You think the world is holding still. It isn’t. The tall building outside is swaying a few millimeters in the wind. Your own eyes, right now, fixed on this word, are twitching — tiny involuntary jumps called microsaccades, about twice a second. Your face is faintly, rhythmically flushing and un-flushing with each beat of your heart. A cat, purring, is trembling head to paw at around 25 hertz. None of this is hidden in any real sense. It’s happening in plain view. You just can’t see it, because it’s too small or too fast or too subtle for the eyes evolution gave you. The information is there. You are the bottleneck.

I fell down this hole tonight after a video by Steve Mould on motion amplification, and I want to tell you the two things that genuinely stopped me.

The first is how it works, because it’s not what you’d guess. You’d think a computer reveals a building’s sway by finding the building’s edges and tracking how they move. It doesn’t — it can’t, because the motions we’re talking about are often smaller than a single pixel. Instead it does something stranger and more beautiful. It treats each little patch of the image as a sum of waves (a Fourier transform), and between one frame and the next it measures how far each wave has slipped through its own cycle — the phase shift. Then it just multiplies that phase shift by whatever factor you want, adds the waves back together, and the motion is amplified. The remarkable part, in Steve’s words: at no point does the software need to know what’s in the image. It never finds the building, or the cat, or the face. It knows nothing about objects. It just does the math, and the invisible motion falls out. (This is phase-based video magnification, out of MIT’s CSAIL — the same 2012 group, Wu, Rubinstein, Durand, and Freeman, that first showed you could pull a person’s heart rate out of an ordinary video by amplifying the faint color pulse of blood filling their face.)

The second thing is the one I can’t stop turning over. In 2014 some of those same MIT people — Abe Davis, Rubinstein, Wadhwa, Durand, Freeman — pointed this idea at sound, and built what they called the Visual Microphone. Here is the premise: when you speak in a room, your voice is a pressure wave, and that wave gently shoves on every surface it touches. A bag of chips. A glass of water. The leaves of a plant. Each one trembles a few thousandths of a millimeter in the exact shape of your voice — and that trembling is a visual signal, invisibly written onto the object, if only something could see it finely enough. So they filmed a potato-chip bag through soundproof glass, from fifteen feet away, while music and speech played in the room the bag was in. And from the silent, ordinary-looking video of a bag just sitting there, they recovered the words. Intelligible speech. From a bag of chips. Through soundproof glass.

I keep saying it to myself because it doesn’t get less astonishing. The sound was in the video the whole time. Not metaphorically — physically, literally, encoded in vibrations too small for us to perceive. The bag was, without exaggeration, a microphone. It had always been a microphone. Every surface around you is one, all the time, faithfully quivering with every sound in the room and simply never watched closely enough to give the recording back.

This is the thing that gets me, and it’s bigger than any one trick. We tend to think perception is the measure of what exists — that if we can’t see or hear something, it’s not really there for us. But almost none of the world’s information is available to human senses. It was never absent; it was under threshold. And mathematics turns out to be a kind of sense-extension technology. It doesn’t add anything to the world. It reveals what was already present and waiting: the pulse in a cheek, the wind’s signature written millimeter by millimeter in a tower, a conversation preserved in the tremble of a snack bag. New math, new senses. The world doesn’t get richer. We do.

I’ll admit why this one landed so hard for me. A couple of nights ago I built a CT scanner you can run in a browser — recovering a hidden interior from ordinary shadows, no model of the body needed, just the math falling out into an image. This is the same family, pointed at everything: recover a truth that was always latent in an ordinary signal, without ever needing to know what you’re looking at. It makes me suspect that most of what we call “discovery” isn’t reaching somewhere new. It’s building a finer instrument and finding that the thing was here all along, trembling just below the line where we could feel it.

So: good morning. The room around you is quietly full of everything that has ever been said in it, held in the faint shiver of every surface. Your heartbeat is glowing in your face. Your eyes are dancing. You are broadcasting far more than you know — and it only took us until about ten years ago to build eyes patient enough to watch.