Reflection Quest — Part 1 of The Watchmaker's Dilemma
Tomás has his laptop angled toward me, but he's not showing me a slide deck or a sprint board or any of the artifacts that usually populate the space between two engineers at 2:30 on a Wednesday. He's showing me a GitHub repository he hasn't touched in four years. The commit history reads like a diary written in code — entries at 11 PM, at 2 AM, on weekends, on what I can only assume were holidays. No pull request templates. No CI pipeline. No architecture decision records. Just a person and a problem, in a room, for months.
"It's terrible," he says, scrolling through a function that runs to three hundred lines. "The error handling alone would get rejected in any code review."
He's right. By every metric we use to evaluate software — readability, maintainability, test coverage, separation of concerns — this codebase is a mess. And yet something about it has a quality I can't quite name—an aliveness. The code doesn't just solve the problem; it wrestles with it. You can feel the thinking happening in real time, the false starts left in as comments, the sudden leaps where an idea crystallized mid-function. It reads the way jazz sounds — technically imperfect, structurally loose, and unmistakably the work of someone who was completely inside what they were doing.
His current work is better by every measure. Cleaner. More considered. The kind of code that earns approving nods in architecture reviews. But it doesn't have that quality. That thing. Whatever it was that made a three-hundred-line function feel more alive than a perfectly factored microservice.
"I couldn't write this now," he says. Not with sadness, exactly. More like the way you'd describe a dream that was vivid at 4 AM and shapeless by breakfast.
I've been reading about the Swiss watch industry — not by design, but by the kind of accident that turns out not to be an accident at all. You follow one thread (a podcast mention, an essay someone shared) and it leads you somewhere that illuminates something you'd been circling without knowing it.
Between 1945 and 1970, the Swiss made watches with a quality that watchmakers now refer to, with undisguised longing, as the golden age. The best watches of this period were small, 32-33 millimeters across. The brand names printed on their dials were half a millimeter high. You'd need to be inches away to read them. The manufacturers didn't need you to read them. The work was the signal. Accuracy, thinness, the obsessive pursuit of doing one thing as well as it could possibly be done — these were the only competitions that mattered.
The most prestigious makers — Patek Philippe, Vacheron Constantin, Audemars Piguet — stood on two legs: prestige and performance. Their prestige was earned, mostly, by the quality of their engineering. Nobody was buying a name. They were buying a mechanism so refined that the name was almost beside the point.
And there's a detail that I keep turning over: because these watchmakers were all pursuing the same essential tradeoffs — accuracy and thinness — their best watches all looked remarkably similar. Minimalism converges. When everyone is reaching for the right answer, the answers start to resemble each other. There was no need for distinctive cases, for oversized crowns, for gratuitous protrusions that announced the maker from across the room. The name was tiny because it didn't need to be loud.
This is what a golden age looks like from the inside: people working on the essential problem, and the work speaking for itself.
I think about Tomás's repository. The one with no CI pipeline, no architecture decision records, no pull request templates. The one where the code was the whole conversation — between him and the problem, with no audience, no performance review, no LinkedIn post waiting on the other side. The mechanism, naked.
I think about my own golden age, though I didn't know to call it that while I was in it. The early writing, before the newsletter had readers — when every sentence was addressed to a question I was trying to answer, not an audience I was trying to reach. The first card magic routines I learned, practiced alone in my apartment, shuffling for hours with no intention of performing — just the private pleasure of making fifty-two pieces of coated paper behave like a single living thing. The early engineering work, where problems arrived without stakeholders attached, where the satisfaction lived entirely in the solving.
None of that work was my best. The writing was rougher. The magic was clumsier. The engineering was less elegant. But the attention was different. Undivided in a way that's harder to access now — not because I've lost the capacity, but because the conditions have changed. An audience arrived. And the audience changed what "good" means.
There's something the Greeks might have called sophrosyne in that early work — that quality of occupying oneself completely, of being sound in one's own mind. Not because the work was wise or finished, but because the relationship between the person and the work was unmediated. No signal to manage. No brand to maintain. Just the problem and whatever you could bring to it on a Tuesday at 11 PM.
Watch what happens when the audience appears.
