prompt-eng-interactive-tutorial

A step by step tutorial for learning to engineer optimal prompts within Claude, with 9 chapters, a playground in every lesson, and an advanced appendix.

The course states its intent in one sentence: a comprehensive step by step understanding of how to engineer optimal prompts within Claude. The stated learning goals are concrete. Students should be able to master the basic structure of a good prompt, recognize common failure modes and the 80/20 techniques for addressing them, understand Claude's strengths and weaknesses, and build strong prompts from scratch for common use cases. It is a first party tutorial from Anthropic, the company behind Claude.

The course is broken into 9 chapters with accompanying exercises, plus an appendix of more advanced methods, and the README recommends working through it in chapter order. Every lesson has an Example Playground at the bottom for experimenting with the examples directly, and there is an answer key. Chapter 9 covers building complex prompts for industry use cases, from scratch prompts and chatbots to legal services, with exercises for financial services and coding. The whole course also exists on Google Sheets using Anthropic's Claude for Sheets extension, which the README recommends as more user friendly.

The tutorial runs on Claude 3 Haiku, described as Anthropic's smallest, fastest, and cheapest model at the time. The README notes two other models, Claude 3 Sonnet and Claude 3 Opus, which are more intelligent than Haiku, with Opus the most intelligent. Using the smallest model keeps experimentation cheap, which fits the interactive format. The repository is listed with Jupyter Notebook as its primary language.

The course is short on theory and long on practice, with a playground and answer key attached to each lesson. The README does not claim more than the course delivers.

• Source 1 · Project repository
• Source 2 · README