How Large Language Models Work Videos

Lightly edited for clarity.

AI makes you smarter, not in theory, but in the way you actually work.

And when I say that, I don't mean it improves your IQ, but it improves the quality of your decisions, your clarity, your options, and how deeply you think through a problem.

There's a quick example.

Let's say you're a front end developer. You know HTML, CSS, and React. You get put on a new project to build a public website based on Optimizely. Since you don't know Optimizely, traditionally you'd spend some time on Google. You'd read some blogs, maybe watch a few videos, read through the documentation, generally figure out what you need to do and how to approach it, and get started coding.

With AI, you have different options. You can use it to plan what you're actually gonna do. So you can ask it questions like: what's the best architecture for this? What are the new React features I should be using? What pitfalls do people usually run into, enough to optimize the implementation?

And even more critically, what can I automate? What can I use AI to do? How can I do this faster than I did in the past?

And it walks you through the things that you would think about, and fills in the details and answers questions, but it also points out things that you wouldn't think about.

You still have to bring your own judgment, but AI expands the surface area of your thinking.

So that's what I mean when I say AI makes you smarter. It doesn't replace you, but it does extend you.

Lightly edited for clarity.

AI doesn't think, but it is really good at guessing the next thing to say.

When people talk about AI, usually what they're talking about is large language models, or LLMs.

Now, LLMs operate on tokens. Tokens and words are not the same thing, but it is really helpful to think of them as the same thing to understand the concept.

So in language, words have relationships to each other. When I say, “uh oh, sounds like somebody has a case of the Mondays,” every word connects. “Uh oh” relates to what follows. It signals that something bad is coming. “Case” relates to “Mondays.” And together they form a phrase.

And even if you've never heard that phrase, you can tell from context it's not good.

LLMs are mathematical models that learn those relationships, not as rules, but as probabilities. When we train a model, we give it a bunch of examples of language. Training teaches it to map which words are likely to come next, based on all the words that came before.

So in that example, “uh oh, sounds like somebody's got a case of the blank,” it might predict Mondays, or it might predict flu. And it really depends on if you were talking about Office Space, or you were talking about the fever that you have.

When AI is responding to you, what it's really doing is predicting the next word, one at a time, based on all the words that came before it in the conversation. And then it predicts the next one, and the next one, and so on.

And that's why it can sound so confident while being completely wrong. It doesn't know. It just guesses well, based on context.

This is the fundamental concept for using AI. Once you understand this, all the weird behaviors start to make a lot more sense.

Lightly edited for clarity.

Most people still think that AI is thinking. It isn’t. A large language model doesn’t reason or understand. It predicts one token at a time. Each word you see is the next most probable continuation of the last one, based on patterns it learned from billions of examples.

When you type a prompt, the model converts everything into tokens, fragments of words, punctuation, or symbols. Then, starting from the first one, it predicts what should come next. Every token it generates becomes new context, more data it can use to predict the following one.

And that difference matters, because misunderstanding it leads to misplaced trust. This is why AI can produce fluent, convincing, even insightful responses and still be completely wrong. It’s coherent, not conscious.

This is where context comes in. The model’s reasoning is defined entirely by what’s inside its context window. If the information isn’t in that window, it doesn’t exist to the model. Understanding this changes how you work with AI. Stop expecting it to know things and start focusing on feeding it the right context.

That’s not prompt engineering. That’s context engineering. In the next video, we’ll look at what a context window is, the space the model works inside when it predicts.