Jacob Luetzow runs Elixir Mentor, where he teaches Elixir through YouTube tutorials, a podcast, and a community. He’s also building Killswitch, a dead man’s switch for your digital assets. On his podcast, he’s had conversations with Jose Valim, Chris McCord, Sasa Juric, and other leaders in the Elixir ecosystem.
I sat down with Jacob to talk about his path into tech, why he started teaching Elixir, and how AI is changing the way developers build and learn.
From Civil Engineering to VP of Engineering
How did you end up in the Elixir world?
I’ve been around computers forever - my parents always had them, from the old orange terminal IBM with floppies to a Mac IIfx. I saved up for my first computer, built my own gaming PC in high school with my cousin before you could buy pre-built gaming rigs. I cut a blow hole in it for cooling and added cold cathode fluorescent tubes for lighting before LEDs were a thing.
I started at MSOE for Electrical Engineering and ended up graduating from UW-Milwaukee as a civil engineer. My family owned a successful business, so I worked there for years learning everything from cleaning bathrooms to building equipment to day-to-day office work. But I always had this itch to build my own thing.
Eventually I left to pursue startups. My first tech job was at Devslopes, a learn-to-code platform fresh off a successful Kickstarter. I started as a junior making like $1200 a month in California, living off savings. I was hired as an Android dev even though I was an iOS guy - I just said “deal, I’ll learn it.” Worked my way up to VP of Engineering. After that sprint I was completely burned out - I moved into the mountains and lived in an RV for a year because I hated computers.
Then I needed money again, took a senior iOS role at Earbuds (a social music app), and eventually left to build my own workout app called PerfectForm. A good friend kept telling me to check out Elixir for like a year and I ignored him. Finally gave it a shot because it was perfect for the fitness marketplace I was building, and I’ve been hooked ever since.
What made you fall in love with Elixir?
It’s the way to build software. The paradigms of Elixir force you to write better code, and if you use the tools properly you end up with a robust piece of software ready for the gauntlet of growth.
About a year into learning Elixir, I started the YouTube channel at ElixirMentor.com. I struggled going from mutable object-oriented programming languages to an immutable functional language, and there weren’t a lot of video tutorials out there. If you weren’t good at reading documentation, learning Elixir was hard - and that’s a bigger learning curve especially for new programmers. So I figured if I struggled, others probably were too. No one else was doing it well, and it was my way to give back to the community.
Was becoming an educator something you planned?
Totally organic. Zero goals when I started - I just did it. I’ve always found that doing it yourself gets you a lot further, faster. You’re usually slowed down at the beginning if you think someone else is going to help. One-man show all the way.
About two years into my channel, I was talking to a company founder who wanted to sponsor me and he said “you should do a podcast.” So I was like okay, let’s do it - a couple weeks later I started the Elixir Mentor podcast and he was my first guest. Now I’ve been fortunate to have conversations with much smarter people than myself. I learn so much and get to network with some of the best in the space.
Lessons From Jose Valim, Chris McCord, and Sasa Juric
What’s the most surprising insight you’ve gained from your podcast guests?
A few things really stuck with me:
From Jose, two things hit me. First, his perspective on staying in the loop as a developer. He said something I really agreed with: “I don’t buy what a lot of people are saying that the human is going to be fully out of the loop.” His whole insight around building Tidewave was about eliminating the boring loop - the copy-pasting, the navigating source code to find files - not removing the human entirely. You still need to understand what’s happening. Second, his philosophy on not setting specific goals really resonated. He thinks about the future as a road, not a destination. He doesn’t set expectations because then you end up managing expectations instead of doing the work, and you can get addicted to chasing goals. That mindset has helped me enjoy the journey more.
From Chris McCord, it was this idea that Elixir/Phoenix is an “accidentally purpose-built platform for writing AI agents.” Every problem Erlang solved in the 80s keeps mapping perfectly to new technological shifts - multicore, the internet, realtime messaging, and now AI agents. The primitives just keep working.
From Sasa Juric, what struck me was hearing about how he came from the Erlang world and immediately saw Elixir’s potential even at version 0.7 - just months after it was born. He called it “basically Erlang with some additions” - like “Erlang++” - but he was blown away by how mature the documentation and developer experience was even that early. It’s a reminder that sometimes you can recognize great foundations before the world catches on. His book Elixir in Action was actually the first Elixir book I ever opened, so talking to him on the podcast felt like coming full circle.
AI and Learning to Code
Do AI coding assistants make it easier or harder for beginners to truly learn programming?
I think if it’s used correctly, it can be a powerful tool. We used to have to read books, blogs, or YouTube to understand things. Now you can tell Claude to break it down and explain it to you. So I think if you’re willing to be a good engineer, you still can be.
Chris McCord said something I really liked: the folks that would have gone on to become amazing programmers are just going to have that gap shortened. AI helps them get there faster. But he also acknowledged the flip side - the copy-paste StackOverflow developers can now fake it a lot longer and get other team members into trouble.
