ai-explained

the curriculum

Pick a path. Climb it. Walk away knowing how AI works.

Four tracks, four reasons to be here. The whole site is structured around your starting point — not ours.

108
articles
57
demos
13
animated walkthroughs
7h 45m
of reading
step 1 · pick your starting line
1
"I want to understand how machines learn at all."
Foundations: linear algebra → probability → neural nets → language modeling.
24 articles 16 demos 1h 25m
first interactive demo ▶ Gradient Descent
start here →
2
"I know ML basics. How does a transformer actually work?"
LLM Internals: tokenization → attention → transformer block → modern architectures.
18 articles 13 demos 1h 9m
first interactive demo ▶ Attention Inspector
start here →
3
"I want to build with LLMs. Show me what works."
Builder track: prompting, RAG, fine-tuning, agents, multimodal — the production toolbox.
39 articles 20 demos 2h 55m
first interactive demo → RAG Visualizer
start here →
4
"I'm shipping AI. How do I keep it shipped?"
Production track: cost, latency, evals, guardrails, observability, applications.
19 articles 8 demos 1h 31m
first interactive demo ▶ Observability Trace
start here →
step 2 · or just spelunk

If you have 30 minutes

Five hand-picked interactive demos, in order. No prerequisites — open one in a tab, lose track of time. Each is built on real models or real algorithms; nothing's a cartoon.

1
Inference Pipeline
The capstone. Question → tokens → embeddings → 12 layers → logits → sampling → next token → answer. Step through a real GPT-2 forward pass end to end.
2
Attention Inspector ▶ animated
Three modes: narrated walkthrough, interactive inspector, and a Composition view that drives one slider through pattern → diversity → entropy → semantic resolution. Real GPT-2 attention.
3
Diffusion Process ▶ animated
Voice-narrated forward + reverse diffusion on a 2D smiley, plus the full interactive sandbox with four shapes. The toy version of every image generator.
4
RAG Visualizer
Two modes: a 3-strategy comparison (dense / BM25 / hybrid) and a Composition view where top-k drives a wave through scores → retrieved → recall → cost-per-query.
5
RLHF Preference Simulator
Pick A or B on paired completions. Watch a reward model emerge from your preferences. The pattern behind RLHF, DPO, and every aligned chat model.
step 3 · or read the whole thing in order

The full curriculum

15 stages, top to bottom. Each stage is a coherent group of articles plus one or more demos that bring the lesson to life. Click any article to start; the demo pills are shortcuts.

1

Foundations

linear algebra → probability → neural nets → language modeling.

01

Math Foundations

Linear algebra, probability, calculus, info theory.

5 articles 23 min 5 demos · 1 ▶
  1. Calculus & Optimization
    5 min ▶ 1 demo
    ▶ Gradient Descent
  2. Information Theory for ML
    4 min → 1 demo
    → Entropy & KL Divergence
  3. Linear Algebra for ML
    6 min → 2 demos
    → Linear Algebra Lab → Eigenvectors Lab
  4. Probability & Statistics for ML
    6 min → 1 demo
    → Bayes Theorem Lab
  5. Stage 01 — Math Foundations
    2 min
02

ML Fundamentals

Supervised, unsupervised, evaluation, regularization.

7 articles 25 min 4 demos · 1 ▶
  1. Classical ML Algorithms
    4 min → 1 demo
    → Linear Regression Lab
  2. Evaluation & Metrics
    5 min → 1 demo
    → Confusion Matrix Lab
  3. Loss Functions & Optimization
    4 min
  4. Stage 02 — ML Fundamentals
    2 min
  5. Regularization & Generalization
    4 min → 1 demo
    → Bias-Variance Tradeoff
  6. Supervised Learning
    3 min
  7. Unsupervised Learning
    3 min ▶ 1 demo
    ▶ K-Means Clustering
03

Neural Networks

Backprop, activations, optimizers, regularization.

