The internet is abuzz about the latest LLM darling: Gemini 3.
Based on the early benchmarks I’ve seen — including those by Google itself in the blog post announcing the new model — we can expect great things. No surprise there. We’re in another cycle whenever a new model is released, which means incremental improvements across the board for all to see.
Sure enough, Cursor was one of the first major dev platforms that I could see support Gemini 3. Under the model picker you can add it and then select Gemini 3 Pro. I was able to successfully test it on the Stark Insider server with no problem, and was getting excellent scripting and simple coding results (I don’t vibe code to create apps, so much as to administer this server in addition to the Vertigo AI lab server).
Soon enough, however, the popularity of Google’s shiny new Frontier model was evident and an overload message popped up in Cursor:
Again, no surprise really, as a lot of us have been excited for this release and to try and compare it to the coding incumbents: Claude Code and Codex (OpenAI GPT). So far, based on my limited use today I’d say it looks the part, and seems to perform as you’d expect a flagship model to perform. Claude Code still appears strongest at overall communication, docs and task management, but it’s far too early for me to say for sure.
And that overload problem disappeared pretty quickly in any case, and I was experiencing no issues running Gemini 3 Pro through its paces, creating scripts, monitoring logs, fine tuning performance, etc.
But: Antigravity!
While Gemini may get all the headlines today… and for the foreseeable future when it comes to LLM discussions, there is a Trojan Horse that’s not to be missed.
It’s called Antigravity and it’s Google’s take on the AI-powered development platform, aka an IDE (Integrated Development Environment). Like we’ve seen with Cursor, Windsurf and other modern, AI IDEs, this one is also a fork of the well established Visual Studio Code (open source, by Microsoft). Only, in this instance it’s entirely built around Gemini.
I installed Antigravity, imported my VS Code settings, chose the slick Tokyo Night theme and was off and running, and connected via ssh to starkinsider.com. I authenticated so it could see my Gemini subscription (Google One). From there, everything was pretty much business as usual with the interface looking nearly identical to VS Code… and Windsurf… and Cursor. Gemini, of course, is the built-in AI Assistant, and I found it worked well.
We want Antigravity to be the home base for software development in the era of agents. Our vision is to ultimately enable anyone with an idea to experience liftoff and build that idea into reality. From today, Google Antigravity is available in public preview at no charge, with generous rate limits on Gemini 3 Pro usage.
— Google Antigravity Team
But, the really big play here, in my estimation, is the enterprise.
RELATED POSTS ON STARK INSIDER:
When Engineering and Development teams make standardization decisions they’ll have to include Google and Antigravity. They’re just too big and important to ignore during and evaluation. So, if it was mainly VS Code, Cursor and a handful of others, it will now be the usual suspects + Google. And that’s daunting news for the likes of Cursor. Things just got a whole lot tougher.
This will be an interesting space to watch. IDEs are an expanding market. Once, purpose built for hardcore developers, with the advent of LLMs and vibe coding, their practical applications and use cases have significantly broadened in recent years, and, especially, I’d suggest, in 2025. With AI on board, these toolkits are quite powerful and the entry to barrier has been reduced.
Keep an eye on Gemini 3, yes, but really watch out for the real killer app: Antigravity.