Will "Real-time LLM Inference on Standard GPUs: 3k tokens/s per request" reach 250 points on Hacker News? | Prophecy