Essay

Never Delegate Understanding

The AI answer came back in 12 seconds. It was good. But I was holding something that didn't belong to me, and I've been thinking about why that matters.

The answer came back in twelve seconds.

I had put real thought into the question. Layered in context, shaped it carefully, submitted it to the AI the way you’d brief a good colleague. What came back was coherent, specific, and, as far as I could tell, right. I pasted it into a working document. Then I stopped.

I was holding something that didn’t belong to me. A conclusion I’d received without doing anything to earn it.

That’s the problem. Not the answer. Where is the earning?


The Library

My junior year of college I enrolled in a stage directing class I’d been looking forward to since freshman year. By the end of the first week I could see it wasn’t going to give me what I wanted. The professor had a method. He was going to teach it, I did think it was valuable.

After class, I walked to the university library and checked out a book on directing that wasn’t on the syllabus.

Over the semester I read every book the library had on directing. The ones they didn’t carry, I special-ordered. Some I read twice. I did the coursework too, showed up, turned things in. But I stopped waiting for the class to teach me.

I learned something in that library I didn’t fully understand until later: my education is mine to build. At the time I had been waiting for someone to hand it to me.


What Newport Knows

Cal Newport argues in Deep Work that learning is itself a form of deep work. You can’t absorb a skill or an idea while half-attending. You have to put your full attention on it, long enough that the relevant neural circuits begin to change.

The mechanism he describes is specific. Focused attention on a cognitive circuit triggers the cells that wrap it in myelin, cementing the pathway. “By focusing intensely on a specific skill,” Newport writes, “you’re forcing the specific relevant circuit to fire, again and again, in isolation.” The wrapping accumulates. The circuit gets faster. The skill compounds.

This only works in one direction. Scattered attention and distracted engagement don’t trigger the same process. Newport is direct about what happens to people who spend too much time in the shallows: “Spend enough time in a state of frenetic shallowness and you permanently reduce your capacity to perform deep work.”

That’s the thing about the twelve-second answer. The answer was fine. What didn’t happen was the circuit firing.


What I Got Wrong at Evergreen

When I was hired as Director of Admissions, I came in knowing things. I had done this work before. I understood the landscape. I had a clear picture of the kind of leader this office needed, and I started acting like that leader immediately.

It took me longer than I want to admit to realize I had it wrong.

Not because my past experience was worthless. It was real. But I had applied my experience without first trying to understand this place, these people, this particular culture. I had delegated the work of understanding Evergreen to what I already knew. And what I already knew didn’t know Evergreen.

I had to slow down and listen. Do a lot more of it than felt comfortable. Adjust my picture of what leadership looked like here before I could actually lead.

The lesson wasn’t that I needed more confidence, or less. It was that I had arrived with an answer before I had done the understanding that makes an answer yours.


The Useful Distinction

AI didn’t create this problem. It’s older than the technology by a long distance.

But AI has made it more available. More frictionless. The answer is faster to get, better on arrival, and easier to act on without asking whether you actually understand it. Newport put it plainly in a 2024 interview: “They use AI to spare them from having to reach high peaks of cognitive engagement or concentration.” Writing, he said, is to cognitive health what steps are to physical health. You can outsource the output. You can’t outsource the capacity.

AI used to push you deeper into a question is different from AI used to skip the question. One accelerates the struggle. The other removes it.

Research on AI tutoring makes this concrete. In a 2024 study, students who used ChatGPT freely during math practice improved their accuracy by 48%. Their concept understanding scores, tested without AI, dropped by 17%. Better at the performance. Less capable of the understanding. The struggle was where the learning lived, and they had traded it away without knowing it.

I still use AI. Most days. But I’ve started noticing the difference between using it to advance something I’m genuinely working to understand and using it to skip to the part where I look like I understand it. The first is a tool. The second is a trade I’m not always aware I’m making.


Never Delegate Understanding

What the Eames family calls an Eamesian parable: Never delegate understanding.

Charles and Ray Eames built their studio on this principle. When they encountered a new material, a new technology, a new problem, they didn’t farm out the exploratory phase. They ran thousands of experiments themselves first, learned the material from the inside, before bringing the results to the team. The understanding they built was the foundation everything else stood on.

That’s not a rule about doing everything yourself. It’s a principle about knowing which understanding is yours to do.

The directing class was mine to do. The understanding of what Evergreen needed was mine to do. The judgment about whether that twelve-second answer is actually right, and right for this, and right for me: that one’s mine too.


What I’m Still Figuring Out

I don’t always know where that line is.

There are domains where I grab an AI answer and move on, and I’m not entirely sure whether that’s reasonable delegation or a slow erosion of something I should be building. I notice I’m more comfortable with that tradeoff in areas where I already have real knowledge, where I can evaluate what comes back and recognize what’s off. In those areas the tool is genuinely useful.

It’s the other areas that concern me. The ones where I don’t know enough to know what I don’t know. Where the answer looks right because I have no way to see that it isn’t. I haven’t developed a reliable method for recognizing when I’m in that situation. I’m working on it.

The library story ends simply: I graduated knowing how to direct. Not because of one class, one professor, or one method. Because I had done the reading myself, brought my own questions to it, built my own understanding from the raw material.

AI can give me a lot of things. That’s not one of them.

Where are you outsourcing the understanding that’s actually yours to do?

Wade Arave
Written by
Wade Arave

Higher education leader, writer, and builder. More essays →

A letter every other Sunday. Drafts, not noise.