The art of prompting

About this course

This course is for people who already use ChatGPT or Claude but find their prompts inconsistent — great one day, off the next. It tackles the problem of random results, where answer quality depends on a lucky phrasing rather than a repeatable method. Instead of scattered tricks, the course gives you a single method for building and debugging prompts for any task.

The lessons move from why some phrasings work and others don't, through few-shot examples and breaking down complex tasks, to debugging: diagnosing failures, testing versions, and spotting antipatterns. Projects have you build a working prompt for a recurring task, harden your most important prompt into a reliable one, and finish by assembling a personal prompt library with an update process.

The course suits people already comfortable with basic AI assistant use who want to move from occasional luck to predictable results. No advanced technical background is needed, though some prior experience framing requests for AI helps.

What you'll learn

• Remove ambiguity from a request and replace it with measurable criteria
• Build few-shot prompts with examples, including edge cases
• Structure a complex prompt using role – task – context – constraints – format – examples
• Break a complex task into subtasks the model can handle
• Extract facts and produce analysis rather than a summary of a source text
• Diagnose why a prompt fails and identify the type of failure
• Test prompt versions and compare them against measurable criteria
• Build a personal prompt library and share it with a team

What you'll end up with

• A hardened, reliable prompt for your recurring task, with before/after versions on record
• A personal prompt library ready to use and hand off to a team
• A diagnostic method: a checklist of common reasons a prompt fails
• A certificate of completion on paid plans

This course isn't for people who have never written a working prompt — start with "AI from Scratch" instead.