THE THREE DISCIPLINES AROUND THE MODEL

CHECK YOUR UNDERSTANDING • Which discipline would you use to change the wording of one request? ◦ Answer: Prompt engineering. • What does context engineering decide? ◦ Answer: What information the model is shown at each step, and what is left out. • Why is a permission check in code safer than a polite instruction in a prompt? ◦ Answer: Because code outside the model enforces the rule regardless of what the model decides. • How do the three disciplines nest? ◦ Answer: Prompt engineering sits inside context engineering, which sits inside harness engineering. • What does it mean to say the harness often matters more than the model? ◦ Answer: Among capable models, the surrounding engineering often explains more of the real-world reliability than small differences between models. PART TWO: INTRODUCTION TO HARNESS ENGINEERING AS A DISCIPLINE Part One taught you what AI and language models are and how they relate to the three disciplines around them. Part Two introduces the harness itself. You will learn what a harness is, why the field emerged, how the seven-layer ETCLOVG taxonomy organizes it, and what the working parts of a production harness look like. These four chapters are your map for the rest of the book.