We Swapped the Motor

I set up a backend and a Stripe account today for a project I started last night. I need you to understand that sentence. Last night this project did not exist. Today it accepts payments.

My day job is product design. Has been for years. The work, on paper, looks like wireframes and flows, user interviews and usability tests, building UI in Figma, adding spec and redlines so engineers know what to build. It is good work. I like it. But it also means I’ve spent most of my career one layer removed from the thing itself. I designed the blueprint. Someone else poured the foundation.

I have always loved to code. I loved the feeling of seeing an iteration realize itself, the moment a thing you imagined starts behaving like a thing that exists. For years, the dream in product design circles has been to “design in code,” to close the gap between what you draw and what ships. I loved the promise of that. But there was never enough time, and when I did try, I’d hit the limits of my technical know-how almost immediately. Not the creative limits. The bureaucratic ones. Dependency conflicts. Environment variables that worked on Tuesday and didn’t on Wednesday for reasons that may involve God. Build failures that sent me down four-hour rabbit holes only to discover I needed to add a single line to a config file I didn’t know existed.

So I coded only when I really had to. Simple local prototypes. Little proof-of-concept sketches that never left my machine. I have old Figma files full of loose product ideas I had a dream about. Landing pages, dashboards, tools I thought were interesting enough to exist. They stayed in Figma. The ambition was there but it remained theoretical, because the friction of building real software was just high enough to keep me permanently in the planning phase. I was a general with a war room full of maps and no army.

All of that is different now.

I don’t want to be precious about this, so I’ll say it plainly: AI has been one of the most freeing creative tools I’ve ever encountered. Not because it writes perfect code. It doesn’t. But because it obliterates the specific kind of friction that used to stop me cold. The “why won’t this build” friction. The “what is this webpack error trying to tell me in its alien tongue” friction. The stuff that had nothing to do with the creative act and everything to do with the bureaucracy of machines. That wall is, if not gone, significantly shorter. I can see over it now. And what’s on the other side is a lot of work I actually want to do.

At work, I’m starting to design in code for the first time. Not as a stunt, but as a real part of how I practice. Outside of work, those old Figma files are waking up. I’m building and iterating on three or four ideas at a time. I’m pivoting product decisions in hours after showing someone a working prototype running on a Vercel domain. Not a mockup. Not a click-through. A living, breathing thing with a URL. The speed is real.

But I’ve also burned entire days chasing automations I could have done manually in one. I’ve made consequential decisions about what to build, only to sunset those same projects weeks later because the tools made me feel like I could do more than I actually could. Because when you can build anything, you will try to build everything. And then you will burn out spectacularly on a Tuesday afternoon, having spoken to five simultaneous instances of Claude Code and dictated to your computer via Wispr Flow for the last six hours straight, before the quiet part of your brain finally breaks through to inform you that you need to go outside and touch grass immediately.

There is a strange upside to that velocity, though, one I didn’t anticipate. When you can go from idea to working prototype in a day or two, you arrive at the hard questions much faster than you used to. Not the technical questions. The product questions. How does someone actually find this? What does the path to sustainability look like? Who am I asking people to trust, and why? Questions that used to live comfortably in the abstract, years away from mattering, now show up on day three with their arms crossed.

I’ve killed projects not because they didn’t work but because I could suddenly see the whole shape of what making them viable would require, and I didn’t want that life. One idea was solid, the concept worked, the prototype was clean. But the only realistic path to reaching people meant becoming a visible authority in a space I had no interest in inhabiting long-term. Building the thing was the easy part. The go-to-market was the dealbreaker. I would have never arrived at that clarity had I not been able to flesh the idea out far enough to see what it would actually cost me, not in dollars but in years and identity. The building itself became a form of product research, and sometimes the research tells you to walk away.

What I’m discovering is that the hard part was never the technical ceiling. The hard part, the part AI can’t shortcut, is knowing what to build. Having taste. Understanding which of your many ideas deserves the investment of your finite attention. That is a human problem, and it has gotten more urgent now that the cost of starting has dropped to nearly zero.

If I had to boil it down: there are really only two questions that matter now. Who are you building for, and how do they find out about it? Execution used to be the bottleneck. Knowing how to build the thing, how to ship it, how to make it work at scale. AI has largely solved that part. What it hasn’t solved, what it can’t solve, is whether the thing should exist in the first place and whether you can put it in front of the people who need it. Distribution and audience. Those are the load-bearing walls now. Everything else is furniture you can rearrange.

