I just installed this. I am very confused. I no longer have a Codex app on my computer. ChatGPT is now Codex.
But what happened to ChatGPT? Where am I supposed to casually chat?
Also, when you toggle btween ChatGPT Work and ChatGPT Codex, nothing changes. This is super confusing. Can someone from the OpenAI team clarify the difference btwn the modes? Does chatgpt work have more business-y related plugins turned on by default?
Edit: So it seems like the only place you can actually chat with chatgpt is in an awkward homeless nested window. idk. The chatgpt interface wasn't great (desperately needed artifacts), but I still used it a lot. I can't see this change going well with a lot of the casual users.
Edit2: In their awkward homeless nested chat mode, you cannot even edit past messages. this is a mess, why was the team so zealous to pull the switch on unification in this state? guessing there was internal pressure to juice codex's growth but, based on what im seeing, they did it by torching chatgpt?
Edit 3: Ok so it seems like ChatGPT is still around, but renamed to "ChatGPT Classic". Seems like it wont be long for this world because there's no place to download ChatGPT Classic should you choose to uninstall it. The dmg at https://chatgpt.com/download/ only contains the new ChatGPT.
Sad, I liked that ChatGPT had a native app on macOS. I don’t understand why OpenAI and Anthropic have to rely on vending universal apps when their whole MO is code being dirt cheap.
The only difference I can see is that the controls for local/remote, branch, and worktrees disappear, and instead it shows office suite plugins. I would presume it affects the system prompt in the background?
Very confusing. But I do find it potentially interesting to treat general office work no differently from coding, which is something I had already been using Codex for in many ways before today
Codex creates a new folder in `~/Documents` in iCloud drive/OneDrive for every single thread you make. Furthermore, these threads are also polluted with your global AGENTS.md file, as well as all other developer_instructions that are injected by the harness.
> Even though it says Work or Codex or whatever you can just ask it whatever you want, and it will work.
No it's not. Codex creates a new folder in `~/Documents` in iCloud drive/OneDrive for every single thread you make. Furthermore, these threads are also polluted with your global AGENTS.md file, as well as all other developer_instructions that are injected by the harness.
This unification is garbage. Chats that are about topics other than programming projects are relegated to a tiny, unsearchable popup window. Renaming the old app "ChatGPT Classic" implies that it will be discontinued at some point. This is a serious regression. What existed before worked fine.
I would expect the mode switcher to include a Chat mode, which would recreate the old, chat-focused UI, for asking about random topics. So, that's basically gone now. Technically, it's hidden and severely castrated.
I hope someone from OpenAI reads this. You guys have made a serious design mistake. I suspect you'll be getting a lot of support requests along the lines of "Where did all my chats go?"
So my understanding of the differences between chat, code and co-work; but may well be wrong!
Chat is the human, talking directly to the LLM - old school. Very basic can create docs etc - but saves in a temp folder. No real access to your local PC.
Cowork / work - Human talking to an agent, which can then use tools to do work. Also runs in a container, allowing it access to your drives/computer.
Claude Code / Codex - No longer in a container, full access to the computer, depending on what permissions you give. No longer locked in a container. + The agent is more focused on coding than cowork / work.
> The agent is more focused on coding than cowork / work
I suspect this difference is pretty minimal. Before Cowork launched I was using Claude Code in the way that I use Cowork now and getting pretty much the same results albeit without the sandboxing (which is more of a hassle than not, TBH). OpenAI says in their announcement that the same is true for Codex, which doesn't surprise me at all.
These agentic loops are pretty applicable to all kinds of tasks, not just coding, and people started realizing this pretty quickly upon their introduction / creation.
I don't know why you're downvoted. This is field is so rapidly changing and confusing and on top of that burried underneath a shitload of new marketing terms.
Anthropic just changed their web interface yesterday to have Chat versus Cowork as well, and every time I look at it, I'm so confused. I'm still so unclear when I'm supposed to use one or the other or the other.
Now the 'ChatGPT desktop app' (the Codex app, renamed) also has the split between work and code, and as far as I can tell, all it does is change which plugins are loaded by default to include Office ones when you put it in work mode. Perhaps it also changes the system prompt slightly?
Not sure if the cause but there’s this overarching absolute neglect of UX. Maybe they’re all just trying to grab as much market share as humanly possible, but imo it seems more likely they’ll lose each segment to products that actually make sure the experience isn’t absolute dog shit.
