scripod.com

⚡ [AIE CODE Preview] Inside Google Labs: Building The Gemini Coding Agent — Jed Borovik, Jules

Shownote

Jed Borovik, Product Lead at Google Labs, joins Latent Space to unpack how Google is building the future of AI-powered software development with Jules. From his journey discovering GenAI through Stable Diffusion to leading one of the most ambitious coding ...

Highlights

In this episode, Jed Borovik, Product Lead at Google Labs, shares insights into the development and vision behind Jules, Google's autonomous coding agent. The conversation delves into the technical evolution of AI-driven software engineering, the strategic role of coding agents in advancing AI capabilities, and how Google is positioning itself in the rapidly evolving landscape of AI development tools.
02:19
The speaker has been at Google for nine years
06:41
Jules can run autonomously in its own environment with API and CLI support
09:39
As models improve, agent scaffolding becomes simpler
11:31
Attention-based search can scale better than RAG for code indexing
22:25
Building agent companies is easier than infra due to faster ramp-up and better margins.
26:15
The best context compression method is still unknown and evolving with models
34:05
Commoditized software work can be delegated to agents, freeing engineers for higher-level design.
38:15
Interactive planning enables real-time collaboration between users and machines to refine code requirements.
40:21
AI coding agents can now process visual inputs like images and videos for enhanced development workflows

Chapters

Introduction and GitHub Universe Recap
00:00
New York Tech Scene and East Coast Hackathons
00:57
From Google Search to AI Coding: Jed's Journey
02:19
Google Labs Mission and DeepMind Collaboration
04:19
Jules: Autonomous Coding Agents Explained
06:41
The Evolution of Agent Scaffolding and Model Quality
09:39
RAG vs Attention: The Shift in Code Understanding
11:30
Jules' Journey from Preview to Production
13:49
AI Engineer Summit: Community Building and Networking
15:05
Context Management in Long-Running Agents
25:06
The Future of Software Engineering with AI
29:02
Beyond Vibe Coding: Spec Development and Verification
36:26
Multimodal Input and Computer Use for Coding Agents
40:20

Transcript

swyx: Okay, Jed Borovik, welcome to Latent Space. Yeah, thanks for having me. So we're sitting here at F.Ink's beautiful podcast studios, and we're actually meeting at GitHub Universe. How's it been so far? Jed Borovik: It's been great. I mean, yeah, the ...