4x faster LLM inference (Flash Attention guy's company) by from Hacker News on 2025-10-12 08:37 (#70PRA) Comments