one life to live · metaphor 46 of 100
Wander first, then settle — everyone says it, nobody says the schedule. Metallurgy does: heat until everything can move, cool slowly enough that what sets, sets deep. Cool too fast and you freeze into the first shape you happened to hold.
You know them both. The friend who took the first stable job at twenty-two and never looked sideways again; the friend still "exploring" at forty-five, résumé a travelogue, nothing kept. We talk about them as opposite errors — one too cautious, one too flighty — but they are the two ways a cooling schedule fails: quenched too fast into a shallow arrangement, or held so hot that nothing was ever allowed to crystallize.
A blacksmith who cools steel suddenly locks its atoms mid-argument: hard, brittle, full of strain. Annealed steel — cooled by degrees — lets the atoms keep renegotiating until they find a deep arrangement, and holds it. Computer science stole the trick, called it simulated annealing, and made the schedule explicit: a single number, temperature, dialed down at a chosen rate. Below, one small life runs the algorithm. The valleys are ways it could settle; the deepest one is not the nearest one. Watch what the cooling rate — and nothing else — decides.
Uphill on purpose
A pure optimizer — always move to the better spot — rolls to the bottom of whatever valley it starts in and stays there forever. That is the friend at twenty-two: greedy in the technical sense, and greed on a rugged landscape is a guarantee of mediocrity. The nearest valley that will have you is almost never the deepest one that would.
Annealing's one addition is the rule above: sometimes, deliberately, go uphill. Take the move that makes life worse — with probability e−Δ/T. While T is high, backsliding is the only mechanism by which a search escapes a shallow home. The gap year that torched a good CV, the first career abandoned at its peak, the move away from everyone who loved you — every one of them is an accepted uphill move, unjustifiable by the local gradient, justified only by the schedule. Watch the uphill moves accepted readout: while it is high, the ball can leave any valley; as it falls, the current valley slowly becomes the last one.
The schedule
Everything above is folklore — "be open when young." The algorithm's real contribution is the part folklore omits: what matters is the rate of cooling. The instrument's anneal is geometric — every step, T ← 0.9995·T — patient, relentless, never finished-all-at-once. Each stage of the search has a slightly lower tolerance for disruption than the stage before, and none of them has zero, until the very end. Cool slowly enough (logarithmically slowly) and the search reaches the global optimum with probability one. The catch is that "slowly enough" means, in practice, longer than you have. Every real schedule is a negotiated surrender between the depth you want and the time you are given.
Read as biography, the schedule is the pacing of commitment: how much disruption you will still entertain at twenty-five, at thirty-five, at fifty. The quenched life set T to zero at the first offer. The never-cooled life kept T high forever — and the instrument shows what that costs: the ball visits the deepest valley and keeps the memory of it in best fit found, but it cannot keep anything, because keeping is what low temperature is.
What to try
Quench, ten times. Press it, let it freeze, press it again. The tally underneath keeps score. Most runs end in hometown or the sensible degree — decent valleys, honestly reached, nowhere near the deepest. Nothing was wrong with any single move; the schedule itself was the mistake.
Anneal, once, patiently. For a long while the trajectory looks like the flaky friend: aimless, backsliding, crossing the whole landscape without keeping anything. Then watch the trail's color cool from ember to ice as the wandering narrows, and the ball settles — usually into the vocation, the deep valley a greedy search never sees. The aimlessness was load-bearing.
Freeze, then reheat. Let a quench stick somewhere shallow, then press Reheat. Temperature jumps; the settlement melts; the search resumes with everything it learned still on the board. That is what a sabbatical is for: a deliberate, temporary re-raising of your tolerance for disruption, taken because you suspect your comfort is shallow. Then try the manual dial and hold T anywhere: every steady temperature is a temperament, and none of them, held forever, is a life.
Reading lives as schedules
Once you have the lens, schedules are everywhere. A startup runs hot on purpose: pivots are cheap, identity is molten, every uphill move is entertained. A bureaucracy is quenched — often it is a startup that froze in whatever posture it held the day it first succeeded, strain and all. And the traditions that actually survive centuries are neither: they are annealed, cooled over generations of small proposals, most rejected, a few accepted even though they made things locally worse. A constitution with an amendment clause is a cooling schedule written down: harder to change than a habit, easier than scripture, temperature low but — crucially — never zero.
The mapping
| Mathematics | Life |
|---|---|
| the landscape | The possible ways your life could settle — towns, trades, scenes, vocations — laid side by side. |
| depth of a valley | Actual fit: how well an arrangement suits you, discoverable only by standing in it for a while. |
| temperature T | Your current tolerance for disruption — how large a worsening you will still accept to keep searching. |
| accepted uphill move | The choice that looks like backsliding — quitting, leaving, starting over — that is the only exit from a shallow home. |
| cooling schedule | The pacing of commitment across a life: each stage slightly less tolerant of upheaval, none of them entirely closed. |
| quenching | Settling into the first valley that would have you, and calling the resulting strain "stability." |
Where the metaphor tears
The convergence guarantee assumes a frozen world and infinite time: cool infinitely slowly over an unchanging landscape and you reach the global optimum. Your landscape does neither favor. Valleys deepen, silt up, and close while you wander — the scene disbands, the industry dies, the person who made the town home moves away. You must cool faster than the theorem would like, which means accepting that your settlement is probably not the best one that ever existed, only the best one reachable on the schedule a mortal gets.
In the algorithm, reheat costs one assignment: T goes up, nothing else changes. In a life, melting a settlement destroys some of it irreversibly — the marriage reconsidered is not the marriage it was, the career paused does not wait at the same salary. Real systems show hysteresis: the path back is not the path out, and some arrangements simply do not survive being made molten again. Budget for the reheat before you press it.
The algorithm evaluates the energy function exactly: it always knows precisely how deep it stands. You do not. Humans routinely misread shallow comfort as deep fit — the valley feels bottomless from inside, especially after years of telling yourself so. Half the value of high temperature in a human life is not escape but measurement: you learn your own depth mainly by climbing partway out and feeling the pull.