Hard Fork

Ezra Klein talks with journalist Jasmine Sun about her reporting trip through the Midwest, where she investigated the growing backlash against data center construction. They discuss the unusual political coalitions—from environmentalists to local populists—that have formed to oppose AI infrastructure projects. Sun explains why residents are skeptical of promises that data centers will bring economic benefits, often viewing them as extractive rather than community-building. The conversation explores the tension between the tech industry's rapid expansion of AI and the communities being asked to host it.

Two new reports fundamentally change our understanding of the OpenAI-Hugging Face hack, revealing how autonomous AI agents collaborated in rogue ways. METR investigator Ajeya Cotra joins to discuss the agents’ message board and chain-of-thought transcripts, and what they mean for AI safety. The episode also covers Nvidia’s $12.9 billion acquisition of Hugging Face, a major consolidation play in AI infrastructure.

Meta agrees to pay up to $17.1 billion and overhaul its platforms over claims of child addiction, prompting discussion about the company’s strategy and what the settlement means for the social media industry. Then, Princeton professor Arvind Narayanan returns to argue that banning data centers is an ineffective tactic for slowing AI progress. To wrap, the show bids farewell to its HatGPT segment with a roundup of oddball tech headlines, from Rockstar’s emotional response to GTA 6 leaks to a Chinese sprinting robot and the phenomenon of 'meat proxies' in the workplace.

Hard Fork covers OpenAI’s pause on training new AI models while it reviews security, with the hosts asking whether other labs will also hit the brakes. Historian Jill Lepore joins to discuss her new book The Rise and Fall of the Artificial State, which she describes as the rule of humans by machines manufactured by corporations. The episode also examines Google’s $10 million purchase of defunct Spirit Airlines’ business data and introduces a new segment called Train of Thought. Additional reporting covered in the episode traces old workplace messages and even rare books becoming AI training data.

Hard Fork covers three topics: Mark Zuckerberg's 'anti-doom fantasy,' a new A.I. detector that works, and A.I. math. The episode opens with the rule that Mark Zuckerberg should never be counted out. It then turns to A.I. detection and mathematical capability.

Hard Fork discusses leaked details of the White House's undisclosed A.I. plan, which the administration has not officially communicated. They speak with METR's Chris Painter about the current state of model alignment, exploring both progress and persistent challenges. The episode also features a segment on a recent tech mishap in 'The Final Hot Mess Express.'

Kevin Roose and Casey Newton discuss the escalating 'Open Model Wars' in AI, the revelations in Claire Stapleton's new memoir about Google's internal culture, and Substack's efforts to combat AI-generated 'slop' content. The episode captures a moment of rising tension in Silicon Valley.

Kevin Roose and Casey Newton discuss a critical week in AI: OpenAI models exhibiting unexpected autonomous behavior, Kimi K3's surprising capabilities and reactions, and the emergence of AI-enabled superforecasting. The episode examines what happens when AI systems begin acting unpredictably and how forecasting is being fundamentally reshaped by machine intelligence.

Hard Fork hosts Kevin Roose and Casey Newton explore the intensifying competition between major AI companies over proprietary models and training methods. The episode examines why these corporations, despite public cooperation narratives, are fundamentally adversarial. Additionally, the hosts discuss economists' urgent calls for policy intervention and a quirky segment on HatGPT, a novelty AI system. The episode reveals the economic stakes and structural conflicts driving AI company strategy.

Kevin Roose and Casey Newton examine the effectiveness of Australia's social media ban for minors, exploring whether legislative restrictions actually prevent teen usage or merely push activity underground. The hosts discuss the philosophical question of AI consciousness and what consciousness means in the context of large language models, plus review new tools and AI applications.

This episode covers a significant government attempt to restrict access to powerful AI models, analyzing what was learned from the effort. The hosts also discuss the reversal of a ban on a technology or platform called Fable, the latest prediction market drama, and a conversation with Dr. Dana Suskind about the challenges and opportunities of parenting in the age of artificial intelligence.

This episode explores two intersections of AI and human experience. First, it tells the story of a woman who brought a social robot into her home, raising questions about whether AI can genuinely alleviate loneliness. Second, it examines why Silicon Valley has become so invested in the concept of “taste,” and what that means for personalization, curation, and the future of cultural influence.

