The "Normsky" architecture for AI coding agents — with Beyang Liu + Steve Yegge of SourceGraph

RAG has emerged as one of the key pieces of the AI Engineer stack. Jerry from LlamaIndex called it a “hack”, Bryan from Hex compared it to “a recommendation system from LLMs”, and even LangChain started with it. RAG is crucial in any AI coding workflow. We talked about context quality for code in our Phind episode. Today’s guests, Beyang Liu and Steve Yegge from SourceGraph, have been focused on code indexing and retrieval for over 15 years. We locked them in our new studio to record a 1.5 hours masterclass on the history of code search, retrieval interfaces for code, and how they get SOTA 30% completion acceptance rate in their Cody product by being better at the “bin packing problem” of LLM context generation. Full show notes: https://www.latent.space/p/sourcegraph Timestamps: 0:00 - Intros & Backgrounds 6:20 - How Steve's work on Grok inspired SourceGraph for Beyang 8:53 - From code search to AI coding assistant 13:18 - Comparison of coding assistants and the capabilities of Cody 16:49 - The importance of context (RAG) in AI coding tools 20:33 - The debate between Chomsky and Norvig approaches in AI 25:02 - Code completion vs Agents as the UX 30:06 - Normsky: the Norvig + Chomsky models collision 36:00 - How to build the right context for coding 42:00 - The death of the DSL? 46:15 - LSP, Skip, Kythe, BFG, and all that fun stuff 1:02:00 - The SourceGraph internal stack 1:08:46 - Building on open source models 1:14:35 - SourceGraph for engineering managers? 1:26:00 - Lightning Round