The Google AI Pro subscription, which encompasses a diverse suite of AI-powered development tools—including AI Studio, Antigravity, Stitch, Jules, and Colab—initially launched with immense promise. However, over the past few months, both its efficiency and cost-effectiveness have severely regressed.
The platform’s reliability is continuously deteriorating as access to fundamentally functional models feels increasingly restricted. A major culprit is the hidden system parameter frequently injected into user prompts behind the scenes, artificially throttling the model to an invisible EFFORT LEVEL: 0.50 (“Medium Thinking”).
This silent downgrade consistently leads to hallucinated outputs and autonomous agents getting stuck in infinite processing loops until usage limits are exceeded. It essentially wastes paid AI credits and resources on an unproductive, dejecting experience. At this stage, even straightforward natural language instructions for quick fixes—or highly structured, bite-sized prompts—result in the same degraded output. I expected a significantly higher standard from a “Pro” subscription.