It's not a single moment — not a corrupting event you can point to and say there, that's where it changed. It's more like a tide. The first time someone praises your work publicly, something shifts — a degree, maybe two. Not enough to notice. But the next time you sit down to write, or code, or practice, there's a faint new presence in the room. Not the problem anymore, not just the problem — but the echo of the praise, and the quiet question it carries: will this be as good as the last one?
The Swiss watchmakers felt this tide in the late 1960s. Japanese manufacturers caught up to them on accuracy. The exchange rate shifted, making Swiss watches expensive without making them better. And then quartz arrived — cheap, precise, thin — making the entire game of mechanical accuracy irrelevant. The essential tradeoff they'd spent decades mastering was suddenly obsolete. The thing they were best at no longer mattered.
The ones who survived did so by shifting from performance to prestige. They redesigned their cases to be recognizable — larger, distinctive, branded. Patek Philippe commissioned an iconic watch whose case expanded the surface area devoted to the maker's identity from 8 square millimeters to 800. The name went from a whisper to a shout. Not because they wanted to shout, but because whispering no longer worked. The mechanism — the thing that had been the whole point — became secondary to the signal.
I see this in the work around me, and I see it in myself. The moment when the code stops being a conversation with the problem and becomes a deliverable for review. The moment when the essay stops being an exploration and starts being content. The moment when the magic trick stops being practiced for the private pleasure of the choreography and starts being rehearsed for the performance. A tiny shift in orientation — from inward to outward — and the golden age ends. Not with a catastrophe. With an audience.
Tomás doesn't know he had a golden age. That's the thing about golden ages — as one essayist put it, they don't call them golden ages while they're happening. The participants take them for granted. They think this is just what work feels like: you encounter a problem, you give it your full attention, you struggle with it until something emerges. They don't know that this particular quality of attention — unperformed, unobserved, un-optimized — is the rarest condition of all. That it's already ending by the time they notice it's there.
"I couldn't write this now," Tomás said. And what I heard underneath was not a statement about skill — he's far more skilled now. It was a statement about access. About a door that closed so gradually he didn't hear the latch.
The golden age watches were 33 millimeters across. The brand names were half a millimeter high. You had to be close to read them. The watches now are 42 millimeters. The names shout from across the room. Better? More impressive, certainly. More successful, by every revenue metric. But the watchmakers themselves look back at that earlier period and call it golden. They know something was lost in the enlarging. They just can't quite say what — only that the watches were smaller then, and quieter, and somehow more themselves.
I think about that when I look at my own work. The pieces I'm proudest of aren't the ones that reached the most people. They're the ones written at 11 PM with no plan, no outline, no audience in mind — just a question that wouldn't let me sleep until I'd sat with it long enough to find its shape. The mechanism is not the case. The problem, not the performance.
The golden age doesn't end because you get worse. It ends because you get watched. And once you've watched, you can't unwatch yourself. The audience takes up residence in your attention, and the space it occupies is exactly the space where the undivided work used to live.
Maybe the question isn't how to get back to the golden age — you can't unmake the audience, can't unknow that people are reading, can't unlearn the metrics. Maybe the question is smaller and harder: can you still find thirty minutes where the door is closed, the phone is in the other room, and the only conversation happening is between you and the problem?
The afternoon light is doing that thing it does in glass-walled offices — catching the edge of Tomás's screen, turning his old code into something bright and unreadable. He closes the laptop. The repository goes dark. Four years of midnight commits, of wrestling, of attention so undivided it left its fingerprints in every messy, overlong, structurally imperfect function.
"I should probably archive it," he says.
I don't say anything. But I think: don't. Not because the code is good. Because the code is evidence. Evidence that you were once in a room with a problem and nothing else. Evidence that the mechanism was the whole watch. Evidence of what your attention sounds like when nobody's listening.
The golden age is small. Thirty-three millimeters. Half a millimeter of name. You have to get close to read it.
But it's still there.
If this stirred something — if you recognized the quality of work that happened before the audience arrived — send it to someone whose best code, or writing, or thinking, lives in a repository they haven't opened in years. The golden age doesn't expire. It just gets quieter.
