From IDE to Multi-Agent Command Center Overnight
Antigravity 2.0 AI marks Google’s most aggressive push yet into AI developer tools. Unveiled at Google I/O as a standalone Google AI coding tool, it sheds its identity as a simple IDE and re-emerges as a multi-agent development suite. Powered by Gemini 3.5 Flash, which Google says surpasses its previous Gemini 3.1 Pro on coding and agentic benchmarks, Antigravity now spans macOS, Linux, and Windows. The release bundles a new command center for orchestrating parallel agents, an Antigravity CLI that consolidates the old Gemini CLI, and an SDK for building custom agents and workflows. Google is clearly positioning Antigravity 2.0 AI against rivals like Claude Code and Codex, pushing beyond single-prompt assistance toward full workflow orchestration. But this strategic leap also means developers must rethink how editing, terminals, and agents fit together in their daily tools.
Automatic Update Shock: When Editors Turned into Chat Windows
Within hours of launch, Antigravity 2.0’s rollout triggered a wave of backlash as existing environments were silently transformed. An automatic update pushed on May 19 replaced the familiar code editor with an agent-first interface, leaving developers staring at chat windows where terminals, file explorers, and core editing tools used to be. Essential features like source control, local terminals, and remote connections vanished from the default install. Google simultaneously split Antigravity into three separate downloads, with the new 2.0 app focused on AI agents and a separate Antigravity IDE required to reclaim traditional workflows. Community forums and subreddits quickly filled with complaints, with some developers calling the move a massive step backward and accusing Google of shipping production changes without technical empathy. What was meant as a bold upgrade in AI developer tools instead felt like an unannounced platform migration executed in the middle of active projects.
Inside the Multi-Agent Development Suite Google Envisions
Behind the chaotic rollout sits an ambitious technical vision. Antigravity 2.0’s new agentic layer runs subagents in parallel, defines them dynamically, and keeps each agent’s context uncluttered through better memory management. Google demonstrated the system building an entire operating system in about 12 hours using 93 subagents, processing billions of tokens before running a Doom clone live on stage. The multi-agent development suite adds a command center for coordinating complex projects, grouping conversations into Projects, and spanning multiple repositories. Asynchronous task execution prevents agents from stalling when compilers or long-running jobs are in play. On the terminal side, the Antigravity CLI replaces Gemini CLI while retaining skills, hooks, subagents, and plugin support. Together with the SDK and managed-agent hosting via the Gemini API, Google is pitching a platform where developers design agent networks, schedule routines, and offload long-running coding or automation tasks to background AI processes.
Competing with Claude Code and Codex on Speed and Scope
Antigravity 2.0 AI is also a statement of intent in the AI coding tools market. Google is explicitly targeting the momentum behind Claude Code and Codex with a broader, faster platform. Gemini 3.5 Flash is pitched as a high-throughput engine, reportedly generating at 289 tokens per second, compared with figures cited for rival models like Claude Opus and GPT-5.5. This performance underpins Google’s demos of large-scale, multi-agent builds and long-running workflows. On the business side, Antigravity now includes subscription tiers starting at USD 20 (approx. RM93) per month, with a higher AI Ultra plan at USD 100 (approx. RM466) that lifts usage caps and adds storage and media perks. By deprecating Gemini CLI and consolidating under the Antigravity brand, Google is clearly framing this as its flagship AI developer tool, aiming to become a primary choice rather than just another experimental assistant in a crowded market.
Enterprise Trust, Governance, and the Cost of Rushed Change
For enterprises evaluating AI developer tools, Antigravity 2.0’s debut raises as many questions as it answers. On paper, managed agents running in isolated Linux environments, SDK-based extensibility, and a split between IDE and CLI surfaces offer a compelling architecture for controlled, large-scale automation. However, the auto-update that broke thousands of setups underscores gaps in rollout governance and change management. Enterprise teams typically demand clear deployment controls, opt-in migration paths, and published security and governance terms before standardizing on a platform. Antigravity 2.0’s agent-first default and sudden deprecation of Gemini CLI highlight how quickly Google can rewire core workflows. Until Google demonstrates a steadier release cadence, stronger configuration controls, and transparent upgrade policies, many organizations may treat Antigravity 2.0 as a powerful but volatile option—impressive in capability, yet risky to adopt at the heart of mission-critical development pipelines.