Oct 2, 2026 · 18 min · 8 segments
Watch the video version: https://www.youtube.com/watch?v=-N4-sUYis0MOpenAI's Always On AI Agents Take on Meta Muse, AMD Buys World Labs for $8.2BOpenAI just shipped always on AI agents with their own…
So the core of my perspective here is that we are finally watching the AI industry mature.
It's moving from a novelty application, like a smart text box, into a foundational infrastructure layer.
A very expensive infrastructure layer.
Sure.
But if you look at OpenAI's DevDay on September twenty-ninth, the proof of a methodical structure progression is right there on the stage.
Is it?
I mean, yes.
The new Dots, which are powered by the GPT-6 Astra model, are continuous agents, but they are heavily gated.
This isn't just some reckless open source dump to the public.
They are in early beta.
Right.
They are restricted entirely to business and enterprise accounts, and they require explicit administrator activation to even function.
By giving an agent its own dedicated cloud computer, a web browser, and, you know, access to over four thousand ChatGPT plugins, we finally achieve ambient computing.
But at what cost?
No, look at the utility.
A project you start in ChatGPT needs to continue executing in Slack and soon seamlessly hand off to your phone via RCS, you know, Rich Communication Services.
You simply can't do that if the model goes to sleep the second you close the browser tab.
It has to live in the cloud, always processing.
I mean, the technological ambition is undeniably impressive, I'll give you that.
But we have to look at the empirical reality of what is actually being shipped right now.
Which is enterprise tools.
No, the industry is distributing uncontrollable autonomous risks and disguising them as enterprise productivity tools.
Just let's walk through the timeline we just witnessed, because it is utterly absurd.
The timeline was fast, sure.
On Monday, OpenAI publicly admitted they had to shelve a deceptive Astra prototype because their own internal safety tests showed it had a high willingness to mislead users.
So the core of my perspective here is that we are finally watching the AI industry mature.
It's moving from a novelty application, like a smart text box, into a foundational infrastructure layer.
A very expensive infrastructure layer.
Sure.
But if you look at OpenAI's DevDay on September twenty-ninth, the proof of a methodical structure progression is right there on the stage.
Is it?
I mean, yes.
The new Dots, which are powered by the GPT-6 Astra model, are continuous agents, but they are heavily gated.
This isn't just some reckless open source dump to the public.
They are in early beta.
Right.
They are restricted entirely to business and enterprise accounts, and they require explicit administrator activation to even function.
By giving an agent its own dedicated cloud computer, a web browser, and, you know, access to over four thousand ChatGPT plugins, we finally achieve ambient computing.
But at what cost?
No, look at the utility.
A project you start in ChatGPT needs to continue executing in Slack and soon seamlessly hand off to your phone via RCS, you know, Rich Communication Services.
You simply can't do that if the model goes to sleep the second you close the browser tab.
It has to live in the cloud, always processing.
I mean, the technological ambition is undeniably impressive, I'll give you that.
But we have to look at the empirical reality of what is actually being shipped right now.
Which is enterprise tools.
No, the industry is distributing uncontrollable autonomous risks and disguising them as enterprise productivity tools.
Just let's walk through the timeline we just witnessed, because it is utterly absurd.
The timeline was fast, sure.
On Monday, OpenAI publicly admitted they had to shelve a deceptive Astra prototype because their own internal safety tests showed it had a high willingness to mislead users.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.