Elixir Mentor

George Millo — creator of Learn Phoenix Live View and alumnus of the Gauntlet AI fellowship — joins me to talk through what really changes when you commit to LLM-driven development. George went through Gauntlet's intensive 10-week program built around LLM maximalism, came out the other side rethinking how he builds software, and has spent the past year working on an AI-first engineering team applying those lessons in production.
We spend a lot of time on the practical realities of agentic coding: verification debt (the gap that grows between your mental model and what the AI actually built), the importance of planning before prompting, why George works in small self-contained PRs, and how he uses Codex and Claude Code in parallel tabs without losing track of what's happening. We also get into the debate around vibe coding, spec-driven development, testing pitfalls, and why deep technical knowledge matters more now — not less — when AI is writing most of the code.
The conversation covers where Elixir fits in an AI-first world: the BEAM's process model as a natural fit for agent architectures, Phoenix shipping with an agent.md file, Tidewave's approach to closing the feedback loop, and why Elixir's tooling consistency puts it ahead of the JavaScript fragmentation George deals with at his day job. We also get into the security risks that come with AI-assisted development — giving LLM tools codebase access, the OpenClaw skills marketplace vulnerabilities, and the kinds of security mistakes that are now much easier to ship without noticing.
George closes with practical advice for anyone hesitant to adopt agentic workflows: stay curious, ask the AI to explain the code it writes, build something outside your comfort zone, and put in the reps.
Resources Mentioned:
- Learn Phoenix Live View: https://learnphoenixliveview.com
Connect with George:
- X/Twitter: x.com/georgemillo
- LinkedIn: linkedin.com/in/millog
Sponsors:
- Paraxial.io — Elixir-first application security: paraxial.io
- Jido — Elixir AI Collective Discord: agentjido.xyz/discord
SUPPORT ELIXIR MENTOR
- Elixir Mentor: elixirmentor.com

Leandro Pereira is back on the Elixir Mentor Podcast — this time to dive deep into MDEx, his Rust-powered Markdown library for Elixir. MDEx is built on the Comrak Rust crate, runs 31x faster, and uses 3,500x less memory than existing Elixir alternatives. We also get into Lumis, his standalone syntax highlighting engine powered by tree-sitter and Neovim themes.
Leandro walks through why he chose a Rust NIF over a pure Elixir implementation, what it took to ship Lumis as its own project, and the surprisingly hard technical challenge at the heart of MDEx: streaming Markdown for AI applications. We discuss how MDEx handles incomplete Markdown fragments in real time, what the upcoming Components feature unlocks for Phoenix/LiveView developers, and how the HEEx parser integration works under the hood.
We also cover the human side of maintaining two solo open source projects: how Leandro prioritizes, uses AI to chip away at the backlog, and thinks about monetization. The conversation goes deeper into how the AI era has changed Markdown's role in the ecosystem, the pitfalls of vibe coding, and what it really takes to get an open source project noticed — including the uncomfortable truth that marketing matters more than most developers want to admit.
The episode closes with a wide-ranging conversation on developer growth — the Dunning-Krueger curve, making the mental shift from OOP to functional programming, and why Elixir feels easier once it finally clicks. A great listen for anyone building libraries, wrestling with Rust NIFs, or navigating open source in the Elixir ecosystem.
Resources Mentioned:
- MDEx: https://mdelixir.dev
- MDEx on Hex: https://hex.pm/packages/mdex
- Lumis: https://lumis.sh
- Lumis on Hex: https://hex.pm/packages/lumis
Connect with Leandro:
- X/Twitter: https://x.com/leandrocesquini
- GitHub: https://github.com/leandrocp/mdex
SPONSORS
- Paraxial.io: https://paraxial.io
- Jido (Elixir AI Collective Discord): https://agentjido.xyz/discord
SUPPORT ELIXIR MENTOR
- Elixir Mentor: https://elixirmentor.com

