Every field has words for what it does. Carpentry has plumb, true, flush. Music has groove, pocket, feel. Print production has registration, bleed, trap. The words aren’t jargon. They’re handles. You grab them when you need to point at something specific and say: that. Right there. That’s the thing.

I work across enough materials that I needed my own handles. Seven words keep showing up. They showed up in the print shop, in the classroom, in enterprise software, in AI system design. I didn’t choose them. They chose me, by being the only words that fit the thing I was pointing at.

Here they are. Not as definitions. As the rooms they came from.

Registration

A print shop in Brooklyn, 2003. Four plates on a press: cyan, magenta, yellow, black. Each one prints independently. Each one is a separate pass through the machine. If the plates are aligned, you get an image. If one plate is off by a fraction of a millimeter, you get a blur. The image exists in none of the plates. It exists in their alignment.

I kept finding registration everywhere I went. At Encore, I was running three independent layers: information architecture, front-end interface, back-end logic. None of them could see the others. The product existed in their alignment, or it didn’t exist at all. In the classroom, twelve students with twelve IEPs, each one an independent layer of accommodation. The class worked when those layers registered against each other. When they didn’t, the room fell apart.

The layers don’t merge. They align.

Decomposition

A construction site on the Grand Concourse, sometime in the mid-eighties. I was a kid. The building had its skin off. You could see the plumbing, the electrical, the framing, the ductwork. All running through the same walls but belonging to completely different systems. Somebody had to figure out where each one went before the walls went back up.

That image never left. When I write an AI skill, it does one thing. One instruction, one objective. When I build a curriculum unit, each lesson targets one capability. When I scope a product feature, each component handles one responsibility. The instinct is always the same: find the seams where the compound thing separates into parts that can each be built, tested, and understood on their own.

Decomposition is breaking a whole into parts that can each stand independently. Structural separation.

Scaffolding

A classroom in the Bronx, 2007. A student who couldn’t organize a five-paragraph essay. Not because he didn’t have ideas. Because the gap between having ideas and producing a structured document was too wide to cross in one step.

So I built a graphic organizer. Boxes. Arrows. Sentence starters. The scaffolding held his ideas in position while he figured out where they went. Six weeks later, he stopped using the organizer. He didn’t need it. The capability was his now, and the temporary structure had done its job by becoming unnecessary.

That’s the test. If the scaffold stays permanently, you’ve built a dependency, not a capability. The goal is always the same: meet the person where they are, hold the weight they can’t hold yet, and come off when they can.

Fidelity

Encore, 2004. I built a content management platform for the entertainment industry. Twelve years later, it was still running. Same architecture. Same information design. Same structural logic. Three complete technology shifts happened underneath it, and the intent survived all of them.

Most projects lose fidelity in months. Not because anyone makes a bad decision, but because the gap between what was meant and what gets built widens with every handoff, every new developer, every quarterly pivot. Fidelity is that gap. Small gap, high fidelity. The original intent survives. Large gap, low fidelity. You end up with something nobody designed.

Twelve years is not normal. I know that. But it proved something: fidelity is holdable. It takes specific structural choices, and it takes someone who refuses to let the standard slide. Which leads to the next word.

Drift

Encore again. Different angle.

Six months into any project, the output starts degrading. Nobody changed anything. Nobody made a bad call. But somewhere around month three, the original intent stopped being the thing everyone was checking against. Good-enough decisions accumulated into something nobody chose.

This is deeper than cutting corners. When you cut corners, you know the standard and you’re choosing not to meet it. Drift is when the standard itself goes invisible. You can’t violate something you’ve stopped seeing.

I grew up in the Bronx in the eighties. I watched it happen to entire neighborhoods. One broken window stays broken. Then two. Then the standard for what “maintained” means shifts without anyone noticing. Nobody decided to let the block fall apart. The standard just went invisible, one reasonable decision at a time.

Convergence

Five people evaluating the same student work in an SVA critique room. Each one is looking through a different lens: typography, concept, craft, communication, cultural awareness. They don’t coordinate. They don’t discuss beforehand. They evaluate independently.

Where three or four of them point at the same thing, that’s signal. Where they disagree, that’s where the interesting decisions live. The agreement isn’t consensus. It’s convergence. Independent evaluations arriving at the same place without being steered there.

I built this principle into how I evaluate anything. Run multiple lenses. Don’t average them. Look at where they agree on their own. The points of convergence are the most reliable data you’ll get. The points of divergence are where the maker has to step in and decide.

Attunement

Three thousand people on a dance floor. One DJ.

The DJ who plays the right track at the right moment isn’t guessing. They’re reading. Energy in the room, where the crowd is in the arc of the night, which section of the floor is locked in and which is drifting. They see something most people in the room can’t see, because they’re trained to read the system before they act on it.

A teacher does the same thing. You scan the room. One student is checked out. Another is about to lose the thread. A third is ready for more than you’re giving. You read all of that before you open your mouth. The response you design comes from what you noticed, not from the lesson plan.

The system might be a person, a room, a codebase, a product. You read first. Then you build.

Seven words. Seven rooms. They came from print shops, construction sites, classrooms, nightclubs, enterprise software, and design critiques. The materials were never the same twice. The operations were.