3 ms·Semantic retrieval vs. grep for LLM code queries: 2.4x faster, 5.6x fewer tokens1 points by vincentastral 6mo agovincentastral 6mo ago[dead]