Learn what happens inside a language model.
LLM Primer begins with the programming and mathematics that make machine learning possible. From there, it builds toward neural networks, language models, Transformers, and complete systems.
Why this publication exists
Language models are easy to try. It is harder to explain how text becomes numbers, how a model learns from examples, or why it chooses one word instead of another. This publication is for readers who want to answer those questions, not only use a finished tool.
The route is deliberately gradual. A small program comes before a training loop; a hand-worked probability before a loss function; a simple neural network before a Transformer. Words, equations, code, and diagrams show the same idea from different sides.
The aim is not to finish pages quickly. It is to reach the point where you can explain an idea, check it with numbers, and build a small version yourself.
Who this is for
The Foundation Path is written for a curious learner who has studied mathematics at high-school level and may be new to machine learning. You may be in school or university, learning on your own, or returning to the foundations after working in another field.
Previous programming experience will help, but the Python subject begins with small programs. The path still takes patience and practice. Difficult chapters are meant to be worked through with a notebook, code editor, and time for the exercises—not skimmed in an afternoon.
How to read the lessons
Pause before running an example and predict what it will do. Work through a small calculation before accepting the general rule. Change an input and explain the result. Attempt an exercise before opening its hint. These small acts turn reading into understanding.
When a step feels mysterious, follow its prerequisite instead of memorizing the sentence around it. Readers with more experience can begin at the first subject whose ideas are unfamiliar.
What you can expect
- Free foundation
- The published core lessons, essential diagrams, standard exercises and projects, and browser examples in the Foundation Path remain free.
- Accounts
- Reading does not require an account. Signing in adds continuity across devices for supported progress, responses, reviews, and notes.
- Additional services
- Paid services may later add managed computing, evaluated projects, or personal guidance. They will not hide prerequisite explanations from the free path.
Some ideas in AI have been studied for decades. Others are still changing quickly. We will not write about both with the same certainty. Established material should be explained and checked; newer material should name the evidence, limits, and open questions.
The Curriculum shows the complete sequence, prerequisites, optional subjects, later directions, and current publication state.
Open the curriculum