---
title: "Index"
---

# 02 — Prompting

Now that you can run a model locally, let's learn how to talk to it effectively.

These three "spells" (techniques) work on **any** AI model — local or cloud. They're simple to learn, but they profoundly change the quality of what you get back.

## The Three Spells

| Spell                          | What It Does                                                                    | When To Use It                                                   |
| ------------------------------ | ------------------------------------------------------------------------------- | ---------------------------------------------------------------- |
| **#1: Clarifying Questions**   | Makes the model ask what you mean before it answers                             | Whenever you're not 100% sure the model understands the task     |
| **#2: Simplest Thing (YAGNI)** | Tells the model to stop over-complicating things                                | When the model proposes complex solutions to simple problems     |
| **#3: One-Shot Prompts**       | Makes your work reproducible by putting everything in one self-contained prompt | When you need to iterate, version-control, or share your prompts |

## Why These Matter

Local models are generally less capable than GPT-4 or Claude. These spells **close the gap**. They give a weaker model the structure it needs to produce reliable, high-quality output.

They also teach you something fundamental about working with AI: **the quality of your output depends on the quality of your communication.** Garbage in, garbage out applies to humans too.

**[03 — Clarifying Questions →](/run-ai-locally/02-prompting/03-clarifying-questions/)**