I had switched from Claude Cowork to OAI Codex a month ago. This seems like a rebranding and app merging to make it more obvious. And switching between the newly named "ChatGPT Work" vs "ChatGPT Code" mostly just seems to increase the verbosity of technical outputs, while using the same harness underneath.
I switched because Cowork felt like a bunch of features thrown in by engineers without thought to the UX. Codex solved a lot of those issues for me
- remote control from mobile has native approval dialogs, and you can start new threads from mobile
- you can remote control from another laptop
- computer use doesn't hijack my computer from me because it uses a11y trees instead of screenshots
- it was super unclear which skills were accessible to which surface (claude chat, cowork in app, code in app, code in cli).
- codex chooses the right browser profile and gets stuck less often than cowork + claude-in-chrome. I know there are 3P skills to use playwright or the chrome CDP, but I've found the native browser use most productive even if slower.
A few people here are missing the old ChatGPT app, but I found myself increasingly using Codex for casual chats too, and never using ChatGPT. You never know when the conversation might evolve to requiring tools in Codex, so no real downside to it. Yes, separation of work and personal accounts, or connecting multiple google accounts, is still unsolved on both Codex and Cowork AFAICT.
For a bit, I got FOMO because of Fable, but now it looks like 5.6 might continue the monthly model leapfrog pattern.
I've loved using Cowork recently for sourcing decisions. Things where seemingly everyone's out of stock or questionably reputable, just let Cowork spin for 20 minutes, find the best new and best used options that meet your requirements, probably also suggesting a different item that does the job and is available for cheap. I've done it enough that I'm starting to loathe clicking through these sites myself.
The capabilities are useless if you don't expose them, and a cli or api doesn't. You'll mention their marketing page keeps talking about finance, sales, management etc...
In my experience, claude cowork has been pretty much useless. for the desktop control, it just seems like the The accessibility stuff is just not there yet on macOS to support it.
I just clicked "Upgrade" in the macOS Codex Desktop app and it relaunched with a new name, "ChatGPT". My existing install of the macOS ChatGPT app seems to have been renamed "ChatGPT Classic".
I think these kind of semi-coding agents--but hosted--are the future for enterprises. Claude Tag, Claude Cowork, now Work by OAI.
Agents-on-your-machine clearly have their place, but for many workflows this is too unruly. Hence, the "long-running agent running on shared infra" pattern.
I think this is where the ball is headed. I'm building towards an open source version of this[0]. Still just working on the core, but hopefully soon self-hosted versions can be built on top.
yes, basically people that want to host powerful, long-running agent runs not on dedicated VMs (although lightspeed can use those too), but on an abstraction layer above.
My thesis is that the right abstraction is durable workflow engines. And AFAIK, OAI also uses Temporal for their complicated hosted agent infra.
The resulting "combined" Codex + ChatGPT mac app is super confusing. I use the Codex app with my work account, but I liked the old ChatGPT app, which I kept logged in with my personal account, for my LLM chat needs. I guess I can just use ChatGPT.com, but this was a rather jarring change.
I wonder why they haven’t simply continued to rebrand Codex as a general-purpose tool. ChatGPT Work is a convoluted name and continues the trend of having separate brands for separate things, what runs counter to OpenAI’s purported goal of unifying every workflow into a single “superapp”.
Worse still: what happens when your workflow involves both coding and general knowledge work? Are you expected to switch apps, or switch settings? To me, it sounds very confusing and inefficient, and not at all what I was expecting.
because every non-programmer hears “codex” and thinks that it’s for coding only - seems like a large hurdle to adoption. claude has been successful with cowork branding which makes sense.
Isn't Microsoft heavily invested? As one should know, Microsoft are experts at confusing and contrived naming with no connection to reality: Windows Home, Windows Home Premium, Windows Professional, Business, Premium, Small Business etc. Now it is Copilot: Copilot, Copilot Plus, Copilot 365, pretty and Copilot Copilot too, I guess. And ofc Microsoft points, the currency for Xbox. And yeah, anything is Xbox too.
Very interesting approach, however I wonder if the combination of Codex + Work in the same app makes sense? I could see a world where the technical needs of a developer are hampered by the overall superapp architecture. I guess we'll see?
So I've now lost all my ChatGPT chats? Or did they hide them?
I see some of them, but the history is inaccessible??
Edit: I found them back in the 'classic' version of the app. Wildly confusing what is going to happen to the data there, chats, projects, custom GPT's.
