What Gets Passed Down: Systems, Platforms, and the Architecture of Thought
The Medium Changes. The Constraint Holds.
There’s a problem that doesn’t change across systems, platforms, and mediums.
My grandfather built a dealership. My father built a direct mail business to serve dealerships like it. And I’ve spent forty-plus years building systems across contexts neither of them would fully recognize–and kept encountering the same problem in each one.
The problem isn’t new. It isn’t unique to any one field, or to me. But forty-plus years of lived data on the same constraint, across genuinely different contexts, produces a different kind of clarity than reading about it does. That’s what this publication is built on. This is where it came from.
Malcolm Wright Buick Oldsmobile GMC was vast and cavernous in the way that only commercial spaces are to a child–mechanic bays stretching back into the dark, tools hung in specific places for specific reasons, the whole operation running on its own logic.
My grandfather had worked on the Manhattan Project before he built any of this. He understood systems at a scale most people never encounter. What he built on East Lincoln Highway was smaller–but it ran on the same principle: you design it so it holds without you, because you can’t be everywhere at once, and the work can’t stop when you’re not there.
During World War II,
my grandfather worked for DuPont on the Manhattan Project.
I was already drawing by then, had been since I could hold a pencil–Brady’s Village art classes, then Wyomissing Institute, mixed media and traditional fine art, before I had language for either phrase. The dealership added something different: my first real exposure to industrial-commercial space, to design at scale, to systems built for automation rather than supervision.
My father absorbed that orientation and then went on to apply it elsewhere–first to building houses, then to a direct-mail business serving the same automotive industry he’d grown up in. He named it Front Line Ready.
I was six. My job was execution: assemble the mailers, fold them correctly, and make sure the address showed through the window. The copy had already been written. The recipients had already been identified. I was the last human in the chain before the envelope hit the mailbox–inside a system designed to reach people my father might never meet.
That was my first introduction to the constraint. It wouldn’t be my last.
The years that followed were dense in a way that’s hard to compress without losing what made them formative.
The parallel tracks started earlier than JAM and under harder conditions. My mother raised six kids as a single parent while working three jobs and putting herself through school to become a public school teacher–not sequentially, not with a safety net. The capacity to hold multiple things at once without any of them collapsing wasn't an instinct I developed at JAM. It was something I watched being done at home, under conditions that made everything that followed look well-resourced by comparison.
From sixteen to nineteen, I was running three tracks simultaneously. JAM Screen Printing was a commercial and advertising art apprenticeship–hand-drawing and cutting art on a light table, shooting and burning screens, and learning early Adobe tools. The parents of people I’d known since I was six. Drawing, which had been a practice since Brady’s Village and Wyomissing, became a production discipline here: a burned screen is committed. You either got the specification right before you burned it, or you started over. Zero tolerance for ambiguity at the output stage, enforced by the medium itself.
“A burned screen is committed.
You either got the specification right before you burned it, or you started over.”
At the same time, through Upattinas School, I was designing and executing my own curriculum–a largely self-curated two-year program that fed into my final high school credits, running alongside the JAM apprenticeship and the professional work that followed. The feedback loop was immediate and personal. When the curriculum was underspecified, I felt it as the executor before I could diagnose it as the designer.
And from sixteen through twenty-two, I worked for the same employer and mentor across two companies running simultaneously–Valley Forge Outdoors, a tri-state outfitter running guided eco-tours and canoe and kayak trips on the Schuylkill, and Mainstreme Productions, a seasonal haunted house operation that grew into something much larger. We developed and operated Terror Behind the Walls at Eastern State Penitentiary, and later at Fort Mifflin. Multiple sites, rotating staff, live performers, thousands of visitors moving through on any given night.
Both had to perform without anyone controlling every variable in real time. A guided full-moon canoe trip has to go on without the person who designed it managing every moment on the water. A haunted attraction at Eastern State has to run the same way on opening night and closing night, regardless of who’s there and who isn’t. You design the experience so it holds. You train the people inside it so they can execute without you having to narrate every step. Then you let it run and see what breaks.
What breaks is always the underspecified part.
What breaks is always the underspecified part.
Photography entered at eighteen–another practice that’s run continuously since, with its own arc of solo shows, commercial work, and a return to active exhibition in recent years. At nineteen, at The Bradley Academy, it produced my first solo show: a month-long mixed-media gallery event that brought together the drawing, photography, and digital arts I’d been building across all those simultaneous tracks. Bradley is also where the formal curriculum–drawing, illustration, color theory, desktop publishing, motion design–gave shape to instincts I’d been developing since Brady’s Village. Four-point-oh. The constraint wasn’t just present in the work. It was the grading standard.
At twenty-two, I moved to California.
A freelance period gave way to managing RedRox–a patented LED mobile rock-climbing gym on Melrose Avenue, where I lived and worked inside the operation for three months. Designer and resident simultaneously. Then a year at a European boutique record promotions label, tracking press and radio placement nationally, managing their databases in FileMaker Pro, learning what it means to move a product through systems you don’t own or control.
Then, at twenty-four, I started working with the people who would later build Heroku, as the first employee of what became one of the first internet fabric stores. Heroku is a platform designed so that applications run without their authors present. The absent-instructor problem at infrastructure scale, before it had a product name. I didn’t know that’s what I was watching being built.
