I am a paid subscriber, and over the last few days I have noticed a clear drop in the agent’s quality inside the course. I write very explicit instructions, yet the agent often returns wrong information, changes constraints I asked it not to change, and ignores steps that I specified. This has started to affect my ability to follow the course and complete the exercises.
Has anyone else seen the same behavior, and is this expected due to any recent changes. If this is intentional, could you clarify the reasoning. I want to be sure this is not a case of the model being limited in a way that nudges users toward a more expensive plan, since I am already on a paid plan.
If this is a bug or configuration issue, could someone share recommended troubleshooting steps. For example, which model or version is currently used inside the course, any known outages or regressions this week, best practices to reduce instruction following errors, and whether there are settings I should adjust.