Prompt engineering's unofficial textbook has 76k stars

What it does

This repository is a curated knowledge base that documents prompt engineering techniques, retrieval-augmented generation patterns, and AI agent design. It powers the promptingguide.ai website and collects guides, academic papers, model-specific notes, risk assessments, and practical examples under an MIT license. A one-hour video lecture and an accompanying Jupyter notebook are also bundled in the repo.

The interesting bit

The project has become a de facto canonical syllabus for a discipline that barely existed before 2022, crossing three million learners and hitting #1 on Hacker News in February 2023. That traction suggests developers wanted structured, vetted guidance rather than scattered blog posts and Twitter threads.

Key highlights

• Concrete technique guides covering zero-shot, few-shot, chain-of-thought, ReAct, tree-of-thoughts, and RAG
• Model-specific sections for ChatGPT, GPT-4, LLaMA, Mistral, Gemini, and others
• Addresses adversarial prompting, factuality failures, and biases rather than ignoring failure modes
• Translated into 13 languages with an open-source content model
• Includes a video lecture, slides, and a runnable notebook

Caveats

• This is documentation and curation, not a framework; executable code is limited to the lecture notebook
• The README heavily promotes paid DAIR.AI Academy courses and consulting services alongside the free material

Verdict

Bookmark it if you build or evaluate LLM applications and need a quick, authoritative reference for prompting patterns. Look elsewhere if you need an installable library or framework.