High-Performance LLM Inference | Dark Hacker News