How far from equilibrium?

The setup

A formula that looked exactly like a mistake

There's a number at the centre of this whole programme, and the first honest thing to say about it is that it looked, for a while, like it was wrong.

Call the number χ. It's a ratio — how much energy your surroundings deliver while you're settling down, divided by how much it would cost you to leave. Above 1, the world is giving you more than enough to break loose. Below 1, it isn't, and you stay exactly where you are.

The trouble is that the shape of this — an exponential fall-off in one number — is also the shape of one of the most famous equations in chemistry, the one that says how fast a reaction goes as a function of temperature. Line the two up and the energy your surroundings deliver sits precisely where temperature sits. Which seems to say something specific and checkable: throughflow is a kind of temperature.

That's not a small claim to get wrong. Temperature is supposed to be one number — a mercury thermometer and an infrared one are supposed to agree. In systems that are being pushed rather than left alone, that stops being true: three different, independently reasonable ways of measuring an "effective temperature" agree for a while and then, measurably, disagree. So if χ really were a temperature, it had wandered into a fight it was never going to win.

The specific way three thermometers can disagree

You could define an effective temperature from how much a particle jiggles in place, from how fast it moves, or from how often it manages to escape a trap it's sitting in. Three reasonable definitions. In a system that's being driven rather than sitting still, those three numbers can agree beautifully across a wide range of conditions — and then, in identifiable regimes, come apart from each other. It was measured, directly, in 2026. A quantity that's supposed to be one number isn't, once something is pushing on the system from outside.

→ So was the formula simply wrong?

The way out

It was never about temperature. It was always about wind.

Not one character of the formula changes. Only what it's a formula of changes.

Go back to the mountain range. A hiker gets blown across it by gusts of wind that don't favour a direction. Whether they cross depends on exactly two things: the height of the lowest pass — not the average height of the range, the lowest point anyone could get through — and how strong the wind is. The chance of crossing falls off exponentially in the pass height, measured in units of wind strength.

Notice what that argument never once needed. It never needed the wind to be thermal. It never needed a temperature, or equilibrium with anything. It only needed the wind to be weak compared to the pass — because if it weren't, the hiker would cross constantly and there'd be nothing here worth writing an escape law for in the first place. The exponential shape comes from the mathematics of rare events, not from thermodynamics. On this reading, the energy in χ was never a temperature. It's a noise scale — the strength of the wind — and a noise scale doesn't have to agree with itself measured two different ways, because it never claimed to be a physical property of anything. It's just how hard the world happens to be pushing.

And the regime where this actually matters turns out to be the regime where the reasoning is actually valid, for free: an escape law is only worth having when escapes are rare, and "rare" and "the pass is high compared to the wind" are the same sentence. You don't have to go check that the assumption holds separately — needing the formula at all is the check.

→ Fine — so what does the wind actually do to you? Time to find out yourself.

Try it

A marble in a dip

Three controls. No right answer — a range of what happens, and you're looking for the shape of it, not a score.

The conundrum

Get the marble over the lip.

(This part needs JavaScript to actually play — here's what you'd find.) A marble sits in a dip with a lip on either side. Shake pushes it symmetrically, flipping direction at random — more shake, more chance of escape, but not by much on its own. Persistence (ψ) controls how long each push holds its direction before flipping, relative to how fast the marble naturally settles. Tilt is a steady lean, not noise at all.

Turning up shake alone barely moves the escape rate. But holding shake steady and sliding persistence finds a narrow window — an intermediate value, not the extreme — where the marble escapes far more often, the same way a swing goes higher when you push it in time rather than harder. Push persistence far past that window, in either direction, and the escape rate falls back down on both sides. Tilt behaves completely differently: past a fairly low threshold, the marble goes over the same lip almost every time, steadily, regardless of shake or persistence — because a lean isn't noise, and no amount of shaking reproduces it.

Shake 0.50

Symmetric noise. As likely to push left as right.

Persistence, ψ 1.00

How long a push holds its direction before it flips.

