Google just dropped something wild alongside Gemini 3 - meet Antigravity, an IDE that's basically rethinking how we code with AI.
What's Different?
Instead of the usual "AI chatbot in your IDE" approach, Antigravity is built agent-first from the ground up. The agents can autonomously work across your editor, terminal, AND browser at the same time. Think of it like having a junior dev who can actually execute full tasks while you work on other stuff.
The Cool Parts
Artifacts Over Noise
Instead of showing you every single tool call and API request, Antigravity agents produce "Artifacts" - task lists, implementation plans, screenshots, browser recordings. It's way easier to verify what the AI actually did.
Two Modes
- Editor View: Traditional IDE experience (like Cursor, Copilot, etc.)
- Manager View: Mission control for spawning multiple agents across multiple workspaces in parallel. This is the game-changer.
Self-Learning
Agents build a knowledge base from past work - code snippets, architecture patterns, successful task sequences. The more you use it, the smarter it gets for YOUR codebase.
The Details
- Free in public preview with generous rate limits
- Supports Gemini 3 Pro, Claude Sonnet 4.5, and GPT-OSS
- Available for macOS, Linux, Windows
- Local operation with async feedback on artifacts
My Thoughts
This feels like the first IDE that's actually designed for the "agent era" instead of just bolting AI features onto traditional tools. The Manager view especially - being able to orchestrate multiple agents while working in the foreground is exactly what we need as these models get more capable.
The Artifacts approach is smart too. Nobody wants to read through 500 lines of tool calls. Just show me what you're planning, what you did, and proof it works.
Worth checking out if you're building anything serious with AI coding assistants.
🔗 Official announcement: https://antigravity.google/blog/introducing-google-antigravity
Top comments (0)