← Courses
Building with AI
← Prev
Module 1 of 8
Module 2 →
Intro
Scenario
Lesson
Context
Lab Skill ~25 min
Intro

The Prompt is the Program

2 min read

For fifty years, programming was the skill that separated those who could instruct machines from those who couldn't. If you wanted a computer to do something, you wrote code.

Prompting is the new programming. And unlike code, everyone can learn it.

But prompting is deceptively simple. You can ask a question and get an answer. That's easy. Getting consistent, reliable, high-quality answers from an AI system — answers that work at scale, that a team can reuse, that you can depend on — requires precision. Structure. Clarity about what you're asking for.

That's the difference between writing a prompt and building a prompt that works.

Your artifact
A documented prompt library with 5 production-ready prompts
  • Write a well-structured prompt with role, context, and constraints
  • Identify why a prompt is producing bad output and fix it
  • Apply three core improvement techniques to iterate a prompt systematically
  • Test and evaluate a prompt against explicit success criteria
  • Document a prompt so another person can use it exactly as you intended
Scenario

The Team That Got Stuck

3 min read

It's day 22 of using AI for content. The marketing team at a fast-growing SaaS startup started three weeks ago, full of energy. The AI could handle outlines, drafts, ideas — everything seemed possible on week one.

By week three, they've hit a wall. The outputs are inconsistent. One post reads like the company's voice, authentic and conversational. The next one sounds like a corporate template. The third one is fine — technically correct, but forgettable. They're spending more time editing outputs than they would have spent writing from scratch.

The team lead calls an emergency meeting. Three people show their prompts. They're all different. Completely different. One person is telling the AI to be a marketing expert. Another is asking it to write a specific post title. A third is pasting in examples of good content and hoping the AI figures it out. Only one of the three outputs is consistently good. None of them know why.

The team that figures this out first — that understands why one prompt works and two don't, that documents the working one so anyone can use it, that learns to tweak it for different use cases — that team is going to own this tool. The others will be guessing forever.

That's the story you're going to be inside of in the lab.

Lesson

Three Parts of Every Prompt

3 min read

Every prompt that produces consistent, reusable output has three parts. Miss one, and the output degrades.

Who is the AI in this conversation? Not just a job title — a specific kind of expert with specific constraints and experience. "You are a copywriter" is vague. "You are a B2B SaaS copywriter who writes for busy executives who spend 7 seconds on an email" is precise.

The role shapes everything. An economist and a storyteller would write about the same topic completely differently. A role tells the AI which perspective to take, what to prioritize, what to ignore.

What does the AI need to know to give you a useful answer? This is information the AI doesn't have. The audience. The goal. The constraints. The background. If you're asking for a product description, the AI needs to know: what product, who's buying it, how it's different from competitors, what problem it solves.

Vague context produces vague output. Specific context produces precise output.

What shape does the output need to take? Is it a list or prose? How long? What tone? What should never appear in the output? Constraints give the AI guard rails. "Write an executive summary" is instructions. "Write a one-paragraph executive summary suitable for a CEO, no jargon, no calls to action, under 200 words" is a constraint that actually works.

Every good prompt has all three. Leave one out, and you'll be guessing why the output doesn't work.

Context

Patterns That Work

2 min read

There are three patterns that separate prompts that work from prompts that fail. These aren't tricks — they're foundations. Master them, and your prompts will be reliable.

Be the expert, not the guesser

Don't say "You are a writer." Say "You are a senior UX writer at a B2B SaaS company with 8 years of experience writing for busy executives." The more specific you are about expertise, constraints, and perspective, the more precise the output.

Show the shape

If you want a bulleted list, say so. If you want exactly 5 items, say it. If you want a specific format, show it. "Return a bulleted list of exactly 5 key features. Format: Feature Name — one-line benefit." That works. "Tell me about features" doesn't.

Use the negative constraint

"Do not include jargon, clichés, or calls to action." This tells the AI what to avoid — often more powerful than telling it what to include. The best prompts have at least one strong negative constraint that clarifies the boundaries.

These patterns compound. Use all three in a single prompt, and the output becomes dramatically better. Your lab will teach you to apply all three and to see why each one matters.

⚡ Skill Lab
Prompt Engineering
~25 minutes · 3 challenges
What you're doing
You'll write prompts to produce specific outputs. Each challenge gets harder. Your goal: write a prompt that produces an output scoring 8/10 or higher on clarity, specificity, and usability.
Roles
🧑‍💼
You — Prompt EngineerYou write the prompt. You're being tested on whether your instructions produce reliable, reusable output.
🔍
AI — EvaluatorI'll run your prompt against the task, show you what it produces, and score it. Be specific. Be clear. Make it reusable.
Challenges
Executive summary · Product description · Content outline
Framework — use all three
The Role — who is the AI
The Context — what info is needed
The Constraints — what shape the output takes
Success criteria
Score 8+ on each challenge. You can revise and resubmit. The output should be reusable by someone who's never seen your prompt before.
Shift + Enter for a new line
✓ Module Complete
You've completed Module 1 of 8.
Next Module →