Better questions. Better answers.

What you’ll be able to do

• Write a prompt with a goal, context, boundaries, and a clear output format.
• Recognize when an example helps more than another adjective.
• Improve a prompt through a controlled comparison.

A prompt is a brief

You are defining a task, not searching for a magic phrase. A strong brief explains what you need and how you will judge it. You do not have to make every prompt long. Include the information that would change the result.

Use four questions:

• Goal: What should the assistant help me accomplish?
• Context: Who is the audience, and what source material matters?
• Boundaries: What must it preserve, avoid, or leave uncertain?
• Output: What form and quality standard do I need?

Watch a vague request become a brief

A vague request says: “Write something about our AI class.”

A clearer brief says: “Draft a 100-word invitation for adults attending their first AI class. Use only these confirmed details: Saturday, public library, free, laptop optional. Use a welcoming tone. Do not invent a time or registration link. Finish with the details I need to confirm.”

Notice that “welcoming” is only one part of the instruction. Audience, source details, limits, and format make the request operational.

Use examples when a label is ambiguous

“Professional,” “warm,” and “concise” mean different things to different people. A short example can make your standard more concrete. Supply your own non-sensitive writing and explain what to preserve: plain vocabulary, short sentences, or direct openings. Do not ask for superficial imitation when the underlying purpose is different.

You can also show a bad example and explain the problem: “Avoid promising guaranteed results. This sentence goes beyond what we can support.”

Try the four-part prompt

Goal: Draft a short invitation to an introductory AI workshop.
Context: The audience is adults with no technical background.
Confirmed facts: Saturday; library meeting room; free; laptop optional.
Boundaries: Do not invent a start time, duration, instructor credential, or registration link.
Output: About 100 words in plain language, followed by a checklist of missing details.
Success: A reader understands who it is for, what to bring, and which details are still pending.

Your practice: compare two prompts

1. Write a vague prompt for a low-risk task you actually do.
2. Define three success criteria before generating anything. For an invitation, use factual accuracy, audience fit, and clarity of next steps.
3. Run the vague prompt and save its output.
4. Run a four-part brief for the same task in a fresh conversation.
5. Score both outputs against the same criteria, using 0 for missing, 1 for partly met, and 2 for met.
6. Revise one instruction that addresses the weakest result. Run it again.

Your finished artifact: Three prompt versions, their outputs, and a six-point comparison. A better score is useful evidence for this task, not proof that one prompt wins for every task.

Give specific revision feedback

Instead of “make it better,” try: “Keep the first two sentences. Remove the invented time. Replace the last sentence with a placeholder for the registration link.” This identifies the failure and preserves what already works.

A common mistake

Adding more instructions after every disappointing output without checking whether the instructions conflict. If you request “explain fully” and “use fewer than 20 words,” decide which goal matters more. Sometimes the right fix is better source material, a different workflow, or a narrower task.

Reflection

Which instruction produced the largest useful change? Which instruction added length without improving the result?

Go deeper

Anthropic’s prompt-engineering overview emphasizes defining success criteria and testing results before tuning prompts.