Breakfast was taking 90 minutes. Four people, different dietary needs, nobody sure what was available, every morning starting with the same set of open questions. So I built defaults. Wrote them down. Monday looks like this, Tuesday looks like this. A set of decisions made once so they don’t have to get made again at 7 AM when nobody has the capacity to make them.

Then I was building this site and the same thing happened. The navigation needed a taxonomy, so I built a three-tier classification. The copy kept drifting, so I built a voice protocol with verification checklists. The AI kept contradicting its own earlier decisions, so I built a governance protocol and embedded it in the project root. Every problem became a structure, and every structure became a tool, and at some point I was looking at all of it and I thought: this is the same thing I did with breakfast. This is the same thing I always do.

I looked at the site. I looked at the kitchen. I looked at the system I built for the plants in the backyard. Same operation. I just hadn’t put them next to each other before.

Where it started

My father was a construction superintendent. He held the entire picture of a building while fifty trades worked their piece. Electricians, plumbers, ironworkers, framers. Each one responsible for their channel, none of them responsible for how it all came together. The whole picture only existed in his head. I watched that for years before I had any language for what I was seeing. Something about watching someone hold the whole while everyone else works the parts. I was learning it before I knew I was learning anything.

Screenprinting is where it locked in as a skill I could name. At SVA, then at USA Tees in Brooklyn, then at Atlas Embroidery in Fort Lauderdale. Separation means you never work on the image. You work on a channel, one color at a time, and the whole composition exists only in your head while you’re building a single screen. You decide how colors interact when they overprint before any of that is visible. Same operation my father was running, just a light table instead of a job site.

The screenprinting vocabulary (separation, registration, overprint, channel) became the vocabulary I use for everything. SCSS cascades. Function chains. Multi-agent workflows. You’re always working on a channel, and the whole only comes together when the layers register.

The naming happened while building this site

I knew I’d been doing this across different kinds of work for a long time. What I didn’t know, until I was trying to organize twenty-five years of it into a portfolio, was that it was one operation with different materials.

I had to sort everything into categories and I kept finding the same thing underneath. Encore, the enterprise platform I maintained for twelve years, is the clearest example. The software ran on a codebase that had accumulated decisions from multiple teams across a long time. Features would drift. Terminology would drift. The thing a module was supposed to do would stop matching what it actually did, and nobody caught it until something broke downstream. My job, for twelve years, was to keep the original intent legible through every handoff. I built the structure that made the software’s purpose holdable, so the people working on it later could find it without having to reconstruct it from scratch. That’s not engineering work in the usual sense. That’s load-bearing documentation and governance. That’s registration. Aiden Jae, the jewelry brand: I built the system that made the craft visible through a product page, so the quality didn’t collapse in transit from bench to browser. SavePoint Syntax: I built the markup that captures where thinking shifts before the context closes, because a session’s value drifts the same way a feature does if you don’t catch it.

Every one of those is the same job: take something fragile, build structure around it, and make that structure holdable by someone other than me.

What the structure protects

Every one of those starts with an intent. A tiny, specific thing someone is trying to do. The creative process is brutal on intent. It has to go through production, through other people’s hands, through time, through entropy. Most of the time the original intent doesn’t survive. Enough small decisions accumulate and nobody catches the drift.

At Encore, I watched it happen in real time. A feature would start with a clear purpose, pass through three teams and two review cycles, and arrive in production stripped of the thing that made it worth building. The structure I spent twelve years maintaining was specifically this: a way to keep the original intent legible through every handoff, so the thing that shipped still resembled the thing someone needed.

The structure holds the intent steady while you carry it through the build. My father was doing that on the job site. I was doing it at the light table. The breakfast defaults do it too, honestly. Protect the thing (a building, a composition, a functional morning) from the accumulation of unstructured decisions.

And it goes both directions. Craft-level attention applied to modern tooling. Modern technological approaches applied to handcraft. The construction superintendent and the screen printer and the software architect are all running the same process: decompose the whole into workable pieces, build each piece with fidelity to the original intent, and make sure the pieces register when they come back together.

What I know now

My father held the whole picture in his head, and that was the problem. When he left a job site, the picture left with him. Everything I’ve built since has been an attempt to solve that: externalize the picture so it doesn’t leave the room when the person who made it does.

So that’s why. My brain turns everything into a system because the alternative is a building that only stands while the superintendent is watching. The breakfast defaults, the voice protocol, the twelve years of platform governance. None of them are the work. They’re the part that lets someone else pick up where I left off, which is the only test I trust for whether the structure actually worked.