Codex and Claude Code are not browser plugins. They are replacing the browser as the primary workspace for knowledge work. That is Dan Shipper's argument from Lenny's Podcast, and this video breaks it down with concrete examples: agent-native document editors, in-app browsers, generative mini apps connecting to Gmail, Notion, and Typefully, and workflows where Codex or Claude controls the browser, reads your context, and completes tasks directly alongside you.
The structural shift is this: AI agents will not live inside your apps. Your apps will be rebuilt so agents can use them with you. Tools like Claude Code, Codex, Cowork, and Proof are early signals of that transition. The video introduces the concept of moving from browser tabs to task tabs, where multiple tabs open per agent thread and agents automatically surface the right tools for each job. That reframing alone is worth the watch.
The presenter is also building toward this future directly: an iMessage-first AI agent running 24/7, connected to external tools, supporting automations via Claude and Codex on mobile and desktop. The predictions here about agent-native SaaS displacing traditional software with AI buttons, and organizing work at the SOP level rather than the app level, give this video shelf life beyond the current hype cycle.
[WATCH ON YOUTUBE →]