Latent Space: The AI Engineer Podcast ⚡ [AIE CODE Preview] Inside Google Labs: Building The Gemini Coding Agent — Jed Borovik, Jules
760 snips
Nov 10, 2025 
Jed Borovik, Product Lead at Google Labs and a seasoned expert in AI, reveals the innovative world of Jules, an autonomous coding agent. He discusses his transition from generative AI to coding agents and how Google collaborates with DeepMind to enhance product innovation. The conversation dives into the technical evolution of agent infrastructure, challenges in managing extensive context windows, and the transformative impact of coding agents on software development. Borovik argues that these advancements are crucial for the future of AGI and the evolving software economy.
AI Snips
Chapters
Transcript
Episode notes
A Stable Diffusion Moment
- Jed describes his Stable Diffusion moment that pushed him to build coding agents rather than fear them.
- He treated generative models as a new brush to amplify his love of programming and joined Google Labs to work on Jules.
Labs' Tight Product-Model Loop
- Google Labs builds new AI-first products by tightly coupling product teams with DeepMind model work.
- That end-to-end control—from UI to model and infra—lets Labs iterate on ambitious agent experiences.
Models Reduce Scaffolding Complexity
- Model quality drastically changes required agent scaffolding; better models simplify orchestration.
- Early systems needed complex sub-agent frameworks that become unnecessary as models improve.