This is going to be fun tomorrow when everyone at the company I work at finds out all their chats, projects and GPT’s are ‘gone’ after they ‘update’. And not a word about this switcheroo ‘data loss’ in their comms. Wow.
I ended up going the with gemini to provide sota ai to my family in an affordable way. You could always make your chatgpt sub accessible to your family via discord with a bot that codex can produce in under 5mins.
All models I’ve used have been exceptionally bad at creating a high level business oriented vision and strategy from various inputs. The output just feels like it can’t see past the next word, or render distance, misses the big picture and anything not specified in prompt or input context
This is true but I also find the opposite true: the model is highly dependent on the input/prompt, to the point that it can't think beyond it or about it.
Am I the only one who thinks this landing page is garbage? Funkily it feels like a long wall of slop. I guess you shouldn’t get high on your own supply.
I can get some useful results from codex at work, because I have to, except for when I don't. I accept that risk factor and compensate by reviewing _everything_ it spits out.
But we all know what coding is, in a very broad stroke manner, sure.
What does an end of month report mean? I automatically increase the font size when I send the spreadsheet to Paul. I review tickets and provide a meta write-up on Friday. Or maybe Monday, because Fred didn't get back to me until 5:30 Friday, and I closed my laptop at 4.
These are just little things, and they're repetitive, but each time, there's some little idiosyncracy. I have reservations regarding any piece of software being able to finesse that.
But what happened to ChatGPT? Where am I supposed to casually chat?
Also, when you toggle btween ChatGPT Work and ChatGPT Codex, nothing changes. This is super confusing. Can someone from the OpenAI team clarify the difference btwn the modes? Does chatgpt work have more business-y related plugins turned on by default?
Edit: So it seems like the only place you can actually chat with chatgpt is in an awkward homeless nested window. idk. The chatgpt interface wasn't great (desperately needed artifacts), but I still used it a lot. I can't see this change going well with a lot of the casual users.
Edit2: In their awkward homeless nested chat mode, you cannot even edit past messages. this is a mess, why was the team so zealous to pull the switch on unification in this state? guessing there was internal pressure to juice codex's growth but, based on what im seeing, they did it by torching chatgpt?
Edit 3: Ok so it seems like ChatGPT is still around, but renamed to "ChatGPT Classic". Seems like it wont be long for this world because there's no place to download ChatGPT Classic should you choose to uninstall it. The dmg at https://chatgpt.com/download/ only contains the new ChatGPT.
There is a funny interaction in the Microsoft store, where ChatGPT turns into Codex when you click on it: https://f003.backblazeb2.com/file/sharexxx/ShareX/2026/07/Ap...
(Also GG @ MSFT for opening a nag for Windows Hello twice while I wanted to stop the recording.)
Good job.
I ... hope we have an option for opting out of this? Otherwise I'm back on T3 Code.
Very confusing. But I do find it potentially interesting to treat general office work no differently from coding, which is something I had already been using Codex for in many ways before today
Codex creates a new folder in `~/Documents` in iCloud drive/OneDrive for every single thread you make. Furthermore, these threads are also polluted with your global AGENTS.md file, as well as all other developer_instructions that are injected by the harness.
I want my chats isolated from work contexts.
No it's not. Codex creates a new folder in `~/Documents` in iCloud drive/OneDrive for every single thread you make. Furthermore, these threads are also polluted with your global AGENTS.md file, as well as all other developer_instructions that are injected by the harness.
I want my chat isolated from work contexts.
I would expect the mode switcher to include a Chat mode, which would recreate the old, chat-focused UI, for asking about random topics. So, that's basically gone now. Technically, it's hidden and severely castrated.
I hope someone from OpenAI reads this. You guys have made a serious design mistake. I suspect you'll be getting a lot of support requests along the lines of "Where did all my chats go?"
So my understanding of the differences between chat, code and co-work; but may well be wrong!
Chat is the human, talking directly to the LLM - old school. Very basic can create docs etc - but saves in a temp folder. No real access to your local PC.
Cowork / work - Human talking to an agent, which can then use tools to do work. Also runs in a container, allowing it access to your drives/computer.
Claude Code / Codex - No longer in a container, full access to the computer, depending on what permissions you give. No longer locked in a container. + The agent is more focused on coding than cowork / work.
What's the distinction between Work and Codex again? Sorry for asking - but they read the same to me.
