Skip to content

Overview

Section 1 · Foundations — A correct mental model: models, tokens, context, the agent loop, and where agents are weak.

6 lessons.

By the end of this section you can

  • Describe what a model actually does when it answers, and why it can be confidently wrong.
  • Explain the agent loop, and why verification must be a tool result rather than an opinion.
  • Judge whether a task suits an agent — and when to keep it yourself.
  • Name the five things competing for space in a context window, and the cheapest lever on each.

How LLMs Work For Coders

beginner · ~12 min — When you ask an agent to add a feature, you are not talking to a mind that held a meeting about your codebase.

From Autocomplete To Agents

beginner · ~10 min — Autocomplete finishes your line. An agent finishes your task.

The Agent Loop

beginner · ~12 min — An agent that finishes in one shot is the exception.

What Agents Are Good And Bad At

beginner · ~12 min — Delegating the wrong task wastes more time than doing it yourself.

A Taxonomy Of Context

intermediate · ~12 min — Two agents on the same model, given the same task, can produce a working change and an unusable mess. The difference is almost never the model. It is what was in the window when they started.

Choosing A Model

beginner · ~10 min — Model names change every few months, so memorising the current best one is wasted effort.

← 0. Start Here · Home · Sidebar · 2. Prompting & Context →