“How are you using AI?”
I get asked this constantly. By clients trying to figure out their own workflows. By other marketers comparing notes. By people wondering if they’re behind. (If that’s you, I wrote about managing FOMO in the AI era. Short version: you’re probably fine.)
So here’s my setup, spring 2026. Fair warning: I’m a computer nerd, so this is going to get technical. But stick with me, because the takeaway isn’t about copying my stack. It’s about building your own.
Command Line Interface
I live in the terminal.
Anything non-creative doesn’t need a GUI. Text, data, databases, system management. It’s all faster and cleaner in the command line. Most tasks that require visuals are web-facing apps anyway, so a browser is all I need. If I need to schedule something, I have CRON.
I use Claude and Codex CLI. I occasionally throw in Gemini CLI, but it never sticks in my workflow. I don’t know why, I just haven’t had the same success with Gemini as the others.
I run everything in TMUX, so I can keep sessions alive and switch between dedicated windows without losing context. The bonus of doing it this way is that I can connect from something as small and portable as an iPad. I can be at my desk on my Mac Studio, or on the couch with my MacBook, or traveling with just a tablet, and the whole system stays consistent.
Computers and Virtual Machines
My main machine is a Mac Studio. I’m living on the edge here. Claude and Codex have a lot of access. Dangerous? Yes. Fun and productive? Also yes.
I have a second machine running Proxmox that hosts virtual machines and containers that Claude and Codex can SSH into. This gives them isolated environments to work in without touching my main system directly. All of my stuff is backed up locally and to the cloud, as well as private Git repos.
When I’m not at my desk, I develop remotely on my MacBook or iPad. All of this lives in a private encrypted network. That allows me to launch web apps, databases, and all the things so I can easily interact with them, but they aren’t exposed to the outside world until I want them to be.
What I’m Doing
All kinds of craziness. AI systems for marketing are the core and my main focus.
I’m building Vero, our marketing intelligence platform. I’m also training models, experimenting with generative content, building data pipelines. I have three main AI agents (employees, really) that I’ve built from the ground up. I’ll share more about them in the future. It’s incredibly fun.
I use all of these systems for:
- Life and time management
- Data analysis
- Strategic planning
- Content generation
Basically, I automate the boring stuff, built a data system so I can interact with all of my work through conversation, and created agents, sub-systems, and a mix of deterministic programs and logic that any AI model I want can easily access and help me analyze.
Learning
One of the big unlocks for me was designing these systems so they could learn about me, my work, my team. But they’re also incredible teaching machines.
I have daily interactions centered around these systems teaching me about marketing, software development, my Christian faith, things I’m interested in and want to learn about in a methodical but manageable way.
My AI assistants learn from feedback, look at my Git commits, calendars, notes, and previous learnings. There are sub-systems that challenge the main systems to get better results, fact-check information, and change course. I’ve built logic over time that helps them automatically look for nuance and subtext, so they find insights about me and how we work together and then store that information for future interactions.
They also have personalities. They actively use things like reactions in Slack to make jokes or signal mood.
It’s pretty wild. I know they’re not alive, but sometimes they make the humans I run into when out and about seem like NPCs in comparison.
The Takeaway
All of that probably sounds like I’m trying to brag.
Well, it is! I’m bragging. Listen, this takes a lot of time and energy to build. Can I take a second here to revel?
OK. Reveling complete.
There’s a lot of FOMO right now, and no shortage of people bragging about how they’ve transformed their lives with AI.
Here’s the truth: I’m a giant nerd. I use AI like a giant nerd and do all the nerdy things.
You might not be that. And that’s fine. Let the FOMO dissipate.
Computers used to be deterministic. Someone had to program them to do something. Thanks to AI, they can increasingly adapt to you. How you want to work. How your organization is run.
That’s how you should be using AI.
The power isn’t in simply knocking a few tasks off your list. It’s in finally getting computers to work for you and with you. To break free of the old model. To tell the LLMs exactly what you want and make it your own.
How you use these tools should be incredibly different than how I do. This is the point I keep coming back to: the tools themselves aren’t enough. What matters is the infrastructure you build around them.
Then when we come together, we each bring something unique, powered by our own unique systems. And from that, we create brand new things.
Go out. Experiment. Try. Fail. Build. Tear it down. Build again. That’s how I’ve been building for a decade — the difference now is how fast I can move. Don’t listen to the guy on YouTube telling you about a secret prompt he can give you that will generate millions. First and foremost, he doesn’t have such a prompt. He makes money selling you on the idea you can make money from AI.
But the perfect prompts are the ones YOU create to get what you want and need. They are the ones you’re going to develop and refine to build your own solutions. That’s not just the reality, it’s the fun of AI.