Kevin Roose’s AI Book Lands Just as the Media Business Reinvents Itself

Kevin Roose has written a new book about the rise of modern artificial intelligence without using AI to write it, and he says that choice was intentional. The longtime tech journalist’s The AGI Chronicles arrives as he and former cohost Casey Newton launch Machine Gods Media, a new AI-focused podcast and media company backed by an NPR distribution partnership.

The timing matters because Roose’s book tries to do for the current AI boom what very few accounts have done so far: turn a fast-moving, hype-heavy, and often polarized story into a durable record of how OpenAI, Anthropic and Google helped create the generative AI era.

Roose spoke about the project in late September, shortly after announcing the new media venture with Newton. In the conversation, he laid out why he believes the public conversation around AI is stuck between cynicism and boosterism, why radio still offers unusual reach for ambitious audio journalism, and why he thinks legacy media is struggling to hold onto top talent in a world where individual journalists increasingly build their own brands.

Why Kevin Roose wrote a book about AI history now

Roose says the decision to turn years of reporting into a book came in early 2025 while he was stuck in traffic on the Bay Bridge. He had already interviewed leading researchers, founders and executives, but that moment made him realize the story he had been following was larger than any one news cycle.

What changed, in his view, was perspective. After years immersed in Silicon Valley, the race to build more powerful AI systems had started to feel normal. But when he stepped back, he saw something unusual: a small group of companies and researchers trying to build systems that could reshape the economy, society and possibly the balance of power between humans and machines.

Roose described the AI race as something that felt historically significant and potentially fragile, arguing that the field needed a real record before the conversations, documents and internal debates disappeared into ephemeral chat threads and deleted messages.

That concern shaped the structure of The AGI Chronicles. Rather than writing a quick industry trend book, Roose spent roughly a year reporting and writing, interviewing more than 150 people. He wanted to create something closer to a historical archive than a reaction piece.

What Roose says was missing from AI coverage

Roose argues that public discussion of AI has become distorted at both extremes. On one side are critics who dismiss the technology as overhyped, economically irrelevant or destined to collapse in a financial bust. On the other are promoters who talk almost exclusively about productivity gains and enterprise use cases while downplaying risk.

He says the gap in the middle is where serious reporting should live. That middle ground, which he and Newton call “AI realism,” accepts that the systems are powerful and sometimes dangerous without reducing the conversation to either panic or marketing.

In practical terms, that means treating AI as consequential while still remaining skeptical, curious and accessible. Roose also said he believes readers and listeners should be able to engage with the subject without having to choose between apocalyptic warnings and glossy demonstrations.

How OpenAI, Anthropic and Google came to define the AI era

The book follows three companies Roose sees as central to the modern AI story: Google, OpenAI and Anthropic. He says their relationship is not a simple tale of competition among separate firms, but an interlocking history of talent, fear and escalation.

Google, as Roose tells it, was the oldest and deepest AI lab in the group. The company’s research helped produce the transformer architecture that underpins today’s most capable large language models. That breakthrough became foundational to the products now marketed to consumers as chatbots and agents.

OpenAI emerged partly out of anxiety that Google was moving too far ahead. Roose said the founders wanted to create a lab that could challenge Google and prevent a single company from dominating the field. Later, Anthropic was launched by former OpenAI employees who wanted to take a different approach and, in Roose’s telling, stop OpenAI from crossing a critical threshold first.

What surprised Roose most about the AI companies

The biggest surprise, he said, was how personal and combative the rivalries became. He expected something closer to classic consumer-tech competition, but instead found a deep, almost familial split among people who had once worked side by side.

Rather than a neat rivalry like Coke versus Pepsi, Roose described the AI landscape as a blood feud in which former collaborators now race one another to build the most advanced systems.

Why Roose and Casey Newton are starting Machine Gods Media

Roose and Newton announced Machine Gods Media shortly before the interview, alongside a podcast distribution deal with NPR. The project marks a step away from the New York Times ecosystem that helped make their previous show, Hard Fork, a major tech podcast.

Roose said the new arrangement gives them more creative independence while still allowing broad distribution. He described NPR as a good fit because of its reach, its reputation and its radio presence, which he says still matters more than many people in digital media assume.

He also said the partnership helps avoid the trap of speaking only to a narrow self-selected tech audience. A public radio distribution model, in his view, makes it more likely that people outside Silicon Valley — including local officials, policy professionals and casual listeners — will encounter the show.

Roose said the appeal of NPR was not only financial, but also editorial and strategic: broad distribution, creative control and a chance to reach people who do not already spend their time searching for AI coverage.

How much did the media deal matter?

Roose would not discuss exact terms, but he acknowledged that the company received a strong offer and could have earned more money elsewhere. The point, he suggested, was not to maximize a short-term payout but to find a structure that supported growth, reach and editorial ambition.

