the second hundred · metaphor 138
Ask how much there is to do, and the honest answer is a question back: at what resolution? The finer you look, the more the list unfolds — every task splitting into sub-tasks, every sub-task into steps — and its true length runs away from you.
"Clean the kitchen" is one line. Look closer and it's counters, dishes, floor, fridge. Look closer still and the fridge is shelves, drawers, the thing at the back you're afraid of. There is no scale at which the list finally sits still and says this is all of me. Zoom in and it keeps producing more of itself — the same jagged structure at every magnification. The length you'd write down depends entirely on how fine a ruler you were foolish enough to pick up.
Coastlines do exactly this. Measure Britain's edge with a long ruler and you get one number; with a shorter ruler you catch more inlets and get a bigger one; shrink the ruler toward zero and the measured length climbs without bound. When length depends on the ruler like that, the thing you should report is not a length at all — it's a dimension, and it lives somewhere between a line and a plane. Below, a coast you can measure yourself.
The runaway length
Walk a pair of dividers set to opening ε along the coast, counting steps. The measured length is L(ε) = steps × ε. On a smooth curve, halving the ruler just doubles the steps and the length holds steady. On a crenellated one, a shorter ruler dips into inlets the long ruler jumped straight across — so you get more than double the steps, and the length grows. Keep shrinking ε and, for a true fractal, the length has no limit.
A number that runs to infinity is useless, so we ask a better question: how fast does length grow as the ruler shrinks? Plot log L against log ε and the points fall on a line. Its slope is a stable, finite number that does describe the coast — and the fractal dimension is D = 1 − slope. A smooth line has slope 0 and dimension 1. A coast so crinkled it starts to fill the plane approaches slope −1 and dimension 2. Real coastlines land in between — Britain's west coast sits near 1.25.
That is the deep move: when a quantity refuses to converge as you refine your ruler, its magnitude was the wrong thing to measure. The invariant — the thing that stays put while the length runs away — is the rate of runaway itself. Dimension is what's left when length falls apart.
What to try
Drag the ruler slider toward finer and keep your eye on two numbers. Measured length climbs, sometimes steeply — the same coast, "longer" every time you look harder. But dimension D, read from the slope of the whole log-log plot, barely twitches. The length is an artifact of your ruler; the dimension is a property of the coast.
Now raise ruggedness. The coast grows more crenellated, the log-log line tilts steeper, and D climbs toward 2 — every measured length inflates faster as you refine. Press New coastline for a fresh random draw at the same ruggedness: different shape, same dimension, because dimension is statistical, not about any one wiggle. And press Koch curve for a shape with an exactly known dimension, log4/log3 ≈ 1.262 — watch the instrument's measured slope land right on it. Nothing here is rigged; the number is dug out of the geometry every time.
The mapping
Your to-do list is a coastline. Its length is not a fact about the work — it's a fact about your ruler: the granularity at which you're willing to break things down. At a coarse ruler ("get healthy") the list is short and the work looks doable; at a fine one ("floss, then book the dentist, then find the insurance card…") it explodes. Neither length is the truth. The truth is the dimension — how fast the work multiplies each time you look closer. A low-dimension task is roughly as big as it first appears. A high-dimension one betrays you: every level of detail spawns another, and the total was never going to sit still.
The same shape haunts grief that reopens at every scale of memory, a house renovation that finds rot behind every wall it opens, a research question that fractures into ten questions each time you answer one. Knowing something is fractal changes how you treat it: you stop demanding a final length, because there isn't one. You choose a ruler deliberately — coarse enough to act, fine enough to be honest — and you accept that a different ruler would have told a different, equally real, story.
Read as life lessons
"How much is there?" has no answer without a resolution. Anyone quoting a single length is quoting a ruler you can't see. Two honest people measuring the same task get different totals — and both are right.
When magnitude runs away under refinement, the invariant is the slope — how fast it runs. Dimension survives what length can't. Ask not "how big" but "how fast does it grow when I look closer."
You can't measure at infinite fineness and you shouldn't try. Pick a scale that fits the decision — coarse to act, fine to audit — and name it. Precision past your ruler is theatre, not truth.
In the wild
National borders and coastlines have no well-defined length; atlases disagree wildly because they used different rulers. What's reported instead — a fractal dimension near 1.1–1.3 — is stable across mapmakers.
Lungs, blood vessels, and neurons pack enormous surface into small volume by being fractal. Their dimension, not their length, characterises them — and changes measurably in disease.
Price charts, coastlike, look the same crinkled at every zoom. Their roughness is captured by a fractional dimension rather than any length, which is meaningless for a path that wiggles at all scales.
The mapping, exactly
| Mathematics | Life |
|---|---|
| the ruler ε | Your chosen granularity — how finely you're willing to break a task, a memory, a problem down. |
| measured length L(ε) | How much "there is to do" at that granularity — a number that depends entirely on the ruler. |
| L grows as ε shrinks | The list unfolding: every item spawning sub-items the finer you look, the total running away. |
| self-similarity | The same jagged structure at every zoom — sub-tasks that look like the task, questions inside questions. |
| the slope of log L vs log ε | The one stable thing: how fast the work multiplies under refinement, the same for every ruler. |
| dimension D between 1 and 2 | The honest measure of the thing — roughly-as-it-looks (near 1) versus explodes-when-examined (near 2). |
The honest model
The coast is a genuine polyline — a random one built by midpoint displacement (a fractal-terrain method whose ruggedness sets its dimension), or, on the Koch preset, the exact Koch curve after seven generations. To measure it, the panel runs a true divider walk: from an anchor it solves for the point exactly ε away along the curve, plants the compass there, and repeats — the orange path you see is that walk, and the step count is real.
It then sweeps eighteen ruler sizes across nearly two decades, computes L(ε) for each by the same walk, and fits a straight line to log L versus log ε by least squares. The reported D = 1 − slope is that fit — recomputed whenever you change the coast, never hard-coded. That's why the Koch preset lands near 1.262 on its own: the instrument is measuring, not reciting. The slider's ruler simply shows one ε from the sweep, so you can watch a single measurement climb while the slope of the whole set holds steady.
Where the metaphor tears
The runaway length is a mathematical idealisation. An actual coast bottoms out at grains of sand; an actual to-do list bottoms out at atomic actions you can't usefully split. Below some scale the self-similarity stops and the length does converge. The dimension is only meaningful across the finite window where the wiggling actually repeats — quote it outside that window and you're inventing detail that isn't there.
Dimension is a slope, and a slope needs a spread of rulers to estimate. One measurement — one length, one glance at the workload — carries no dimension at all, and a short or noisy scaling range gives a wobbly, untrustworthy fit. The metaphor sharpens how you think long before it hands you a reliable number; treat any single "D" from thin data with suspicion.
The coast makes every scale equally substantive, but a to-do list doesn't. Much of what a finer ruler "finds" is busywork, anxiety, and duplicated framing — length without weight. Measuring your life more finely can inflate the apparent load while adding nothing that matters. High dimension can be a property of the thing, or an artifact of how compulsively you subdivide it.