This assistant is a large language model (LLM) that runs entirely in your browser on WebGPU — no server, no API call, so nothing you type leaves your device. A small Startup model becomes ready first so you can chat right away, then a stronger Final model downloads in the background and takes over automatically. Weights are cached after the first visit; loading progress is shown at the bottom. A WebGPU-capable browser is required for live chat — without one, the topic buttons above still work (they return curated content).