Why naive RAG fails for coding agents—and what Satori does instead

Why fixed text chunks are weak evidence for code changes, and how Satori combines symbol-aware chunking, hybrid retrieval, freshness checks, and exact bounded reads.