Back to catalog
Prompt engineeringBeginner

AI Prompt Mastery

How to phrase a task for a language model so the answer is actually usable

AI Prompt Mastery

About this course

A practical course about writing requests to language models. Not about one particular chat product and not about secret wordings, but about the few things that reliably change the result: a clear task, the context the model does not have, an explicit output format, examples where they help, and a way to check what came back.

Every lesson shows prompts side by side: a weak one and a stronger one on the same task, with an explanation of what exactly the change buys you. You will also learn where the technique stops working, why a model invents facts with complete confidence, and what must never be pasted into someone else's service.

No programming required. The techniques are not tied to any model version and work in any chat interface.

Course curriculum

  1. What you are actually talking to

    How a language model produces text, why it is not a search engine or a fact database, and why a long polite request is not better than a short precise one.

    • 1.What the model actually does12 min
    • 2.Long and polite is not the same as precise11 min

  2. Anatomy of a workable request

    The parts a prompt is made of: task, context, format, constraints, audience. What a role really changes, and when examples in the prompt help more than any explanation.

    • 1.Task, context, format, constraints, audience14 min
    • 2.Role and tone: what they actually change10 min
    • 3.Examples in the prompt, and when they help13 min

  3. Big tasks and long context

    How to split work the model cannot do in one move, when asking for reasoning helps, and how to work with your own documents without the model drifting away from them.

    • 1.Step by step and splitting the work13 min
    • 2.Long context and your own documents12 min

  4. Fixing, checking, not trusting

    How to repair a bad answer instead of starting over, why a model invents facts with full confidence, and the specific things it is reliably bad at.

    • 1.Repairing an answer instead of starting over12 min
    • 2.Hallucinations: why it invents, and how to catch it14 min
    • 3.What models are reliably bad at11 min

  5. Everyday tasks and safety

    Ready patterns for the requests people make most often, and the rules about what must never go into someone else's service — including instructions hidden in text you did not write.

    • 1.Patterns for the tasks you actually have15 min
    • 2.What not to send, and instructions hidden in text13 min