The Ruby on Rails Podcast

JP Camara joins the show to explore background job coordination, real-time collaboration, and the opportunities he sees for Ruby and Rails as AI changes how developers work. First, JP shares the story behind bringing batches to Solid Queue, from an early proposal to the conference conversations and community interest that helped move the feature forward. He explains how batches let developers treat a group of jobs as a unit, track success and failure, and trigger callbacks when processing finishes—making it easier to coordinate workflows that span many background jobs. JP also discusses the implementation challenges behind the feature, including database transactions, jobs deferred until a transaction commits, and reducing the performance overhead of tracking batches. He describes his related work on Mission Control Jobs, which gives developers a visual way to inspect batch progress, investigate failures, and find the jobs involved. The conversation then turns to YRBY, JP’s project for adding real-time collaborative editing to Rails applications. He explains how conflict-free replicated data types, or CRDTs, support Google Docs-style collaboration, and how Ruby bindings to a Rust library let him integrate with Action Cable or AnyCable without requiring a separate Node.js server. He also discusses editor integration, encrypted collaborative text, and his work on allowing AI agents to participate alongside people in shared documents. Finally, JP reflects on how AI helps him maintain momentum on open source projects while balancing a full-time job and family responsibilities. He shares how his team at Wealthbox uses RubyLLM, why its expressive interface makes working with different AI providers more approachable, and why he believes there is still plenty of valuable work ahead for the Ruby and Rails ecosystem. YRBY - https://yrby.dev lexxy-realtime - https://github.com/jpcamara/lexxy-realtime/tree/main RubyLLM - https://rubyllm.com/ Sponsor AppSignal: https://appsignal.com/rubyonrailspodcast

What does it mean to be a software developer when writing code is no longer the central part of the job? Jessica Kerr and Avdi Grimm join David Hill for a conversation about the changing work of software engineering—and how to navigate that change without losing the curiosity, satisfaction, and human connection that drew us to it. They begin by examining why AI-assisted programming differs from AI-generated art and entertainment, including the role of tests and measurable feedback in evaluating software. From there, Jessica and Avdi discuss the challenge to developer identity when tools can handle more of the implementation, and the opportunities that opens for building useful things, exploring neglected projects, and focusing on product decisions. The conversation gets practical with a look at overseeing AI-generated code. Avdi describes his use of specialized reviewer agents and repeated auditing, while Jessica explains how developers can define correctness, provide feedback loops, and get clearer visibility into changes. They also explore why inconsistent code and technical debt create friction for agents, and how token usage and evaluations can help measure the cost of that friction and compare different development approaches. Beyond implementation, David, Jessica, and Avdi discuss using AI as a tutor, expanding into design and product work, and the renewed value of design patterns and domain-driven design as shared vocabulary for communicating with agents. They also consider what developers can learn from people who began coding with AI, rather than assuming how those newcomers will learn. Throughout the episode, one question keeps returning: how can these tools create more time for people? From board games and conferences to conversations about what customers actually need, Jessica and Avdi make the case for using AI to support richer human collaboration—and keeping it from becoming a substitute for it. Sponsor: AppSignal: https://appsignal.com/rubyonrailspodcast

Recorded at Rails World, David sits down with Buzzsprout co-founder and CEO Tom Rossi for a wide-ranging conversation about AI, software development, product thinking, and the changing role of developers. Tom shares his perspective on the rapid pace of change in the software industry and why today's AI transition reminds him of the early days of the internet—only moving much faster. They discuss how AI is already changing development inside Buzzsprout and Higher Pixels, from adding AI-powered features for podcasters to making it dramatically easier to understand, maintain, and extend older Rails applications. But as the cost of building software falls, Tom argues that another skill becomes increasingly important: deciding what is actually worth building. David and Tom discuss the shift from thinking primarily as coders or engineers toward thinking as makers—understanding customers, identifying valuable problems, and choosing work based on the impact it can have rather than simply because it can be done. The conversation also turns to podcast analytics and David's experiments building his own analytics application with AI. They explore the difference between collecting more data and answering genuinely useful questions, why analytics should help creators make better decisions rather than send them down distracting rabbit holes, and how the same principle applies to product development. Ultimately, the conversation asks what developers uniquely contribute in a world where AI can generate increasing amounts of code: judgment, prioritization, customer understanding, and the ability to recognize which problems are actually worth solving. Sponsor Judoscale - https://judoscale.com/

Recorded at Rails World, Marco Roth joins the podcast to discuss the rapidly evolving future of the Rails view layer. Marco explains Herb’s integration into Rails 8.2, how its parser and tooling improve the experience of working with HTML+ERB, and how Herb creates opportunities that weren’t previously possible with traditional Rails templates. The conversation then turns to ReactionView, Marco’s approach to bringing fine-grained reactivity to Rails applications while allowing developers to continue writing familiar ERB. Marco explains how ReactionView can identify dynamic portions of a template, track dependencies, and update the browser without requiring developers to manually wire together Turbo Frames, Turbo Streams, and Stimulus controllers. He also traces these ideas back through StimulusReflex, Hotwire, Stimulus LSP, Turbo LSP, and Prism, explaining how those projects helped make ReactionView possible. David and Marco also discuss what increasingly capable AI coding agents mean for developer tooling. Rather than making tools such as Herb, RuboCop, and Prism less important, Marco argues that agents need strong feedback loops and validation tools to recognize incorrect output and follow established application conventions. Finally, they discuss RubyEvents and the value of making the history of the Ruby community more accessible, from lowering the barrier to open-source contributions to preserving and indexing years of conference talks that can inspire the next generation of Ruby developers. Sponsor Judoscale - https://judoscale.com/

What does the future of Ruby and Rails look like as AI begins to fundamentally change how software gets built? Recorded at Rails World, David sits down with longtime Rails community member Obie Fernandez to talk about the rapid rise of agentic development and what it means for programmers, teams, and the craft of software engineering. Obie argues that while AI can dramatically accelerate development, the industry is still figuring out how to apply the discipline, constraints, and guardrails necessary to use these tools effectively in production systems. The conversation explores how software development may be moving to a new level of abstraction—away from carefully crafting every line of source code and toward maintaining precise descriptions of a system's behavior, contracts, constraints, and intent. They discuss code archaeology, non-functional requirements, specification-driven development, regenerative software, and the possibility that generated source code could eventually become a disposable implementation detail. Along the way, Obie connects these ideas back to one of Rails' foundational principles: convention over configuration. In a world where AI gives developers an almost unlimited number of implementation choices, conventions and well-designed guardrails may become more valuable rather than less. The discussion ultimately asks whether the framework developers need for the agentic era might not be a replacement for Rails at all—but an evolution of the ideas Rails has represented for more than twenty years. Sponsor AppSignal: https://appsignal.com/rubyonrailspodcast