Experimental validation: Agentic Search (context trees) vs traditional RAG for code retrieval. Agentic wins with 99% fewer tokens and 2× better accuracy. Fully reproducible with automated pipelines. - View it on GitHub