Games that see.
Brains that learn.
Beat a neural net at rock-paper-scissors, outsmart an image generator, and keep up with vision research — all in your browser. Every model runs on your device. No photo ever leaves your camera.
The arcade
Each game hides a real computer-vision lesson. Play first — the "how it works" card appears when you win or lose. Your camera frames are processed locally and thrown away instantly.
Rock-Paper-Scissors vs. the Net
Hand-landmark model reads your throw off a webcam. Check its confidence bars — that's a softmax in the wild.
classificationhand landmarks 🕵️Spot the AI
Real photo or diffusion model? Train the oldest neural net there is: your visual cortex, with feedback.
forensicsgenerative models 🧠CV Daily
Today's vision papers from Hugging Face, filtered to what actually sees things. One link each, zero fluff.
research feed🔒 Privacy is the whole architecture
We use MediaPipe's on-device models (WebAssembly + WebGPU). There is no upload endpoint to begin with — which is why we can honestly promise your face and your room stay yours. Nothing is stored, nothing is streamed.
Fuel the arcade
Education stays free forever — every "how it works" card, always. Donations unlock new games, fresh weekly Spot-the-AI rounds, and leaderboards. Current community goal: 20 coffees → Pose Charades ships 🕺
Prefer paying with your phone's camera? Scan the QR →
Scan to donate