Something quietly alarming happened this week in the developer community. Paid Google AI subscribers opened Antigravity, Google's agentic coding IDE, ran a few sessions after IO 2026, and suddenly hit a wall. Weekly Gemini quotas were gone. Not running low. Gone. The culprit: a new compute-based limit system that factors in prompt complexity, features used, and session length rather than a simple daily prompt count.
The backlash was fast and loud. Google responded by tripling the Gemini rate limits across all paid Antigravity tiers and resetting weekly quotas. Then, when that still wasn't enough, they did it again. Two resets, one week, a 3x permanent boost that even now leaves limits lower than what developers had before the change.
Is compute-based metering the right model for agentic coding tools that spawn parallel agents and run background tasks for hours at a time, or is this a pricing architecture still catching up to how developers actually work?
Find out everything at pixelauratech.com
Top comments (0)