Learn

AI Agents

From 'what is an agent?' to shipping reliable, evaluated, secure agentic systems.

Deep, first-principles command of AI agents as a software engineer, so I can design, build, debug, and lead work on agentic systems. Built visual-first and practical, for long-term retention: model, loop, tools, memory, orchestration, evaluation.

Written by Claude Opus 4.8 28 lessons, 9 modules updated Jul 2026

Module 0 Orientation

start here

What is an AI agent?

Agent vs workflow, the autonomy spectrum

Anatomy of an agent

Model, loop, tools, memory, orchestration

Module 1 The Reasoning Core

the engine

LLMs as the engine

Tokens, context window, sampling

Reasoning and test-time compute

CoT, reasoning models, thinking budgets

Prompting for agents

System prompts, few-shot, structure

Module 2 The Agent Loop and Acting

how agents do things

The agent loop and ReAct

Perceive, reason, act, observe

Tool use and function calling

Schemas, tool calls, error recovery

Structured outputs

JSON mode, constrained decoding

Planning and decomposition

ToT, plan-and-execute, subtasks

Reflection and self-correction

Reflexion, self-critique loops

Module 3 Memory and Knowledge

giving agents recall

Context engineering

Managing the context window

Embeddings and vector search

Semantic similarity, vector DBs

Retrieval-augmented generation

The RAG pipeline, chunking, reranking

Agent memory

Working, episodic, semantic memory

Module 4 Multi-Agent Systems

many agents, one goal

Multi-agent patterns

Supervisor, hierarchical, swarm

Orchestration and handoffs

Routing, delegation, shared context

Module 5 Protocols and Frameworks

the plumbing

Model Context Protocol (MCP)

The USB-C of tools

Frameworks landscape

LangGraph, CrewAI, AutoGen, SDKs

A2A and interoperability

Agents talking to agents

Module 6 Reliability, Evaluation and Safety

making it trustworthy

Evaluating agents

Evals, benchmarks, LLM-as-judge

Observability and tracing

Seeing inside the agent

Guardrails and failure modes

Loops, cost blowups, drift

Security and prompt injection

The lethal trifecta, defenses

Module 7 Applied Agents

agents in the wild

Coding agents

How Claude Code and friends work

Computer and browser use

Agents that click and type

Training agents and agentic RL

RLVR, GRPO, tool-use fine-tuning

Module 8 Production and Frontier

ship it, then what is next

Shipping to production

Cost, latency, HITL, reliability

The landscape and the future

Where agents are heading

Reference shelf

Glossary

Every agent term, crisply defined and cross-linked.

Patterns cheat sheet

The core agent and workflow patterns on one page.

Frameworks comparison

LangGraph vs CrewAI vs AutoGen vs SDKs, side by side.

How to use this course

Do the widgets. The quizzes and flashcards are not decoration: trying to recall before you reveal is what makes it stick.
Space it out. Come back a day later and answer them cold.
Read critically. An LLM wrote these, so verify anything before you rely on it.