Jed Borovik from Google Labs is building Jules, Google’s autonomous coding agent, after nine years working on search. Jed has lived through the shift from classic ML ranking to long-running AI agents with their own computers. His “Stable Diffusion moment” convinced him that AI wouldn’t erase his craft as a software engineer, but become a new brush for writing code which led him into Google Labs to help build Jules. We caught up with him at GitHub Universe to dig into what truly makes Jules “autonomous” (dedicated cloud environments, API/CLI triggers, sessions that run for days), how better base models are letting teams delete over-engineered scaffolding and persona trees, why naive code RAG are giving way to grep + tools + huge context windows, and how coding agents are reshaping software engineering. @jedborovik @swyx