Learn prompt engineering by doing
Short modules on the patterns that move AI output from generic to useful. Each module has a concept, a worked example, and a try-it panel that runs your prompt against Claude live. No code. No signup. Twenty runs a day per browser.
Available now
- Module 1
Clear instructions
The single biggest lever in prompt quality. Tell the model what you want, in what format, for whom, and what to leave out.
- Module 2
Role and persona
Telling the model who it is, and who it is talking to, narrows the universe of possible responses to the one you actually want.
- Module 3
Structured output
When you need to use AI output downstream, ask for it in a structure, not prose.
Coming soon
- Few-shot examples
- Chain of thought
- System and user composition
How runs work. Each try-it panel sends your prompt to a Claude model running on our server. We do not store your prompts or outputs. Saved variants live in your browser only.
Limits. 20 runs per browser per day. Output capped at 500 tokens to keep costs predictable.