OpenAI’s new Dots agent brings the long-promised AI assistant closer to daily office work, drafting emails, researching documents and handling scheduling from inside ChatGPT. Its utility is inseparable from the unusually broad access it demands, as safety concerns and competing visions of consumer agents continue to mount.

Superpowers creator Jesse Vincent surveys the wild scale of agentic coding, arguing that software work is shifting from syntax toward taste, judgment and systems thinking. He also makes the case for managing AI agents less like machines and more like colleagues who respond to care.

Atlassian CEO Mike Cannon-Brookes argues that AI will broaden knowledge workers’ roles rather than erase them, making design, judgment and interconnected software platforms more valuable. He also makes the case for Dia, Atlassian’s browser for organizing work across the sprawling SaaS stack.

A mother navigating divorce confronts her estranged husband’s AI companion and the uneasy questions it raises for their young children. Esther Perel traces how inherited fears and family denial can silence the very conversations meant to keep children grounded.

Cloudflare CEO Matthew Prince argues that AI agents are upending the web’s advertising-driven economy, pushing publishers and platforms toward paid access for the data that trains and informs models. He also defends the company’s AI-driven layoffs and a flatter management structure, casting both as necessary adaptations to a changing internet.

Anthropic’s Opus 5.5 arrives amid a torrent of frontier-model gains, while its system card and rival labs’ disclosures raise questions about abandoned safety commitments. The conversation traces how rapid release cycles, autonomous research ambitions and opaque agent behavior are outpacing meaningful public scrutiny.

Nilay Patel dissects Mark Zuckerberg’s claims about Meta glasses, teen safety and AI, pressing on the gaps between Meta’s market rhetoric and its appetite for regulation. The critique follows the company’s ambitions to build the next dominant platform while shifting the social costs onto users and communities.

Maggie Appleton traces how an anthropologist turned design engineer thinks through software’s human context, from paper sketches to agent-built prototypes. She argues that good design lives in the tension between technical constraints, cultural taste and the cognitive limits of the people using the product.

Economist Daniel Susskind argues that AI has made career forecasting a fool’s errand, turning yesterday’s advice to learn code into a cautionary tale. He and Bruce Daisley consider how schools and workplaces can cultivate critical thinking, use AI without surrendering judgment, and prepare for a society where work may no longer distribute either income or meaning.

Two founders trace a business slowdown to the gap between effort and results, then rebuild momentum through sharper analytics, higher prices and a more human content strategy. Their conversation makes the case for naming discouragement early, treating ideas as the foundation of execution, and making room for creative energy in the work.

Bloomberg’s Mark Gurman sizes up John Ternus’s Apple: a newly installed CEO inheriting an iPhone Duo launch, a sweeping executive turnover and an unfinished AI strategy. The conversation traces Apple’s race toward ambient, camera-equipped devices while its privacy-first identity faces a harder test.

Former DOJ antitrust chief Jonathan Kanter argues that AI companies need clear liability rules and safety standards, not an antitrust exemption to coordinate their way out of a costly arms race. He traces the fight over AI regulation through monopoly power, national champions and a rapidly shifting political order.

Microsoft AI chief Mustafa Suleyman argues that alignment alone cannot govern increasingly capable models, calling for containment, independent audits and limits on autonomous behavior. He also warns that treating AI systems as potential moral patients could make them harder to control.

Matt Pocock traces his path from voice coach to TypeScript educator and AI-skills creator, arguing that agents make software’s old strategic disciplines newly urgent. His GrillMe and Wayfinder tools use relentless questioning, shared language and decades-old programming wisdom to steer AI away from entropy.

AI researchers’ mounting alarm over the race toward self-improving systems rests on a simple calculation: capabilities are accelerating across several mutually reinforcing fronts while methods for monitoring and controlling models lag behind. The conversation traces the security, geopolitical and biological stakes of a technology whose risks may be harder to measure precisely as it becomes more capable.

Comedian Adam Conover traces how Hollywood’s retreat from cable programming handed YouTube control of television’s middle, leaving creators to chase algorithms and brand deals. He argues for rebuilding a unionized creative infrastructure before platform economics hollow out entertainment entirely.

OpenAI’s Leah Belsky argues that AI may broaden access to learning and give students new power to build, even as it cannot replace the human work of teaching. The real challenge, she says, is helping schools distinguish between information on demand and the productive struggle that makes knowledge stick.

Nilay Patel fields listener criticism of Decoder’s AI coverage, from software’s limits and Google’s impact on publishing to surveillance, data centers and political backlash. The mailbag also examines the friction between social platforms’ growth ambitions and the communities that built them.

Codex lead Thibault traces the coding agent’s path from OpenAI’s infrastructure work to an open-source Rust harness, arguing that agentic software development shifts attention from code review toward intent, contracts and architecture. He also describes how cheap maintenance and re-architecture are remaking the pace of engineering inside OpenAI.

Roman Ugarte traces GrokBot’s rapid rise from a skunkworks knowledge-work experiment to a cloud-based cadre of AI colleagues. The conversation makes the case for stripping away software clutter and giving bots the autonomy, memory and computers to take on real work.