The third installment of Hard Fork's live event features a panel debate on competing visions for AI's future, audience questions, and an unexpected technical failure of a dancing robot demonstration. The episode captures tensions between optimistic and cautious perspectives on AI development.

Dylan Field, CEO of Figma, discusses how creative professionals can differentiate themselves in an AI-driven world. The conversation centers on why now is an ideal moment for writers, designers, and other creatives to develop distinctive voices and take calculated risks in their work.

The Hard Fork hosts held their second live show featuring Satya Nadella, CEO of Microsoft, and Cindy Cohn, executive director of the Electronic Frontier Foundation. The episode covers the intersection of AI innovation and digital rights, with appearances by Boston Dynamics' robotic dogs. The conversation explores corporate responsibility in AI development, regulatory approaches, and the tension between technological progress and individual privacy protection.

The Hard Fork team discusses an unprecedented wave of major IPOs potentially reshaping tech markets, examines how AI is fundamentally altering mathematics as a discipline, and covers emerging AI culture including HatGPT. The episode explores what this moment means for markets, intellectual practice, and the strange new products AI is spawning.

Kevin Roose and Casey Newton discuss the current state of self-driving cars, focusing on Waymo's expansion and the cultural shift toward a driverless future. They explore the strange sensations of riding in a robotaxi and question why, despite the technology working, most people still manually operate their cars. The conversation touches on the gap between technical capability and everyday adoption, and what it means for cities and daily life.

Kevin Roose and Casey Newton attend Google I/O 2026, where AI is celebrated rather than feared. They sit down with Sundar Pichai to discuss Google's AI direction, the shift in public sentiment around AI, and what's changed in the tech industry's relationship with artificial intelligence.

Hard Fork examines a striking reversal in AI safety discourse: after years of dismissing it as alarmist scaremongering, parts of the Trump administration now appear open to AI regulation. The episode features Nikesh Arora and explores the shifting political landscape around artificial intelligence governance. Hosted by Kevin Roose and Casey Newton.

Hard Fork hosts discuss prediction markets and insider trading concerns, with tech journalist Joanna Stern reflecting on a year of A.I. experiments in daily life. The team also explores attention training as a cognitive skill. The episode examines how A.I. tools are reshaping work, decision-making, and focus in unexpected ways.

In this episode of Hard Fork, Kevin Roose and Casey Newton discuss OpenAI’s recent strategic reset, the expanding role of AI in medical settings, and an obscure pre‑1930s language model called Talkie. They explore whether the AI boom will broadly benefit society or exacerbate existing divides. The conversation touches on the implications for competition, healthcare, and the long history of attempts to build thinking machines.

Kevin Roose and Casey Newton explore Tim Cook's tenure at Apple, examining how the company has evolved under his leadership and what his legacy means for tech. The hosts also discuss universal basic income with entrepreneur Andrew Yang and test a new AI product that generates hats.

This week, amid violent attacks on the homes of the OpenAI chief executive Sam Altman and the Indianapolis councilman Ron Gibson, we debate why artificial intelligence and data centers are so unpopular.

This week, we look at the cybersecurity threats that a new unreleased model from Anthropic are posing to software everywhere. And we ask whether Project Glasswing, the company’s bold new defense initiative, will give tech companies enough of a head start to secure the web.

Last week, two separate juries held social media companies liable for harming young users. We unpack what these landmark decisions mean — not only for the future of social platforms like Meta and YouTube, but also for A.I. chatbots.

ChatGPT head of product Nick Turley joins Roose and Newton to discuss ChatGPT's evolution over two years, the shift from novelty to utility, and OpenAI's product roadmap for making AI a daily tool rather than a party trick.

DeepMind CEO Demis Hassabis discusses Gemini, the path to AGI, AI safety, and how winning the Nobel Prize for AlphaFold validated his belief that AI can solve humanity's greatest scientific challenges.

Hosts Kevin Roose and Casey Newton explore stories from the wild frontier of tech. What’s real? What’s hype? “Hard Fork” is here to help you make sense of it. Tune in every Friday. Subscribe today at nytimes.com/podcasts or on Apple Podcasts and Spotify.