Prompt Engineering Basics - How to Get Better AI Results

AI models are only as good as the instructions they're given. Prompt engineering isn't magic — it's simply the practice of communicating clearly.

Rule 1: Be Specific, Not Vague

❌ "Make my code better"

✅ "This function runs duplicate database queries — combine
   them into a single query, and make the variable names
   more descriptive."

The more specifically you describe the desired outcome, the more targeted the result. Vague requests produce vague answers.

Rule 2: Give Context

❌ "Write an email"

✅ "Write a professional follow-up email to a client who
   said they'd review a proposal 3 days ago. Keep the tone
   polite but with a bit of urgency."

A model only knows what you tell it — if the audience, tone, or purpose is missing, it has to guess, and that guess isn't always right.

Rule 3: Specify the Format

"Put these 5 features into a comparison table,
columns: Feature, Pros, Cons."

If you want output in a particular structure (a table, a bullet list, JSON), say so clearly — the model will follow the structure you give instead of guessing a default format.

Rule 4: Give Examples (Few-Shot Prompting)

Follow this tone:
"Order shipped! 📦 Track it here: [link]"

Now write: a message for an order cancellation.

Giving one or two examples ("write it in this style") makes your exact expectation clear to the model — often more effective than explaining in words alone.

Rule 5: Iterate

If the first response isn't perfect, don't throw it away — refine it: "this is good, but make it shorter" or "make this part more technical." Working with AI is a conversation, not a one-shot trick.

Key Takeaways

• Specific instructions plus context lead to better results.
• Explicitly state the output format when structure matters.
• Giving examples (few-shot) is the most effective way to match a tone or style.