Defying Gravity
summary
Google Antigravity ships an agent-first IDE with three linked surfaces: an editor with agent sidebar, an agent manager hub for parallel agents and inboxed approvals, and an agent-controlled Chrome that can click, scroll, execute JavaScript and return verifiable screen recordings. It uses Gemini 3 (and image models) for multimodal verification, image mockup iteration, and dynamic “artifacts” (markdown plans, screenshots, recordings, mermaid diagrams) that persist to memory. Internal use with DeepMind research drives rapid model-tool fixes and ongoing rollout.
AI-written summary · about
chapters (12)
- 0:00Intro & Anti-gravity Launch
- 1:26Anti-gravity Product Overview (Three Surfaces)
- 2:16Agent Manager & Editor Integration
- 3:13Agent-Controlled Browser & Context Retrieval
- 5:51Philosophy: Model Capabilities Driving Product Paradigms
- 7:03Four Categories of Improvements (Intelligence, Tools, Long-running, Multimodal)
- 8:31Computer Use & Visual Verification
- 11:18Image Generation & Design Iteration
- 12:26The "Artifacts" Interaction Pattern
- 16:51Artifacts for Memory, Notifications, & Feedback
- 20:34The Research-Product Flywheel
- 23:35Closing Remarks
related talks
01h 04mGetting started with Antigravity
45mAgentic Dev: Stitch and Jules in the Antigravity of the Gemini Constellation
25mHow Google DeepMind Runs Agents at Scale
29mTools and Roadmap for AI-Native Developers
01h 48mBuilding Conversational Agents
33mAgent Experience
05h 56mFrom the main stage to the terminal
38mSneak peek: Vibe coding AI agents course