What Happened to Codex? GPT-5.6, ChatGPT Work, and Buda
Codex is not gone. OpenAI is merging the Codex app into the new ChatGPT desktop app. Here is what GPT-5.6 and ChatGPT Work mean for AI agent workflows.
If you opened your machine looking for the Codex app and wondered whether Codex disappeared, the short answer is: Codex is not gone.
OpenAI says the Codex app is merging with the new ChatGPT desktop app. In the new ChatGPT desktop experience, Chat, Work, and Codex sit inside one work surface. GPT-5.6 provides the new model layer; ChatGPT Work provides the new work entry point; Codex remains the execution layer for code, files, terminals, and engineering workflows.
For Buda, the important point is not a product-name change. It is the direction of work: AI is moving from chat answers to reviewed execution.
Is Codex Gone?
No. Codex is being folded into the broader ChatGPT desktop workflow.
That matters because many people understood Codex as a separate app: open Codex, connect a repo, ask it to change code, run commands, and prepare a pull request. That workflow still matters. What changed is the surrounding workspace.
The new pattern is closer to this:
| Search intent | Practical answer |
|---|---|
| Where did Codex go? | It is moving into the new ChatGPT desktop app experience. |
| Is Codex discontinued? | No. OpenAI is repositioning it as part of ChatGPT Work and desktop execution. |
| Does GPT-5.6 work with Codex? | GPT-5.6 is positioned across ChatGPT, Codex, and the API. |
| What should teams watch? | How coding execution, business context, files, and review come together in one workspace. |
The interesting shift is that Codex is no longer only a place to write code. It becomes one execution mode inside a larger AI workbench.
GPT-5.6, Codex, and ChatGPT Work
GPT-5.6 arrived as a family: Sol, Terra, and Luna. That is useful because real agent work needs routing. Some steps need strong reasoning. Some need balanced everyday work. Some need low-cost repeated execution.
ChatGPT Work points to the other side of the same shift. OpenAI's examples are not framed as casual chat. They are framed as finished work: prepare a slide, analyze business data, read connected context, build an artifact, and use desktop apps and local files.
Put together, the message is simple: models are becoming workers inside workspaces.
Buda Turns This Shift Into an Agent Workspace
Buda turns this shift into a managed Agent Workspace: a place where teams can give agents context, tools, files, and review paths instead of scattering AI work across separate chats and apps.
A real AI agent workflow needs more than a strong model. It needs:
- shared context instead of scattered prompts;
- files and memory the agent can actually use;
- tools that execute rather than only suggest;
- model routing for cost, risk, and judgment;
- human review before important actions leave the workspace;
- a record of what happened and why.
That is why Buda is organized around Spaces, Agents, Drive, tools, sessions, channels, and reviewable outputs. The question is not only which model answers best. The question is where the work happens, who reviews it, and how execution is controlled.
From Coding Agent to Work Agent
Codex started in many users' minds as a coding agent. The new ChatGPT Work direction makes it clearer that coding is only one part of agentic work.
A coding agent may edit files and open a pull request. A work agent may also prepare a customer report, summarize a data room, update a spreadsheet, draft a sales follow-up, or turn research into a page a human can review.
This is where Buda's human-in-the-loop pattern matters. AI can move faster, but the human still owns direction, taste, approval, and responsibility.
FAQ: Codex, GPT-5.6, and ChatGPT Work
What happened to the Codex app?
OpenAI says the Codex app is merging with the new ChatGPT desktop app. Codex is not disappearing; it is becoming part of the broader ChatGPT Work and desktop workflow.
Why can't I find Codex as a separate app?
If you are looking for the old standalone Codex entry point, the new direction is to look inside the ChatGPT desktop experience where Chat, Work, and Codex are grouped together.
Is Codex still useful for coding agents?
Yes. Codex still matters for code, repositories, files, terminal workflows, and pull-request style execution. The difference is that it now sits closer to business context and finished-work workflows.
How does GPT-5.6 relate to Codex?
GPT-5.6 is the new model layer. Codex is one execution surface where that model capability can be used for coding and agent workflows.
How does Buda help teams use this pattern?
Buda gives teams a managed AI Agent Workspace where agents can work with files, tools, memory, channels, model routing, and human review inside one controlled workspace.
Start from the Buda dashboard, or read more about the Buda Agent Workspace.