The teacher who learned it last week

What makerspaces taught me about how we’ll all learn AI

I’m certified on most of the industrial machines at my makerspace. The CO2 lasers. A CNC router that will take a full 4×8 sheet of plywood and carve it into whatever the file tells it to. These are machines that can take a finger off, throw a piece of stock across the room, or catch fire if you walk away at the wrong moment.

Here is everything the instructor-led certification training covered: how not to hurt myself, how not to hurt anyone standing nearby, and how not to hurt the machine. That was the curriculum. The rest was a list of YouTube links.

The first time, I kept waiting for the real class to start — the part where someone teaches you to actually use the thing. It never came. And once I got over the strangeness of it, I realized the instructor had done something precise. He hadn’t taught me less than I needed. He’d taught me exactly the part I couldn’t afford to learn by trial and error, and then he handed the rest back to me. The irreversible stuff is the only learning worth requiring. Everything else, I could build my way into.

Sew first

I teach sewing for other makers, and the makerspace forced me to throw out everything I knew about how sewing is taught.

A normal sewing class starts with orientation. You learn about patterns and grainlines and fabric. You learn why you never, ever cut paper with your fabric scissors. You do not get to touch the machine until you’ve earned it by understanding the things around the machine.

In my class, the first thing we do is sew. People sit down in front of a machine that’s already threaded, with fabric already under the foot. I tell them two things: keep your fingers away from the needle, and the foot pedal works like the gas pedal in your car — the harder you press, the faster you go. Then they sew. They’re terrified and exhilarated and laughing, and they make a wobbly, crooked line of stitches across a scrap of cotton. Then I show them how to make the line straighter, and we go from there.

They learn what they need to know by doing a curated set of actions and picking up the why along the way. Grainlines still matter. They just matter more once you’ve felt the machine pull fabric through your hands.

This is not a clever trick. It’s constructionism — Seymour Papert’s old idea that people build durable understanding by building actual things, not by being told about them first. A makerspace runs on it whether anyone there has heard the word or not. So, increasingly, does the way I think the rest of us are going to have to learn at work.

There are never enough experts

The other unavoidable truth of a makerspace is that there are not enough experts to go around, and there never will be.

The people who know things are generous with their time, but their time is finite, and it is genuinely wasted teaching you something you could have learned from a fifteen-minute video. So the culture sorts itself out. Experts spend their scarce hours on the things you can’t safely figure out alone or on challenging things they want to learn too. Everything else gets learned sideways — from the person at the next bench who tried it last week.

That’s the part that can break people raised on the traditional model. Most of us were taught that you earn the right to teach by becoming an expert first. You put in the years, and then you instruct. In a makerspace, the person showing you how to do something often learned it seven days ago. They don’t know all the ways to do it. They know the one way they tried that worked — and they know how not to hurt themselves, others, or the machine. That turns out to be enough to get you moving.

The three ways to get hurt

Here’s the frame I keep coming back to. In the makerspace, mandatory instruction collapses down to three things: don’t hurt yourself, don’t hurt others, don’t hurt the machine. Teach those, then get out of the way.

I think AI at work has its own version of those three, and almost nobody is teaching them on purpose.

Don’t hurt yourself. Trusting a confident, wrong answer costs you your own time and your own credibility. The skill that protects you isn’t prompt-engineering wizardry — it’s the habit of checking, of knowing which tasks you can verify at a glance and which ones you can’t.

Don’t hurt others. This is other people’s trust, and the colleague or customer on the receiving end of your work. In my work, harm here almost always comes from inattention — letting something you didn’t really read go out under your name to someone who assumed you had.

Don’t hurt the machine. Call it the system, the org, the governance layer. It’s the stuff you genuinely can’t undo — information that needs to be protected or the thing that can’t be un-shared once it’s shared. Pasting the wrong thing into the wrong tool might seem like the AI equivalent of cutting paper with the fabric scissors, but scissors are easily replaced.

Teach those three with real seriousness. Then stop pretending you also have to teach the rest the slow way, because you can’t. The tools change every few weeks. Nobody gets to be an expert before it’s time to help the next person.

What this means for workforce of the future

When AI moves this fast, the expert-first model becomes a bottleneck. By the time someone has earned the traditional right to teach, the thing they mastered has shifted underneath them. And the rare people who are ahead shouldn’t be spending their afternoons explaining, for the fortieth time, that an AI assistant doesn’t behave like the software we’re all used to.

So a workforce that learns the way a makerspace does needs two kinds of people, and it needs to stop apologizing for both.

It needs colleagues with the courage and vulnerability to share what’s working for them right now, while it’s still a little wobbly — the ones willing to be the person who learned it last week.

And it needs colleagues who are comfortable learning this way: self-directed, curated into a first real action, picking up the why as they go, asking the bench neighbor instead of waiting for a class that is never going to be scheduled.

I was raised on the other model. I believed you earned the right to teach by becoming an expert. The makerspace taught me a different way, and now I‘ve begun to teach AI at my company the way I teach sewing — I sit people down in front of something that already works, point out the few ways to get hurt, and let them make a wobbly first line.

A lot of the time, I’m the one who learned it last week. I’ve stopped treating that as a confession. It’s the qualification.