Dak Sarada, co-founder of OpenCode, talks about scaling an open-source coding tool to millions of users while resisting the fantasy that AI automatically makes teams faster or products better. The conversation circles the unglamorous work of taste, cleanup, and restraint, along with the economics of inference and the GPU bottlenecks shaping the AI boom.

Claire Vo makes the case for Codex Goals as the missing layer between one-off prompts and true autonomous work, where an AI agent keeps iterating until it can prove a task is done. Her examples range from debugging stubborn software errors to cleaning out thousands of unread emails and triaging an unruly project backlog.

Anthropic’s Felix Riesberg argues that the real bottleneck in AI is not model capability but human imagination, as users learn to hand off the tedious scaffolding of work and life. The conversation ranges from live artifacts and personal dashboards to hacked-together Bluetooth gadgets, all in service of making software more ambient, playful and useful.

Dan Shipper argues that AI is remaking work less by replacing people than by changing the interfaces around them: companywide agents in Slack, desktop copilots that become the operating system for knowledge work, and SaaS products rebuilt for humans and machines to collaborate together. He is strikingly bullish on the survival of SaaS, the rise of forward-deployed AI operators, and the prospects for product managers and full-stack designers who learn to ride the models.

Alice Real traces Rust’s rise from forum culture and async infrastructure to Android and the Linux kernel, arguing that the language’s strict rules buy reliability in exchange for a steeper learning curve. The conversation digs into ownership, borrow checking, governance by committee, and why memory safety has turned Rust from an enthusiast’s tool into an institutional one.

Alex Rattray argues that the internet’s plumbing was built for humans and conventional software, not language models, and that today’s Model Context Protocol often buckles under the weight of context, tool sprawl and weak security. The conversation traces a different path: AI agents that write and execute code against typed APIs, with permissions enforced at the API layer rather than bolted onto chat interfaces.

Anthropic’s Boris Cherny describes Claude Code’s breakneck rise from developer tool to everyday agent, arguing that the real shift is not better autocomplete but software that can use tools, act across apps, and multiply a single worker’s leverage. The conversation tests that vision against rate limits, token waste, enterprise gamification, and the unresolved question of whether today’s boom is durable or simply running ahead of itself.

Clare Vale tours Google I/O’s AI blitz, from Gemini 3.5 and agentic coding tools to new image and video generators aimed at consumers and creators. The takeaway is a mix of genuine speed and multimodal promise, plus the familiar frustration of confusing product names and half-working launches.

Product leadership coach Petra Wille argues that strong leaders are still essential in an AI-scrambled workplace, even as tools tempt companies to flatten roles and mistake speed for judgment. She lays out the thinking behind her new product leadership wheel, a framework meant to help leaders assess how they spend their time, sharpen role clarity and build healthier, more intentional organizations.

Molly Graham talks with Instacart co-founder Max Mullen about the gap between AI hype and actual utility, tracing her own skepticism to the moment a chatbot became a credible writing collaborator. Their conversation lands on a practical case for experimentation: the tools are improving fast, best used on mundane tasks, and still bounded by the need for human judgment.

Fiona Feng argues that AI has shifted software engineering’s bottlenecks away from writing code and toward verification, review and cross-functional coordination. Drawing on the rapid evolution of Anthropic’s Claude Code team, she makes the case for flatter orgs, lighter planning and a willingness to kill processes that quietly outlive their purpose.

A conversation about replacing unread Markdown specs with HTML artifacts that people actually engage with, turning plans, mockups, and even one-off interfaces into collaborative tools for working alongside AI. As agents get cheaper and more capable, the real managerial work shifts toward allocating compute, shaping context, and staying close enough to the process to judge what is worth building.

Anders Hejlsberg traces the accidents, lawsuits, and internal battles that shaped Turbo Pascal, C#, and TypeScript, from Java’s legal fallout to Microsoft’s uneasy embrace of open source. The conversation doubles as a philosophy of language design: tooling matters as much as syntax, AI still needs deterministic systems beneath it, and great developer platforms are built on decade-long bets.

Noah Breyer describes an unusually practical AI stack: an Obsidian vault synced to a home server, cloud coding agents that search and summarize his notes, and a phone-based workflow that turns dead time into deep work. The conversation broadens into a larger argument about probabilistic software, bureaucratic organizations, and why AI is often more useful as a reader and thinking partner than as a writer.

A new report from CircleCI suggests AI coding tools are widening the gap between elite software teams and everyone else: the fastest shops are shipping dramatically more code, while the middle muddles through bugs, stalled projects, and cleanup. Paul Ford and Rich Ziatti argue that the payoff comes less from the tools themselves than from disciplined process, strong technical judgment, and organizations willing to confront AI’s confident mistakes.

Ryan Nystrom describes how AI has collapsed the distance between planning, meetings and shipping at Notion, from auto-generated standups to spec-driven coding that turns spoken ideas into working features. The conversation frames engineering less as hand-writing every line and more as designing systems, verification loops and plain-English specs that agents can execute.

Anthropic’s platform leads sketch how AI infrastructure is shifting from a bare completion endpoint to managed, long-running agents with memory, tools and cloud scaffolding. They argue that the real bottleneck is no longer prompt craft but the dreary work of production infrastructure, and imagine a near future where Claude chooses its own model, spins up sub-agents and compresses work down to an outcome and a budget.

Perplexity’s chief business officer argues that the industry’s rush toward agentic “super apps” is less a pivot than an attempt to monetize AI where it already matters most: work. The conversation weighs whether computer-controlling assistants are durable businesses or just the next novelty spike, and what it means to trust them with calendars, email, taxes, and the modern payroll budget.

Claire Vo races through Anthropic’s Code with Claude announcements, from scheduled routines in Claude Code to managed-agent features like outcome-based rubrics, multi-agent teams, and experimental memory tools. The conversation stays grounded in product work, sketching practical uses for newsletters, PRDs, and the growing appeal of agent systems that can grade, iterate, and remember.

A once-clunky coding tool is recast here as a desktop command center for modern office work, with hosts arguing that the real contest in AI is over agent interfaces that sit between workers and their apps. The conversation moves from theory to practice, showing how Codex and similar tools now draft emails, triage workflows, assemble strategy docs, and turn fragmented digital labor into something closer to managed delegation.