Some of the best learning opportunities are painful. If you’re willing to actually learn from the LLM response, read the code, and understand it - you can still develop real skills. But if you just copy-paste without thinking, you’re not learning anything.
If AI can write code, what’s the point of teaching people to code?
Teaching is about the why, not just the what. When you create courses, you can’t just do something because it’s how you do it - you have to explain why you do something. That forces a deeper understanding.
You still need to understand the paradigms, the architecture, the trade-offs. AI is amazing at writing Elixir code, but someone needs to know if that code is good or bad. Someone needs to architect the system. The educator’s role is helping people develop that judgment and understanding - teaching them how to think about problems, not just how to solve specific ones.
What skills will become more valuable as AI handles more routine coding?
Seeing the larger picture and knowing how everything fits into place. You need to be a systems architect now. You need to sense bad code smell because it can still creep into your codebase - AI absolutely produces sketchy code sometimes - and you need to know what is good and bad.
The orchestration skills matter more than ever. I don’t write a ton of code anymore, but I find my knowledge orchestrates everything very well. You need to be able to review code, create the overall architecture plans, and catch the things AI misses.
Building Killswitch
What’s the origin story behind Killswitch?
Probably growing up watching Hackers and other whistleblower movies from the 90s. I always thought it would be cool to build a dead man’s switch platform that anyone could use without being a “hacker.” I’ve had the idea for years, and a recent conversation brought it back to life in my mind. I just had to build it.
You can check it out at killswitch.app - it’s a platform for securing your digital assets with dead man’s switch triggers, share links, and full zero-knowledge architecture.
What was the hardest technical challenge?
Learning how to truly handle zero-knowledge architecture. I researched how 1Password and Bitwarden do it, then figured out how to apply that to sharing files via share links and dead man triggers. It was difficult, but I’m really glad I dug in. Super happy with my implementation - you have full control to revoke any shares created individually.
Are you using AI to build Killswitch?
I use AI in my day-to-day workflow now. I’m always reviewing code and creating the overall architecture plans. I don’t write a ton of code anymore, but my knowledge orchestrates everything very well. The AI does a lot of the implementation, but I’m constantly reviewing, correcting, and steering the direction.
Balancing Content, Community, and a SaaS
How do you balance creating educational content, running a community, hosting a podcast, AND building a SaaS product?
I think most people like to say they don’t have time, but honestly I have even more time I can spend on things. I do enjoy some down time - I try to get to the gym 4-5 times a week, and I make sure I get out for skiing, mountain biking, rock climbing, canyoneering, and backpacking.
But here’s the thing: you’re choosing to binge Netflix instead of building your dream product. You do have the time.
In my “free time”? It’s Elixir Mentor, Review Pro, and Killswitch.
My advice: do it yourself. One-man show gets you further faster at the beginning. You’re usually slowed down if you’re waiting on someone else. Also - life kicks you in the nuts. Everyone usually thinks life is perfect because they only see the filtered version on social media. Hard things hit you and you still gotta push through. You can’t pause your life.
With AI now, you are capable of building a full product pretty quickly as a solo team. That wasn’t true even two years ago.
The Future of Phoenix and Learning
What excites you most about where Phoenix is heading?
Phoenix 1.8 introduced agents.md and a lot of tools to help integrate LLMs and agents. What excites me most is what Chris called Elixir being an “accidentally purpose-built platform for writing AI agents.”
The problems Erlang solved in the 80s for telephone switches just keep mapping to every new technological shift - multicore, the internet, realtime messaging, and now AI agents. The concurrency model, the OTP behaviors, the fault tolerance - all of it maps perfectly to building AI agent systems. Phoenix.new is a great example - you can write a Phoenix app in the browser, build multiplayer games, realtime chat apps, whatever - without even having Elixir installed locally.
Five years from now, how will people learn to code?
I have no idea, honestly. Wild times.
What I do know is that the barrier to entry is getting lower and lower. You can ask Claude to explain something to you at whatever level you need, which is something a static YouTube video can’t do. But YouTube and video content has that human connection, the personality, the context that pure AI tutoring doesn’t have.
I think both will exist, but the role of each might shift. Educators might become more about curating learning paths and providing the “why” context, while AI handles the “explain this specific thing to me” on demand.
Advice for Developers in the AI Era
Learn how to do everything. Don’t shy away from tools like AI because you think you’re too good of an engineer for it. You don’t need to use it the same way as others, but it definitely makes you a more efficient engineer.
And here’s the thing - you are now capable of building a full product pretty quickly with a solo team. That’s insane. Take advantage of it.
The more you know, the more you realize you don’t know. Imposter syndrome never goes away - you just have to ignore it and keep moving. You’re forever a student.
Jacob teaches Elixir through tutorials, a podcast, and a community at ElixirMentor.com. He’s also building Killswitch, a dead man’s switch for your digital assets. If you’re exploring Elixir or other programming languages, CourseShelf can help you find the best courses based on real reviews from learners.