Let’s talk about the machine in that workshop that nobody is allowed to open.
There’s a guy I think about. Small workshop, two bays, a kettle that’s never been off since 2011. He fixes things. Tractors, mostly. Bakkies. Pumps. The kind of work where you hear the fault before you see it.
He’s good. Been doing it thirty years.
He’s also going out of business, and it has nothing to do with AI.
Here’s what happens when a machine comes in now. He opens the bonnet. He finds the fault in twenty minutes, because he’s been inside that engine a hundred times. And then he stops, because the part he needs is paired to the control board. It has a serial number that has to be married to that machine by software he can’t buy. The tool that does the marrying is licensed to one dealer in the province.
So he tells the farmer the truth. “It’s a sensor. Two thousand rand. But I can’t put it in.”
The farmer takes it to the dealer. Same sensor. Four times the price. Six weeks wait. And a line item on the invoice for a diagnosis he already paid for once.
Nothing about that machine got worse. It still works the same way it always did. The door just got welded shut.
*
Now I want to be fair here, because this is where people start shouting about greed and it muddies the argument.
There are real reasons things get sealed. Safety. Security. A pacemaker should not be repairable by a guy with a spanner and good intentions. A phone that anyone can open is a phone anyone can get into. Some of the welding is justified, and pretending otherwise makes the whole case look stupid.
But look at what’s actually being decided when a lid gets welded shut. It isn’t whether the thing can be repaired. It’s who is allowed to repair it. And that is a different question, and nobody ever asked you.
The right to repair fight has been running for years now. Europe passed a directive in June 2024 telling manufacturers they have to repair what they sell — washing machines, vacuums, phones — at a reasonable price, with parts available. Several American states pushed through their own laws. And every one of those fights was about the same sentence: the owner of a thing should be able to keep it alive.
That fight is not over. But the ground has already shifted, and most people haven’t noticed where it moved to.
Because the thing that decides more about your life than your car does now has no bonnet at all.
*
Think about what you actually run your business on.
Pricing. Scheduling. Your books. The thing that reads your email and decides which supplier gets a reply first. The model that tells you which customers are about to leave.
Now: open it.
You can’t. There’s no panel. No screws. When it’s wrong, you can’t reach in and adjust the part that’s wrong, because you can’t see the part. You can file a ticket. You can wait. You can leave. That’s the whole menu.
And when it’s wrong in a way that matters — a price that shouldn’t have gone out, a client who got flagged, a number that got misfiled — you will not be able to say why. Not because you’re not smart enough. Because there is nothing to open.
That’s a strange kind of dependency. Not the dependency of a man who can’t fix his own car. The dependency of a man who can’t even look under the bonnet to agree that yes, that’s the problem.
And here’s the part people skip past. It isn’t only the machine that gets sealed. It’s the knowledge.
The old system had a byproduct nobody priced. To fix a thing you had to understand it. The mechanic who fixed your bakkie for ten years knew its history — the previous owner, the bad repair in 2019, the reason the second gear whines on a cold morning. That knowledge lived in a person, in a place, and it cost nothing to consult.
You didn’t buy a service. You had a relationship with a specific human who knew a specific thing.
Now you have a subscription to a general thing that knows nothing about you in particular.
That’s the trade, stated plainly. You gave up the man who knew your machine for a system that knows every machine on average.
*
I’m not going to pretend this is all loss, because that would be a lie and you’d stop listening.
The average is often better than the local guy. At three in the morning, the average answers. The average doesn’t go on leave, doesn’t have a bad month, doesn’t get too old to crawl under the bakkie. For a lot of work — the boring, repeatable, well-understood work — the average is a straight upgrade, and pretending otherwise is how people talk themselves out of good tools.
But the average can’t be maintained. It can only be replaced. And a thing that can only be replaced is a thing whose life is decided by somebody else’s release schedule.
Which brings me to the actual fear, and it isn’t about machines.
When nothing can be fixed, everything becomes disposable. That’s the logic. You don’t repair, you replace. You don’t maintain, you migrate.
And once a business gets used to that logic applied to its tools, it starts applying it to the other thing in the building that used to be maintained.
People.
Not with a memo. With a habit. If the systems get swapped every eighteen months and nothing is ever nursed back to health, then the idea of nursing anything back to health starts to look sentimental. The guy who knows why the process works the way it works looks like overhead next to a dashboard that’s already producing the same number.
He gets replaced too. Same logic. Same ease. Same silence.
That’s why I don’t think this is a hardware story. Repair is a way of relating to a thing that has history. Somebody opened it, understood it, and kept it alive on purpose. Every time we design that out, we get a little more used to the idea that anything old and inconvenient can just be swapped.
*
So what do you actually do. Four things. None of them heroic.
One. Ask before you buy whether a third party can look inside. Not whether they’ll support it — whether a competent stranger with the right papers could open it and fix it. If the answer is no, you’re not buying a tool. You’re renting access to one, and you should know that on the day you sign.
Two. Keep the logic in a format you can read. The rules for how your pricing works. The list of what your support team is allowed to promise. The test cases that prove the thing still behaves. Write them down somewhere outside the vendor’s system, in plain words, in your own file. If a new hire could read it in an afternoon, you can survive a bad quarter.
Three. Name the person who can rebuild it. One name. Not a team. Somebody in your company or outside it, who has actually done it once, on a copy, while it didn’t matter.
Four. Price the exit before the pitch. What would it cost, in money and weeks, to move this work somewhere else? If you can’t answer that number, you don’t know what you signed. That number is your real contract term, not the one on the page.
And for the ones who don’t run a company — the smaller move. When something of yours breaks, try to get it fixed once before you replace it. Not for the money. Because the attempt is how you find out what you actually own.
*
Because underneath all of it there’s one idea, and it’s older than any of this.
Ownership was never the receipt. It was the ability to keep a thing alive without permission.
A man who can’t open his own tractor doesn’t own a tractor. He has a licence to use one until the manufacturer decides otherwise. Same for the phone. Same for the software. Same for the model that knows your margins better than you do.
The world is going to keep sealing things. Some of it will be worth it. But the seal is always presented as a feature, and the cost is always the same — a little more of the world becomes something that happens to you instead of something you can reach into and change.
So the principle, plainly: you have the right to open what you own. Not everything, and not without care. But the ability to look inside, understand what you find, and put it right — that has to stay with the person who carries the risk.
Because a thing you cannot fix is not yours. It’s lent. And the lender sets the terms.
Have a look at one thing you depend on this week. Ask whether you could keep it alive without asking permission.
If the answer is no, you’ve found the thing to work on.
You’ve got this.