OK, so Gemini 3 and Antigravity are very cool — particularly for folks trying to vibe code things into production. Antigravity can take control of a browser and actually do the intricate manual work that newbies to coding can’t do when working with production services.
That last part is the point. For a beginner, the code is the easy half. The hard half is the hundred small steps between the code and a service that’s actually running — dashboards, permissions, keys, the form that has to be filled in exactly right. That is where most people who’ve vibe coded something run aground, and it’s the part an agent with a browser can now do for them.
Some context on what arrived this week.
Google shipped Gemini 3 on November 18 and calls it “our most intelligent model that helps you bring any idea to life.” It went out on day one in the Gemini app, in Search, in AI Studio and inside a new product called Google Antigravity — an agent-first development environment where, as Google describes it, agents work directly in the editor, the terminal and the browser. Antigravity’s own launch post promises agents that “autonomously plan and execute complex, end-to-end software tasks,” is free in public preview, and lists Gemini 3, Anthropic’s Claude Sonnet 4.5 and OpenAI’s GPT-OSS as the models available inside it.
The browser is what separates it from one more chat window bolted onto an editor.
And then Opus 4.5 came out two hours ago, and I am back to using Claude on everything.
Anthropic released Claude Opus 4.5 today, November 24, calling it “the best model in the world for coding, agents, and computer use,” priced at $5 per million input tokens and $25 per million output, and available from day one in Anthropic’s apps, its API and all three major clouds. Claude Code now runs inside the desktop app as well.
I would use Claude Opus 4.5 in Antigravity, once that is an option. It isn’t on the launch list.
Been an exciting week.
