It looks like ChatGPT finally has their own form of artifacts, which is quite welcome. I can't believe it's taken them two years to copy Anthropic on this. Artifacts were a huge differentiator btwn claude and chatgpt. I wonder why it took them so long?
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.
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.
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 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.
I think they're grappling with the fact that their positioning needs to be as the operating system but they don't have a leg in currently and destabilising the existing OS players is a monumental task. So they're stuck with the super app direction that just doesn't make sense for how the future of AI UX is envisioned.
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
> 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 folder in `~/Documents` with every one of these threads. 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 don't want my chats that are isolated from work contexts.
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.
Because there is more competition to render sites in a pleasing way, switching cost is lower. The operating systems have a high switching cost, very few people will learn a new OS because the text is a little fuzzy.
I look forward to John Gruber’s screed on this given his recent rampages against Electron apps, and his preference for OpenAIs native macOS app (relative to Claude).
Seems the old ChatGPT is still on computer but renamed ChatGPT Classic (no longer downloadable), so clearly being deprecated. But just don't understand how the whole CHATGPT Chat History Chat feature is essentially gone other than seeing at the past 3 or 4 in a tiny popup window. Glad everyone is as confused as I am, so hopefully an announcement from ChatGPT clarifying all this.
It really is a mess. An hour or so ago, the download for macOS was called ChatGPT.dmg. However the app in there was still called Codex. But it installed ChatGPT.
But it seems to be fixed now.
Also, subtitling Work with "For getting work done" and Codex with "For developers" is a bit aggressive towards developers ;)
>Also, when you toggle btween ChatGPT Work and ChatGPT Codex, nothing changes. This is super confusing.
This is the weirdest thing. It's completely unclear even if I was using the app for "work" rather than "coding" why I wouldn't just use the "Codex" option. Why even bother toggling? Can Codex actually not do any of the things in "Work"?
Lol what a shit show. It was fine the way it was. Why'd you break it, OpenAI? What product manager talked you into thinking this was a good idea? Don't they have access to ChatGPT to talk them out of it? Of course, I'm a random on the Internet who doesn't get it, but what a dumb move. ASI can't come fast enough.
So will the new ChatGPT (Codex) refuse to respond at all after the 5-hour quota or weekly quota is used up? It always felt that 5.5 on the website was a lot more generous than the Codex version. And at least the website/old app could downgrade you to a lower model if you used up that quota.
1) If you have the Codex desktop app installed, and you update from within the app, it replaces itself with the new chat-demoted-to-an-overlay app called "ChatGPT" and renames your existing ChatGPT app to "ChatGPT Classic."
2) If you get the new ChatGPT app and replace your current app with it, you don't get a "Classic" app and there's no way to download it.
3) Naming your app Classic is putting a bright red sign on it that says "this app will be abandoned soon." They might as well have called it "ChatGPT Sunset."
If this is all true, that is certainly...one way to do product management.
The best way to access ChatGPT "chat" was always on ChatGPT.com via a web browser, not even the iOS app seems to have all the little things you can do on the web version.
The Mac desktop ChatGPT app had weird issues like not being able to see the model being used by a chat inside a project etc. I just ended up installing ChatGPT Atlas and using that as an AI-only browser for all LLMs including Claude and Grok lol
I feel like the web is under attack. They want chatgpt in desktop app so they can start moving people away from chatgpt via web. Web being suspectible to automation is driving everything into locked down apps. The very toxic reddit.com a prime example of this -- forbidding mobile web access behind a timed wall.
Yes, the side effect has been me using Reddit via Incognito, which means I don’t see it every time I open iOS Chrome, and also I’m not signed in, so after 14 years I’ve basically stopped using it.
I guess with their IPO “the line must go up”. Very stupid, dark pattern.
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.
They have this, it's essentially ChatGPT Business. And if you use an Amex Plat card, it's basically free (plenty of ChatGPT 50% off links going around, or at least there were a few weeks ago).
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?
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.
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.
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 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.
I came to the exact same conclusion recently. I am building something like lightspeed that can run on “pay per second” infra like fly.io sprites because I wanted to have an open claw style agent but not have to deal with paying for another VM.
Nowadays, depending on the project’s complexity I either run just 1 sprite, several sprites, or in one case 1 permanent “CTO” VM (with a dedicated GH App identity) that specs work that a fleet of worker sprites pull and implement.
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.
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".
just commented ton Tibo's post; they should have made things clearer and told us if the macOS ChatGPT app was going away. So I could either delete it or not be surprised when it rolls into something else :(
161 comments
[ 2.9 ms ] story [ 91.8 ms ] threadThis looks like OpenAI catching up to Anthropic's Cowork.
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.
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.
That does sound like what would tend to happen if they used their own product exclusively.
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.
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
But what happened to ChatGPT? Where am I supposed to casually chat?
Also, ChatGPT Work seems to be the same thing as ChatGPT Codex
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
No it's not. Codex creates a folder in `~/Documents` with every one of these threads. 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 don't want my chats that are isolated from work contexts.
What functionality are you seeking from a native app “for chatting”?
Plus they have lots of embedded web content these days and managing webviews is a relative pain vs it all being one webview
I ... hope we have an option for opting out of this? Otherwise I'm back on T3 Code.
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.
what is badly broken about it, and what feature does it not support? im a pretty casual user, so i guess i havent noticed this bad breakage.
But it seems to be fixed now.
Also, subtitling Work with "For getting work done" and Codex with "For developers" is a bit aggressive towards developers ;)
This is the weirdest thing. It's completely unclear even if I was using the app for "work" rather than "coding" why I wouldn't just use the "Codex" option. Why even bother toggling? Can Codex actually not do any of the things in "Work"?
By the way, it seems ChatGPT is confused as well:
https://imgur.com/a/O1Aj6jz
ASI is a myth used by these companies to justify their pursuit of regulatory capture.
1) If you have the Codex desktop app installed, and you update from within the app, it replaces itself with the new chat-demoted-to-an-overlay app called "ChatGPT" and renames your existing ChatGPT app to "ChatGPT Classic."
2) If you get the new ChatGPT app and replace your current app with it, you don't get a "Classic" app and there's no way to download it.
3) Naming your app Classic is putting a bright red sign on it that says "this app will be abandoned soon." They might as well have called it "ChatGPT Sunset."
If this is all true, that is certainly...one way to do product management.
The Mac desktop ChatGPT app had weird issues like not being able to see the model being used by a chat inside a project etc. I just ended up installing ChatGPT Atlas and using that as an AI-only browser for all LLMs including Claude and Grok lol
I was just about to pull the trigger, and upgrade my sub, but maybe I'm better off, moving over to Claude, even though it means rebuilding my history.
the website for downloading the new chatGPT ( https://learn.chatgpt.com/docs/app ) has a big download button in the middle of the page that links to "https://persistent.oaistatic.com/codex-app-prod/Codex.dmg"
wtaf OpenAI... pick a freaking lane! what did you vibe code this ?
I guess with their IPO “the line must go up”. Very stupid, dark pattern.
https://help.openai.com/en/articles/12703475-retiring-group-...
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.
Admittedly, I was already using Codex a bit like Claude Cowork. I'm just surprised they decided to merge threads.
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?
Nowadays, depending on the project’s complexity I either run just 1 sprite, several sprites, or in one case 1 permanent “CTO” VM (with a dedicated GH App identity) that specs work that a fleet of worker sprites pull and implement.
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.
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.
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 understand what you mean by this but in the end you decide what's locked in a VM / container. Not OpenAI. Not Anthropic.