Course · intro

Prompt Design

Write prompts that get reliable, repeatable output from any language model: the anatomy of a prompt, clear instructions, roles, and in-context learning, from first principles.

0 of 15 read · 15 lessons · ~85 min

Chapter 1 · Anatomy of a prompt

1. 1.1 What is a prompt A prompt is the complete text input you send to a language model, and its contents determine everything the model can do in response. 5 min
2. 1.2 The slots of a prompt A well-structured prompt has up to five distinct slots, each serving a different purpose in guiding the model's response. 6 min
3. 1.3 Primacy and recency Models attend more strongly to information at the beginning and end of a prompt, so slot placement is not arbitrary. 5 min

Chapter 2 · Writing instructions

1. 2.1 Writing clear instructions An instruction that leaves no room for interpretation produces more consistent output than one that relies on the model to fill in gaps. 6 min
2. 2.2 Positive and negative framing Instructions that tell the model what to do are more reliably followed than instructions that tell it what not to do. 5 min

Chapter 3 · Giving the model a role

1. 3.1 Giving the model a role Assigning the model a role before the instruction shifts its default vocabulary, tone, and assumed knowledge level before it reads a single word of your task. 5 min

Chapter 4 · In-context learning

1. 4.1 Zero-shot and few-shot prompting Providing examples inside the prompt shifts the model toward a specific output pattern more precisely than instructions alone can. 6 min
2. 4.2 Formatting your examples The quality of few-shot prompting depends as much on how examples are structured and selected as on how many you provide. 5 min

Chapter 5 · Decomposition and chaining

1. 5.1 When one prompt is not enough Tasks with multiple distinct steps, conflicting constraints, or dependent intermediate outputs are better handled by breaking them into smaller, focused prompts. 5 min
2. 5.2 Chaining prompts Prompt chaining connects a sequence of prompts so that the output of each step becomes the input to the next, enabling complex tasks that no single prompt could handle reliably. 6 min

Chapter 6 · Reasoning techniques

1. 6.1 Chain-of-thought prompting Asking the model to show its reasoning before giving an answer improves accuracy on tasks that require multiple logical steps. 6 min
2. 6.2 Tree of thoughts Tree of thoughts extends chain-of-thought by generating and evaluating multiple reasoning paths, allowing the model to explore alternatives before committing to an answer. 6 min

Chapter 7 · Output format

1. 7.1 Controlling output format Explicitly specifying the structure of the model's response removes the last degree of freedom the model has to produce something you did not intend. 5 min

Chapter 8 · Reliability and testing

1. 8.1 Prompt sensitivity Small changes in wording, order, or framing can produce different outputs, and understanding why helps you write prompts that stay stable across variations. 5 min
2. 8.2 Testing your prompts A prompt that works on one input is not necessarily a prompt that works reliably. Systematic testing across varied inputs is the only way to know. 5 min
3. 8.3 Course quiz Six questions covering the key concepts from the Prompt Design course. 4 min