I wrote a 1-bit WebGPU runtime to run a 1.7B LLM in the browser

Aidekin is a browser-based runtime that uses WebGPU to run a 1.7 billion parameter large language model with 1-bit quantization. The implementation enables on-device LLM inference directly in the browser without requiring server-side computation. It leverages 1-bit weight representation to reduce memory and compute requirements enough to fit within browser constraints.