Running an LLM in the Browser: Verifying WebGPU, and Local Inference (capytoolkit.com)