Ruby AI: Interview with Obie Fernandez

This week I had the privilege of interviewing Obie Fernandez, Principal Engineer at Shopify. He is the author of the book Patterns of Application Development Using AI, creator of Ruby AI frameworks Raix and Roast, founder of Hashrocket and Olympia AI, and drops sick beats on Soundcloud as an electronic dance music DJ. So here it is, presenting Obie on:

• Anthropic

• Patterns of Application Development Using AI

• Raix & Roast

• Future Ideas

• Closing Thoughts

Anthropic

You recently met with Anthropic to discuss improving their models’ understanding of Ruby and Rails. What can you tell us about that meeting and their plans going forward for the Ruby community?

Obie: Can’t tell you too much in detail due to non-disclosure agreements. However, I can say that the models are not struggling, and the effort I am trying to help with is simply to give additional base knowledge to the models about the full spectrum of Ruby and Rails codebases that are out there. For instance, I’m trying to get a multi-million line Rails 6 codebase accepted as training material. It has a ton of RSpec and StimulusReflex code in it, along with other more esoteric gems and libraries like AcidicJob, etc. All of this helps the new models work better with Ruby and Rails. Overall I love that Anthropic cares about the Ruby community as much as they do.

Patterns of Application Development Using AI

Last year you released Patterns of Application Development Using AI, a phenomenal Ruby-focused book about working with large language models and generative AI, and one of the best resources for getting started with Ruby and artificial intelligence. What do you feel has aged well about the book, and what hasn’t aged so well? If you had to write another chapter for the book, what do you think would be the most important topic right now?

Obie: I purposely wrote the book in as evergreen style as possible. There are some small details especially related to state-of-the-art model capabilities that are slightly out of date, but most of the rest of the book is still perfectly valid.The book doesn’t go into agentic stuff and that feels like it would merit a chapter in the second edition. It may even be the topic of a new book if I can find the time, which right now due to workload I do not have.

Raix & Roast

Raix just hit version 1.0, and Roast continues to get better every day. What is your favorite workflow you’ve built so far with Roast? What is your favorite Raix abstraction? What are the most exciting plans in the pipeline for these libraries?

Obie: Raix deserves to have more of the patterns from my book implemented in it, like the Predicate. Roast has a lot of attention internally at Shopify and I’ve seen some amazing workflows built for it, for example one that can automatically add and improve existing Sorbet type annotations to both production and test code. I also recently collaborated on a Roast workflow that can answer questions about lines of code based on their history, by digging into PR comment threads and commit messages, even across moves. That one is called “Chesterton” in honor of the famous fence.

As for the pipeline, there’s a sizable backlog of issues constituting the Roast roadmap at https://github.com/Shopify/roast/issues.

Future Ideas

Considering everything that Anthropic, Shopify, and the Ruby community are currently developing, what would you like to see built in AI software, developer tooling, and / or resources to support the Ruby AI ecosystem?

Obie: There’s a need for more evals for RL’ing models. It’s not a trivial undertaking at all, and I think individuals and companies currently have an interesting window of opportunity to step up and provide that kind of material to Anthropic, OpenAI and other interested parties. Other than that, I’d love for the Ruby community to lead the charge on new kinds of tooling like Roast that takes full advantage of all this cool new LLM-based technology. There’s tons more we can be doing to create the “Ruby on Rails” of this era, beyond just integrating with existing tech.

Closing Thoughts

What skill sets and disciplines do you think Rubyists should pursue to level up their careers and expertise in the age of AI? Is there anything else exciting you would like to share?

Obie: I don’t have much concrete advice other than to join the community of Ruby+AI enthusiasts in our Discord. There you will find lots of other intellectually stimulating people and discussions amongst those of us that are really pushing the envelope of what’s possible. It’s also one of my main sources of knowledge about new developments.

I want to thank Obie Fernandez for taking the time to answer these questions, and for all of his contributions to the Ruby and AI communities. If you haven’t already, be sure to check out all of the latest updates from Obie:

• Announcing Raix 1.0: Production-Ready AI Components for Ruby

• Roast v0.4.1 Released!

• Shopify Introduces Roast: Structured AI Workflows Made Easy

• Roasting Ruby AI Workflows on the Ruby AI Podcast

• Discussing AI, Rails, and pairing on the Distributed podcast

• SF Meetup Group Discussion on Structured AI Workflows at Scale with Roast

• Upcoming Talk at Euruko on Roasting Code for Fun & Profit with Structured AI Workflows

• Book on Patterns of Application Development Using AI

• Course on Patterns of Application Development Using AI

• New Progressive House and Trance Album “The Purest Light”