I’ve been using AI code assistance with Antigravity lately and once a skeptic, I should say I’m impressed by how much it can achieve. I’ve first tried Gemini and then, switched to Claude Code. In my experience, Gemini is pretty good and gets the job done but Claude Opus seems to be on another level when it comes to coding. I was beyond impressed when it laid out the implementation plan and the measures it “thought” to eliminate possible bugs and errors. Needless to say, the end result for the generated code was way beyond my expectations.
So far, I’ve tried two different approaches:
- describing the whole game at once,
- separating logic.
After watching many videos online, I tried the whole game approach at first but somehow, it ended up locking Gemini into an error mode. After the incident, I switched to Claude without losing more time and also chose the “separating logic” approach. It appears to be working flawlessly for me at the moment. I am hoping to release the game and the source code soon and see if it works 