🧭 The Complete Journey · 15 Modules

AI Fundamentals

From a single pixel to building your own working language model. Every concept behind modern AI — taught visually, one interactive module at a time. Start at the top and click Continue through to the end.
▶ Start with Module 1
15modules
4parts + capstone + bonus
50+interactive games
0→LLMfull arc
PART A

Foundations — How a Neural Network Works

Before language, the basics: how a computer sees, what a neuron is, and how a network learns from its mistakes.

MODULE 1
A Picture Becomes Numbers
How an image turns into a flat list of numbers a network can read.
pixelsvectors
→
MODULE 2
The Neuron & the Network
Weights, biases, layers — the building blocks of every model.
neuronslayers
→
MODULE 3
How a Network Learns
Cost, gradient descent & backprop — "turning the knobs" to improve.
trainingbackprop
→
↓
PART B

Words & Language — Teaching Machines to Read

Now apply networks to text: turn words into meaning-vectors, predict what comes next, and judge sentiment.

MODULE 4
Words as Numbers
word2vec & embeddings — why king − man + woman ≈ queen.
word2vecembeddings
→
MODULE 5
Predicting the Next Word
Bigrams, trigrams, n-grams & neural prediction — the core LLM task.
n-gramsprediction
→
MODULE 6
Reading the Mood
Sentiment analysis — classifying text as positive or negative.
sentimentclassify
→
↓
PART C

Sequence & Attention — The Road to Transformers

The heart of the course: how models handle order & memory, why the old way failed, and the breakthrough that fixed it — attention.

MODULE 7
Memory & Context (RNNs)
The first models that remember earlier words via a hidden state.
RNNmemory
→
MODULE 8
The Vanishing Gradient
Why RNNs forget long context — the wall that forced a new idea.
the limit
→
MODULE 9
Attention: The Big Idea
"Attention Is All You Need" — every word looks at every word.
attention2017
→
MODULE 10
Query, Key, Value
The mechanism inside attention — a search engine per word + multi-head.
Q·K·Vmulti-head
→
↓
PART D

Transformers & LLMs — The Modern Machine

Assemble everything into the full Transformer, look inside a decoder as it writes, and run a model yourself.

MODULE 11
The Transformer: Encoder vs Decoder
The full architecture; GPT (writer) vs BERT (understander).
encoderdecoder
→
MODULE 12
Inside the Decoder
Tokenize → embed → position → layers → output → auto-regression.
tokensgeneration
→
MODULE 13
Running LLMs Yourself
Ollama, system prompts, temperature & top-p, activation functions.
Ollamatemp/top-p
→
↓
🏗️ CAPSTONE

Build an AI — Put It All Together

The finale. Assemble a working text model from raw data to generation — using everything you learned.

MODULE 14 · CAPSTONE
The Build Simulator
Dataset → Tokenize → Embed → Architecture → Train → Generate. Build an LLM end-to-end in a sandbox.
full pipelinehands-on
→
↓
🎓 BONUS

How Models Really Learn

The deep-dive everyone asks about: how a real LLM actually gets trained to become a helpful assistant.

MODULE 15 · BONUS
Pretraining → Fine-tuning → RLHF
Where knowledge comes from, how it learns the assistant format, and how human feedback polishes it.
pretrainingRLHF
→

🎓 By the end, you'll understand — and be able to build — the whole pipeline:

✓ Pixels → vectors ✓ Neurons & learning ✓ Word embeddings ✓ Next-word prediction ✓ Sentiment ✓ RNNs & memory ✓ Vanishing gradient ✓ Attention ✓ Query/Key/Value ✓ Transformers ✓ Inside the decoder ✓ Running LLMs ✓ Building one yourself ✓ How models are trained (RLHF)
How to use this map: each module is self-contained and builds on the last. Click any card to jump in, or just hit "Start with Module 1" and use the Continue → button at the bottom of every page to walk the whole path. You can always come back here to jump around.
▶ Begin the journey — Module 1
🚀 Finished the fundamentals? Continue to Applied AI — the hands-on sequel where you build it all in real PyTorch: train a GPT from scratch, fine-tune with LoRA, and set up RAG.
⚙️ Go to the Applied AI course →
AI Fundamentals · a 14-module visual course · rebuilt from your notes 🏁