Prompt Engineering From Scratch in 5 Days

What a Prompt Is, and Isn't: How the Model Reads Instructions; the Four Elements of Role / Task / Format / Constraints

First get clear on how the model actually reads what you write, then use four elements — role, task, format, constraints — to turn a fuzzy requirement into a work ticket the model can execute reliably.

Few-Shot, Chain of Thought, Step-by-Step, and Self-Checks; When None of These Work

What problem each of the four most common prompting techniques solves, what it costs, and under what circumstances none of them can save you.

Structured Output: JSON Schema, Templates and Variables, Multilingual Output

Make the model's answer directly consumable by code: constrain the output shape with a schema, turn a prompt into a reusable function with templates and variables, and get the same prompt to reliably output multiple languages.

Iteration and Evaluation: Small Test Sets, A/B Testing, Version Control, Common Anti-Patterns

Turn "this version feels better" into "this version got three more of ten test cases right": build a small test set, run an A/B comparison, version your prompts, and recognize the most common anti-patterns.

Migrating Across Models: Differences Between Claude / GPT / Domestic Chinese Models, Organizing the System Prompt; Where to Go Next — the Claude Course or the Codex Course

Why the same prompt breaks when you switch models, exactly where it breaks, how to organize a system prompt into a portable structure, and whether to study the Claude course or the Codex course next.