Remember when we thought generative AI would replace human creativity overnight? That was about five minutes ago in tech time. We’ve been testing every 3D modeling tool, every AI architecture generator, every “just describe it and we’ll build it” platform that promises to turn words into worlds. And yet here’s Joe Macken, a truck driver from Queens, who spent over two decades carving a scale model of every building in New York City out of balsa wood, and suddenly all our fancy toolkits feel a bit hollow.
Macken started in 2004 with 30 Rockefeller Plaza. Not with Blender. Not with SketchUp. Not with some AI prompt that spits out a passable approximation. With balsa wood, a blade, and what I can only assume is the patience of someone who actually enjoys traffic on the Long Island Expressway.
This is the kind of story that makes me question what we’re actually reviewing here at agntbox. We spend our days testing AI tools that promise to automate everything, to make creation faster, to eliminate the tedious parts of building things. And most of them work, technically. They do what they say on the tin. But none of them would produce what Macken produced, because none of them have to sit with a project for 20 years.
The Problem With Fast Tools
The AI toolkit space is obsessed with speed. Generate a website in 30 seconds. Create a 3D model from a photo. Turn a sketch into production code. We review these tools, we rate them, we tell you which ones actually deliver. But we’re always measuring the same thing: how quickly can you go from idea to output?
Macken’s miniature metropolis went viral on TikTok with 10 million views, which is its own kind of absurd. The platform built for 15-second dopamine hits becomes the stage for work that took two decades. The comments probably said things like “bro could’ve just used AI” or “imagine if he learned Unreal Engine instead.” Missing the point entirely.
The point is that some things are supposed to take time. Some projects are the project. The truck driving was the day job. The model was the life.
What This Means For Tool Builders
Here’s what bothers me about most AI toolkits: they assume the goal is always to be done faster. Ship faster. Iterate faster. Fail faster. But Macken’s work suggests there’s value in the opposite direction. In accumulation. In the kind of intimate knowledge you only get from spending 20 years with something.
Could an AI generate a 3D model of New York City? Sure. Feed it enough data and it’ll spit out something impressive. But it won’t know which buildings matter. It won’t have opinions about the skyline. It won’t remember when it got the Chrysler Building’s spire wrong the first time and had to start over.
The tools we review are getting better at mimicking output. They’re getting worse at understanding why anyone would want to spend 20 years on anything. That’s not a criticism of the tools, exactly. It’s a criticism of what we’re asking them to do.
The Honest Assessment
If Joe Macken walked into our review lab and asked which AI toolkit could help him build his next model, I’d probably recommend a few. Maybe something for reference photos. Maybe a tool to help with measurements. But I’d also tell him the truth: the tools would make it faster, but they wouldn’t make it better. They’d help him finish, but they’d rob him of the 20 years.
And maybe that’s fine for most projects. Most of us aren’t building scale models of entire cities. We’re shipping products, meeting deadlines, trying to get something out the door before the market moves on. The tools we review are built for that reality, and they’re good at it.
But every once in a while, a truck driver from Queens reminds us that the best work isn’t always the fastest work. And no toolkit, no matter how smart, can automate the decision to spend 20 years caring about something.
That’s not a feature request. That’s just the truth.
🕒 Published: