Prompt Engineering Guide — 73K-Star Learning Hub

What it is

The Prompt Engineering Guide is an open-source educational resource covering prompt engineering techniques for large language models. It includes written guides, research paper summaries, Jupyter notebooks, and course materials on topics like chain-of-thought prompting, retrieval-augmented generation, AI agents, and multi-modal models. The content is available in 13 languages.

This resource targets developers, researchers, and AI practitioners who want to understand and apply prompt engineering systematically rather than through trial and error.

How it saves time or tokens

Understanding prompt engineering techniques directly reduces token waste. A well-structured prompt that uses chain-of-thought or few-shot examples gets better results on the first try, avoiding the iterative cycle of prompt refinement that consumes tokens.

The guide covers token-efficient patterns: when to use system prompts vs user prompts, how to structure few-shot examples for maximum effect, and when chain-of-thought helps vs hurts (simple tasks do not benefit from verbose reasoning).

How to use

1. Clone the repository or read online:

git clone https://github.com/dair-ai/Prompt-Engineering-Guide.git
cd Prompt-Engineering-Guide

1. Start with the fundamentals:

guides/
  prompts-intro.md        # Introduction to prompting
  prompts-basic-usage.md  # Basic techniques
  prompts-advanced.md     # Advanced patterns
  prompts-chatgpt.md      # ChatGPT-specific tips
  prompts-applications.md # Real-world applications

1. Explore technique-specific notebooks in the notebooks/ directory for hands-on practice.

Example

Chain-of-thought prompting pattern from the guide:

# Without CoT (often wrong on complex tasks)
Q: Roger has 5 tennis balls. He buys 2 more cans
   of 3 tennis balls each. How many does he have?
A: 11

# With CoT (step-by-step reasoning)
Q: Roger has 5 tennis balls. He buys 2 more cans
   of 3 tennis balls each. How many does he have?
A: Roger starts with 5 balls.
   He buys 2 cans of 3 balls each = 6 balls.
   5 + 6 = 11.
   The answer is 11.

Explicit reasoning steps improve accuracy on math, logic, and multi-step tasks.

Related on TokRepo

• Prompt library -- Curated prompts and templates for various AI tasks
• AI tools for research -- Research tools and resources for AI practitioners

Common pitfalls

• Applying chain-of-thought to every prompt. Simple factual questions and classification tasks do not benefit from verbose reasoning. CoT adds tokens without improving accuracy on straightforward tasks.
• Ignoring model-specific prompt formats. Claude, GPT-4, and Gemini have different system prompt conventions and capabilities. Techniques that work on one model may underperform on another.
• Over-engineering prompts when a simple instruction suffices. Start with the simplest prompt that works and add complexity only when needed.