From Fastest AI Coder to Broken Dev Environments Overnight
Antigravity 2.0 was unveiled as Google’s most ambitious AI coding tool yet, powered by Gemini 3.5 Flash and positioned directly against Claude Code and OpenAI Codex. In benchmarks highlighted by Google, Gemini 3.5 Flash not only tops Gemini 3.1 Pro on coding and agentic tasks, it also pushes aggressive performance numbers, streaming tokens far faster than rival models. On stage, Google showcased Antigravity orchestrating dozens of parallel subagents to assemble an operating system from scratch, then booting Doom on it. But the excitement evaporated when the automatic Antigravity update landed on existing installations. Overnight, many developers opened their machines to find familiar terminals, file explorers, and editors missing, replaced by an agent-centric chat interface. Active projects were suddenly unusable without manual intervention, turning what should have been a flagship Antigravity 2.0 release into an avoidable outage for thousands of developer setups.
An Agent-First Suite: Command Center, CLI, and Managed SDK
Beneath the rollout chaos, Antigravity 2.0 represents a strategic shift from single-assistant coding helpers to a multi-agent development suite. The new command center can coordinate several agents in parallel across projects and repositories, scheduling routine actions and long-running automation. A reworked Antigravity CLI consolidates the now-deprecated Gemini CLI into a single, branded toolset that keeps core concepts like Skills, Hooks, subagents, and plugins while emphasizing multi-agent workflows. Google is also pushing an SDK and managed-agent hosting via the Gemini API so teams can prototype custom agents and run them as isolated Linux jobs in the background. The vision is clear: one coordinated system handling editor-based coding, terminal automation, and agent orchestration. Yet this centralization also concentrates control in Google’s hands, raising uncomfortable questions about lock-in, flexibility, and how quickly missing developer ergonomics will be restored.
The Gemini CLI Shutdown and New Access Gatekeepers
The most contentious change is not just new features, but who gets to use the old ones. Google is sunsetting Gemini CLI and related IDE extensions for most users, cutting off free and many individual tiers. Starting June 18, Gemini CLI and Gemini Code Assist extensions will no longer serve requests for Google AI Pro and Ultra users, or for those relying on the free Gemini Code Assist for individuals. New installations of Gemini Code Assist for GitHub will be blocked, and existing setups will soon stop receiving responses. Enterprises, however, largely retain access: organizations with Gemini Code Assist Standard or Enterprise licenses, and enterprise GitHub integrations via Google Cloud, can keep using Gemini CLI. For everyone else, the path forward is either Antigravity CLI—still missing full feature parity—or direct access via paid Gemini and Gemini Enterprise Agent Platform API keys.
Developer Backlash: Closed Components, Forced Migration, and Lost Control
The community reaction has been swift and blunt. Forums and subreddits filled with reports of broken environments and criticism of an “agent-first” pivot that feels premature for daily production work. Many see the automatic update, which split Antigravity into three separate downloads and stripped out the IDE defaults, as a top-down decision shipped without regard for real workflows. Source control, local terminals, and remote connections disappeared from the main install, forcing developers to hunt down a separate Antigravity IDE just to restore basic capabilities. At the same time, Antigravity CLI is more closed than its Gemini predecessor, eliminating the comfort of an open-source core that many teams had built tooling around. For developers accustomed to open AI coding tools and transparent CLIs, the combination of forced migration, access restrictions, and opaque components has turned a promising enterprise AI development platform into a cautionary tale about platform risk.
What Developers Need to Know About Migrating to Antigravity 2.0
Teams relying on Gemini CLI or the original Antigravity editor now face a staged but unavoidable migration. First, restore baseline productivity: install the standalone Antigravity IDE to recover editing, file management, and terminal features removed by the default 2.0 agent app. Second, audit any scripts, hooks, and plugins currently bound to Gemini CLI. While Antigravity CLI supports core concepts like Skills, Hooks, and subagents, Google has already warned that there will not be one-to-one feature parity at launch. That means testing each workflow in a non-critical environment before cutting over. Third, clarify your access model: enterprise customers with Code Assist Standard or Enterprise licenses can keep Gemini CLI for now, buying time for a controlled transition. Individual developers, by contrast, must either embrace Antigravity CLI’s closed, multi-agent future or budget for paid API keys to retain a measure of flexibility and independence.