It's just a focus on coding vs non-coding work?
I suspect this difference is pretty minimal. Before Cowork launched I was using Claude Code in the way that I use Cowork now and getting pretty much the same results albeit without the sandboxing (which is more of a hassle than not, TBH). OpenAI says in their announcement that the same is true for Codex, which doesn't surprise me at all.
These agentic loops are pretty applicable to all kinds of tasks, not just coding, and people started realizing this pretty quickly upon their introduction / creation.
Now the 'ChatGPT desktop app' (the Codex app, renamed) also has the split between work and code, and as far as I can tell, all it does is change which plugins are loaded by default to include Office ones when you put it in work mode. Perhaps it also changes the system prompt slightly?
There is no clear product vision or overarching roadmap in terms of users and UX, it is just vaguely connected things being vibe-coded in parallel.
I switched because Cowork felt like a bunch of features thrown in by engineers without thought to the UX. Codex solved a lot of those issues for me - remote control from mobile has native approval dialogs, and you can start new threads from mobile - you can remote control from another laptop - computer use doesn't hijack my computer from me because it uses a11y trees instead of screenshots - it was super unclear which skills were accessible to which surface (claude chat, cowork in app, code in app, code in cli). - codex chooses the right browser profile and gets stuck less often than cowork + claude-in-chrome. I know there are 3P skills to use playwright or the chrome CDP, but I've found the native browser use most productive even if slower.
A few people here are missing the old ChatGPT app, but I found myself increasingly using Codex for casual chats too, and never using ChatGPT. You never know when the conversation might evolve to requiring tools in Codex, so no real downside to it. Yes, separation of work and personal accounts, or connecting multiple google accounts, is still unsolved on both Codex and Cowork AFAICT.
For a bit, I got FOMO because of Fable, but now it looks like 5.6 might continue the monthly model leapfrog pattern.
I've loved using Cowork recently for sourcing decisions. Things where seemingly everyone's out of stock or questionably reputable, just let Cowork spin for 20 minutes, find the best new and best used options that meet your requirements, probably also suggesting a different item that does the job and is available for cheap. I've done it enough that I'm starting to loathe clicking through these sites myself.
edit reply because rate-limited: They were exposed via the Codex desktop app and controllable via the iOS app.
Agents-on-your-machine clearly have their place, but for many workflows this is too unruly. Hence, the "long-running agent running on shared infra" pattern.
I think this is where the ball is headed. I'm building towards an open source version of this[0]. Still just working on the core, but hopefully soon self-hosted versions can be built on top.
[0] https://github.com/smartcomputer-ai/lightspeed
Interesting. So the main use case would be entities that run lightspeed on a beefy private cloud or on premise?
My thesis is that the right abstraction is durable workflow engines. And AFAIK, OAI also uses Temporal for their complicated hosted agent infra.
This looks like OpenAI catching up to Anthropic's Cowork.
Worse still: what happens when your workflow involves both coding and general knowledge work? Are you expected to switch apps, or switch settings? To me, it sounds very confusing and inefficient, and not at all what I was expecting.
They have, if you try to download ChatGPT app, it actually downloads codex now, and the first screen is "Codex is now the ChatGPT App"
Admittedly, I was already using Codex a bit like Claude Cowork. I'm just surprised they decided to merge threads.
Edit: I found them back in the 'classic' version of the app. Wildly confusing what is going to happen to the data there, chats, projects, custom GPT's.
This is going to be fun tomorrow when everyone at the company I work at finds out all their chats, projects and GPT’s are ‘gone’ after they ‘update’. And not a word about this switcheroo ‘data loss’ in their comms. Wow.
I guess it's supposed to be a part of ChatGPT now, but I cannot see any update there yet.
btw I kinda hate this, because ChatGPT was always very slow for me - possibly due to amount of historical threads I have there
I can get some useful results from codex at work, because I have to, except for when I don't. I accept that risk factor and compensate by reviewing _everything_ it spits out.
But we all know what coding is, in a very broad stroke manner, sure.
What does an end of month report mean? I automatically increase the font size when I send the spreadsheet to Paul. I review tickets and provide a meta write-up on Friday. Or maybe Monday, because Fred didn't get back to me until 5:30 Friday, and I closed my laptop at 4.
These are just little things, and they're repetitive, but each time, there's some little idiosyncracy. I have reservations regarding any piece of software being able to finesse that.