LLMs & Generative AI · 2.3

✍️ Prompting as a Skill

How to actually get good results out of AI tools⏱ ~2 min

🗺️A vague request gets a vague answer

Asking an LLM 'tell me about dogs' is like telling a taxi driver 'just drive.' You'll go somewhere, but probably not where you wanted. Good prompting is giving clear directions: where you're going, how you want to get there, and what a good arrival looks like.

The Ingredients of a Good Prompt

IngredientWhat to IncludeExample
RoleWho you want the AI to act as'You are a patient biology tutor for a 10th grader.'
TaskExactly what you want done'Explain how vaccines work.'
ContextBackground it needs'I already understand what a virus is.'
FormatHow you want the answer shaped'Use a simple analogy, then 3 bullet points.'
ConstraintsLimits and rules'Keep it under 150 words. No jargon.'

Techniques That Reliably Improve Results

  • Be specific — 'a 200-word summary for beginners' beats 'summarize this'
  • Give examples — show one or two examples of what you want (called 'few-shot prompting')
  • Ask it to think step by step — for reasoning or math, this measurably improves accuracy
  • Iterate — treat it as a conversation; refine your request based on what you get back
  • Ask it to ask you questions — 'What do you need to know to do this well?' surfaces missing context

Prompting Is Not Cheating — Verification Is Your Job

💡 TipUsing AI to learn, draft, and explore is a genuine 21st-century skill. But you own the output. If you submit AI writing as your own for school, that's usually plagiarism — check your school's policy. The best use: let AI help you understand and draft, then make the work truly yours and verify every fact.
🔗
DeepLearning.AI — ChatGPT Prompt Engineering
Andrew Ng's platform offers short, high-quality courses on prompting and building with LLMs — several are free.
https://www.deeplearning.ai
🧠Quick Checkfirst try = +5 XP

Which makes the BEST prompt?

0 XP🔥 0 days