Skip to content

Getting started · Beginner

Prompting fundamentals

Turn a vague question into a clear brief, choose useful context, and check an AI answer before you use it.

3 lessons · About 20 minutes including practice · Free to read

What you will learn

  • Write a prompt with a task, context and output format.
  • Use remembered messages deliberately.
  • Check an explanation against the facts you supplied.

Lesson 1 of 3

Write a clear brief

A useful prompt tells the model what to do, what information to use and what the result should look like. Start with an action such as compare, explain or draft. Then name the audience and include the facts the answer must preserve.

Use Prompt for an open-ended task. Keep instructions separate from source material so a quoted email or document is clearly something to work on. When information is missing, ask the model to identify the gap instead of filling it with a plausible detail.

Worked example

Task: Draft a project update for our design team.
Facts: The prototype is ready. Testing starts on 12 October. Maya owns recruitment. The release date is not decided.
Format: Three bullets: progress, next step, open question.
Constraints: Use only these facts. Do not invent a release date.

Try it yourself

Write an update using the example. Change the audience to a client and rerun it. Compare which words changed and which facts stayed the same.

Check your result

  • The response contains exactly three bullets.
  • Maya and 12 October are preserved.
  • The undecided release date stays undecided.

Lesson 2 of 3

Choose your context

A follow-up works best when it carries the specific information it needs. In Nevatal, Reply targets an exchange and Remember selects messages to include in later requests. Use Memory & history to review what you have selected before starting an unrelated task.

Remember a stable preference, such as the intended audience, when it applies to the next request. Avoid remembering an entire conversation just because it exists: old facts can conflict with a new brief. If an answer goes off track, inspect the selected context and restate the current task.

Worked example

Useful context: Our readers are new project managers. Explain unfamiliar terms.
New task: Explain the difference between a milestone and a task in two sentences, followed by one example.

Try it yourself

Send the audience preference in Prompt and mark it Remember. Ask the new question, then forget that preference and ask for a version for experienced project managers.

Check your result

  • The remembered preference is visible in Memory & history.
  • The explanation fits the intended reader.
  • Unrelated facts from the project update do not appear.

Lesson 3 of 3

Check the answer

A fluent answer can still contain a mistake. Use Explainer to unpack an unfamiliar idea, then compare the explanation with your source material. Asking the same model whether it is correct is not independent verification.

Before using a result, check names, dates, calculations and claims against something you can inspect. If the brief only supports a limited answer, a useful response should say what is unknown. Save a version of the prompt that worked so you can reuse its structure.

Worked example

Explain this statement to a beginner: “The team completed 8 of its 10 planned tasks.”
Distinguish what this tells us about task count from what it does not tell us about effort or project completion.

Try it yourself

Ask Explainer to explain the statement. Calculate the completed share yourself and identify one conclusion the numbers do not support.

Check your result

  • Eight of ten tasks is 80% by count.
  • Task count is not treated as a measure of hours worked.
  • The response does not claim the whole project is 80% complete.

Put it into practice

Repeat one exercise with your own material. Keep the original beside the result, use the lesson’s checklist and revise the input if something is missing.