Hi DEV Community! 👋

Since the release of Antigravity 2.0, all I'm seeing on social media is AI Gurus or tech influencers glossing over the agent management feature we already had. It was physically ripped from the previous Antigravity, where the agent management feature and the IDE were bundled as one. They are talking about what agent management can do like it is a brand-new thing. But beyond that, there are really cool new features and some drawbacks too. Let's talk about what Google did, probably why they did it, and the pros and cons.

Split Antigravity and Antigravity IDE

What made Antigravity different from general VS Code? It was the separated agent management environment that it provided, where we could just press Cmd + E and see all the repos we worked on with agents along with the chat history. We could also multitask using agents, even if it was kind of limited. Other than that, it wasn't any different from late VS Code updates. When Antigravity first released, we didn't have that many features on VS Code, but it caught up.

So in my opinion, Google just ripped them apart so vibe coders and any other non-coding professionals can use Antigravity 2.0, where they don't really need to see or edit code or use a terminal. They left Antigravity IDE for the people who stuck with it without going back to VS Code. Slowly, they will discontinue the agent manager on Antigravity IDE and it will just be another VS Code clone. Instead, they'll double down and develop their agentic AI coding environments on an Antigravity 2.0 type of service.

This is where Google is operating quite differently from the general market like Claude Code, Cursor, and Codex. I personally haven't used Codex and Cursor much, so there might be a bit of falsehood there. If there is, please don't mind it—I'm just sharing my opinion after seeing people use them.

Project-Level Scope (Instead of Repos)

This is another thing AI Gurus are glossing over. In 2.0, we can go beyond individual repos and look at or work on a whole project.

Personally, I have already worked on whole projects as a monorepo since I started using AI coding assistants. The reason we didn't choose monorepos before for our projects was simply because they were too hard to manage, too hard to document, and much more. But after mid-2025, I didn't have any of those problems in monorepos. I don't think Antigravity 2.0 will be used in projects large enough to require multiple monorepo-sized repositories to manage. Personally, this wasn't much of an upgrade for me.

The Real Upgrade

• Parallel Execution & Backgrounding: 1.0 would block or pause you if an agent was executing a long task. 2.0 lets you kick off multiple agents asynchronously in the background. You can have one agent writing tests while another spins up a separate service.
• The New CLI and SDK: Google is deprecating the old Gemini CLI and replacing it with a Go-based Antigravity CLI. They also launched an SDK, meaning you can write code to build custom agent hooks using a unified backend.
• Scheduled Tasks: You can now configure crons directly in JSON. For example:

  {
    "cron": "0 2 * * *",
    "task": "Run this agent every night at 2 AM to check for security vulnerabilities and refactor deprecated code"
  }

What the Future Will Look Like
Google is forcing us developers to dual-wield Antigravity 2.0 and their new CLI rather than use something built directly into VS Code. Since the average person does not want to debug a CLI or maintain a local server terminal just to have an autonomous assistant or to vibe code, Google will most likely develop Antigravity to stand alone and build.

I know practically vibe-coded projects can be a mess, bringing a lot of issues with debugging, deploying, security, or scaling that vibe coders can't handle right now. Those are the new problems the dev world has today. Since our job is problem-solving, sooner or later there will be solutions.

Let's wait and see.

(Disclaimer: Any of the articles written by me are not AI-generated. These are my honest thoughts. I don't see a point in using AI to write articles, but I will use AI tools to just format the text.)