I am a Google AI Pro subscriber and I used Antigravity actively before the 2.0 update for real development work, not just small toy prompts.
Before Antigravity 2.0, Gemini Flash was a practical daily-driver model for my workflow. I could work on a real local project: edit files, run terminal commands, test code, rebuild an EXE, inspect logs, and continue iterating. I rarely hit the 5-hour limit, and when I did, waiting for the next refresh was acceptable.
After Antigravity 2.0 and Gemini 3.5 Flash, the experience changed dramatically.
Today, after one normal development session, Gemini models moved to a weekly cooldown of around 6–7 days. This was not some huge enterprise workload. It was normal agentic development: fixing code, running checks, rebuilding an executable, and testing a local automation workflow.
This makes the Google AI Pro plan feel almost unusable for real development.
The core value of Antigravity is that the agent can work with files, terminal, browser, logs, tests, and builds. If users have to ration every step because one normal session can burn the weekly quota, the product loses its main advantage.
I do not need the smartest model for every small task. I need a reliable low-cost daily-driver model that can be used for normal development without hitting a 6–7 day lockout.
Please consider one of these fixes:
- Bring back Gemini 3 Flash as a low-cost daily-driver model.
- Give Gemini 3.5 Flash Medium a separate generous 5-hour quota pool.
- Increase the Google AI Pro weekly quota substantially.
- Make quota usage transparent before users unexpectedly hit a multi-day lockout.
- Clearly separate “short-term 5-hour quota” from “weekly quota” in the UI.
Right now the UI and messaging make it feel like the Pro plan was upgraded, but in practice the experience feels much worse than before for active users.
If this is intentional, then Google AI Pro is no longer a practical plan for real Antigravity development workflows.