Engineering notes on making gesture control work on the phone already in your pocket — including the parts we got wrong. No launch-hype, no invented metrics.
The wedge · July 2026
Why your hands, and not a ₹3 lakh headset
Spatial computing arrived with a price tag most of the world was never going to pay. The pitch was compelling and the hardware was genuinely impressive — and it asked you to strap a computer to your face and spend more than many people earn in a year.
We started from a different question. Not "what's the most advanced way to do this", but "what's already in the pocket of a billion people?"
A front-facing camera is a depth-blind, low-cost sensor that happens to be everywhere. That constraint isn't a compromise. It's the whole strategy.
Every WICKFURY product runs on hardware you already own. No headset, no controller, no depth sensor, no wearable, no subscription to a cloud service that has to stay online for your game to work. The camera watches your hands, the phone does the maths, and nothing else is required.
What that buys you
Zero hardware cost. The barrier to trying gesture control drops to "open a link."
Zero latency to a server. There's no round trip, because there's no server in the loop.
It works on a train. Or a plane, or a village with patchy signal.
What it costs us
Honesty matters more than polish here, so: a single RGB camera has no depth information. We recover hand pose from a 2D image, which means bright even lighting helps a lot and a dark room hurts. Fast motion blurs. A hand at the very edge of frame gets lost. We spend a great deal of engineering effort on exactly these failure modes — filtering, dropout grace periods, re-acquisition — and we still lose your hand sometimes.
That's the trade. A headset would solve those problems and introduce a much larger one: almost nobody would ever use it.
"The camera never leaves your device" — what that actually means
It's an easy sentence to write on a landing page and a hard one to earn. So here is the specific, checkable version of the claim.
Hand tracking runs entirely in the browser or on the phone, using an on-device model. Your video frames are never uploaded, never stored, and never sent anywhere. They are read from the camera, turned into a set of coordinates describing where your hand is, and discarded — frame after frame, in memory, on your machine.
The strongest privacy guarantee isn't a promise in a policy. It's an architecture where the data physically cannot go anywhere.
How you can verify it yourself
Open your browser's developer tools, go to the Network tab, and use any gesture feature on this site. You will see model files being downloaded. You will not see anything being uploaded.
Turn off your WiFi after the page loads. The gestures keep working.
On our fitness app, go further: check the Android manifest. It declares no internet permission at all — the operating system itself would refuse to let it phone home.
Why we build it this way
Partly because it's right. Mostly because a camera-driven product that uploads video is a product nobody should trust, and we'd rather not ask you to. There's also a practical benefit: an architecture with no server in the request path is one that can't leak, can't go down, and doesn't get more expensive as more people use it.
Building WICKFIT: fitness with no internet permission
WICKFIT is a gym and a runner you play with your whole body. The camera counts your reps and checks your form. There's no band, no watch, no weights, and — the part we're most pleased about — no internet permission in the app at all.
It's built on a different engine to our hand-tracking games: where those read 21 points on a hand, this one tracks your entire body. Same principle, more joints.
Where it is today
Built and running on Android. Not released yet. It covers 11 exercises across 4 movement styles. We're not going to give you a launch date, because we don't have one we'd bet on, and a date we miss is worse than no date at all.
The decision we're proudest of
Seated and wheelchair-playable modes were designed in from the first build — not added later after someone asked. A fitness product that only works if you can stand up excludes an enormous number of people for no good technical reason. Full-body tracking doesn't require standing; it just requires that somebody bothered to think about it.
Accessibility isn't a feature you bolt on at the end. It's either in the foundation or it's a retrofit that never quite fits.
That's also the honest reason this company exists. Gesture control started as a way to make a phone usable for people who can't reliably use a touchscreen. The games are how we prove the technology works and fund the rest of it.