Repository navigation
Persistently display whether each conversation uses ChatGPT Work or Codex #33079
Description
What variant of Codex are you using?
App
What feature would you like to see?
I frequently use ChatGPT Work and Codex within the same project. Some conversations involve research, analysis, documents, or other deliverables and belong in Work, while others involve repository changes, implementation, testing, or debugging and belong in Codex.
The current interface makes this confusing because Work or Codex is selected globally rather than being clearly identified at the conversation level. When I return to an older conversation, I may not remember which environment I used to start it. If the global selector is currently set to the other environment, I can unintentionally continue the conversation using the wrong agent experience.
My current workaround is to manually prefix every conversation title with “[WORK]” or “[CODEX]”. This works, but it places the responsibility on the user, is easy to forget, and duplicates information that the application should already know.
Suggested improvement:
- Store and visibly display the originating environment for every conversation or task.
- Add a persistent “Work” or “Codex” badge in the sidebar and conversation header.
- When opening a conversation, automatically restore its originating environment or offer a clear prompt such as: “This conversation was created in Work. Switch to Work?”
- If switching an existing conversation between Work and Codex is supported, make that an explicit conversation-level action and show which environment is currently active.
- Optionally allow users to filter project conversations by Work, Codex, or Chat.
The conversation does not necessarily need to be permanently locked to one environment. The important requirement is that its originating and currently active environment remain clearly visible.
This would be especially valuable in projects that intentionally contain a mixture of research, writing, data analysis, and software-development tasks. It would reduce accidental use of the wrong tools, eliminate manual labeling, and make the relationship among projects, conversations, Work, and Codex much easier to understand.
Thanks.
Andres Castaneda
Additional information
Platform: ChatGPT desktop app
Area: Projects / conversation navigation / Work–Codex selector
Activity
calibrakem commentedon Sep 4, 2026
I am encountering a broader identification problem among the three conversation experiences in the ChatGPT desktop app:
I tried to create a ChatGPT Work conversation inside a ChatGPT project. The assistant gave me specific instructions: select Work, send a typed message first, verify Work, and then start Voice in that same conversation. I followed those directions exactly. When I returned, the assistant inspected an internal conversation registry, treated its Codex classification as authoritative, and repeatedly told me I was in a Codex conversation.
When I challenged that conclusion because I had explicitly selected Work, the assistant became inconsistent. It speculated that I might have a Work conversation with a separate Codex backend, then retracted that explanation because it was unverified.
The problem is therefore not merely that the user cannot see the conversation type. The assistant also could not reliably identify it and gave incorrect or contradictory answers as fact. We still cannot determine whether this is truly a Codex conversation, a ChatGPT Work conversation misclassified internally, or whether Work and Codex share an underlying representation that the interface and assistant do not explain.
This creates a practical problem inside projects: users cannot know which conversation to reopen for ordinary discussion, general Work, or coding work.
Please add persistent Chat, Work, or Codex labels to the sidebar and conversation header; expose a reliable conversation-type field to the assistant; ensure that internal classification agrees with the mode the user selected; and document clearly whether ChatGPT Work and Codex are distinct conversation types or different interfaces over shared underlying tasks.
NeoLeo928 commentedon Oct 12, 2026
I would like to emphasize that this requirement should include ordinary Chat versus ChatGPT Work, not only Work versus Codex.
Within a ChatGPT Project, ordinary Chat conversations and Work tasks can appear in the same list without a sufficiently persistent visual distinction. Users should be able to identify the mode of every item without opening it or manually modifying its title.
Requested behavior:
ChatorWorkbadge or icon on every conversation row.Acceptance criterion:
A user looking at any conversation list can immediately determine whether each item is an ordinary Chat conversation or a Work task, without opening it, hovering over it, or inferring its type from the title.
This visual identification should work together with Project-level grouping and sorting by mode.
Related Project organization request: #48910.