Feature Request: Separate Conversation from Execution in the New ChatGPT Desktop App

ChatGPT Feature requests
1

I like the direction of unifying ChatGPT, Work, and Codex into a single desktop application. However, I think the current navigation conflates two fundamentally different activities:

  • Execution (Work, Codex, Computer Use, etc.)

  • Conversation (asking questions, planning, reviewing, steering work)

I suggest restructuring the desktop UI around three persistent regions.

Proposed Layout

Left Sidebar

  • Projects

  • GPTs

  • History

  • Workspaces

  • Files

Center Workspace

  • Work

  • Codex

  • Browser / Computer Use

  • Artifacts

  • Live task execution

Right Sidebar (Persistent Chat)

  • Always available, regardless of the active workspace

  • Can be collapsed if desired

  • Uses the selected Project as its context

  • Used to ask questions, change direction, explain results, and review ongoing work

The key idea is that Chat should always be available while Work or Codex are executing.

Why this is better

1. Better mental model

Execution and conversation are different activities.

The center pane performs work.

The right pane helps the user think about the work.

Users no longer need to switch modes simply to ask a question.

2. Projects become true knowledge contexts

Projects contain:

  • conversations

  • uploaded files

  • instructions

  • long-term context

Those should be selected from the persistent chat panel, not hidden behind search.

Today, Project conversations are searchable (Ctrl+K), which suggests the underlying data is already available. The missing piece is exposing the Project hierarchy in the desktop UI.

3. More scalable architecture

As new capabilities are added—Work, Codex, Deep Research, Browser, Computer Use, Images, Voice, etc.—they can all share the same interface:

  • Left = organization

  • Center = execution

  • Right = reasoning

The chat becomes the orchestrator instead of another mode.

4. Better multitasking

While Codex is generating code or Work is running a long task, users should be able to ask questions like:

  • “Why did you choose this implementation?”

  • “Compare this with our previous design.”

  • “What assumptions are you making?”

  • “Summarize progress.”

without leaving the execution view.

Additional suggestion

Allow the persistent chat to choose its context independently.

For example:

  • Project: Opt42

  • Workspace: Codex

  • Model: GPT-5.5

This would make it clear which knowledge base the assistant is using while allowing the execution workspace to remain unchanged.

Overall, I believe separating organization, execution, and conversation would make the desktop application significantly easier to understand, especially for technical and professional users who spend hours working inside it.

2

Hey metahex, I agree that would be a valuable addition. Thanks for taking the time to share your idea and explain the use case.

I'll pass this feedback along to the product team so they can review the request and decide whether it's something they'd like to implement.

Avinash