Agentic AI Chapter 16 / 16 · Learning Path

16 Your Learning Path

Now go build one.

You've covered an enormous amount of ground. Here's the most important thing anyone can tell you: you learn agentic AI by building agents — not by reading about them or watching videos. This is a 10-week roadmap from zero to something real, and every phase links back to the chapter that teaches it. Start small, ship end to end, then add complexity.

10 weeks Build, don't watch Zero → production

The plan

Ten weeks, six milestones.

Each phase is a thing you build, not a thing you read. Tap any phase to see what to make — and jump straight to the chapters behind it.

Weeks 1–2Fundamentals

Get the mental model and your hands dirty: understand how LLMs actually work, learn the basics of prompt engineering, make your first real API call, and build a simple chatbot. Get comfortable with the dev tools.

Weeks 3–4Your first agent

This is where it clicks. Implement the ReAct loop yourself, from scratch. Add a web-search tool. Add code execution. Build something that uses tools to answer questions it couldn't answer alone.

Weeks 5–6RAG & memory

Set up Chroma locally (5 minutes). Build a RAG pipeline over documents you actually care about. Experiment with chunking strategies. Add memory so your agent remembers across sessions.

Weeks 7–8Architectures & frameworks

Learn a framework properly. Implement Plan & Execute for a structured workflow. Add long-term memory. Explore MCP servers so your agent can plug into real tools.

Weeks 9–10Multi-agent systems

Build a two-agent system where one agent checks the other's work. Implement an orchestrator–worker pattern. Try a role-based framework for a team of agents. This is where the real power becomes obvious.

Final stretchProduction

Add guardrails and safety checks. Set up observability so you can see exactly what your agent is doing at every step. Measure success rate, step efficiency, latency, and cost. Then deploy something real — start small: one agent, one task, a few tools, working end to end.

The whole journey

What you now understand.

Sixteen chapters, one arc. You started with what AI even is and ended knowing how to build and ship an agent safely. Here's the shape of it:

Ch 1–2

The engine

Neurons, transformers, and how LLMs predict text.

Ch 3–5

The leap to agents

Autonomy, the loop, and the tools that reach the world.

Ch 6–10

Making them capable

Memory, RAG, vector DBs, embeddings, and MCP.

Ch 11–15

Scale & ship

Architectures, multi-agent, frameworks, safety, applications.

The beginning, really

Close the tab and go build something.

The shift from AI that answers questions to AI that completes goals is as big as the shift from static pages to real applications — maybe bigger. The people who understand not just what these tools are but how they fit together, and how to build safely, are going to be extraordinarily valuable. You now have that foundation. What you do with it is up to you.