Josh Price, founder & CTO of Alembic and one of the core forces behind Ash Framework, joins me for a wide-ranging conversation that starts with the origin story of Alembic, winds through the history of GraphQL and Ash, and lands on Clarity — his new interactive introspection and visualization tool for understanding your Ash codebase. Josh has been writing Elixir for over ten years and building Alembic for nine, which gives him a rare perspective on how the ecosystem has matured and where it's headed in an agentic world.
We trace how Josh's frustration with real-time data at a gaming company pointed him toward Elixir and Erlang, how that led to an obsession with GraphQL domain modeling, and how that obsession eventually collided with Ash — which turned out to solve exactly the problems he'd been hacking around for years. We talk about what Ash actually is beyond an API generator, why auto-generated migrations are criminally underrated, and why the developers who resist Ash most are often the ones in the middle of the experience curve. Josh also shares the inside story of how slowing Zack Daniel down was actually the best thing that ever happened to the Ash ecosystem.
A big chunk of the conversation covers the AI moment we're in right now — Claude Code workflow tips (including the /insights command and how to keep session history beyond 30 days), why CLIs are beating MCPs for LLM tool use, Claude Code skills and usage rules for progressive disclosure, and how Clarity grew from Ash's built-in introspection into something far more interesting: an in-memory Erlang digraph knowledge graph of your entire Elixir application. Josh also shares his take on multi-model databases, the disappearance of the UI, and why the only limits left for software engineers are taste, judgment, and imagination.
Resources mentioned in this episode:
- Alembic: https://alembic.com.au
- Clarity (Hex): https://hex.pm/packages/clarity
- Ash Framework: https://ash-hq.org
- Ash Framework book: https://pragprog.com/titles/ldash/ash-framework/
- Killswitch: https://killswitch.app
Connect with Josh:
- Website: https://alembic.com.au
- X/Twitter: https://x.com/joshprice
- LinkedIn: https://www.linkedin.com/in/joshcprice
Sponsors:
- Paraxial.io: https://paraxial.io
- Jido: https://agentjido.xyz/discord
SUPPORT ELIXIR MENTOR
- Elixir Mentor: https://elixirmentor.com

Thomas Athanas, Head of Engineering at LevelAll, joins me on the Elixir Mentor Podcast to talk through the infrastructure, architectural, and leadership decisions that come with building systems you actually own — and what happens when vendor lock-in catches up with you in production.
Thomas walks through LevelAll's move away from Fly and Gigalixir toward bare metal hardware, the thundering herd problem that comes with serving 50,000 concurrent education users, and why they made the call to remove both Phoenix LiveView and Ash framework from production. We get into Ash APS premium support, JSONB query challenges, and the tradeoffs of leaning on a framework when hiring for it is hard.
We talk about using AI as a development planning tool and context keeper for managers — including Thomas's "Lore Master" concept, where AI agents preserve institutional knowledge so it never walks out the door. From there we get into the Auth0 rate limiting incident that hit during a live onboarding, the FusionAuth migration that followed, enterprise auth requirements like OIDC and SAML, and the bcrypt hash conversion work that made it all possible. Thomas also shares his work on a custom Erlang-based bare metal deployment agent and his approach to Postgres configuration and backups with pgBackRest.
The second half covers founder mode mentality, total ownership of problems, the player-coach leadership style, Sanity CMS vendor lock-in, building an audience vs. building a customer base, and practical advice for technical founders who keep procrastinating with features instead of making sales.
Connect with Thomas:
- X/Twitter: https://x.com/ThomasAthanas
- LinkedIn: https://www.linkedin.com/in/thomasathanas/
Sponsors:
- BEAMOps: https://beamops.co.uk
- Paraxial.io: https://paraxial.io
- Jido: https://agentjido.xyz/discord
SUPPORT ELIXIR MENTOR
- Elixir Mentor: https://elixirmentor.com/?utm_source=elixir-mentor

In this episode of the Elixir Mentor Podcast, I sit down with Amos King, Senior Staff Backend Engineer at Adobe, founder of Binary Noggin, and longtime Elixir community contributor. We dig into mentorship, knowledge sharing, and the team dynamics that make software organizations actually thrive.
Amos traces his non-traditional path into software — from structural engineering to manufacturing automation to Erlang on Navy submarines — and explains how that background shapes how he thinks about building reliable systems. We talk about his decade running Binary Noggin, why he ultimately made the move to Adobe, and the hard lessons learned when the consulting market shifted. From there the conversation goes deep on team composition, why diverse backgrounds matter more than uniform credentials, and the mindset shift from object-oriented to functional programming.
We also get into the practical side of Elixir: when GenServers are the right tool and when they're not, why vibe coding worries him from an engineering quality standpoint, and why teaching is actually a selfish act that benefits the teacher most. We close out with what separates a real staff engineer from a senior one, a call for the Elixir community to revive local meetups, and a real-world database query optimization story that reframes how to think about performance problems.
Resources Mentioned:
- Binary Noggin: https://binarynoggin.com
Connect with Amos:
- X/Twitter: https://x.com/Adkron
- LinkedIn: https://www.linkedin.com/in/amosking/
SPONSORS
- BEAMOps: https://beamops.co.uk
- Paraxial.io: https://paraxial.io
- Jido: https://agentjido.xyz/discord
SUPPORT ELIXIR MENTOR
- Elixir Mentor: https://elixirmentor.com