Skip to content
Atlas
Start
Back to AI agent engineering

Module 01 · Base

LLM fundamentals

What a model actually computes, why tokens and the context window are the constraint that binds first, how a sampler turns a distribution into a token, and why a conversation is a list you own and resend.

4 lessons · 3 graded exercises·3.5 hours·beginner·Prerequisite: Foundations
01

What a model does

25 min
A model scores every token in its vocabulary and something else picks one. The confident wrong answers, the missing memory and the cost of an agent loop all fall out of that single fact.
02

Tokens and the context window

35 min · exercise
The model reads subword tokens, not words or characters, and everything you send plus everything you get back has to fit one shared budget. In an agent loop that budget is what breaks first.
03

Sampling

30 min · exercise
The model returns a distribution and the sampler turns it into a token. Temperature, top-p and top-k are three ways of reshaping that distribution, and none of them make a wrong answer right.
04

The conversation loop

30 min · exercise
A chat API is stateless, so the conversation is a list you own, grow, and resend in full every turn. What you keep, what you pin and what finish_reason tells you are all your decisions.