Give you the map, not just the tools
Most AI material is a pile, not a path — each article quietly assuming the thing that would have explained it. So people read about attention before embeddings, get lost, and conclude the material is too hard.
The concepts have a real dependency order. These 150 articles are laid out in it, across 17 layers, with 16 interactive explainers for the mechanisms prose explains badly.
A pile versus a path
The pile
- Attention explained before embeddings
- RAG before hallucination, so the fix arrives before the problem
- Every article written as if you already read the rest of the pile
- No signal for whether you have the background yet
- You do not know what to read next, so you stop
The path
- Ordered by what genuinely depends on what
- Every page shows its layer, its prerequisites, and its next step
- Three routes: use them well, build with them, or know how they work
- Machine-checked dependencies — no article assumes something missing
- You always know exactly where you are
Three routes, one map
Depth decides what you study; time decides only how long it takes. Pick the depth that matches what you actually want to do.
The 17 layers
The first seven are the main path. The rest go deeper — most people need only some of them.
- 1 · Entry 4
The AI Jargon Decoder · In What Order Should You Learn AI? · What AI Can and Cannot Do (Yet) · +1 more
- 2 · ML Foundations 12
Probability for AI · What Is Machine Learning? · How Data Becomes Numbers · +9 more
- 3 · Classic NLP 8
Recurrent Neural Networks · LSTM and GRU: Memory That Lasts · Sequence to Sequence and Encoder-Decoder · +5 more
- 4 · Foundations 13
Logits and Softmax · What Is a Token in AI? · What Is an Embedding? · +10 more
- 5 · Internals 13
Decoding Strategies: Greedy Search and Beam Search · The Geometry of Meaning · How Does a Transformer Work? · +10 more
- 6 · Prompting 10
What Is a Context Window? · Why Do AI Models Hallucinate? · Prompt Basics · +7 more
- 7 · RAG 7
Why RAG Exists · How RAG Works, Step by Step · How to Split Documents for RAG · +4 more
- 8 · Agents 9
What Is an AI Agent? · Agent Architectures: From Rules to Autonomy · How Does Tool Calling Work? · +6 more
- 9 · Application 6
Your First LLM API Call · How Do You Know It Works? · From Demo to Production · +3 more
- 10 · Evaluation 11
The Bias-Variance Tradeoff · Why Evaluation Is Harder Than Training · Benchmark Contamination and Goodhart's Law · +8 more
- 11 · Multimodal 8
How Do Models See Images? · How Image Generation Works · Multimodal Embeddings · +5 more
- 12 · Training and Adaptation 12
How Fine-Tuning Actually Works · Model Distillation · Hyperparameter Tuning · +9 more
- 13 · Advanced Architecture 8
Attention Variants: Sparse, Linear, and Sliding Window · How Embedding Models Are Trained · How Inference Serving Works · +5 more
- 14 · Advanced RAG 7
Agentic RAG · Contextual Retrieval · GraphRAG · +4 more
- 15 · Advanced Agents 7
Evaluating Agents · Agent Memory · Coding Agents · +4 more
- 16 · Security 7
Model Supply Chain Risk · Handling PII and Sensitive Data · AI Governance and Compliance · +4 more
- 17 · Production 8
Latency Optimization · Observability for LLM Systems · Cost Monitoring and Attribution · +5 more
Start where everyone should start
Not with the tools, and not with a definition. With what the words actually mean — starting from a single token.