Edition 367 | August 24, 2026

The Dyslexic AI Newsletter by LM Lab AI

Reading Time: 7 minutes | Listening Time: 9 minutes

Quick honesty check before I get into this one. What follows is a way of thinking about something, not a technical claim about exactly how a model works under the hood. I want to be upfront about that distinction, because the idea is genuinely useful even though I am using it as a metaphor, not a spec sheet.

Here it is.

Computing has always been described in one image. On or off. One or zero. A switch, flipped one way or the other, billions of times a second.

Modern AI does not really work that way, at the level that actually matters for meaning.

Underneath everything, yes, the hardware is still ones and zeros. That part is true and it is not what I am talking about.

I am talking about how meaning actually gets represented once you get up to the level where a model understands that "bank" means something different in "river bank" than it does in "bank account." That distinction is not stored anywhere as a fixed, isolated fact. There is no dictionary entry sitting in the model saying "bank equals this, definitively, in all cases."

Instead, the meaning of a word in these systems comes from its relationships to everything around it. Where it tends to appear. What other words tend to sit near it. How its position shifts depending on the sentence, the context, the conversation. Meaning is not a fixed point. It moves, depending on what's near it.

Here is a way to picture it that I find genuinely useful, even though it is closer to metaphor than technical description.

Think of meaning less like a dot with fixed coordinates, and more like a wave. A wave does not exist at one single point. It exists as a shape moving through a space, shifting as it interacts with whatever else is nearby. Two waves overlapping can reinforce each other or cancel each other out, depending on how they meet.

That is closer to how meaning actually behaves, both in these AI systems and, I would argue, in a lot of human thought. A word does not carry one fixed meaning that gets looked up like a dictionary entry. It carries a field of possible meanings, and context is what determines which part of that field actually lights up in a given moment.

This is not a new idea, by the way. Go back to Edition 359, to Feynman and the bird. If you know the name of a bird in every language on earth, you still don't know anything about the bird. That is exactly this. The name is the fixed dot. The bird itself, the actual thing, only shows up in the relationships. What it eats. Where it lives. How it moves. The word was never carrying the meaning by itself. It was always pointing toward a field.

Here is the part that stopped me when I first started connecting these dots.

A lot of dyslexic thinkers describe their own thinking in almost exactly this language, without ever having heard the technical version. Not thinking in isolated words, one after another, in a fixed sequence. Thinking in images. In patterns. In relationships. In stories and associations that connect across contexts instead of staying locked inside one definition.

That is not a coincidence, and I do not think it is even that surprising once you see it clearly. Traditional education was built almost entirely around symbol-based intelligence. Learn the definition. Memorize the fixed meaning. Retrieve the correct isolated fact, in order, under time pressure.

A mind that naturally works in relationships instead of fixed symbols is going to feel constant friction inside a system built entirely around fixed symbols. Not because the thinking is wrong. Because the system was measuring the wrong layer the entire time, the same argument I made back in Edition 361, just arriving here from a different direction.

Modern AI may actually be moving closer to relationship-based intelligence, the same kind a lot of dyslexic minds have been operating in the whole time, while traditional education is still measuring almost entirely for the symbol-based kind.

If understanding is genuinely relational instead of symbolic, a test built entirely around retrieving fixed, isolated facts was never actually going to measure it well.

You can know a concept deeply, be able to apply it, extend it, connect it to three other things nobody else in the room thought to connect it to, and still bomb a multiple choice question that only rewards recalling one fixed definition, worded the exact way the textbook worded it.

That gap is not a small assessment problem. It is a sign that the entire measuring instrument was built for a different kind of intelligence than the one it is trying to measure. I am going to come back to that directly in the next edition, because I think it points toward something specific and buildable, not just another abstract critique of standardized testing.

One thing to try this week.

The next time you are trying to understand something new, do not start by looking up the definition. Start by asking how it relates to something you already understand. What is it similar to? What is it the opposite of? What does it remind you of from a completely different part of your life?

You are not being lazy by skipping the dictionary. You are working the way your mind actually processes meaning in the first place.

Use this the next time you're trying to actually understand something, not just define it.

Understand the relationships. The definition will take care of itself.

Matt "Coach" Ivey

Founder, LM Lab AI | Creator, The Dyslexic AI Newsletter

Dictated, not typed. Obviously.

Edition 368: If meaning is relational and traditional tests only measure the symbolic layer, what would an actual learning environment look like if it was built for the mind we've been describing this whole series? Edition 368 opens with something a man named Thomas Budd wrote in 1685, walks through why traditional college worked, and lands on the six pillars and the AI-Native Microcollege model I've been building toward this entire series.