.

The full machinery — trajectories through the space of systems, with fixed points as destinations — is called the renormalization group flow, and you have now run it with your own hands. One more idea completes the toolkit: what happens when two different machines flow to the same place. That is the next section, and it is the punchline the whole module exists for.

Honest finite-size numbers On an infinite lattice the critical trajectory would hold perfectly level — that is what "fixed point" means. On our 128×128 lattice with a 2×2 majority rule, the measured critical trajectory drifts slowly downward (mean over seeds: a = 0.730 → 0.641 → 0.560 across two squints, versus the hot run's immediate crash toward 0.5 and the cold run's immediate saturation at 1.0). Two honest reasons: the lattice runs out of levels, and the majority rule is an imperfect (truncated) squint — both standard, both stated in every textbook treatment. The qualitative signature survives the imperfection: near the critical point the system takes conspicuously longer to commit, and the pictures stay domains-within-domains rather than solid or speckle. The critical temperature of this lattice family is known exactly: T_c = 2/ln(1+√2) ≈ 2.269.