Chapter 3: The Art of Prompt Engineering (Or: Teaching Your Digital Intern Who Knows Everything and Nothing)
Today’s Training Session with Claude
Me: Write me a marketing email.
Claude: Subject: Marketing Communication
Dear Valued Customer,
We are pleased to inform you about our product which offers features
and benefits. Please consider purchasing it for reasons.
Thank you for your time.
Me: That's... terrible.
Claude: You gave me the equivalent of "make food" as a cooking instruction.
What did you expect, a five-course meal?
Me: Fair point. Let me try again. Write a marketing email for our new
coffee subscription service targeting tired parents, emphasizing
convenience and quality, with a playful tone and a 20% off coupon.
Claude: Subject: "Because Your Coffee Maker Shouldn't Judge Your Pajamas at 2 PM"
[Proceeds to write actually good email]
Me: So the secret is being specific?
Claude: The secret is treating me like a brilliant intern who was raised
by Wikipedia and has never actually drunk coffee. Specific is good.
Context is better. Examples are chef's kiss.
Welcome to the weirdest skill of the 21st century: teaching a know-it-all digital entity how to do exactly what you want. It’s like having an intern with a photographic memory of the entire internet who somehow doesn’t know what you mean by “make it pop.”
The Prompt Engineering Paradox
Here’s the thing nobody tells you about AI: It’s simultaneously the smartest and dumbest tool you’ll ever use. It can write a sonnet about quantum physics in the style of Shakespeare, but ask it to “make something nice” and you’ll get the digital equivalent of a shrug.
This is where prompt engineering comes in. Despite its fancy name, prompt engineering is basically the art of explaining things to a very talented alien who’s read every book but never lived a day on Earth.
The Five Stages of Prompt Grief
Stage 1: Denial
“I’ll just ask it simply. How hard can it be?”
Proceeds to get a response that’s technically correct but completely useless
Stage 2: Anger
“Why doesn’t it understand what I obviously mean?!”
Writes angry follow-up prompt in ALL CAPS
Stage 3: Bargaining
“Okay, if I just add seventeen more clarifications…”
Creates a prompt longer than the output you wanted
Stage 4: Depression
“Maybe I should just do this myself.”
Stares at blank document for 20 minutes
Stage 5: Acceptance
“Fine, I’ll learn to speak Robot.”
Discovers the magic of context, examples, and specific instructions
The SPACE Framework (Because Everything Needs an Acronym)
After extensive research (read: messing around until something worked), here’s your guide to prompt engineering:
S - Specific Context
Don’t: “Write about productivity”
Do: “Write a 500-word blog post about productivity tips for remote workers with ADHD”
P - Purpose and Audience
Don’t: “Explain quantum computing”
Do: “Explain quantum computing to my grandmother who thinks the cloud is actually in the sky”
A - Action and Format
Don’t: “Help with my presentation”
Do: “Create 5 bullet points for a slide about quarterly sales, focusing on growth metrics”
C - Constraints and Requirements
Don’t: “Make it good”
Do: “Keep it under 200 words, use a professional tone, and include at least one statistic”
E - Examples (When Possible)
Don’t: “Write like Malcolm Gladwell”
Do: “Write like Malcolm Gladwell - start with an unexpected anecdote that connects to a larger truth about human behavior”
Real-World Prompt Disasters (And How to Fix Them)
The Vague Request Disaster
What I Asked: “Help me with my resume”
What I Got: A generic template that looked like it was from 1995
What I Should Have Asked: “Review my software engineering resume and suggest improvements for applying to senior roles at startups, focusing on demonstrating leadership and technical innovation”
The Missing Context Catastrophe
What I Asked: “Write a thank you note”
What I Got: “Dear [Name], Thank you for [thing]. Sincerely, [Your name]”
What I Should Have Asked: “Write a warm but professional thank you note to my mentor who spent 2 hours giving me career advice about transitioning from engineering to product management”
The Content Creator’s Comedy of Errors
The Problem: Jessica, a social media manager, was spending 6 hours a day creating posts.
The Failed First Attempt: “Write social media posts for my company.” Got generic corporate speak that made her brand sound like a robot having an identity crisis.
