Fractals
August 12, 2026 · 3 min mins read
There's an old question that sounds like a trick: how long is the coast of Britain? Measure it on a map with a hundred-kilometer ruler and you get one number. Switch to a one-kilometer ruler and the number goes up, because now you're tracing every bay and headland the first ruler skipped over. Switch to a one-meter ruler and it goes up again, tracing every rock. There's no ruler small enough to give you a final answer, because the coastline is rough at every scale you look at, and the roughness never smooths out. That's a fractal: a shape where zooming in doesn't get you closer to simple, it just gets you more of the same kind of complicated.
The strange part is how little you need to produce that. Take a straight line, replace the middle third with two sides of a triangle, and you've made a zigzag. Now do the exact same replacement to every straight segment in that zigzag. Then do it again. And again, forever. The rule fits in one sentence. After enough repetitions you get the Koch snowflake, an infinitely detailed edge wrapped around a finite area, and every piece of it, no matter how far you zoom, looks like a smaller copy of the whole. Nothing about the rule mentions a snowflake. The complexity isn't designed in, it just falls out of repetition.
I notice the same shape in things that have nothing to do with math. A codebase where one small convention, like how errors get handled, gets applied consistently at every level ends up with a kind of coherence that's hard to fake by planning it top-down. A habit repeated daily compounds into something that looks, from far enough away, nothing like a single instance of itself, and yet is made of nothing but that one instance over and over. You don't need the grand design. You need the rule to be worth repeating, and then you need to actually repeat it.