Which brings me to the broader thing happening right now, the thing everyone is either ecstatic or terrified about, often in the same breath.

A fifteen-year design veteran, unemployed for a year and a half, writes plainly that their profession is no longer sustainable. A twenty-year practitioner says they hate their industry. Someone in their late forties with two decades of experience says they can barely get interviews. These aren’t people on the margins. They’re the people who helped build the thing. And alongside them, a growing cohort feels like they’ve been handed a skeleton key. Shipping faster than ever, building what used to take teams, operating at a scale that would have been fantasy two years ago.

Both groups are telling the truth. That’s what makes this so disorienting.

Someone put it well recently: the people thriving with AI right now are selecting for a specific neurotype. High novelty seeking, high conscientiousness, a tolerance for rapid context switching. Maybe ten to fifteen percent of the population. For the rest, the experience isn’t energizing. It’s overwhelming. And even within that narrow band, the cognitive toll is real. You can run fifty, a hundred micro-cycles a day. Ramp up, grind, feel the rush when it works, repeat. By evening you are a different species of tired. Harvard Business Review published a study on this and found that workers using AI experienced significantly more decision fatigue but reported less burnout overall. The fatigue is cognitive, not emotional. Your brain runs out of working memory, not out of heart.

And a related study found that workers voluntarily did more because AI made more feel doable. Work bled into lunches and evenings. More output and more exhaustion, twinned together like a gift you can’t unwrap without cutting yourself a little.

There’s a framework that helps explain why this all feels the way it does. An economist named Carlota Perez studied every major technological revolution since the industrial era and found the same pattern each time. First comes installation: new technology floods in, speculation runs wild, old professions decline, new ones form. Then a turning point. Then deployment, where institutions actually redesign themselves around the new reality. And if it goes well, a golden age.

When factories got the electric motor in the 1880s, they swapped out the steam engine and plugged in the new one. Everything else stayed the same. Layout, processes, organization. New tools doing old things. For thirty years, those early-adopter factories saw almost no increase in output. The gains came only when companies redesigned the entire operation around the technology. New building designs. Machines arranged by workflow. Ford’s assembly line.

We have swapped the motor. We have not yet redesigned the factory.

That gap between installation and deployment is where the dissonance lives. “This is incredible for some of my work” coexisting with “my entire workflow is broken” is not a contradiction. It’s the signature of a revolution in its awkward teenage years. The tools iterate on monthly cycles now, not decades, so the thirty-year gap may compress into three. But the gap is real, and we are deep inside it.

The data on what’s coming is all over the map. Anthropic’s own research suggests a possible severe contraction for white-collar workers. Stanford found a fourteen percent drop in job-finding rates for young workers in AI-exposed fields. The World Economic Forum projects a hundred and fifty million jobs displaced by 2027 alongside a hundred and seventy million new ones. The Brookings Institution titled their analysis “Research on AI and the labor market is still in the first inning.” The different measures of AI exposure don’t even agree with each other. Nobody knows. Not the people selling doom, not the accelerationists, and certainly not me.

What I do know is that both things are true simultaneously. The technology extends human capability in ways that feel like genuine freedom. An individual with a laptop and a wifi password can now do things that used to require a team, a budget, and a decade of specialized knowledge. That is extraordinary, and I am living proof of it. I’m building things I couldn’t have built a year ago, not because I got smarter but because the floor rose to meet me.

And real people are getting hurt. Right now, this week. The veteran who can’t land an interview. The designer whose entire team was cut while she was offered a raise to stay and operate the tools that replaced her colleagues. She quit the next day.

These truths coexist. The technology opens doors and the transition crushes people underneath them. Holding both is the work right now. Choosing just one is comfortable but incomplete.

We swapped the motor. We have not yet redesigned the factory. But I think I can see the blueprints starting to form. The redesigned factory won’t optimize for who can execute fastest. It will optimize for who knows what’s worth executing at all. The people who understand their audience, who can see a distribution path before they write a line of code, who know when to kill a project at prototype stage instead of life-support stage. Taste, conviction, and the willingness to walk away. Those are the load-bearing skills of whatever comes next. The factory floor is going to look nothing like it did, and most of us are still standing in the old one, squinting at the new motor, wondering why the output hasn’t changed yet.

The chaos is structural, not personal. Even though it feels deeply, unmistakably personal. And if you’re struggling to keep up, it’s probably just because you’re human. That is the correct explanation.