After using ChatGPT professionally for years, I never thought I would write this.
My problem is no longer just bugs. Bugs may get fixed. My concern is the direction the product is taking.
The new macOS app feels like ChatGPT without the Chat.
The chat—the very concept that made ChatGPT successful—has been pushed into the background. Finding previous conversations has become more difficult. Resuming work exactly where you left off weeks ago is no longer the natural workflow it once was.
And that, to me, is the real regression.
For years, ChatGPT was built around conversations. I could structure my work in chats, return to them days or weeks later, continue thinking, refine documents, revisit decisions and preserve the context of an entire project.
That conversational workflow was incredibly powerful. Now I increasingly get the impression that everything is expected to become an agentic workflow.
Projects.
Work.
Agents.
Tasks.
Those are useful additions. But they should be additions, not replacements.
Sometimes I don’t want an agent.
Sometimes I don’t want to create a project.
Sometimes I simply want to open a conversation, think through a problem, and come back to it next month.
That’s what ChatGPT was about. Ironically, the one thing that now seems to receive the least attention is the “Chat” itself.
OpenAI appears to be racing toward becoming an AI operating system. I understand the vision, and in many ways I even like it. But in doing so, I fear they are neglecting the workflow that attracted millions of users in the first place.
Not every user is a software engineer.
Not every task needs an agent.
Not every thought belongs in a structured workflow.
Sometimes a conversation is simply the best interface.
If OpenAI wants to build the future of AI work, that’s great.
But please don’t forget why so many of us came here in the first place. It wasn’t because of agents. It was because of the chat.