Who Knows the Layer Below?

By Mark Sentell · November 15, 2026 · 4 min read ·

A doctor demonstrates the muscles of a forearm while a group of men lean in close to watch and learn.
The Anatomy Lesson of Dr. Nicolaes Tulp, Rembrandt van Rijn, 1632. Public domain, via Wikimedia Commons.

In software we have a name for what every new tool does. We call it a layer of abstraction.

An abstraction hides the details underneath so you don’t have to think about them. Early programmers gave the machine instructions it understood directly, one tiny step at a time. Then came programming languages that let you write something closer to English. Then frameworks that handled the hard parts for you. Now you can describe what you want in plain words and an AI writes the code.

Every layer lets more people build more things with less knowledge. That’s the whole point.

It isn’t just software. Most of us have no idea how our food gets from the farm to the shelf. Money went from cash to cards to tapping a phone. Cars went from a carburetor you could adjust with a screwdriver to a computer you plug into.

Most of that is good. I’m grateful for it. I don’t want to grow my own wheat.

But every layer makes a trade. We gain leverage. We give up understanding.

This is the twelfth article in the series on the human side of the AI shift. If you’re just starting, go back to the first one, The Computer Works for Us.

Abstractions leak

Back in 2002, a software writer named Joel Spolsky put it this way: “All non-trivial abstractions, to some degree, are leaky.”

In other words, every layer eventually breaks. The thing it was hiding comes up through the floor. And when it does, you need somebody who understands the layer below.

We see this at The Garage all the time. A modern car will tell you something is wrong. You plug in a scan tool and it gives you a fault code. But a code isn’t a diagnosis. A misfire code could mean a bad plug, a bad coil, a vacuum leak, a fuel problem or a dozen other things. The scanner tells you where to start looking. It takes a mechanic who understands the engine to know what it actually means.

The tool is a gift. But it’s no substitute for the person who knows what’s under the hood.

This layer is different

Here’s what I think is new about AI.

Every earlier layer hid how the machine did the work. You didn’t need to know how the compiler worked, or how the payment network moved your money. But you still did the thinking. You still decided what to build, weighed the options and judged whether the result was any good.

AI is the first layer that starts to hide the thinking itself. It drafts, summarizes, weighs options and recommends. The layer underneath it isn’t a machine anymore. In a real sense, it’s us.

And it leaks too. AI makes things up, confidently. It misses context. It produces something that looks right and isn’t. When that happens, the only thing that catches it is a person who can think through the problem without the tool.

Where will the next expert come from?

This is the question that worries me most as a leader.

People become experts by doing the low-level work for years. Engineers learned by writing code by hand, breaking it and fixing it. Writers learned by writing bad first drafts. Mechanics learned by getting their hands dirty on engines that didn’t run. That slow, inefficient work is what built the judgment.

If AI does the first ten years of the work, where does the person come from who understands the layer below twenty years from now?

I don’t think that’s only a career question. It’s a discipleship question. Paul told Timothy, “what you have heard from me in the presence of many witnesses entrust to faithful men, who will be able to teach others also” (2 Timothy 2:2). Faith has always been handed down person to person, one generation forming the next. Skill works the same way. Somebody has to be formed. And formation doesn’t come from a summary. (I’ve written about this for pastors in Using the AI Inside Logos Without Letting It Do Your Study.)

Know the condition of your flocks

There’s an old proverb I keep coming back to: “Know well the condition of your flocks, and give attention to your herds” (Proverbs 27:23).

A shepherd with a lot of sheep can’t carry every one. But he can’t afford to lose track of them either. He has to know what’s happening down at ground level.

That’s what this moment asks of leaders. You don’t have to do every task at the bottom layer. You do need to understand it well enough to know when something’s wrong. As a leader, I don’t need to write every line of code. I do need to be able to read it.

Pick your layer

I’m not saying throw out the tools. Nobody can know every layer, and nobody should try.

But pick one or two that matter most to your calling, and keep your hands in them. For a mechanic, it’s the engine. For a pastor, it’s the text and the people. For an engineer, it’s the code and how the system really works. For a parent, it’s your kids, not the app that tracks them.

Let the machine handle the layers you don’t need. Stay close to the ones you do.

What to do this week

Do one thing by hand. Write the first draft yourself. Read the passage before you ask for a summary. Work the problem before you ask for the answer.

Ask, “How would I know if this were wrong?” If you can’t answer that about something AI gave you, you’ve found a layer you need to learn.

Teach someone the layer below. If you lead people, let someone do one task the slow way with you beside them. It’s inefficient. That’s how it works.

Pick your layer. Name the one or two things in your work you never want to lose touch with. Then protect them.

See also

  • The Law of Leaky Abstractions by Joel Spolsky, Joel on SoftwareThe 2002 essay behind the idea that every abstraction eventually breaks, and someone has to understand what's underneath.

Enjoyed this article?