Google Antigravity, the Gemini 3 Pro-powered VS Code fork launched in November 2025, offers developers two operational modes that represent fundamentally different approaches to AI-assisted coding.
Planning mode generates a detailed implementation plan before writing any code. The agent analyzes the task, documents its proposed tech stack, outlines step-by-step actions, identifies file changes, and waits for review. Developers can comment on the plan, suggest alternatives, or request changes before execution begins. After completion, it generates a walkthrough documenting what changed.
Fast mode executes prompts immediately without generating plans or waiting for approval. It interprets the request and takes action, prioritizing speed over visibility.
The practical difference matters for enterprise teams. Planning mode suits complex multi-component tasks (authentication system refactors, CI/CD pipeline creation), production code changes requiring oversight, or situations where architectural alignment needs verification before implementation. Fast mode works for straightforward fixes, quick iterations, or tasks where the approach is obvious.
What's notably absent: production readiness data. Antigravity remains in public preview, three months post-launch. Enterprise architects evaluating it against Cursor or GitHub Copilot face limited information about multi-repository context handling, a critical gap for distributed systems. Alternative tools like Continue.dev offer established enterprise features (SSO, on-premises data, JetBrains support) that Antigravity hasn't addressed publicly.
The modes themselves aren't revolutionary, they're workflow choices. Planning mode resembles standard code review processes, automated. Fast mode is autocomplete with more initiative. The real question for CTOs: does a three-month-old experimental tool warrant production evaluation, or is this vendor positioning ahead of enterprise-grade delivery?
For teams experimenting with AI coding assistants, the mode choice is straightforward. Use Planning when you need to see the strategy first. Use Fast when you trust the agent to execute without supervision. The tool's maturity for production workloads remains unproven.