Stage 1 of 10 · 6 lessons · ~2 weeks

How agents work

Understand what makes a program an agent, the loop every agent runs, how agents differ from fixed workflows, and how to decide whether a task needs an agent at all.

Start: What an agent is

Lessons· one lesson for each topic, in this order

Do this

Read an overview of agents versus workflows from a model provider.
Sketch the same problem as a workflow and as an agent, and compare their cost and risk.
Write the agent loop in pseudocode, including a step limit.
List the tools your first agent needs and the ground truth each returns.
Mark every irreversible action and decide how it will be approved.
Write the one-page design note for your agent.

Start with· follow one of these

CoursePaid
Grokking Modern AI Fundamentals
Design Gurus · Beginner · $72 one-time
Chapter 8, Agentic AI, covers reasoning, planning and memory in agents and ends with a small agent project
CourseFree
Hugging Face AI Agents Course
Hugging Face · Beginner
Free: Unit 1 explains the thought, action, observation cycle and builds a first agent step by step

Practise

AI Agents for Beginners
Short lessons with runnable code: work through the lessons on agentic design and tool use

Keep handy

Building Effective Agents
The clearest description of workflows versus agents and the common patterns; reread it at every stage
roadmap.sh: AI Agents
A one-page map of agent topics, useful to see where each stage of this roadmap fits

Also worth your time

What Is Agentic System Design? (opens in a new tab)Read
Levels of autonomy, the five parts of an agent system and why agent loops must be bounded
Berkeley Large Language Model Agents CourseGo deeper
University lectures on agent reasoning, planning and tool use, for the research background

Before moving on: A one-page design note for an agent you want to build: why it needs an agent rather than a workflow, its tools, what ground truth each step returns, which actions are irreversible and what limits it will have.

Tauseef's advice

Most agent tutorials start with a framework and a demo. Start with the question of whether you need an agent at all, and with the loop itself. If you already built tool calling in the AI engineering roadmap, skim this stage, but still write the design note: it will save you weeks later.