Tilt 0.00

A steady lean. Not noise — try it with shake at zero.

→ Did you find the window — the persistence where it crosses often, with the shake turned no higher than before? Keep it in mind. What you just found by hand has a name.

What you just found

A resonance — and why one number in a formula can never produce it

If you found the window, you found something the formula as first written can't hold.

A resonance is a relationship between two timescales — how long a push holds its direction, and how fast the thing being pushed settles on its own. A formula with a single number, χ, has no second timescale to be in a relationship with. So the resonance you just found by sliding a control back and forth is, by itself, proof that something's missing from the one-number version — not an edge case, the actual structure of the problem.

The fix is a second number, ψ: the drive's persistence divided by the system's own settling time. χ tells you whether escape is affordable at all. ψ tells you which regime you're actually in — the shop or the auction, same money, completely different rules for what it buys you. And the tilt you tried is the cleanest demonstration of why the two numbers can't be collapsed into one: no matter how you turn up shake, it never reproduces what even a small, steady tilt does on its own.

The 1-in-a-million surprise that comes from the same distinction

Deliver the same total energy two ways — constant small jitter, or rare enormous kicks — and the escape rates aren't close. Under jitter, escape needs an improbable run of small pushes lining up in the same direction. Under kicks, the system doesn't bother waiting for that; it waits for one big one. In one measured case — a tracer particle in a bacterial suspension — the actual escape rate was about one in a million. A calculation from the particle's variance alone, ignoring this distinction, predicted one in 10³². Twenty-six orders of magnitude off: the gap between "once in the age of the universe" and "every few seconds."

Full derivation, the published paper, and three more ways the formula's insides break in interesting directions: go deeper on the escape law.

→ There's a second, older question this same number answers — not just whether you escape, but whether you can tell why, afterward.

The payoff

Two ways to end up somewhere, that look identical from the inside

Cool a metal slowly and it anneals — the accidents get shaken out, and what's left is close to the best arrangement available. Cool it fast and it quenches — whatever configuration happened to be there gets frozen in, accident and all. From inside the metal, both processes end in something that doesn't change any more. You cannot tell, by looking at the final piece alone, which one happened to you.

The same ambiguity sits inside χ. Turning down the world's temperature (annealing) and deepening your own rut so the same push no longer moves you (canalisation) are the numerator and denominator of one ratio — and they produce indistinguishable endpoints. Read a genome as a record of frozen optima, the way you'd read slow-cooled steel, and you get one story. Read the same genome as a record of frozen accidents, the way you'd read a quench, and you get a different one — and nothing about the sequence itself tells you which reading is correct. Get the read-mode wrong and you don't get a vague answer. You get a confident, precise, wrong one — the worst failure mode there is, because it looks exactly like a result.

Figure 3 — Annealing vs canalisation. Brief for Rhys Muirhead: same escape law; one panel moves ε, the other moves ΔΦ. The two end states must look identical.

There's a way to break the tie, and it isn't a cleverer read of the genome — it's a second, independent record. The rock underneath a lineage measures how hard the world was actually pushing (ε) without caring at all what survived it. The genome measures the depth of the rut (ΔΦ) without caring what the weather was doing. Neither ledger alone can tell you the read-mode. Held next to each other, the geological ledger tells you which read-mode the genomic one is in — not two separate answers, one setting the interpretation rule for the other.

Figure 2 — Two ledgers. Brief for Rhys Muirhead: one event, two records, shared unknowns. Must show the sharing.

→ What happens on the far side of the line, when a push is too strong to just anneal or canalise around? Sometimes a system fragments. Sometimes it finds a whole new way of holding together instead. Nobody has a good account yet of what decides which — tracked, unresolved, at the open loops page.

A formula that looked exactly like a mistake

It was never about temperature. It was always about wind.

A marble in a dip

Get the marble over the lip.

A resonance — and why one number in a formula can never produce it

Two ways to end up somewhere, that look identical from the inside

The same resonance, in places that have nothing to do with marbles.

Where this chapter landed