7 articles 22 min 4 demos · 1 ▶
  1. Activations & Initialization
    4 min → 1 demo
    → Activation Functions Lab
  2. Architectures: CNNs and RNNs
    4 min
  3. Backpropagation
    3 min ▶ 1 demo
    ▶ Backpropagation Walkthrough
  4. Optimizers
    3 min → 1 demo
    → Optimizer Race
  5. Perceptrons & MLPs
    3 min → 1 demo
    → Neural Network Trainer
  6. Stage 03 — Neural Networks
    1 min
  7. Regularization Techniques
    4 min
04

Language Modeling

n-grams to RNNs to why transformers won.

5 articles 15 min 3 demos · 1 ▶
  1. n-gram Models
    3 min → 2 demos
    → n-gram Language Model → Perplexity Calculator
  2. Neural Language Models
    3 min
  3. Stage 04 — Language Modeling
    1 min
  4. RNNs and LSTMs as Language Models
    4 min ▶ 1 demo
    ▶ RNN Unrolling
  5. Why Transformers Won
    4 min
2

LLM Internals

tokenization → attention → transformer block → modern architectures.

05

Tokens & Embeddings

How text becomes vectors.

5 articles 17 min 3 demos
  1. Contextual Embeddings
    4 min
  2. Stage 05 — Tokens & Embeddings
    1 min
  3. Semantic Geometry
    4 min → 1 demo
    → Embedding Playground
  4. Static Embeddings
    3 min → 1 demo
    → Word Arithmetic
  5. Tokenization
    5 min → 1 demo
    → Tokenizer Surgery
06

Transformers

Self-attention, multi-head, KV caching, GPT from scratch.

6 articles 22 min 6 demos · 2 ▶
  1. GPT From Scratch
    3 min
  2. Multi-Head Attention
    3 min → 1 demo
    → Head Specialization Gallery
  3. Positional Encoding
    4 min ▶ 1 demo
    ▶ Positional Encoding Lab
  4. Stage 06 — Transformers
    2 min
  5. Self-Attention (KQV)
    6 min ▶ 1 demo
    ▶ Attention Inspector
  6. The Transformer Block
    4 min → 3 demos
    → KV Cache → Layer Norm → Inference Pipeline
07

Modern LLMs

Scaling laws, MoE, reasoning models, long context.

7 articles 30 min 4 demos · 1 ▶
  1. Field report: DeepSeek-R1 — reasoning from pure RL, in the open
    8 min
  2. Frontier Architectures
    5 min
  3. Long Context
    4 min → 1 demo
    → Long Context Attention
  4. Mixture of Experts (MoE)
    3 min ▶ 1 demo
    ▶ MoE Routing Visualizer
  5. Stage 07 — Modern LLMs
    2 min
  6. Reasoning Models
    4 min → 1 demo
    → Reasoning vs Non-Reasoning
  7. Scaling Laws
    4 min → 1 demo
    → Scaling Laws Calculator
3

Builder track

prompting, RAG, fine-tuning, agents, multimodal — the production toolbox.

08

Prompting

Few-shot, CoT, structured outputs, sampling.

6 articles 22 min 5 demos · 1 ▶
  1. Advanced Prompting Techniques
    4 min
  2. Few-Shot & Chain-of-Thought
    3 min → 2 demos
    → Chain-of-Thought Effect → Few-shot vs Zero-shot
  3. Prompt Fundamentals
    4 min
  4. Stage 08 — Prompting
    2 min
  5. Sampling & Decoding
    5 min ▶ 2 demos
    → Sampling Knobs ▶ Beam Search vs Sampling
  6. Structured Outputs
    4 min → 1 demo
    → Structured Outputs
09

RAG

Chunking, embeddings, hybrid search, reranking, evals.

8 articles 36 min 3 demos · 1 ▶
  1. Advanced Retrieval Patterns
    4 min
  2. Chunking Strategies
    5 min → 1 demo
    → Chunking Strategies
  3. Embedding Models for Retrieval
    5 min
  4. Evaluating RAG
    5 min
  5. Hybrid Search & Reranking
    5 min ▶ 1 demo
    ▶ Reranker Lab
  6. RAG Fundamentals
    5 min → 1 demo
    → RAG Visualizer
  7. Stage 09 — Retrieval-Augmented Generation (RAG)
    2 min
  8. Vector Databases
    5 min
10

Fine-Tuning

SFT, LoRA, RLHF, DPO, GRPO, embedding fine-tuning.

