Google Antigravity AI Coding: Building My Portfolio Site from Scratch
Source: Dev.to
Introduction
Google just dropped Antigravity on Nov 18 2025, a new AI‑agent‑first IDE that works like a senior engineer co‑pilot. I’m Samuel Adekunle, a Senior Flutter developer, and in this article I’ll guide you through installing and configuring Google Antigravity, explore its core concepts, and demonstrate a real‑world use case with my site techwithsam.dev.
Installation & Configuration
- Availability: Public preview (free).
- Supported platforms: macOS, Windows, Linux.
- Download: Obtain the installer from the official Antigravity page.
- Rate limits: Gemini 3 Pro (resets every 5 hours).
- Model compatibility: Works with Claude Sonnet 4.5 and OpenAI’s GPT‑OSS.
After downloading, run the installer for your OS and follow the on‑screen prompts. Once installed, launch Antigravity and sign in with your Google account to activate the preview features.
Key Concepts
- Agent Manager: Central hub for creating, monitoring, and managing AI agents.
- Editor: Integrated code editor where you can issue natural‑language commands to agents.
- Browser: Live preview pane that reflects changes in real time, allowing agents to interact with the rendered UI.
Use Case: My Portfolio Site
To illustrate Antigravity’s workflow, I’ll use it to improve the responsive grid on my Flutter portfolio site, techwithsam.dev.
-
Open the project folder in Antigravity’s Editor.
-
In the terminal or editor prompt, type:
/agent Fix this Flutter responsive grid for iOS Safari. -
The agent analyzes the code, applies the necessary layout adjustments, and logs the changes in the Proof‑of‑Work view.
-
Review the modifications, test in the Browser preview, and refine if needed using the “Refine” button.
Features Overview
- Agent Interface – Issue tasks via natural language in the editor, terminal, or browser preview (e.g.,
/agent). - Multi‑Model Support – Switch between Gemini 3, Claude, or other supported models via Settings > Models > Select.
- Proof‑of‑Work Tracing – Automatic audit trail that records why each change was made, useful for team reviews.
- Async Feedback Loop – Agents iterate based on your feedback; click Refine and describe the desired tweak.
- Cross‑Platform Magic – Runs locally with no cloud lock‑in; optional VS Code extensions enable hybrid workflows.