I've spent 20 years delivering software and leading the teams that build it. Today I deliver with teams of AI agents, and I'm having more fun than I've had in my whole career. Here's how I got here.
After 20 years of software development, I was at a job where I loved the people I worked with, but I didn't love the technology. My days were spent pushing through legacy third-party platforms built decades ago, with layers of tech debt stacked on top.
Then I got real exposure to agentic software practices, and I knew instantly: this is going to change everything.
I worked hard days, then spent my nights exploring how far these new agent tools could go before they broke. Each night I was a little more impressed with where things were.
Then I'd bring what I learned to work and watch AI solutions fail the same way humans do. I was having tremendous success at home, and struggling to deliver the same outcomes at my day job.
AI fails the same way people do: on bad code, bad processes, and missing requirements, documentation and standards.
I was fortunate to be at a stage of life where I could step away. I could recharge with my family, and learn everything I could about this new era of product delivery. So I left my employer to build something of my own.
Like every software project I've been part of, it would either succeed, or I'd learn from the failure, pivot and repeat.
For the past few months I've been building Upwardly Math, a math practice platform inspired by real nights and weekends helping my three boys through their own math struggles.
I deliberately built it with technologies I had never used, to prove that the underlying code isn't what matters.
It also pulled me out of my development comfort zone. I learned how hard marketing is, and what it takes to run a business and everything that comes with it.
In a rare free moment, I remembered telling myself at eight years old that I wanted to make video games when I grew up. I never pursued it. I know my limits.
But after this year with Upwardly Math, I realized that today it's all about ideas, process and diligence, not know-how. Agents can write the code. Agents can create art that far exceeds my stick figures. I've played games my whole life, and I know the kind of game I'd make.
So I founded Late Light Games and started building Every King Dies. It's a completely different experience from my 20-year career, and yet game development problems and corporate line-of-business problems have more in common than you'd think.
Between Upwardly Math and the game, I found my limits. Juggling multiple projects and active threads across Claude, Codex and even Grok, I was the bottleneck. That's the same signal I look for in every team I've led when figuring out how to make it work better.
So, following a lot of talented developers exploring agentic workflows, I started a third project: my own Software Factory.
Building a game was fun. This was the part of my career I loved most: building and tuning the processes and tools that make delivery better for a team. It turns out the same processes, workflows and routines help agents finish work just as much as they help people.
I'm a software engineering manager again. The only difference is that my team is a crew of agents, checking in every few seconds for the next card on the board. Each takes the card back to its desk, does the work, and moves it to the next column for the next agent.
A year ago I was in a mid-life crisis, asking myself whether I still had a passion for this kind of work. I'd cornered myself into technology I couldn't find fulfillment in, during seven years at an otherwise great company. It was a dark place: uncertainty, long hours and a constant uphill climb.
Today I've found myself again, and I realized I really do love this sh*t.
Today's tools let me focus on the parts of product delivery I love most and skip the part that was never truly my passion: the code. I'm building things people love, appreciate and even pay for. I work long days and nights, and a twelve-hour day feels like six.
Back in the saddle. Let's go.