Elixir Wizards

← Elixir Wizards23 jul · 1 u 05 min

What Changed and What Remains True with James Gray

What Changed and What Remains True with James Gray23 jul1 u 05 min

In the final episode of Season 15, Charles Suggs and Emma Whamond are joined by James Gray, co-author of Designing Elixir Systems with OTP, to talk about what has changed in software development and what still holds true. Years after the book’s release, many of its core ideas around architecture, boundaries, supervision, and system design remain deeply relevant, even as AI tools reshape how developers learn, build, and collaborate.

James brings a unique perspective to the conversation: he is an experienced Elixir and OTP educator who only recently began using LLMs in his own workflow. He shares what surprised him, where the tools have been useful, where they have created risk, and what happened when Claude accidentally wiped his local development database. The conversation explores how AI can speed up parts of the work while making foundational knowledge, code review, testing, and clear system boundaries even more important.

To close the season, James and the hosts reflect on what developers should hold onto as the stack continues to shift. They discuss responsible LLM usage, the changing role of pair programming, who owns the mental model when AI helps write code, and why Elixir’s long-standing strengths may matter even more in an AI-assisted development world.

James is scheduled to speak at ElixirConf 2026, September 10–11 in Chicago, and the Elixir Wizards will be there too! Join us and the broader Elixir community, and use promo code Elixirwizards for 10% off in-person or virtual tickets at https://elixirconf.com/

Key topics discussed in this episode:

James Gray’s work on Designing Elixir Systems with OTP

What still holds true in Elixir system design

Layered architecture and durable software fundamentals

Functional core, imperative shell in modern applications

Supervision trees as application lifecycle maps

Why boundaries still matter in AI-assisted development

What AI changes about learning and building software