Just the Road: Notes on perspective in an industry being rebuilt under our feet.
My dad lives in southern Alabama near the ocean and I stay a few miles away near a beautiful community set on a golf course. I go every…
Just the Road: Notes on perspective in an industry being rebuilt under our feet.
My dad lives in southern Alabama near the ocean and I stay a few miles away near a beautiful community set on a golf course. I go every quarter, which means I see this neighborhood in discrete snapshots rather than as a continuous reel. The magnolias. The fairways cut to a ridiculous shade of green. The houses lay out on the lots as if someone arranged them from a catalog. Every visit, I notice it all over again.
I was on mile two of a (two mile) Saturday morning run when I passed the same man I pass every quarter. He walks the route early, before the heat lands. We’d nodded at each other maybe eight or nine times over the years, the quiet recognition between two people who share a road a few weekends a year and otherwise know nothing about each other.
That morning I stopped him and told him the place was beautiful, that I ran here every few months and never got tired of it.
He nodded slowly. Said he sometimes has to make himself step back and actually recognize what he has. That if he doesn’t, it all just becomes the road.
I’ve been thinking about that exchange.
Going blind to it
The thing that got me was how honest he was about the work of it. He was describing effort. The neighborhood looked the same as it always had; his ability to notice it didn’t, and pulling that ability back was its own job.
This is what sustained immersion does. It pushes the scenery into the background, and once it’s there, it stops registering at all. You can live inside something striking long enough that you stop seeing it.
This is just how brains work. Novelty gets the spotlight, repetition gets filed away, so you can function. If your brain lit up every time you saw your own living room, you’d never get anything done. The same machinery that keeps you functional strips the wonder out of whatever you live inside long enough.
The dangerous part is that the process leaves no trace. You walk past the magnolias and don’t see them, and you don’t notice you didn’t see them.
Perspective as a choice
Years ago, I was in New York for work, staying at the DoubleTree in the Financial District. I was somewhere around the 20th floor I think. The elevators were famously slow. Guests complained about them constantly, and fairly, because they really were slow.
I took the stairs.
This was not a heroic act. I was running at the time, the stairs were right there, and climbing twenty flights beat standing in a lobby in line. But I watched people around me pour real energy into being frustrated about something with a free alternative sitting ten feet away.
I’m not trying to moralize about elevators. The point is smaller and weirder than that.
A lot of what we experience as friction has a workaround sitting next to it that we never look at. The slow elevator had a stairwell ten feet down the hall. The cold room has a sweater on the chair. The brutal commute could be carrying a podcast queue you keep meaning to listen through. The workarounds rarely advertise themselves, and looking for them is a separate skill from the one used to complain about the friction.
The failure to notice is one layer. The deeper layer is what you see instead. Most of us walk around with a frame already loaded, and the frame decides what registers before we’ve really looked at anything. Confirmation feels like thinking. It’s the mind finding what it already brought with it, and it feels most like truth in exactly the situations where actually looking would have changed your answer.
The man walking in Alabama had to work at seeing the neighborhood. The hotel guests never looked at the stairwell. The same mechanism is running while you’re doing engineering work.
Wait. Don’t you usually write about AI, engineering, and management?
Yes. This is where I’m going with it.
If you’ve been in the industry for any length of time, you have a whole set of internalized norms about how software actually gets built: how long a feature takes, who does which job, what “senior” means, the reasonable estimate for a migration or a prototype or a production rollout. Those norms are the neighborhood you’ve been running through for years. You stopped seeing them a long time ago. They’re just the road.
That road is being repaved underneath us.
The estimation timelines I grew up with don’t hold. A task that looked like a sprint sometimes collapses to an afternoon. Other times, what looked like an afternoon’s work turns out to be the actual hard problem and takes a month. Job descriptions are in flux too. The work of writing a first draft, reviewing it, owning the spec, running the research pass before a design doc, all of that is being redistributed unevenly across teams and companies.
If you can’t step outside your own frame and actually rethink how the work gets done, your estimates will be off, your scoping will drift, your staffing plans will reflect a world that no longer exists, and you’ll invest in learning the wrong things. You can know the new tools cold and still ship the wrong plan if your sense of how the work itself happens hasn’t been updated.
