In Chapter 1 you talked to your AI sidekick. Now you'll learn to get great answers on purpose, every time — not by luck.
The secret is a simple recipe. Almost every strong prompt has four ingredients:
- Context — who you are and what's going on
- Task — exactly what you want done
- Format — how the answer should look
- Examples — a sample of what "good" looks like (optional, but powerful)
Remember it as C-T-F-E. Let's add them one at a time and watch the answers get better.
Step 1 — Add context
Context is the background the AI needs. Compare these two prompts:
Give me tips for a presentation.
I'm 12 and giving a 3-minute class presentation about volcanoes to kids my
age. I get nervous. Give me tips.
The second one gets advice that actually fits you. Without context, the AI guesses. With it, the AI helps.
Tip
A quick formula for context: "I'm ___, I'm trying to ___, for ___." Fill in the blanks before your real request.
Step 2 — Ask for a format
If you don't say how you want the answer, you get a wall of text. So ask:
... Give me exactly 5 tips, each one short sentence, as a numbered list.
You can ask for almost any shape:
- A table with columns you name
- 3 bullet points, no more
- A step-by-step list
- "In under 50 words"
- A specific tone: "friendly", "formal", "funny"
Naming the format is one of the biggest upgrades to any prompt.
Step 3 — Show an example (few-shot)
The most powerful trick: show, don't just tell. Giving one or two examples is called few-shot prompting. Watch:
Turn each word into a fun fact. Follow this example exactly.
Example:
Word: Octopus
Fun fact: An octopus has three hearts! 🐙
Now do these: Volcano, Saturn, Honey
The AI copies your style, length, and even the emoji. Examples are like showing a worker a sample of the job instead of describing it.
Step 4 — Iterate (the improve loop)
Pros almost never get the perfect answer on the first try. They iterate — make it better with follow-ups:
- "Make it shorter."
- "That's too hard — explain it for a 9-year-old."
- "Keep #2 and #4, rewrite the rest."
- "Add an example to the last one."
Each message builds on the chat. Tweaking is not failing — it is prompt engineering.
Check yourself
- What do the letters C-T-F-E stand for?
- Which ingredient means "show an example"?
- True or false: if your first answer isn't great, you should always start a brand-new chat. (False — iterate!)
Project — Build your "Explainer" helper 🛠️
Let's combine all four ingredients into one reusable helper that explains any topic clearly.
-
Open a new chat and build your prompt with C-T-F-E:
You explain tricky topics to curious kids aged 11–13 (Context + Task). When I give you a topic, explain it in: - 1 simple sentence - 3 bullet points with the key ideas - 1 real-world example - 1 "mind-blowing fact" with an emoji (Format) Example topic: Gravity → Gravity is the pull that keeps things from floating away. • It pulls everything toward big objects like Earth. • The bigger the object, the stronger the pull. • It's why a ball comes back down when you throw it. Real-world: It keeps your feet on the ground and the Moon near Earth. 🤯 The Sun's gravity is so strong it holds planets billions of km away! (Example) My first topic is: Black holes -
Read the answer. Did it follow the format exactly?
-
Iterate twice: try
Make the fact funnierandExplain it for a 9-year-old instead. -
Save it in your "My AI Helpers" doc as Explainer v1.
Your turn
Turn the Explainer into something for your life. Swap the topic type for:
- A recipe helper ("explain how to cook ___ in 5 steps")
- A sports-play explainer
- A vocabulary helper that gives a kid-friendly meaning + a sentence
Keep the C-T-F-E shape — just change the job.
Make it simpler · ages 9–11
Use just two ingredients to start: Task + Format. Try:
Write 3 silly animal facts as a bullet list with emojis. Then add one more
ingredient (an example) and see the answer change. Two is plenty to begin.
Level up · ages 13–16
Experiment with structured output. Ask the AI to "reply only as a JSON
object with keys topic, summary, and facts (a list)." This is exactly how
real apps get predictable data out of an AI — you'll use it for real in
Chapter 7. Notice how a tight format makes the output easy for a program to
read, not just a human.
Stay safe
Your prompts are still messages to an AI service. Keep using made-up or general examples — never put your real personal details, your friends' info, or anything private into a prompt. And keep fact-checking: a clear format can make a wrong answer look very confident.
What you learned
- Strong prompts use Context, Task, Format, Examples (C-T-F-E).
- Examples (few-shot) are the fastest way to control style.
- Iterating beats restarting.
- You built a reusable Explainer — real prompt engineering.
You've earned the Prompt Smith badge. 🏅
Next: you'll give your AI a job title and a thinking process, so it acts like a specialist. That's Chapter 3: Roles, Personas & Step-by-Step.