Google Antigravity: An Agent-First IDE for Autonomous, Trustworthy Coding

Google Antigravity is a new agentic development platform that pairs an AI-powered IDE with a mission-control-style Manager to orchestrate autonomous agents across editor, terminal, and browser. It emphasizes trust (task-level Artifacts and verification), autonomy (asynchronous multi-surface operation), feedback (live, comment-based iteration), and self-improvement (a knowledge base). Available free in public preview, it supports macOS, Linux, and Windows and offers multiple model options with generous rate limits.

Key Points

Antigravity is an agent-first development platform combining a powerful IDE with a Manager surface to spawn, orchestrate, and observe multiple agents working across editor, terminal, and browser.
Trust is built through task-level context and verified Artifacts (plans, walkthroughs, screenshots, recordings) rather than raw tool calls or opaque code diffs.
Asynchronous autonomy is core: users can interact with agents across surfaces, switch seamlessly between Manager and Editor, and run background tasks in parallel workspaces.
Feedback is built in via comment-style annotations on text and visual Artifacts, which agents incorporate automatically without interrupting execution.
Self-improvement is a first-class primitive: agents learn from work and feedback via a knowledge base; Antigravity launches in public preview (free for individuals), cross-platform, with multiple model options and generous, refreshable rate limits.

Sentiment

The overall sentiment of the Hacker News discussion is predominantly negative and critical, largely driven by distrust in Google as a company and immediate frustrations with the product's initial rollout. While some acknowledged the pragmatic strategy of forking VSCode and the potential of multi-model AI agent integration, these points are heavily outweighed by strong criticism regarding Google's lack of transparency about the VSCode fork, its perceived monopolistic practices, its history of abandoning products, and significant functional issues like pervasive rate limits, a broken website UI, and login problems. Many also expressed skepticism about the current real-world utility of AI agents for complex coding tasks, suggesting a gap between marketing and reality.

In Agreement

Forking VSCode is a pragmatic and often correct solution for a new IDE, as it provides extensions, familiarity, and avoids wasting time on already implemented core editor features.
The inclusion of non-Gemini models (Claude, GPT-OSS) is seen as a positive, potentially allowing Google to subsidize API calls for data and reward signals without sharing data with other providers.
Browser automation becoming a mainstream part of the AI-assisted development loop for testing is a very interesting and valuable advancement.
The product might benefit from the 'brain-drain' of developers who previously worked on other Google IDE projects like Windsurf, potentially making it a strong contender.
The availability of the tool for Linux users is appreciated.
Offering a public preview at no charge allows broad access and testing.

Opposed

Google's failure to explicitly mention that Antigravity is a VSCode fork is seen as disrespectful and an attempt to obfuscate the product's true origins.
Strong concerns were raised about Google's perceived monopolistic practices, including fast-following competitors, cloning products wholesale, suppressing wages, and stifling market competition, with calls for antitrust enforcement.
Many users expressed deep skepticism regarding Google's long-term commitment to its products, citing a history of shutting down services and making others proprietary or user-hostile.
Numerous users reported immediate 'overload/quota exceeded' errors and hitting rate limits within seconds of trying the tool, contradicting the claim of 'generous rate limits' and indicating unreadiness.
The Antigravity website itself was heavily criticized for its custom 'smooth scrolling' script, which caused nausea and made basic navigation difficult for many users.
Users encountered significant difficulties with account setup, particularly for Google Workspace accounts, often requiring the use of personal Gmail accounts instead.
Skepticism was voiced about the current practical capabilities of AI agents (even advanced models like Claude Sonnet 4.5) for complex coding tasks, suggesting they still require extensive hand-holding and often produce 'insane' bugs.
The product is viewed by some as merely 'another VSCode fork with AI junk slapped on' rather than a true innovation, especially compared to editors built from scratch like Zed.
Concerns were raised about data privacy, specifically whether user code and interactions would be used for model training, particularly given the 'free' preview status.
The name 'Antigravity' is criticized as overly dramatic and inappropriate for a product perceived as derivative, with some calling it 'vibe-coded' marketing.
The necessity to fork VSCode instead of developing an extension is attributed to Microsoft's gatekeeping of certain functionalities for Copilot, preventing extensions from achieving similar integration.
Initial UI bugs, inconsistent element sizing, and ignored settings (like Vim bindings) indicate a lack of polish in the public preview.