Lesson 4 — Prompting Well and Checking the Output
Two skills decide how much you get out of any AI assistant: writing a clear prompt, and verifying the answer. They apply to Claude, ChatGPT, Gemini — all of them.
Writing a good prompt
A prompt is just your request. A good one usually has four parts, though not every request needs all four:
- The task — what you want done, stated plainly.
- Context — the background the AI needs that it cannot see on its own.
- Constraints — length, format, tone, what to avoid.
- The shape of the answer — a list, a table, a paragraph, code only.
See the difference for yourself — pick the prompt that will get a better answer.
This activity needs JavaScript enabled. The lesson still covers everything.
Iterate — the first answer is a draft
You rarely get exactly what you want on the first try, and that is normal. Treat the reply as a draft and steer it: "shorter," "more formal," "you misunderstood — I meant the beginner version, not the advanced one." Refining in a few rounds is faster than trying to write one perfect prompt.
Verifying the output — the non-negotiable habit
From Lesson 1: an AI can be confidently wrong. So nothing it produces is finished until you have checked it. How you check depends on what it gave you:
- A fact or number — confirm it against a source you trust. Be especially careful with citations, dates, and statistics.
- Code or a page change — look at the actual file and the preview. Does it do what you asked? Did it break anything nearby?
- A claim about your own project or files — check the real thing. The AI's knowledge can be out of date; your project is the source of truth.
- An explanation — if a step does not make sense, ask. A gap in your understanding is often a real error in the answer.
Test the habit — in each case below, has the output really been verified?
This activity needs JavaScript enabled. The lesson still covers everything.
Watch for hallucinations
If the AI references a file, function, setting, or source you have never heard of, do not assume it exists — go look. Made-up details are stated just as confidently as real ones. Verifying takes seconds; a wrong detail shipped into a live project costs much more.
Protect private information
Be thoughtful about what you paste into an AI tool — passwords, keys, and personal data about real people do not belong in a prompt unless you are certain the tool is approved for it. When unsure, ask before pasting.
Quick check
This activity needs JavaScript enabled.
You now have the mental model from Lesson 1 and the two core habits — prompting and verifying. The next lesson puts them to work: how to take a real idea and turn it into a small, working app.