What keeps you oriented through all of this is a north star you don’t compromise on. For me, that’s customer obsession, and the discipline has been carrying me through this entire shift. Process, tooling, and speed will keep changing, but each change stays in service of the customer as long as the customer stays in front of you. Without the customer in view, every speed-up just gets you to the wrong place faster.
The patterns worth questioning
Let me name some specific ones, because abstractions about “changing how we work” don’t help unless you point at things.
Estimates sized in engineer-weeks. The unit assumed a roughly stable productivity curve. That curve isn’t stable anymore. The same engineer ships three times faster on some kinds of tasks and half as fast on others, and your gut calibration was trained on the old unit. If you’re still quoting engineer-weeks the way you did in 2023, you’re giving numbers with false precision.
Documentation as tax. For a decade, docs were maintenance overhead. You wrote them for the future human who would read them once on the way into the team. The audience changed. Agents read documentation constantly now, and the ADR nobody used to look at gets pulled in every time the model touches that subsystem. What used to be overhead has become active leverage on the team’s velocity, and most teams haven’t noticed the economics shifted.
The senior engineer as the most prolific coder. The best senior I’ve worked with in the last year writes less code than the mid-level on her team. She spends her time on specs, interfaces, and the parts of the system the models can’t yet reason about coherently. Her contribution moved even though the org chart describing her role didn’t.
“We’ll add tests later.” This was always a shortcut. With AI-assisted refactoring it’s worse. The tests are the only specification that survives the model rewriting around them, and without that spec, every refactor it attempts is happening with nothing underneath.
None of these patterns were wrong when we adopted them. They were load-bearing for the world we used to live in. The problem is the ground shifted and we’re still distributing our weight like it didn’t.
Sit a spell
The other half of this, and the part most of the AI-and-engineering conversation skips, is patience.
Patience. It’s a word that sounds like sitting still while time passes, and the version of patience that actually matters in engineering is closer to active vigilance, the willingness to let the clock run on something whose duration cannot be compressed even when something upstream just got dramatically faster.
There’s a real difference between vibe-coding something into existence on a Saturday afternoon and having that same thing serve millions of customers reliably, day after day, through edge cases no prototype will ever hit.
The prototype is fast now. Genuinely, dramatically fast. You can get a working thing in front of your face in an hour that would have taken a team a week two years ago. That speed is real and I’m not going to pretend otherwise.
But the prototype skips everything that makes something trustworthy at scale. Edge cases, observability, the failure modes nobody thought of in the happy-path demo, the rails that keep the whole thing standing when conditions aren’t ideal. All of that has a required duration that no amount of speed upstream is going to compress.
In practice this means letting the operationally hard part take its time even when the prototype made it look fast, and not allowing a Friday demo to compress the scope of a Monday commitment. The middle of any project is where almost all of the actual engineering happens, and the middle photographs poorly, which is why so much of the AI-and-engineering conversation skips it.
The data that surprised me
Back when I was in school, I spent a semester exploring Medicare and Medicaid claims data. End of summer to Christmas. Most of that time was ETL work, pulling records, reconciling schemas, normalizing codes, writing enough D3.js to produce charts that weren’t insulting. By the time I got through it, I had three months' worth of opinions on data I hadn’t actually examined yet. The mental model was wrong, and the data surprised me almost every time I looked.
I’ve been redoing a version of that project recently, and what used to take a semester now takes a week. The collapse buys back something the old workflow was eating up before I ever got near a chart: the space to look at the data without three months of preconceptions already baked in. The plumbing was always going to manufacture those preconceptions, regardless of who was doing the work.
Back to Alabama
Every quarter, when I go back, I try to run that route with the same attention I had the first time. I fail often. Some mornings I’m already in my head, and the magnolias don’t register at all. The attempt is what matters. Noticing is something you have to practice, and the practice is the kind that doesn’t get easier.
The same practice carries over. Estimates, assumptions about who does which job, the gap between a prototype and a product, and the data you think you already understand. All of it benefits from the muscle that pulls your attention back to what’s actually in front of you. The route looks the same on the thousandth run as on the first. You’re the variable.
The man will be walking his route next quarter. I’ll nod. He’ll nod back. And if I’m paying attention, I’ll notice the fairways are a color that doesn’t exist anywhere else, and I’ll remember he told me he has to make himself do this, and I’ll do it anyway.
By Joshua McDonald on May 6, 2026.
Exported from Medium on August 26, 2026.
Reader discussion