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

OpenAI Codex Update Adds Computer Use, Image Generation, and Memory on Mac

JMalone said:
Do not fall for this “computer use” trap.

It’s a way to make you give your job away completely and erode your privacy at the same time.

At some point it will be a third world worker using your computer and you will think it is “AI”.

No they will not compensate you for it. Free money doesn’t exist. Nothing is free. You will simply be discarded and pushed slowly out of existence.
Click to expand...

It’s the greatest Scam Altman trick of all time. YOU pay THEM while you do all the work training the AI models.
 
Ishimura said:
OpenAI don’t literally control your computer. An LLM does, and you choose where in your computer and how it takes actions.
Click to expand...
Yeah, I realize that. I was being hyperbolic. But the point is that with the LLM running on OpenAI hardware, and tokens constantly moving back and forth, it's anybody's guess as to exactly what can happen with the info. On a paid account they're not supposed to use it for anything, but hey... we're trusting a company that is absolutely desperate for revenue, which seems unwise.

I'm not totally against it, tbh. But it feels like a huge risk. For any agent that has autonomous control over my machine I'd be much more comfortable with it being local. On most hardware that's a huge limitation on LLM capabilities, but things are improving rapidly (look at Gemma 4 as an example).
 
The ChatGPT Mac app can't even set thinking effort to high or extra high, it's so sooo stupid that the PHONE can. Some of my research queries take an hour to run (with Pro).

I don't understand how they haven't fixed this yet. Codex is not a replacement for research work.

I'm glad the pressure is on though, hopefully Anthropic follows through and gets a native app instead of the really janky Electron one that loses data sometimes if you click around too fast. This is not a joke, sadly.
 
netwalker said:
What could possibly go wrong, especially coming from such a trustworthy company like Open AI.
Click to expand...
If you've ever needed to delete all your important emails, this will help.

Just tell it not to delete all your emails. Then it will delete all your emails.

You don't want to miss out on this amazing new technology. It's basically God now.
 
  • Like
Reactions: netwalker
buckwheet said:
On most hardware that's a huge limitation on LLM capabilities, but things are improving rapidly (look at Gemma 4 as an example).
Click to expand...
I’m still not sure what are the capabilities of Gemma 4 on a Mac with, let’s say, an M5 chip with 32GB of RAM. Some people say it’s more than enough, other people tell me that I’ll need, eventually, 80GB of RAM… so I’m a bit confused.
 
  • Like
Reactions: Luke MacWalker and buckwheet
Could AI fix "CoreData: Unable to create token NSXPCConnection. NSXPCStoreServerEndpointFactory 0x7f7b35209160 -newEndpoint returned nil" ??
 
Good to know about this but don't think I have a need for this. Think it will be quickly rolled out in other countries too.
 
  • Like
Reactions: mganu
gaggle said:
Sounds like parity with Claude features. Nice! Competition good etc etc.

MCP and its derivatives have been transformational in my life, both professionally and privately. There’s a future on the horizon of broad integrations of systems thats effectively not been able to talk together, and it’s very enticing (for those of us leaning into AI).
Click to expand...
I understand the individual words in this post, but in the particular order they are written, they make no sense at all.
 
Septercius said:
I understand the individual words in this post, but in the particular order they are written, they make no sense at all.
Click to expand...
The poster is likely someone who speaks another language as their main language, and not English as their first language. They're talking about how using AI improves communications between multiple devices.
 
Not using. Not going to support people and companies like that ever. It's not like they invented fire and they're the only ones who have it. I'm quite happy with Mistral and Lumo. I'm also quite happy not using AI 99% of my day.
 
Doru Oprisan said:
Not using. Not going to support people and companies like that ever. It's not like they invented fire and they're the only ones who have it. I'm quite happy with Mistral and Lumo. I'm also quite happy not using AI 99% of my day.
Click to expand...
Curious, how do you use Mistral? Locally? Lumo as far as I know is the online AI from Proton
 
delsoul said:
Again. Anything with Scam Altman involved is a NO GO. I won’t touch anything associated with him or the brand. Are they finally turning a profit yet, or are they still losing billions and being propped up by shady money from shadowy figures?
Click to expand...
Google? Yeah. Thanks to search and cloud foundational business model
Meta? Yeah. Thanks to AI optimizing their ad engine.
Perplexity? Getting there fast. By pivoting to agentic platform. ARR in '25 ->$100M, in '26 ->$500M
Anthropic? Not yet but getting there. Not expected to be cash flow positive until 2028.
OpenAI? Nope. Project $14B loss for this year due to extreme computing and training costs.
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back