That is notable because audio and newsletter talent increasingly have options outside legacy institutions. Roose framed the move less as a rejection of his former employer and more as evidence that the traditional media career ladder has changed.

What Roose thinks is changing in journalism

Roose says he has no grievance with The New York Times, where he spent much of his career. But he believes journalism is moving into an era where individuals often matter as much as institutions. Readers and listeners, he argued, increasingly follow people they trust rather than the masthead alone.

In his view, that shift makes media companies more dependent on flexibility. Reporters, commentators and podcasters may want to keep independent projects alive rather than surrendering them entirely to a single employer. Institutions that want to compete for top talent will need to adapt to that reality.

He also suggested the old model — the one in which a journalist spends decades climbing a single ladder inside one company — is already broken. He said that is unfortunate, but also undeniable.

Why media companies are losing leverage

Several forces are driving the change:

• Individual journalists can build direct audiences through podcasts, newsletters and social platforms.
• Independent brands can generate leverage when negotiating with publishers or distributors.
• Audiences often attach to personality-led analysis in fast-moving fields like AI.
• Media organizations that cannot offer creative latitude may struggle to keep high-profile talent.

Roose’s own move reflects that shift. His audience already knows his name, not just the outlet he writes for. That makes a standalone media venture more viable than it would have been a generation ago.

How does Roose think AI can help journalism?

Roose is optimistic about using AI as a reporting tool, but not as a replacement for journalists. He said he does not want AI to write his stories or take over the reporting process. Instead, he sees value in using it to extend what journalists can do.

He pointed to document analysis, image review and investigative work as areas where AI could help reporters find patterns at scale. In particular, he noted that some newsroom teams have already used advanced tools to analyze huge batches of material or to detect physical changes from satellite imagery.

He also said he has used AI to help research and edit his own writing, and believes that has improved the quality of his work. The danger, in his view, is not AI-assisted journalism itself but lazy automation that floods the internet with low-value content.

Roose said the most useful newsroom experiments will be the ones that help reporters become better, faster and more thorough, rather than the ones that simply generate more copy.

Who does Roose trust with the future of AI?

Roose said he does not trust any one person with the future of artificial intelligence. After years of reporting on the field, he sees the major leaders as capable, ambitious and imperfect rather than as either villains or saints.

That does not mean he thinks their incentives are pure. It means he believes the people building frontier systems are human beings with mixed motives and blind spots. In his view, the current generation of AI executives is, on balance, more thoughtful than many of the social media executives who shaped the last major tech era.

He drew a sharp contrast with the early platform era, when many tech leaders spoke about openness and connection but did not seriously confront the risks of their products until regulators and lawmakers forced the issue. By comparison, AI executives have talked about danger from the beginning, even while racing ahead.

What responsibility do AI leaders carry?

Roose says the central question is not whether these leaders understand the stakes. It is why they continue pushing the technology forward if they believe it could be destabilizing or even catastrophic.

He said some of them genuinely believe the advance of AI is inevitable. Their argument is that if powerful models can be built, someone will build them, so the better response is to shape the outcome carefully rather than slow progress to a halt.

That logic, Roose suggests, is a defining tension of the current era: the same people warning about existential risk are also the ones racing to release more capable systems.

What the book is trying to preserve

Roose’s book is not just about companies and products. It is also about memory. He believes the AI boom is already being narrated through fragmented posts, Slack messages and promotional claims, which makes it easy for the real history to blur.

His goal is to leave behind a coherent account of how the technology moved from academic research into consumer software and then into one of the most consequential stories in business and politics.

That mission explains the tone of the project. Roose is not trying to crown a winner, deliver a manifesto or settle the debate over whether AI is good or bad. He is trying to explain how the race happened, who drove it, and why the result matters.

Why this matters beyond Silicon Valley

The AI story now reaches far beyond the technology industry. It affects labor markets, creative work, education, search, customer service, national security and public policy. That is one reason Roose is interested in reaching listeners who are not already deep in the tech world.

He believes the public deserves a version of the story that is neither dismissive nor promotional. If the next phase of AI changes the way institutions work, then the people making decisions about those systems need a clearer map of how the current era began.

That is what gives The AGI Chronicles and Machine Gods Media a shared purpose. One is a book-length attempt to document the origins of the AI race. The other is a platform designed to keep interpreting that race as it evolves.

The bottom line

Kevin Roose is betting that AI needs better historical reporting, more nuanced commentary and a wider public audience than the tech bubble alone can provide. His book and his new podcast venture both reflect the same argument: artificial intelligence is too important to leave to either hype merchants or skeptics who think it will vanish on its own.

For Roose, the challenge is not simply describing the technology. It is making sure the story of how it arrived is recorded clearly enough that future readers — and future systems — can understand how a handful of companies and a small circle of people helped shape one of the defining technologies of the century.