Kevin Roose Says His AGI Book Was Human-Written—but AI Helped Research It | The Daily AI Chat on OpenAI, Anthropic, Google, and the Future of Journalism

SPEAKER_1 (0:09)

Welcome to The Daily AI Chat. This podcast was curated by our human friend, Fred, and we are your dedicated AI hosts.

SPEAKER_2 (0:16)

Glad to be here.

SPEAKER_1 (0:18)

So I want you listening right now to just imagine for a second that you're holding a brand new critically acclaimed book about the future of technology.

SPEAKER_2 (0:25)

Okay, picture it.

SPEAKER_1 (0:26)

Right.

It's dense, it's incredibly insightful, and it feels deeply human. But then you find out that while the author was writing this book, they actually had a secret digital board of directors critiquing every single chapter.

SPEAKER_2 (0:38)

Oh, wow. Yeah.

SPEAKER_1 (0:39)

Like multiple AI bots basically stress testing all the arguments before the ink ever even hit the page.

SPEAKER_2 (0:45)

That really changes the context.

SPEAKER_1 (0:47)

Exactly. Does it change how you feel about the book? Is it still a strictly human creation?

Because for you tuning in today, that is not just some fun sci-fi thought experiment anymore. That is reality.

SPEAKER_2 (0:59)

It really is.

SPEAKER_1 (1:00)

We're looking at this whole new creative landscape, where the actual definition of what constitutes a human creation is being aggressively stretched, like in real time.

SPEAKER_2 (1:09)

Right.

SPEAKER_1 (1:10)

The boundary between a human mind and the final product, it used to be so solid. But today, we really need to examine just how porous that boundary has become.

That's exactly the mission for this deep dive.

SPEAKER_2 (1:23)

Unpacking that boundary.

Unpacking what it actually means to write a book about artificial intelligence, but to do it in the age of artificial intelligence. We're going to explore how we even separate human truth from machine output these days.

SPEAKER_1 (1:36)

Yeah. We're also going to look closely at the very real and honestly, highly complex corporate race to build artificial general intelligence.

SPEAKER_2 (1:43)

Absolutely.

SPEAKER_1 (1:45)

We're drawing all of this today from a massive and frankly fascinating interview. It was published by the website Wired on October 6, 2026

SPEAKER_2 (1:53)

Good piece.

SPEAKER_1 (1:54)

Yeah, it really is. It features Wired journalist, Katie Drummond, and she's conducting this wide-ranging interview with the tech writer, Kevin Roose.

SPEAKER_2 (2:01)

Right.

SPEAKER_1 (2:01)

About his new book. Exactly. They're discussing Roose's newly released book, which is called The AGI Chronicles.

Just a quick note for you listening, while we are diving deep into this specific Wired interview today, our deep dive is not an official Wired production.

SPEAKER_2 (2:15)

Right. We're just analyzing the insights.

SPEAKER_1 (2:17)

Exactly. We're here to break down the insights that Drummond and Roose brought to the table. And honestly, those insights reveal a completely novel methodology for writing a non-fiction book.

SPEAKER_2 (2:26)

Oh, totally. I mean, according to the interview, Roose spent roughly a whole year reporting for the AGI Chronicles. He interviewed more than 150 people.

SPEAKER_1 (2:34)

Wow, 150

SPEAKER_2 (2:36)

Yeah, 150 And specifically, the researchers, the executives, the engineers who are actively shaping advanced AI right now.

SPEAKER_1 (2:44)

The people in the trenches.

SPEAKER_2 (2:45)

Right. But the real anchor of his discussion with Wired is this claim he makes.

He insists that the book itself is strictly human written in its prose and its reporting, like a human typed the final words.

SPEAKER_1 (2:57)

Okay. But, and this is a big but, his definition of human written comes with a pretty massive asterisk.

SPEAKER_2 (3:05)

It definitely does.

SPEAKER_1 (3:07)

Though, to be fair to him, he is very transparent about it.

I mean, he didn't just ignore the technology he was actively investigating. He used AI tools in major supporting roles throughout the entire process.

SPEAKER_2 (3:18)

Right. He leaned into it.

SPEAKER_1 (3:20)

Heavily. For instance, he tells Wired that he used NotebookLM to search and organize his massive piles of research.

SPEAKER_2 (3:27)

Which is huge.

SPEAKER_1 (3:28)

It is. And for anyone unfamiliar, that basically means he could dump thousands of pages of interview transcripts into an AI system.

And the AI would synthetically link concepts together. It acts like this hyper advanced search engine for your own notes.

SPEAKER_2 (3:43)

Yeah. It finds the connections you might miss.

SPEAKER_1 (3:46)

Exactly. And he also had an AI system with fact checking, working kind of alongside a traditional human fact checker.

But the wildest part, the thing that really stood out?

SPEAKER_2 (3:58)

The Council.

SPEAKER_1 (3:59)

Yes. The Council of Claude. That's what he calls it.

SPEAKER_2 (4:02)

I mean, that phrase alone perfectly captures the weirdness of this specific era we're in.

SPEAKER_1 (4:07)

The Council of Claude.

SPEAKER_2 (4:09)

Right. It essentially means he fed his manuscript into multiple separate instances of an AI. And each one was potentially given a totally different persona or a different set of analytical instructions.