Running an LLM in the Browser: Verifying WebGPU, and Local Inference | Dark Hacker News