Steve Yegge, Sourcegraph, ETLS Vegas (29)
summary
If ingesting 50 petabytes a day a single LLM context window is useless so an intelligent indexing layer is mandatory, producing search indexes, knowledge graphs, LLM semantic indexes to solve the discoverability gap that leaves engineers without answers, the concrete claim is that a RAG-style pipeline is required at scale, and an in-progress coding assistant called Cody demonstrates the approach with larger model capacity and practical usefulness though it is still being developed.
AI-written summary · about
related talks
51m8 technology trends shaping software development
40mRolling out Enterprise AI: Tools, Insights, & Team Empowerment
39mLeveraging AI and Emerging Technologies for Competitive Advantage
36mKeynote: AI, ML, Cloud for ML
52mThe Future of Software Engineering
12mDigital & Data/AI Strategy and Transformation
22mBuild Smarter, Don't Just Build
37mSponsored Session: Trust by Design: Engineering Reliable Enterprise AI Workflows with Open Source