10 articles 54 min 4 demos
  1. Data & Tooling
    5 min
  2. Distillation
    8 min → 1 demo
    → Distillation Lab
  3. Embedding Fine-Tuning
    4 min
  4. Field report: Llama 3 — frontier post-training, in 92 pages
    8 min
  5. LoRA & QLoRA
    4 min → 2 demos
    → LoRA Lab → Quantization Lab
  6. Field report: Phi-3 — synthetic data and distillation, in the open
    7 min
  7. Stage 10 — Fine-Tuning
    3 min
  8. RLHF, DPO, GRPO — Preference and Reward Training
    6 min → 1 demo
    → RLHF Preference Simulator
  9. Supervised Fine-Tuning (SFT)
    4 min
  10. When to Fine-Tune
    5 min
11

Agents

Loops, tools, memory, planning, multi-agent, browser/vision.

8 articles 35 min 5 demos · 2 ▶
  1. Agent Loop & Architecture
    5 min ▶ 1 demo
    ▶ Agent Trace Viewer
  2. Browser & Vision Agents
    4 min
  3. Guardrails & Safety for Agents
    5 min
  4. Memory Systems
    4 min → 1 demo
    → Memory Systems
  5. Multi-Agent Orchestration
    5 min ▶ 1 demo
    ▶ Multi-Agent Orchestrator
  6. Planning & Reflection
    4 min → 1 demo
    → ReAct & Reflection
  7. Stage 11 — Agents
    2 min
  8. Tool Use & Function Calling
    6 min → 1 demo
    → Tool Use Builder
12

Multimodal

CLIP, VLMs, diffusion, video, speech, synthetic data.

7 articles 28 min 3 demos · 1 ▶
  1. Multimodal Embeddings (CLIP)
    4 min
  2. Stage 12 — Multimodal AI
    1 min
  3. Speech & TTS
    5 min
  4. Synthetic Data
    5 min
  5. Text-to-Image Diffusion
    4 min ▶ 1 demo
    ▶ Diffusion Process
  6. Video Generation
    4 min
  7. Vision-Language Models (VLMs)
    5 min → 2 demos
    → ViT Patches → CLIP Zero-Shot Classifier
4

Production track

cost, latency, evals, guardrails, observability, applications.

13

Production

Deployment, evals, guardrails, observability, cost, data systems.

9 articles 48 min 6 demos · 1 ▶
  1. Cost & Latency
    6 min → 1 demo
    → Cost & Latency Calculator
  2. Data Systems for AI Products
    10 min
  3. Deployment Architectures
    4 min
  4. Enterprise Considerations
    6 min
  5. Evaluation & Benchmarks
    6 min → 2 demos
    → LLM-as-Judge → Calibration Lab
  6. Guardrails
    4 min → 1 demo
    → Guardrails Lab
  7. Hallucination Mitigation
    5 min → 1 demo
    → Hallucination Mitigation
  8. Observability & Tracing
    5 min ▶ 1 demo
    ▶ Observability Trace
  9. Stage 13 — Production
    2 min
14

Applications

Text-to-SQL, code, browser agents, finance, case studies.

6 articles 27 min 2 demos
  1. Browser Agents
    4 min → 1 demo
    → Browser Agent
  2. Case Studies
    6 min
  3. Financial Reasoning
    5 min
  4. Stage 14 — Applications
    2 min
  5. Text-to-Code
    6 min
  6. Text-to-SQL
    4 min → 1 demo
    → Text-to-SQL Playground
15

Career

Roles, learning roadmap, staying current.

4 articles 16 min
  1. AI Engineer Roles
    5 min
  2. Learning Roadmap
    5 min
  3. Stage 15 — Engineering & Career
    1 min
  4. Staying Current
    5 min