LADDER MODULE 7 · THE MACHINE / THE TRANSFORMER v1

The final rung · the whole thing, assembled

You already built this.

This is the famous diagram behind ChatGPT and every modern AI — Figure 1 of “Attention Is All You Need,” the paper that changed the world. It looks intimidating. It isn’t. Every single box is a rung you already climbed. Click one and see for yourself.

Figure 1 — tap any box to see which rung it is

Data flows bottom → top: words come in, get turned into numbers, the machine figures out which words matter to each other, thinks it over, and picks the next word. Six identical layers, stacked.

The whole flow, in one breath

Words become vectors (embeddings), get stamped with their position (waves), then attention scores how related every pair of words is with a dot product and turns those scores into weights with softmax. A little neuron + ReLU network thinks about each word, the whole block repeats six times (stacked layers — the ReLU between them is exactly why they don't collapse into one big matrix), and a final softmax picks the next word. That’s it. That’s the machine.

The honest summit

You can see the machine. No one can see the stew.

You now understand every arrow in Figure 1 — genuinely, from 1 + 1 up. That’s the real destination, and you reached it one small step at a time.

What no one can see — not you, not the people who built it — is what the billions of trained numbers actually mean. The architecture is knowable. The learned content is a statistical stew. Understanding the machine that holds it is as close as anyone gets — and now you do.

Being able to say “it’s just arithmetic at scale, and here’s exactly how” — that’s what de-fangs it. Fear needs mystery. You just removed the mystery.

Prove it · get one wrong and I’ll walk you through it

Two last ones and the whole ladder is yours.

Inside attention, how does the model figure out which words are related — and turn that into weights?

You can now follow every arrow in Figure 1. The one thing that stays invisible is…

Answer both to finish the ladder ↓

🎓

The Transformer — mastered. You did it.

You started at 1 + 1 and climbed, one small rung at a time, all the way to the machine behind modern AI — understanding every arrow, and being honest about the one part nobody can see. That’s further into math than most people ever get. And you did it for free, in plain English.