Google just destroyed all vibe-coding apps (Antigravity)
By David Ondrej
Anti-Gravity: A Deep Dive into Google’s AI-Powered IDE
Key Concepts:
- Anti-Gravity: Google’s AI-powered IDE, a competitor to Cursor and Windsurf.
- Agent Manager: A core feature of Anti-Gravity enabling orchestration and control of multiple AI agents.
- Agent-Driven Development: A development approach where AI agents have high autonomy in building and modifying code.
- VS Code Fork: Anti-Gravity and Windsurf are both built upon the open-source VS Code editor.
- Browser Sub-Agent: An AI agent within Anti-Gravity capable of controlling a browser for testing and interaction.
- Artifacts: Autogenerated outputs from AI agents, including implementation plans, walkthroughs, and screenshots.
- Nano Banana Pro Integration: Built-in access to Google’s Nano Banana Pro for asset generation.
- Gemini & Open AI Models: Support for multiple LLMs including Gemini 3, Gemini 3 Flash, and Open AI’s Opus.
I. Introduction & Installation
David Andre introduces Anti-Gravity, Google’s new IDE designed to leverage AI agents for software development. He positions it as a competitor to Cursor and Windsurf, highlighting its unique “Agent Manager” feature. Installation is straightforward, accessible via anti-gravity.google. A crucial setup step involves selecting an agent configuration, with “Agent-Driven Development” recommended for maximum AI autonomy, while a more cautious “Guided Development” option is available for existing projects.
II. Anti-Gravity vs. Windsurf: Origins & Similarities
Anti-Gravity shares significant similarities with Windsurf due to two key factors: both are forks of the open-source VS Code IDE developed by Microsoft. Furthermore, Google acquired the top developers and founders of Windsurf in July 2025 for $2.4 billion, effectively bringing the Windsurf team under the Google umbrella. This acquisition, described as a “questionable and controversial move,” mirrors Meta’s acquisition of Scale AI founders.
III. Core Functionality: Beyond Basic AI Assistance
While Anti-Gravity can be used similarly to Cursor – utilizing AI chat alongside a project – Andre emphasizes that this approach only unlocks 10% of its potential. The true power lies in the “Agent Manager,” a novel feature allowing users to orchestrate multiple AI agents simultaneously. This enables a paradigm shift where developers can remain largely disconnected from the codebase, file structure, and project details. This contrasts with Cursor, where even with auto-acceptance of changes, developers still need to monitor the codebase.
IV. The Agent Manager: A Detailed Walkthrough
The Agent Manager provides a dedicated UI for managing AI agents. Users can select from various models, including Gemini 3, Gemini 3 Flash, and Open AI’s Opus. Agents are dispatched with simple, plain-language instructions (e.g., “Read spec and implement it like a 10x engineer would”). Multiple agents can run in parallel, each performing a specific task. The “Inbox” provides a centralized overview of agent activity, allowing users to review results and intervene as needed. The UI dynamically adjusts to focus on the active agent, providing a chat interface and a real-time view of the agent’s actions, including terminal commands.
V. Practical Example: Building an Accountability Dashboard
Andre demonstrates building an accountability dashboard for participants in his accelerator program. He begins by creating a spec.md file outlining the project requirements. He then dispatches several agents:
- Agent 1 (Opus): Reads the
spec.mdand begins implementation. - Agent 2 (Gemini 3): Reads the
spec.mdand identifies missing requirements. - Agent 3 (Gemini 3 Flash): Summarizes the core idea of the project.
The agents work concurrently, generating artifacts like implementation plans and walkthroughs. The system automatically sets up a Next.js project and organizes files, demonstrating a high degree of autonomy.
VI. The Browser Sub-Agent: AI-Powered Testing
Anti-Gravity includes a “Browser Sub-Agent” capable of controlling a Chrome browser for automated testing. This agent can perform actions like clicking, scrolling, typing, and taking screenshots. It requires a browser extension and a restart of both Chrome and Anti-Gravity. While still in its early stages, this feature represents a significant step towards fully automated software development, allowing AI to not only build but also test applications. Andre cautions against complete reliance on the browser sub-agent, emphasizing the importance of human oversight.
VII. Customization & Advanced Features
Anti-Gravity offers extensive customization options, including the ability to set “reasoning effort” for agents and define custom rules using a syntax similar to Cloud Code commands. The system supports the gemini.md file for defining custom instructions, and integrates seamlessly with Google’s Nano Banana Pro for asset generation. Artifacts generated by agents, such as implementation plans and walkthroughs, can be commented on directly to provide feedback and guide the AI.
VIII. Cost & Model Availability
Anti-Gravity offers a generous free plan designed to attract users from competitors like Cursor, Lovable, and Windsurf. It supports a wide range of LLMs, including Gemini models and Open AI’s Opus, all available for free. The usage limits are synced with a user’s Google AI account, potentially providing nearly unlimited access for Gemini Ultra subscribers. Andre recommends using Opus 4.5 for backend development and Gemini 3 Pro for frontend design.
IX. Future Implications & Conclusion
Andre envisions a future where software is built for AI agents, by AI agents, with minimal human intervention. Anti-Gravity is presented as a key step towards this future, enabling a new paradigm of agent-driven development. He acknowledges that Anti-Gravity is still in its early stages but believes its potential is immense. He encourages developers to experiment with the Agent Manager and explore the possibilities of AI-powered development. He concludes by recommending Anti-Gravity for those willing to embrace experimentation, while suggesting Cursor for more production-ready solutions.
Notable Quote:
“Anti-gravity is very flexible. I would say it's like giving an AK-47 to a child. If you don't know what you're doing, you will shoot yourself in the foot. But if you do know what you're doing, this tool is very, very powerful.” – David Andre.
Chat with this Video
AI-PoweredLoad the transcript when you're ready to chat so the initial page stays lighter.
Related Videos

Stanford CS153 Frontier Systems | Building the Frontier Ecosystem
Stanford Online

'Things are going to be okay, in Canada and the U.S.': Thorne
BNN Bloomberg

I'M OUT: The $11 Trillion AI Bubble is Breaking!
Steven Van Metre

South Korea bets big on AI with nearly a trillion dollars of investment • FRANCE 24 English
FRANCE 24 English

The Bubble is Bursting... (Emergency Update)
Bravos Research

The AI Bubble Just Ended - Without Popping
Heresy Financial

AI Market Volatility, Europe Heat Wave, Venezuela Quakes Damage | Bloomberg This Weekend: June 27
Bloomberg Television