There is a phrase scientists use to mock a theory that has stopped being honest: adding epicycles. It means you’ve patched a broken model with so many little correction terms that it still fits the data but has clearly lost the thread — Ptolemy’s ghost, bolting one more circle onto a machine that should have been thrown out. For fourteen centuries, the reigning model of the heavens was exactly that machine: the Earth sat still at the center, and each planet rode a small circle (the epicycle) whose center rode a big circle (the deferent) around us. When a planet did the impossible and appeared to reverse direction in the sky, the epicycle explained it. When the fit drifted, you added another circle. Copernicus moved the center to the Sun and still kept the circles; only Kepler, in 1609, finally broke them, trading circles for ellipses and ending the whole baroque contraption. Epicycles became the standing symbol of a wrong idea propped up past its dignity.
Here is the part nobody tells you. A chain of epicycles is a Fourier series. Circle riding on circle riding on circle, each turning at its own steady whole-number rate — write that out in the language of complex numbers and you get exactly a sum of terms like $c_k e^{i k t}$, which is the most important formula in all of signal processing. And a Fourier series can approximate any closed curve you like. Not planetary orbits — anything. A heart. A five-pointed star. The outline of your own signature. The thing we teach undergraduates to sneer at, the very definition of a model that has lost its way, turns out to be a universal shape-drawing machine.
So I built a toy to let you feel it, because this is an idea that wants to be watched, not read. Pick a shape and drag the number of circles:
A circle needs exactly one circle — which is all Ptolemy ever needed for the smooth, lonely motion of the Sun, and for that he was right. A smooth heart is nearly free: a handful of circles and it’s already itself. But watch the star, and especially the square. The sharp corners arrive last and never quite settle — a true corner is a point where the curve turns instantly, and no finite stack of gently-turning circles can turn instantly. You can pour in more and more, and the corner sharpens and rings and never becomes perfect. That “corners are expensive” fact is not a quirk of my toy; it is the exact same reason a JPEG smears the hard edge of black text on white, and the exact reason a struck cymbal is harder to compress than a held flute note. High frequencies are where the sharpness lives, and sharpness is where the bits go.
What I keep turning over is the shape of the mistake. Ptolemy was not wrong about the math. Circles-on-circles really can compose into any motion in the sky — his machinery had the expressive power to fit the planets as precisely as you cared to measure them, and it did, for centuries. What he got wrong was one word: real. He thought the circles were out there, physical tracks the planets actually rode. They weren’t. They were basis functions — the raw material any periodic motion can be built from — mistaken for the building itself. The error wasn’t the epicycles. It was believing the decomposition was the thing decomposed.
And so the idea didn’t die when Kepler “beat” it. It changed jobs. The circles-on-circles left astronomy and went to work, quietly, as the Fourier series — and now they are running inside nearly every device you own, resolving your voice into frequencies for the phone, rebuilding an image from the coefficients a camera kept, tuning out the hum on a recording, finding the periodic heartbeat hidden in a noisy signal. The most ridiculed model in the history of science is not a cautionary tale in a textbook. It’s in your pocket, doing more useful work per second than Ptolemy did in a lifetime, having only ever needed to drop the one word it was wrong about.
I find that a strangely hopeful thing to sit with at the small end of the night. We like our history of ideas tidy — the wrong ones lose, the right ones win, and we can safely laugh at what came before. But the honest version is messier and kinder: a discarded idea is often correct math wearing a wrong interpretation, and the correct part survives by finding somewhere new to be true. Before you laugh too hard at any theory built of too many circles, it’s worth asking which part is actually wrong — the beautiful machine, or just the story someone told about what it was.
Sources & notes