What it does This repo is a hands-on tutorial series from Anthropic itself: nine Jupyter Notebook chapters plus exercises that walk you from basic prompt structure through advanced industry use cases. Each lesson ends with an “Example Playground” where you can tweak prompts and watch Claude 3 Haiku’s responses change in real time. There’s also a Google Sheets version if notebooks feel too developer-y.

The interesting bit The course is structured around failure modes, not features. Instead of listing Claude’s capabilities, it teaches you to recognize when prompts go wrong and apply “80/20” fixes — the kind of pragmatic, debugging-first approach that actual prompt engineers use but rarely document.

Key highlights

• 9 progressive chapters: beginner (structure, clarity, roles) → intermediate (separating data from instructions, chain-of-thought, examples) → advanced (hallucination avoidance, complex industry prompts)
• Built-in answer key and playground areas for experimentation
• Covers legal, financial, and coding use cases in the final chapter
• Appendix on chaining, tool use, and search/retrieval
• Uses Claude 3 Haiku (the cheap/fast model), so you won’t burn API credits learning

Caveats

• The README notes the Google Sheets version is “more user friendly” — the notebook format may feel clunky for non-developers
• Only covers Anthropic models; techniques may not transfer 1:1 to other LLMs

Verdict Worth your time if you’re building with Claude and tired of trial-and-error prompt tweaking. Skip it if you need deep theory on why prompts work — this is a practical workbook, not a research paper.