AI, Agents & the Future of Ruby: Chad Fowler, José Valim, Obie Fernandez, Armin Ronacher

The closing session of day one at RubyConf Austria 2026, recorded live on May 29, 2026 at Haus der Musik in Vienna. Moderator Hans Schnedlitz brings together four people with very different vantage points on AI and agentic engineering: • Chad Fowler — co-founder of Ruby Central, longtime RubyConf/RailsConf organizer, now VC • José Valim — creator of Elixir, former Rails core, now building Tidewave • Obie Fernandez — author of The Rails Way and Patterns of Application Development Using AI, CTO at ZAR • Armin Ronacher — creator of Flask, longtime Sentry engineer, now building in agentic engineering An hour of candid, often hilarious disagreement about what AI agents are doing to our craft, our communities, and our job descriptions. Highlights: • Has agentic coding made you happier? Four takes on joy, mania, and "being as reckless with AI as we were with Ruby in 2004" • Disposable code vs. "single-use plastic software" — regenerating implementations inside stable architectures, and the flood of software with no users • "Social coding is dead": José on what agent-written pull requests are doing to open source, Chad on wanting "your tokens, not your pull requests", and whether the GitHub model is over • Is Rails viable for agentic coding? Obie: "If you're having trouble doing Ruby or Rails with AI, you're holding it wrong" • The language question: Chad's whichlang experiment (0% Ruby), Armin on local reasoning, training-set gravity, and the Goldilocks language that doesn't exist yet, José on why Elixir scores well — and why "I don't know" is the honest answer • Why agents write code we hate: reinforcement learning rewards forward progress, producing try/catch slop and "malicious backwards compatibility" • Architecture over language: constraints, boundaries, conceptual-level linting, and compiling intent into architectures • A year from now: evals, executable specifications, LLMs as reusable building blocks — and the uncomfortable question of who can afford tokens at all Chapters 00:00 Introductions 03:30 Has agentic engineering made you happier? 08:10 Disposable code and single-use plastic software 12:45 Can AI write beautiful code? Where careful code still matters 14:44 The organization downstream falls apart 16:37 Open source in the agent era: "social coding is dead" 25:33 Do we still know how to build together? 28:22 Is Rails suited for agentic coding? 33:02 New frameworks, new languages — whichlang and 0% Ruby 35:38 Local reasoning and the Goldilocks language 38:13 Is Elixir in the Goldilocks zone? 41:16 Architecture over language: constraints and boundaries 46:17 The machine feels no pain: why agents write code we hate 53:32 Reinforcement learning and malicious backwards compatibility 56:22 A year from now: evals, executable specs, and access 64:45 Wrap-up Links Chad Fowler: https://chadfowler.com José Valim / Tidewave: https://tidewave.ai Obie Fernandez: https://obiefernandez.com Armin Ronacher: https://lucumr.pocoo.org whichlang: https://github.com/chad/whichlang RubyConf Austria: https://rubyconf.at RubyConf Austria 2026 took place May 29–31 in Vienna. Thanks to our speakers, sponsors, and community partners. #ruby #rails #ai #rubyconf #softwareengineering