DSpark: Speculative decoding accelerates LLM inference [pdf]
(github.com)
view on HackerNews
Author: aurenvale
Date: 6/27/2026