A Four-Week Agentic AI Reading Challenge, Completed with Harness Engineering

[4-Week Complete Reading Challenge] Building Agentic AI with Harness Engineering

AI agents: How to make them work properly, not just build them!

You can build an AI agent with just a few lines of prompts. But in real-world environments, it may repeat tool calls, lose track of context, or cause unexpected errors. What you need now is the ability to control execution flows and safeguards in code so AI works reliably.

『Agentic AI, Built with Harness Engineering』 is a hands-on guide covering everything from implementing ReAct yourself to memory, RAG, LangGraph, multi-agent systems, three real-world projects, security, and deployment.

In this challenge, we’ll read a book together over 4 weeks, work through the key examples ourselves, and learn the principles and design methods of agent engineering. A 420-page hands-on guide that may have felt daunting to read alone—let’s finish it together!

✅ Activity period: November 23 (Mon), 2026 ~ December 20 (Sun)

• Week 1: 11/23 (Mon) ~ 11/29 (Sun) | Core principles of agents

• Week 2: 11/30 (Mon) ~ 12/6 (Sun) | RAG, LangGraph, and Multi-Agent Systems

• Week 3: 12/7 (Mon) ~ 12/13 (Sun) | Three hands-on projects

• Week 4: 12/14(Mon) ~ 12/20(Sun) | Security, deployment, operations, and a look back at the journey

• Author livestream: Scheduled for Tuesday, 12/15, at 8 PM

✅ How to Participate

1. Read and study the book according to the weekly curriculum.

• Read the assigned sections from Monday to Friday.

• On the weekend, complete the mission based on that week’s learning materials.

• You can find each week’s reading assignments and mission in the curriculum.

2. Take on the weekly challenge. (Once a week)

Please write a review of at least 100 characters and include a screenshot of your practice results or a summary of what you learned.

It’s okay if you don’t succeed in the hands-on exercise. Documenting the error screen, how you tried to resolve it, and what you learned is also a meaningful learning experience.

✅ Participation Benefits

🎁 Completers receive up to 20,000 Hanbit+ mileage points

• Book verification (once during the activity, applicable to all participants)

• Complete the Week 1/2 missions – Earn 10,000 Hanbit+ Mileage points

• Complete the Week 3/4 missions and receive 10,000 Hanbit+ mileage points.

🎤 Author Online Live Seminar & Q&A

✅ Frequently Asked Questions (FAQ)

Q. Can Python beginners participate?

This book is for beginner-to-intermediate developers who can write and run code in Python. If you know the basic syntax and how to use a terminal, you can take part. Experience calling LLM APIs is helpful but not required. However, the exercises may be somewhat challenging for those who are just starting to learn Python.

Q. Do I need to subscribe to a paid Claude plan?

A paid subscription to a chat service like Claude Pro is not required. This book primarily uses the Anthropic API to implement agents in Python. A Claude chat subscription and API usage fees are separate, and charges may apply depending on your API usage. We recommend setting an API usage limit before starting the exercises.

Q. What do I need to prepare for the hands-on exercises?

• The book *Agentic AI with Harness Engineering* (paperback or e-book)

• A PC capable of running Python exercises

• Python 3.12 development environment

• Anthropic API key, Voyage API key

• An OpenWeatherMap API key for the Week 1 weather exercise

You can find the example code in the repository below.

https://github.com/javafa/agentic_ai_engineering

Q. Do I need to run all the projects myself?

No. In Week 3, you only need to choose one of the three projects and complete the mission. If your practice environment isn’t ready or errors prevent you from running the project, you can still participate by summarizing its design structure or learning materials.

Q. Do I need a separate high-performance GPU?

Most of the basic examples can be studied in a standard Python development environment.

However, the hands-on local serving exercise with vLLM in Chapter 13 requires an NVIDIA GPU environment. If you don’t have a GPU, you can instead study the concepts and example code in that chapter.

Q. Are there any additional costs for the hands-on exercises?

Some examples use paid LLM APIs or external services such as AWS, so costs may vary depending on usage. Cloud deployment and GPU-based exercises are optional, so you can choose whether to do them based on your setup.

Go beyond writing good prompts and learn how to design AI agents that work reliably!

Learn agent engineering over four weeks by reading, building, and documenting together.