AI on the web doesn’t have to mean Python servers, GPU clusters, and your users’ data leaving the device. Modern runtimes ship with serious ML horsepower: WebGPU/WebGL, ONNX Runtime Web, Transformers.js, WebLLM, and even built-in AI APIs you can use from plain JavaScript. In this session we’ll build three small but real demos—a browser-based web app, a desktop Electron app, and a Chrome extension—that all run vision, text, and a tiny LLM locally with no custom backend. You’ll see how to load models efficiently, stream results, and keep each UI responsive. We’ll walk through a practical checklist: when client-side AI makes sense. No servers, no API keys, no usage limits — just raw device powered intelligence.