Antigravity has been behaving really weird today. It keeps forgetting the context of the project and instructions. Even worse, it keeps generating code that is either bugged or logically untrue. Also, I noticed tighter limits on the quotas and usage, which was the only advantage over Claude Opus.
It’s been heading this direction for some time now. It seems they’re starting to ramp up tighter limits.
time to end my journey and switch to OPUS
In my case it does all the work in its “thoughts” and it always goes into a loop “I’ll go. I’ll run. I’ll act. I’ll execute. I’ll do” which is funny because it does not do anything and exhausts the model output token limit just by repeating this. I have tried everything even a new workspace, new project.
1 Like
I was wondering this issue is related to a change of the model. Maybe when the usage is high, the servers can not hold it, and the agent switches to a simpler model.
This might be happening also if the user has exceed a daily quota, even in paid accounts.