I’ve been using JetBrains IDEs for years. I’ve always preferred full blown IDEs to editors like vsCode as it had more powerful features. (though vsCode has come a very long way!) Now companies are expanding on vsCode creating fully AI driven tools like Cursor and now Google’s Antigravity.
I installed Cursor a while back, but never really tried to build anything out of it preferring to stick with JetBrains or even vim in the right situation. I’ve used Tabnine’s plugin for AI and enjoyed it. It’s a shame they are deprecating my legacy account in favor of the more expensive Enterprise licenses, but I suppose that is okay. AI tools have leaped far beyond what the subscription offered and they graciously cancelled my subscription and refunded the unused portion.
I tried Aider and Gemini’s CLI application. I know a lot of people really like this format, but it feels very disjointed to me. If I’m going to use AI, it should be in my IDE. Not outside of it.
Cursor now has version 2.0 and Google has now released Antigravity. So, I decided it was time to give both of them a spin. Testing Google’s Antigravity and Cursor 2.0 offers a fascinating glimpse into the current state of AI-assisted coding tools, particularly when compared to established solutions like JetBrains’ suite. Having spent years with JetBrains products, I was eager to explore how Google’s latest offering stacks up against the competition.
While Cursor 2.0 showed a lot of promise in the creation of a multiplayer Tetris game. I found its user interface clunky and unintuitive, with panels that shifted unpredictably, making navigation frustrating. That said, this post focuses on Antigravity, which presents a more polished interface that feels more natural to use.
The project I tackled this time was a chess application, something more complex than Tetris, and one that would reveal how well Antigravity handles real-world development challenges. The UI indeed improves on Cursor’s shortcomings, offering a cleaner and more predictable layout. However, when it comes to code generation, the IDE shows both promise and significant limitations.
While the initial output looks useful, it often fails to function correctly due to a lack of situational awareness. For instance, during user registration in the chess application, the IDE repeatedly reported that “User does not exist” despite the fact that I was trying to register a new user! (of course the user didn’t exist!)
In truth, the issue was the backend FastAPI server wasn’t even running. It took more than a dozen attempts before it even considered checking server connectivity, highlighting a lack of its AI (Gemini 3 Pro High) productive troubleshooting. Now, I don’t blame Gemini 3 Pro for this as I find it quite good. I just fine that Antigravity’s currently implementation of it severely lacking. (somewhat to be expected on it’s initial release) I fully expect it to get a lot better as Google tends to be pretty good about that.
Once the backend was up and running, further issues arose with login functionality, again accompanied by generic error messages that failed to guide the user (or the AI troubleshooting) toward the root cause. After finally successfully logging in, I started a game of chess only to find that no pieces were displayed on the board. It took several iterations to realize that the chess pieces were never even created / added to the project. It only figured out the missing assets after 8+ troubleshooting iterations. Even after it created the chess pieces, it still could not render them on the board.
Upon inspecting the code, I found that the chess board existed but lacked the essential components to render pieces on the board. The entire experience felt like watching Keystone Kops attempting to solve a crime. At this point, I decided to stop experimenting, as the tool seemed to struggle with basic logic both on the project development side and troubleshooting.
It’s clear that Antigravity is still in an early stage of development, and while it has the potential to become a powerful tool, it’s not quite ready for production use yet. Its combination of a superior UI with the current limitations in AI reasoning means it may not be a viable daily driver for developers just yet. That said, I remain optimistic, especially if Google continues refining its approach. If Antigravity can achieve the kind of seamless AI-assisted development that Cursor demonstrated while maintaining a more intuitive interface, it could very well become a compelling alternative for rapid prototyping.
ps. Google, please don’t cancel Antigravity like you do so many other great products.
Meanwhile, JetBrains remains a solid choice for those who prefer traditional IDEs, though they certainly trail behind in AI integration. Hopefully, the momentum from tools like Antigravity will push JetBrains to catch up, bringing powerful AI features to their already beloved platform.