The Breakthrough Moment: She learned to give AI her brand voice, target audience, and specific goals: “Write Instagram posts for a sustainable fashion brand targeting eco-conscious millennials, using a playful but informative tone, include a call-to-action, and reference current trends.”
The Unexpected Benefit: AI became her brainstorming partner. She’d feed it trends and get back dozens of angles she never would have thought of.
The New Normal: Jessica now creates a week’s worth of content in 2 hours, with higher engagement than her manually-created posts. She spends the saved time on community engagement and strategy.
The Tone-Deaf Response
What I Asked: “Write an apology email to my boss”
What I Got: A formal letter that sounded like a legal document
What I Should Have Asked: “Write a sincere but not overly dramatic apology email to my boss for missing the morning meeting, acknowledging the inconvenience while maintaining professionalism”
Advanced Prompt Techniques (For When You’re Feeling Fancy)
The Role Play
“You are a experienced UX designer reviewing a junior designer’s portfolio. Provide constructive feedback on these three designs…”
The Step-by-Step
“First, analyze the problem. Then, list three possible solutions. Finally, recommend the best option with reasoning.”
The Iteration Loop
“Generate 5 options for [thing]. I’ll pick one and you’ll create variations of that one.”
The Socratic Method
“Instead of giving me the answer, ask me questions that will help me figure out the solution myself.”
The Prompt Engineer’s Toolbox
Keep these templates handy:
The Analyzer:
“Analyze [thing] and identify [specific aspects]. Focus on [criteria].”
The Creator:
“Create [specific output] for [audience] that accomplishes [goal].”
The Improver:
“Here’s my [thing]: [paste it]. Make it better by [specific improvements].”
The Explainer:
“Explain [complex topic] as if you’re [specific person] talking to [specific audience].”
The Devil’s Advocate:
“Argue against this position: [your position]. Help me see the weaknesses.”
When Good Prompts Go Bad
Sometimes, even with perfect prompting, AI will confidently tell you that the moon is made of cheese or that Benjamin Franklin invented the internet. This is called “hallucination,” which is a polite way of saying “making stuff up.”
Rules for dealing with hallucinations:
- Trust but verify (especially for facts)
- If it sounds too weird to be true, it probably is
- When in doubt, ask for sources
- Remember: AI is a tool, not an oracle
Your Prompt Engineering Evolution
Week 1: “Why won’t it just read my mind?”
Week 2: “Okay, I’ll add more detail… lots more detail… all the detail…”
Week 3: “Wait, there’s a balance?”
Month 2: “I speak fluent Robot now.”
Month 6: “I can get AI to do 80% of my work with 20% of the effort.”
Year 1: “I’m teaching others how to talk to the machines.”
The Bottom Line
Prompt engineering isn’t about becoming a technical wizard. It’s about learning to communicate clearly with a very literal partner who has access to vast knowledge but no common sense.
It’s the difference between shouting “HELP!” into the void and saying, “I need specific assistance with this particular problem, here’s the context, and this is what success looks like.”
Master this, and you’ll have the equivalent of a genius assistant who never sleeps, never complains, and only occasionally suggests that the best solution to traffic is to “simply teleport.”
The Bottom Line
Meta moment: While writing this chapter, I used all these techniques on my AI co-author. The difference between my early drafts (“write about prompt engineering”) and the final chapter (“write a practical guide to prompt engineering using the SPACE framework, include real failures, make it conversational and slightly sarcastic”) is night and day.
Prompt engineering isn’t about becoming a technical wizard. It’s about learning to communicate clearly with a partner who has infinite knowledge but zero intuition.
The AI revolution isn’t creating a world where machines understand humans better. It’s creating a world where humans who can clearly articulate what they want have superpowers.
Remember: The AI isn’t getting smarter at reading your mind—you’re getting better at explaining what’s in it. And in a world where everyone has access to the same AI tools, the person who can best direct them wins.
Now go forth and prompt responsibly. Specifically. Always specifically.
*P.S. - If you’re still getting generic responses, remember: AI is like a really smart intern who’s never left the library. Give it context, examples, and specific instructions. And never, ever ask it to “make it pop.”
Trust me on this one.*