How Anthropic wants you to talk to Claude

What it does Nine chapters of interactive Jupyter notebooks take you from basic prompt structure to advanced industry use cases like legal, financial, and coding workflows. Each lesson pairs explanations with exercises and an “Example Playground” for unstructured experimentation. The entire course runs against Claude 3 Haiku, keeping API costs minimal while you learn.

The interesting bit This is vendor documentation done as executable homework instead of blog posts: you get notebooks with an answer key and playgrounds for breaking things. The authors even note that the same material is available in Google Sheets and is “more user friendly,” which is either refreshing honesty or a very soft upsell for their Sheets extension.

Key highlights

• Covers the full arc from beginner basics through advanced industry prompts for legal, financial, and coding workflows
• Every chapter pairs theory with hands-on exercises and a playground for unstructured experimentation
• Runs against Claude 3 Haiku, the cheapest and fastest model in the lineup
• Includes an answer key and an appendix covering prompt chaining, search & retrieval, and tool use
• Also available as a Google Sheets workbook via the Claude for Sheets extension

Caveats

• The README explicitly recommends the Google Sheets version as more user-friendly, so the notebook experience may feel secondary

Verdict Grab this if you want structured, authoritative prompt engineering practice from the source. Skip it if you already treat LLMs like a finished compiler and just want API reference docs.