AI Knowledge Atlas
AI, explained properly from first principles to the frontier
Eight domains, forty-eight core concepts. Each one runs from a precise definition to the intuition behind it, from how it works to where it breaks down — with hand-drawn SVG figures and interactive charts. Whether you have never heard of a neural network or read papers daily.
- Domains
- 08
- Core concepts
- 48
- Figures & charts
- 124
- Interface languages
- 12
STRUCTURE
The eight domains
Read top to bottom and you get a natural dependency chain: mathematics grounds machine learning, machine learning frames deep learning, deep learning carries language, vision, reinforcement learning and generative models — and engineering and governance carry them all into products.
Knowledge MapPATHS
Pick your entry point
One site, three ways to read it. No fixed order — jump sideways whenever you like.
I'm starting from zero
No maths, no code. Begin with the plainest questions — what is a vector, how does a model learn — and build intuition through analogy.
Begin with Math & StatisticsI know some, and want structure
You have heard of gradient descent and Transformers, but the knowledge is fragmented. Read domain by domain and connect the loose pieces into a net.
Begin with Machine LearningI work in this field
You need to recall a method's exact formulation and its boundary conditions in ten minutes. Go straight to entries, formulas and further reading.
Browse all conceptsDOMAINS
Domains
Math & Statistics Foundations
Vectors, matrices, probability and gradients — the language of AI
Machine Learning
Letting programs infer rules from data instead of being hard-coded
Deep Learning
Handing feature engineering over to stacked layers
NLP & Large Language Models
From tokenisation to Transformers to models that can talk
Computer Vision
Turning pixels into objects, scenes and 3D structure
Reinforcement Learning
Learning sequences of decisions from trial, error and delayed reward
Generative AI
From judging what something is to synthesising what it should be
AI Engineering, Safety & Ethics
Turning a working model into a usable, trustworthy, accountable product
SELECTED
Start here
If you read only six entries, read these six.
Vectors & Vector Spaces
AI turns everything — words, images, sounds, users — into one thing: a list of numbers
Gradients & Gradient Descent
All of deep learning comes down to one thing: take a small step downhill
Transformer Architecture
Replacing word-by-word relay with a room where everyone speaks at once, so long-range dependencies are one hop away
Diffusion Models
Learn a thousand tiny denoising steps, and you can build an image from pure noise
Reinforcement Learning from Human Feedback
When the good answer cannot be written as a formula, let humans stand in as the reward function
Retrieval-Augmented Generation
Rather than cramming knowledge into parameters, leave it outside and look it up on demand — an open-book exam instead of a closed-book one
USAGE
How to use this site
Read domain by domain
Each domain is ordered by increasing difficulty, so finishing one leaves you with a complete block of understanding.
02Jump to the knowledge map
The map draws all forty-eight concepts as one relation network, so you can see at a glance who supports whom and what replaces what.
03Use the glossary as a dictionary
When you hit an unfamiliar term, look it up directly — far faster than scrolling back through earlier chapters.
Ready?
Start with Vectors & Vector Spaces — the true foundation of modern AI.
Explore the knowledge map