I’ve been an Antigravity Pro user since January, though I recently took a break in April because the usage limits lately were just too restrictive for my workflow. I checked back in yesterday on the latest build to test some refactored code, and I have to start by saying, congrats on the new infrastructure. I’m barely seeing any “Server Busy” errors anymore, which makes the experience much smoother.
It’s clear that moving the family plans into a shared pool was a smart move, it effectively soft bans the resellers in some developing country who were abusing the system, and it’s finally fair to those of us who actually follow the TOS.
However now that the backend is stable and the loopholes are closed, you really need to address the lazy model performance and the lockout policy. While Gemini 3 Flash is incredibly fast now, it feels like it’s stopped thinking.
It used to manage complex, interconnected modules, but now it feels like a modified Gemma 3 with context caching, it reads 3-5 lines, ignores the global context, and just guesses. I spent 90 minutes trying to fix a bug, and it just spat out incomplete code until I hit my limit, leaving me with more bugs than I started with.
The cascade of limits is the real killer:
- Gemini 3.1 Pro cleaned up Flash’s mess but maxed out in 30-60 minutes debugging.
- Sonnet Fast overthought the debugging until I was hit with the dreaded 7-day lockout.
Being locked out for a week after two 5-hour sessions is just absurd for a Pro plan, especially when the models are currently struggling and error to find bugs.
Now that the family plan is a shared pool and the system is more secure, please restore the original usage limits. It would be much fairer to return to a 5-hour session with a 1-day lockout while taking peak-hour usage into account. We have the stability now, let’s bring back the utility so we don’t have to keep jumping back to AI Studio just to get work done.