Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

MacRumors

macrumors bot
Original poster


OpenAI has a Mac "superapp" in development that unifies its ChatGPT app, Codex coding platform, and Atlas browser, reports The Wall Street Journal ($).

openai-logo-moody.jpg

The idea behind the all-in-one app is to simplify the user experience, following the launch of several standalone products, some of which haven't resonated with OpenAI's customers. The company is also trying to bounce back after the recent successes of its main rival, Anthropic.

OpenAI executives are said to be looking at areas it can deprioritize while it focuses on creating agentic AI capabilities within the new superapp that can work autonomously on a user's computer to carry out various tasks like writing code and analyzing data.

In an all-hands meeting last week, OpenAI's chief of applications Fidji Simo reportedly told employees they couldn't afford to be distracted by "side quests" given Anthropic's rapid success winning over enterprise and coding customers. From the report:
An OpenAI spokeswoman said the new "superapp" will enable teams inside OpenAI to work more closely together, and help the research division focus its efforts around improving one central product. Over the coming months, the company expects to add new "agentic" capabilities within its Codex app so it can help with productivity-related tasks beyond coding before merging ChatGPT and the Atlas browser into the superapp as well.
OpenAI unveiled a series of major initiatives last year, like its Sora video app and the acquisition of Jony Ive's AI hardware venture. Since then, however, Anthropic has gained strong momentum with the success of its Code Claude and Cowork offerings.

The WSJ report gave no timeline for the launch of OpenAI's so-called superapp, but it said the company's mobile ChatGPT app will remain unchanged.

Article Link: OpenAI 'Superapp' to Merge ChatGPT, Codex, and Atlas Browser
 
The company is also trying to bounce back after the recent successes of its main rival, Anthropic.

Genuinely curious what makes Anthropic the ‘main rival’, when that whole ‘code red’ drama just a few months ago seemed to be triggered by Gemini 3.

I happen to use Claude for coding, and think it’s marginally better than Gemini and ChatGPT, but that’s just me—I can’t speak for others or the industry as a whole.
 
  • Haha
Reactions: UpsideDownEclair
So at the first sign of market adversity they go straight to the Microsoft model and cram all the stuff nobody wants into the one thing people do want.

The ChatGPT app is already getting bloated enough, and the browser is a privacy nightmare.

The novelty has worn off anyway and their recent changes after it went mainstream has made it less useful.

At this point it’s best to use it through the API.
 
  • Like
Reactions: sos47
Genuinely curious what makes Anthropic the ‘main rival’, when that whole ‘code red’ drama just a few months ago seemed to be triggered by Gemini 3.

I happen to use Claude for coding, and think it’s marginally better than Gemini and ChatGPT, but that’s just me—I can’t speak for others or the industry as a whole.

Anthropic has just had positive marketing/news/spin the last few weeks, combined with OpenAI having the opposite.

They’ve both had mainstream commercials lately for example and Anthropic’s seemed more effective in reaching an audience who was receptive vs one who already knew about it.

That and people have noticed an uptick in Claude quality and a decline in ChatGPT quality.

Google is kind of doing their own unassailable thing cartel style with Apple as per usual.
 
  • Like
Reactions: kalsta
Glad they are switching to Claude's approach of having ChatGPT and the coding agent UI in one app. I switched to Claude temporarily for that feature, since it was more approachable, and Codex wasn't as good. I have since switched to Codex because while people mention Claude being the best for that, when attempting a project on both, Codex was far more effective and much more generous allotments. Never hit the limits vs Claude where I constantly ever would every couple of hours. When looking at the details Codex was so much better, and somehow much smarter than standard ChatGPT whom constantly fails in many ways. I assume because Codex as much longer context window so you can keep taking and doesn't start failing the way ChatGPT does.

In regards to puting the website in the app, ChatGPT is already in the app, so assume really they just mean they are going to stop offering a website interface so they can focus just on the app. Same with the browser, I imagine it's not a full fledge browser coming to the main app but rather they are abandoning a dedicated browser since they quickly realized people are happy with Chrome and such, and a dedicated doesn't make sense. But yes, please add Codex to ChatGPT like Claude and Claude Code does.

And while I haven't hit limits yet, because Codex offers extra until April, I hope they offer a middle plan between 20 and 200 like Claude does. That is the main thing that I imagine will keep people from switching to Codex.
 
Genuinely curious what makes Anthropic the ‘main rival’, when that whole ‘code red’ drama just a few months ago seemed to be triggered by Gemini 3.

I happen to use Claude for coding, and think it’s marginally better than Gemini and ChatGPT, but that’s just me—I can’t speak for others or the industry as a whole.
The unified Claude Desktop app (albeit only recently released on Windows, but I’ve been using it on macOS forever now) - full MCP capabilities, CoWork local with agentic mode, Claude Code, and Chat/Project - is really the core gap I think OpenAI is trying to address, but to include the browser is a little extra IMO. There’s already chrome MCP plugins and Claude handles it pretty well so far.
 
My understanding is that Anthropic got a big boost in usage due to them declining to work with the US DoD(W)...

I don't think a superapp targets that reason. But it does make sense to bring capabilities together that are powered by the same underlying technology.

Once macOS has an API for on-screen awareness, wouldn't that allow any AI to "use" any app rather than duplicating efforts?
 
The unified Claude Desktop app (albeit only recently released on Windows, but I’ve been using it on macOS forever now) - full MCP capabilities, CoWork local with agentic mode, Claude Code, and Chat/Project - is really the core gap I think OpenAI is trying to address, but to include the browser is a little extra IMO. There’s already chrome MCP plugins and Claude handles it pretty well so far.
My own experience with Claude desktop is not perfect. I like the AI side of it but accessing it through the browser is more reliable in my case.
 
  • Haha
Reactions: UpsideDownEclair
So they're building their own version of Copilot?
They already have had Codex (like Claude Code), just Codex is separate app, and they are merging it into the main app, since presumably like you many don't know about it if its not the main app.

Copilot is Microsoft's AI assistant that uses ChatGPT as a backend, that they use to integrate into the software, so that the comparison of Copilot would be more like ChatGPT than Codex or Claude Code. But yes, Claude and ChatGPT already easily fully integrate in things such as VS Code, etc using your $20 plan rather than having use the API.
 
I could see the browser being useful if it’s built in with codex. But they’re falling behind Claude which is so good at programming, especially for scaffolding and prototyping rapidly. Or finding obscure bugs.
 
  • Haha
Reactions: UpsideDownEclair
While there will be more options for customers, I would prefer to use the simple app. In fact, I prefer to use the browser version most of the time.
 
  • Like
Reactions: mganu
Not sure the programmers are using it for the unified app experience. I use it from the terminal mostly as it’s the most frictionless implementation.
Xcode integration has promise but, as always with Xcode, too much bloat and constant crashes. Let’s see if they can make any progress at WWDC.

And I think the race is already over for developer trust and usage. I don’t know anyone in the industry who would pick OpenAI over Anthropocene for coding.
Possible just a successful marketing campaign. But from my experience there is a massive word of mouth force which is even getting ai doomers to give it a go (and subsequently get pleasantly surprised)
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.