Learn AI, visually
Understand how AI models actually work.
No jargon, no math walls. Each concept is a short lesson with an animation you can play with — watch the mechanism move, poke it, and it sticks.
51 lessons, one color each — hover to preview, tap to jump
Foundations · 16 lessons
Linear Regression
Fitting a line before fitting a brain
Begin →The Neuron
The atom of every model
Begin →Backpropagation
Blame, flowing backwards
Begin →Gradient Descent
Rolling downhill to a better model
Begin →Loss Functions
Putting a number on wrongness
Begin →Training
How a model learns from mistakes
Begin →Train / Validation / Test
Grade on questions never seen
Begin →Data & Datasets
The exam you must not study from
Begin →Synthetic Data
Training on a model's own words
Begin →Curriculum & Data Mixes
A model is what it eats
Begin →Layers
How neurons stack into a network
Begin →Normalization
Why deep nets need thermostats
Begin →Seeing Images
How models recognise pictures
Begin →Tokens
The chunks a model actually reads
Begin →Embeddings
How words become numbers
Begin →Sequence Memory
Reading one step at a time
Begin →Transformer Core · 8 lessons
Positional Encoding
How attention knows word order
Begin →Attention
How models weigh what matters
Begin →The Transformer Block
Attention, residuals and FFN assembled
Begin →Language Models
Predicting the next word
Begin →Context Window
The model's working memory and its cost
Begin →KV Cache & Inference
Remember, don't recompute
Begin →Batching & Serving
One GPU, many conversations
Begin →Temperature
How randomness shapes output
Begin →The Modern Era · 25 lessons
Fine-Tuning
Turning a predictor into an assistant
Begin →RLHF & Alignment
Teaching a model what people prefer
Begin →LoRA
Fine-tuning without touching the giant
Begin →Reasoning
Thinking before answering
Begin →Agents
Models that take actions
Begin →Tool Use
How a model calls the outside world
Begin →MCP & Agent Protocols
A universal port for tools
Begin →RAG
Grounding answers in real documents
Begin →Embedding Search
Find by meaning, not keywords
Begin →Embedding Search
Find by meaning, not keywords
Begin →Mixture of Experts
Big models that run cheap
Begin →Diffusion
Making images from noise
Begin →Multimodal
Seeing and reading together
Begin →Vision Transformers
An image is worth 16×16 words
Begin →Quantization
Shrinking models to run anywhere
Begin →Transfer Learning
Start smart, not from scratch
Begin →Distillation
A big teacher, a small student
Begin →Speculative Decoding
A small drafter, a big verifier
Begin →Autoencoders
Compress, then reconstruct
Begin →Model Merging
Two models, one set of weights
Begin →State Space Models
The sequence model after transformers
Begin →Video & Audio Generation
Diffusion learns to move
Begin →Voice Agents
Hear, think, speak — fast enough
Begin →World Models
Imagining before acting
Begin →GANs
Two networks in a duel
Begin →Practical Literacy · 17 lessons
Scaling Laws
Bigger is predictably better
Begin →Overfitting
Memorising versus learning
Begin →Regularization
Keeping models honest about noise
Begin →Prompting
How wording steers the answer
Begin →Evaluation & Benchmarks
Why the best model depends on the test
Begin →Hallucination
Why models invent confidently
Begin →Precision, Recall & the Confusion Matrix
Accuracy can lie
Begin →Bias & Fairness
Models inherit the world's skew
Begin →Explainability (XAI)
Why did the model say that?
Begin →AI Detection & Watermarking
Can you tell what wrote it?
Begin →Prompt Injection & Safety
When data becomes instructions
Begin →Red-Teaming & Jailbreaks
Attack it before someone else does
Begin →Privacy & Local Models
Where does your prompt actually go?
Begin →Federated Learning
The model travels, not your data
Begin →Deployment & Drift
Models rot in a changing world
Begin →Cost & Energy
What a token really costs
Begin →Choosing a Model
Capability, cost, speed — pick two
Begin →Classic ML · 9 lessons
Decision Trees
Learning by asking questions
Begin →Ensembles & Boosting
Many weak learners, one strong vote
Begin →kNN & SVMs
Classify by neighbours or by margin
Begin →Naive Bayes
Every word is evidence
Begin →Reinforcement Learning
Learning from reward, not examples
Begin →Time Series Forecasting
Trend + season + noise
Begin →Clustering
Finding groups nobody labelled
Begin →PCA & Dimensionality
Find the axis that matters
Begin →Recommender Systems
People like you also liked...
Begin →