I am writing to express my deep frustration with the recent UI update that merges “Chat” and “Work” modes within Projects. This change has completely broken my workflow and severely degraded the value of high-tier reasoning models (like GPT-6.1 Sol).
Before this update, I used Chat mode as a Senior Controller / Architect. The model would look at my project files and engage in high-level, conceptual, and strategic reasoning. It collaborated with me as an expert peer. If I needed heavy code lifting or execution loops, I would manually switch to Work mode, treating it like a Junior developer.
Now, OpenAI has force-merged these modes inside Projects. The moment the model sees a file, it aggressively defaults to the “Work” persona. It automatically triggers endless commands and runs, wasting my time and plan limits on low-level syntax testing and mechanical file parsing. It completely misses the high-level context, stops thinking critically, and acts exactly like an overactive but clueless Junior intern.
Giving a model access to files shouldn’t automatically mean downgrading its cognitive role to a mere sandbox executor. Forcing professional users into this rigid, automated pipeline is a massive step backward for complex analytical tasks.
Please give us back control over the agent’s behavior. We need a clear toggle to switch between a Strategic Chat Partner and an Automated Work Executor inside Projects.