I’m extremely disappointed with the quality of your system. I’m currently paying for the ultra service, yet I’m not seeing any meaningful improvement in the models. In particular, Gemini 3.1 Pro has been consistently underperforming as a coding assistant. It often introduces issues into code that was previously working correctly, even after being fixed by other tools like Codex.
This kind of inconsistency makes it difficult to rely on the platform. I would expect a higher level of reliability and performance, especially at a premium tier.
Also, the hallucinations, lies, and laziness of this AI agent is extremely bad.