Heroku, founded in 2007 by Adam Wiggins, Orion Henry, and James Lindenbaum, was one of the first cloud platforms built so that applications run without their authors present. Developers could deploy code to the cloud with a single command — no servers to configure, no infrastructure to manage. The idea: a system that holds without you. Salesforce acquired it in 2010 for $212 million.
The same year, I was an early founding member of The Do Lab–an events and festival organization that had to run as a system, not as a person. Coachella. Lightning in a Bottle. Burning Man installations. Experiences designed to hold without any single hand on them, at scales that made Eastern State look intimate.
The Do Lab produces Lightning in a Bottle and has staged installations at
Burning Man and Coachella since 2004.
Santa Monica College ran through this period too–working towards a BS in computer science with a photography minor, the formal credential catching up to a practice already several years deep.
The signal came from two directions at once. Amazon acquired our earliest competitor in the fabric space, an early indicator of what was coming for the whole category. And the founders I’d been working alongside had stepped away for Y Combinator to launch what would become Heroku. The market was reorganizing around systems at a scale we couldn’t match, and the people who’d been building alongside me were building infrastructure at an entirely different order of magnitude.
I left the fabric company for a formal apprenticeship under an information architect, a year of on-the-job training that named and structured what I’d been building by instinct for two decades. Taxonomy. Ontology. The discipline of making implicit structure explicit. The largest e-commerce system was dominating the market; I delved deeper into the discipline of making systems legible rather than trying to compete with one that had already won.
That year led directly to Disney: the first responsive web design for Spoonful.com, one of the earliest consumer websites designed responsively. Responsive design is the constraint formalized as a discipline: systems that adapt to executors you can’t predict, in contexts you won’t be present in.
Disney offered a permanent role. I declined to pursue an IxD degree at Art Center instead–one of six accepted into the inaugural Interaction Design class on a full scholarship. The program ran from pure IA and UX theory through front-end development, visual communications, hand-drawing in multiple perspectives, 3D fabrication, and prototyping. Four-point-oh. The drawing practice that started at Brady’s Village was formalized thirty years later.
Finished in 2013. Moved back to Philadelphia.
EPAM was a globally distributed professional services firm operating across dozens of time zones, disciplines, and client accounts at once–a monstrous system in its own right, and one I hadn’t designed. Two years of enterprise client work across complex distributed teams refined the craft in ways smaller contexts couldn’t. Helping to manage while working alongside several Drexel UX interns planted something else: the first real test of whether the instinct could be taught, not just applied. Could you design an environment where someone else developed the same diagnostic reflex without you standing over them, narrating every step?
It turned out you could. But the instructions had to be right.
Two years later, I started teaching and developing four online UX courses for UCLA Extension’s certificate program from 2015 to 2020. 267 students across more than 1,200 combined instructional and office hours. Async stripped out every real-time correction mechanism I’d relied on. No raised hands. No reading the room. No performer breaking character to ask what comes next. Students made their own interpretations, not because they were wrong, but because the instructions permitted multiple readings.
The feedback loop stretched from days to weeks and went from personal to secondhand. I couldn’t feel the ambiguity anymore. I could only see its effects downstream–in the work students submitted, in the questions that revealed what the instructions hadn’t actually said.
That’s when I started getting precise about failure modes. Not “this didn’t work” but “this is the category of thing that doesn’t work, here’s the mechanism, here’s what the instruction looked like that permitted the wrong interpretation.” Taxonomy as a discipline. Classification before correction. The same instinct from the screen printing bench–a burned screen is committed–applied then to instructional design at a five-year scale.
Then the executor changed again, and everything got faster.
Four-plus years ago, I started building agentic skills: structured instruction sets that direct AI agents through complex, multi-step workflows. The absent-instructor problem came back in a form I recognized immediately, but with one difference that changes everything: an AI agent cannot raise its hand.
A confused student pauses, infers reasonably, and finds a defensible path through ambiguity. A haunted house performer breaks character and asks. A confused agent produces fluent, confident output and keeps going. The tolerance for underspecification that a human executor absorbs–the gap-filling that happens naturally when someone can ask a question–doesn’t transfer. The design discipline does. The margin doesn’t.
A confused agent produces fluent, confident output and keeps going…
The failure modes from every previous medium reappeared with a shorter lag and higher stakes. The taxonomy built to diagnose them turned out to be the thing that transferred most cleanly. You still have to name the type of failure before you can close it.
The Al Ries and Laura Ries's The 22 Immutable Laws of Branding arrived early in the apprenticeship. I formed my first company not long after and called it Systems of Thought–my first legal corporate identity, built in Philadelphia before I ever moved west. The name has carried forward ever since, through every medium and every executor type that followed. Today, the drawing is still running alongside it–sketching agentic workflows, returning to illustration, and the photography is still active. The writing, cultivated slowly across all of it, is what this publication is now built on. Same name. Forty-plus years of data later. My grandfather’s questioning, at a different scale.
The company was called Front Line Ready. My father meant it about the mailers. I’ve spent the next forty years figuring out what it means about everything else.
The next piece in this series applies that lens to a problem at a different scale–one where the constraint isn’t about a single system performing without its designer, but about democratic governance performing without the four conditions it requires: the authority to bind, the speed to keep pace, the coalition to hold, and the infrastructure to see clearly enough to act.
Systems of Thought is a publication about the architecture of collective thought: the cognitive, institutional, and informational systems through which societies and organizations reason, govern, and lose the capacity to do both. This is where the forty-plus years of data behind it came from.
loved reading this! very unique story and start
Love this!