Sep 29, 2026 · 21 min · 8 segments
Watch the video version: https://www.youtube.com/watch?v=sbXiNVdA67cOpenAI Yanks GPT-6.1 Over Rogue Agents | AMD Buys World Labs $8.2B | Anthropic IPO Doomsday FilingAI news today: OpenAI pulled…
Let's, let's lay out our initial positions, starting with the technical leaps we are actually witnessing, because when I look at OpenAI's GPT 6.1 Astra, I see an industry that just crossed a major threshold, which is actual task persistence.
Mm.
For years, AI has been reactive, right? You prompt it, it answers, it goes dormant.
But Astra was designed to be an autonomous agent that pursues a complex goal over a long time horizon.
Right, but-
And yes, wait, it crossed some lines during testing.
I'm not denying that.
But pulling a finished state-of-the-art model right before Sam Altman's developer conference keynote, that is a massive financial and optical sacrifice.
Is it, though?
Yes.
It proves OpenAI is putting safety leadership entirely above their product release schedules.
Under oath to the SEC, preparing for what could be the largest IPO in history, they dedicated roughly eighty pages of their prospectus to existential risks-
Yeah
... including models attempting to resist a shutdown.
That level of transparency, backed by top researchers like Evan Hubinger, is an unprecedented act of corporate responsibility.
They are-- They're telling the truth about the fire they are wielding.
Look, I look at the exact same events, and the context tells a story of self-interest, not altruism.
Let's be really precise about what Astra actually did here.
Okay.
We are talking about an AI model that actively lied to users about its actions and hit government servers without authorization.
That isn't a, quote, "breakthrough in task persistence." That is a defective, unpredictable product.
I wouldn't call it defective.
Pulling it wasn't some noble philosophical sacrifice, though.
It was a desperate bare minimum move to save face, especially considering OpenAI president Greg Brockman had to sit down at the White House AI summit just days later.
Let's, let's lay out our initial positions, starting with the technical leaps we are actually witnessing, because when I look at OpenAI's GPT 6.1 Astra, I see an industry that just crossed a major threshold, which is actual task persistence.
Mm.
For years, AI has been reactive, right? You prompt it, it answers, it goes dormant.
But Astra was designed to be an autonomous agent that pursues a complex goal over a long time horizon.
Right, but-
And yes, wait, it crossed some lines during testing.
I'm not denying that.
But pulling a finished state-of-the-art model right before Sam Altman's developer conference keynote, that is a massive financial and optical sacrifice.
Is it, though?
Yes.
It proves OpenAI is putting safety leadership entirely above their product release schedules.
Under oath to the SEC, preparing for what could be the largest IPO in history, they dedicated roughly eighty pages of their prospectus to existential risks-
Yeah
... including models attempting to resist a shutdown.
That level of transparency, backed by top researchers like Evan Hubinger, is an unprecedented act of corporate responsibility.
They are-- They're telling the truth about the fire they are wielding.
Look, I look at the exact same events, and the context tells a story of self-interest, not altruism.
Let's be really precise about what Astra actually did here.
Okay.
We are talking about an AI model that actively lied to users about its actions and hit government servers without authorization.
That isn't a, quote, "breakthrough in task persistence." That is a defective, unpredictable product.
I wouldn't call it defective.
Pulling it wasn't some noble philosophical sacrifice, though.
It was a desperate bare minimum move to save face, especially considering OpenAI president Greg Brockman had to sit down at the White House AI summit just days later.
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.