Where Is Your Mental Model?

A well-made AI explanation can make a subject feel clear in seconds, and that feeling is easy to mistake for learning. But recognising an explanation is not the same as understanding the system, and the comfort of almost-understanding is what quietly stops you from exploring and reasoning on your own. Over a career, the difference between the diagrams you consumed and the mental models you actually built is the difference that decides how far you go.

You open your feed and find a clean, well-designed diagram of a kernel subsystem. You read it. It makes sense. You nod, feel a little sharper, and move on. It felt like learning. Most of the time, it was not.

Recognising an explanation is not the same as knowing the system behind it.

What you felt was recognition, not understanding. Recognition is the sense that something is familiar, that it fits, that you have seen it before. It is real, and it is useful, but it is thin. It lasts only as long as the material is in front of you. Close the tab and most of it is gone. Understanding is different. Understanding is what remains when the explanation is no longer on the screen.

Why the clear explanation is the trap

Modern tools are very good at producing fluent, confident, well-organised explanations. That fluency is exactly the problem. The psychologist Robert Bjork described what he called the illusion of competence: the smoother and more familiar material feels, the more we overestimate how well we have actually learned it. Ease of reading is not evidence of learning. It often means the opposite.

The smoother an explanation feels, the more it flatters you. That ease is the illusion, not the learning.

Bjork’s larger finding runs against instinct. The conditions that make learning feel harder and slower, such as struggling to recall something, working a problem before seeing the solution, and returning to it after a gap, are the ones that build durable understanding. He called these desirable difficulties. A perfect AI diagram removes every one of them. It hands you the finished picture and takes away the struggle that would have taught you.

The comfort that costs you

Here is the part that does the real damage over time. Because the explanation felt clear, you believe you have learned it, so you never go and work it out for yourself. You do not open the source. You do not trace the path on your own board. You do not sit with the question long enough to form your own account of it.

Comfort is quiet. It removes the very discomfort that would have pushed you to explore.

This is how a person ends up inside a small bubble of things that felt clear for a minute and were never truly understood. Nothing in that bubble tells you to leave it, because at no point does it feel like you are falling behind. The days pass, the feed keeps supplying clarity, and your own habit of exploring and reflecting slowly goes unused.

One diagram costs nothing you can see. A few years of them is a career that quietly stopped growing.

The bill arrives later, and always at a bad time. A real problem lands on your desk, one no diagram covered. The editor opens, the cursor blinks, and nothing comes. You have seen the subject many times. You still cannot reason about it.

What a mental model actually is

It is worth being precise, because this is easy to get half right. A mental model is not the ability to reproduce the picture from memory. Redrawing a diagram you memorised is still recall, and recall of a surface is shallow. A mental model is the understanding underneath it: knowing why the system is built the way it is, being able to predict what happens when one part fails, and being able to apply it to a situation the original explanation never showed.

A mental model is not the picture you can redraw. It is the reasoning that still works on the case the picture never showed.

This is why Richard Feynman’s test is so exact. “What I cannot create, I do not understand,” he wrote, a line found on his blackboard at the time of his death in 1988. Not “what I can recognise,” and not even “what I can repeat,” but what I can build from an understanding of my own.

Building that kind of model is not about avoiding AI. It is about changing what you do with it. A few habits help:

Predict before you read. Before you open the explanation, work out your own account of the subsystem from what you already know. Reading it afterwards then corrects your reasoning instead of replacing it.
Interrogate the diagram, do not absorb it. Ask why it is arranged this way, what happens when one block fails, and what it quietly left out. A model is made of these questions, not of the boxes.
Go to the source it summarised. Read the actual code, the register map, the specification. Notice what the clean summary dropped. That gap is where your understanding grows.
Explain it without the picture. Describe the system to someone, or to a blank page, with the diagram closed. Where you stall is exactly what you had only recognised.
Sit with the hard part. When something is difficult, spend real time on it before reaching for the answer. That discomfort is not the obstacle to learning. It is the learning.

This is also why practice and reflection sit at the centre of how we teach at TECH VEDA: the understanding that lasts is the kind you build by reasoning through the system yourself, with guidance, not the kind you scroll past.

So the next time an explanation makes a hard subject feel easy, treat that feeling as a question rather than an answer. Ask what you could actually do with it, on a real problem, with the screen switched off.