{"type":"rich","version":"1.0","provider_name":"Transistor","provider_url":"https://transistor.fm","author_name":"YAGNI","title":"Retiring from programming w/ Justin Searls","html":"<iframe width=\"100%\" height=\"180\" frameborder=\"no\" scrolling=\"no\" seamless src=\"https://share.transistor.fm/e/192793fc\"></iframe>","width":"100%","height":180,"duration":4914,"description":"Matt just got back from Rails World, where DHH's keynote left a lot of people thinking Rails was over. Justin Searls of Breaking Change, who gave his own retirement talk at Rails World two years ago, joins him to sort it out. They cover why the keynote hit people's fears so hard, whether \"the one-person framework\" still means anything when agents skip the coding step, and why Justin thinks Rails is heading into another stable era. Justin also explains why taste is the skill that lasts, how he runs seven iOS apps with agents fixing bugs every hour, and what he'd tell someone who's scared right now.\nMentioned in this episode\nDHH's Rails World 2026 keynote: The \"pencils down\" keynote with barely any Rails in it. Justin watched it at 2x and heard things he's been saying for two years. For a recap, see Global Nerdy's \"most confusing funeral I've ever experienced\".\nThe Empowered Programmer: Justin's Rails World 2024 talk in Toronto, the last one before he retired from speaking. He built the app behind it by hand, and even then he could see an end coming for framework conferences.\nBrooks's law: Justin and Scott Werner's octopus math. Eight engineers, each running eight agents, act like 64 people when it comes to coordination. That's why Justin thinks the ideal team size is now one.\nWebKit's performance policy: Safari got fast because any change that made it slower failed the build. Justin's example of a clear numeric target that agents can chase without stopping.\nXcode 27's Device Hub: Apple's replacement for Simulator. Combined with Xcode's MCP tools, Justin runs 16 simulators at once with agents testing each one, and his Mac's fans running flat out.\nMr. Meeseeks: Justin's model for AI code review. His Opus subagents exist only to review the diff and keep going round after round, sometimes 12 or 15 times, until they can't find anything to fix. As he puts it, no human was ever really reading pull requests anyway.\nWhy Elixir is the best language for AI: José Valim's case...","thumbnail_url":"https://img.transistorcdn.com/UMMB8JkX91ih6GoZqlbBWDXY7T-jmGQ8Ufj2jHpLxMw/rs:fill:0:0:1/w:400/h:400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS9zaG93/LzMyODAyLzE2NTk0/MDE1MTYtYXJ0d29y/ay5qcGc.webp","thumbnail_width":300,"thumbnail_height":300}