CHAPTER 1 · ~30 MIN · CINEMATIC

What Generative AI
Actually Is.

Prediction, not magic: what these models are, how they differ from every AI that came before, and why 2022 changed.

5toolsone beating heart

3pillarsprobability · learning · building

70yearsrules → statistics → deep → generative

0mathsanalogies, not equations

ENkaraoke subs

CT-GENAI-900 · Generative AI: From Tokens to Agents

▶ VIDEO · CHAPTER 1 · ~30 MIN · CINEMATIC

From a single token to a working application

The player screen is a live animation stage with real diagrams — the map from token to application draws itself left to right, the critic and the creator split the screen, the three pillars rise one at a time, and a real softmax rolls a weighted die fourteen times on camera. Every idea lands exactly as the narration reaches it. Karaoke subtitles in English, fullscreen.

That was the first 10 minutes.

The rest of this chapter — and chapters 2 to 5 — come with lifetime access.

$11.99one payment · lifetime access

Get the full course Already bought it? Sign in

Chapter 1 · Foundations · Generative AI: From Tokens to Agents

Foundations: five tools, one beating heart

0:00 / 0:00

☰

REFERENCE · THE MESSAGE CATALOGUE

The concept catalogue →

Every idea the chapter builds — what it is, what it does, and where it lands in the story.

REFERENCE · THE CHAPTER 1 CONCEPT CATALOGUE

Every idea in the chapter

Each concept the video builds — what it is, what it does, and where it lands in the story. Click any row to open its full breakdown. Filter to find any of them.

Idea	Category	What it does	In one line	Lesson

Chapter 1 · Foundations — What Generative AI Actually Is

REFERENCE · THE MAP · TOKEN → APPLICATION

The whole course, on one page

Everything Chapter 1 promises, as a reference you can scan: the journey from a single token to a working application, the two kinds of AI, the three pillars under every generative model, and the four eras that led here.

The journey — piece by piece

1 · Tokenthe little chunk of text a model actually sees, instead of words

2 · Embeddingthe trick that turns tokens into points in a vast space of meaning

3 · Attentionthe mechanism that lets a model weigh what matters in a sentence

4 · Transformerthe architecture those pieces assemble into — it powers everything

5 · Trainingthree stages: a newborn model that knows nothing → a polished assistant

6 · Samplinghow it chooses its words, one at a time

7 · Hallucinationwhy it sometimes confidently makes things up

8 · Beyond textimages from pure noise, music, code, one model that sees and hears

9 · Using it wellprompting, retrieval, agents, and a real reliable product

The two kinds of AI

	Discriminative	Generative
the question	“What is this?”	“Make me a new one.”
what it learns	the boundary between things	the entire shape of the data
output	a label on something that exists	something that did not exist before
examples	spam filter · face tagging · star rating	ChatGPT · Midjourney · Suno · Sora · Copilot
the metaphor	a critic — judges what exists	an artist — makes what does not

The three pillars — under every generative model on Earth

ProbabilityIt never declares “this is the answer.” It says “given everything I have seen, this is the most likely answer.” Every output is a roll of beautifully weighted dice.

LearningThose odds were not written by any human hand. They were learned automatically, from enormous mountains of training data.

BuildingIt assembles its answer by stringing those probabilities together, one small piece at a time. Token after token. Pixel after pixel.

The four eras

Era	What AI was	What it could do
1950s	hand-written rules	thousands of brittle if-then statements
1990s	statistical	learned patterns from data — but only in narrow, specific tasks
2010s	deep learning	machines could finally see and hear — but they still mostly classified
2020s	generative	the same network that learned to recognise a cat was taught to paint one

Every idea in this chapter — searchable

Filter by idea, category or lesson.

Chapter 1 · Foundations — the one sentence for the whole chapter: a generative model learns the shape of its data, and then invents brand new examples that fit.

QUIZ · 10 QUESTIONS

Foundations check