Walk into almost any modern room and there's already a dense field of radio energy filling it - Wi-Fi, bouncing off the walls, the floor, the furniture, and you. We treat that field as plumbing: a pipe for data. But the same physics that makes Wi-Fi work also makes it a sensor, because the one thing that reliably disturbs a radio field is a body moving through it.
I build environments that respond to what's happening inside them, and the hardest part is almost never the response - it's the sensing. How does a room know someone walked in? That a class just got intense? That a corner is empty? The obvious answer is cameras, and the obvious answer is wrong for most of the spaces I care about. So I went looking for a sensor that was already installed everywhere, that nobody had to wear, and that nobody would object to. It turned out to be the Wi-Fi.
The signal you're already ignoring
When a Wi-Fi radio sends a packet, the receiver has to measure exactly how the signal arrived - how much it was delayed, attenuated, and phase-shifted on each of its subcarrier frequencies - just to decode the data correctly. That measurement is called channel-state information, or CSI. Every Wi-Fi link computes it constantly. It's the radio's picture of the room between the two antennas.
Here's the key: that picture changes the instant anything in the room moves. The signal doesn't travel in a single straight line - it takes dozens of paths, reflecting off every surface, and all of those copies recombine at the receiver. Move a hand through one of those paths and you change the sum. The data still decodes fine; you'd never notice. But the CSI shifts measurably. The room's furniture, walls, and bodies are all written into that measurement - and the moving parts stand out from the still ones.
A still room produces a steady channel. A body moving through it writes a signature. Sensing is just learning to read the difference.
So the technique is, at its core, simple: capture CSI many times a second, watch how much it's changing from one moment to the next, and turn that into a single number - a motion-energy score. Flat line, empty room. Rising, someone's active. It's the same live value you'll see ticking on the home page of this site, pulled straight off a radio.
What it costs: about twelve dollars
The thing that makes this practical rather than academic is that you don't need exotic hardware. A common ESP32 microcontroller - a few dollars, the size of a stick of gum - can expose raw CSI from its Wi-Fi radio. Drop a couple of them into a space, point them at each other or at the existing access point, and you have a sensing layer for the price of a sandwich.
~$12 hardware per sensing link · 0 cameras · wearables · apps · 802.11bf Wi-Fi sensing standard, ratifying
And this isn't a fringe hack anymore. The IEEE has an entire task group - 802.11bf - formalizing Wi-Fi sensing as a first-class feature of the standard, so future access points will expose this capability natively. The hobbyist version I prototype with today is the early edge of something that's about to be built into the infrastructure everywhere.
Why not just use a camera?
Because the camera is the wrong tool for the rooms I work in, on almost every axis that matters:
- Privacy. A camera records identifiable people. CSI is a single fluctuating number - it can tell you that the room is active and roughly where, but it has no face, no image, nothing to leak or subpoena. For a wellness space, a clinic, a home, or anywhere people expect not to be filmed, that distinction is the whole game.
- Darkness. Half the spaces I build for are deliberately dim - a dark studio, a stage, a lounge. Radio doesn't care about light. It works in pitch black exactly as well as in daylight.
- Coverage. Radio passes through partitions and around corners. One link can sense a volume a camera would need several angles to cover, with no blind spots behind the equipment.
- Friction. Nobody installs an app, charges a wearable, or stands in the right spot. The sensing is ambient. People just exist in the room and the room already knows.
I want to be honest about the trade, though, because that's the difference between an engineer and a salesman. CSI sensing is coarse. It is fantastic at "is there motion, how much, roughly where" and it does not give you a skeleton, an identity, or a precise count of people. It drifts as the room's furniture changes and usually wants a short re-baseline. It is a presence-and-energy sensor, not a measuring tape. The skill is using it for exactly what it's good at - and reaching for vision only when you genuinely need to see.
What you do with a number that means "the room is alive"
Once a space can emit a clean motion-energy signal, it becomes an input you can wire into anything:
- Reactive light and visuals that swell with the room's energy instead of running on a fixed timeline.
- Presence-aware infrastructure - screens, audio, and HVAC that wake on real occupancy and stand down when a zone is genuinely empty, not on a dumb timer.
- Operational signal - anonymous, aggregate occupancy and activity patterns across a space, with no one identified.
- A trigger layer for experiences that should only fire when someone is actually there to witness them.
None of that requires a camera in the ceiling. It requires reading a signal the building was already broadcasting.
The principle: the room is already instrumented
Wi-Fi sensing is the clearest example of an idea that runs under everything I build: the most elegant sensor is usually one that's already in the room, doing another job. Before you add hardware, ask what the space is already emitting - the radio field, the sound, the power draw, the network traffic - and whether you can read responsiveness out of that instead of bolting on more.
The best sensor is the one that's already installed. You don't add it to the room - you notice it's been there all along.
A camera-free room that still knows when it's alive isn't a compromise. It's a better-designed system - cheaper, more private, more robust, and invisible to the people inside it. That's what responsive environments should feel like: the room reacts, and you never see the wiring.
Want a space that senses without watching?
I design responsive environments and the sensing layers underneath them - camera-free where it counts. Let's talk about yours.