01Mental Models for Prompting

The way you think about the AI shapes the prompts you write. In this lesson, you'll pick up three mental models — simple but powerful analogies — that will permanently change how you approach every prompting task. You'll finish with a clear mental picture of what you're really doing when you write a prompt.

THE INTERN MODELThe AI is a brilliant intern with amnesia —it knows everything except your situation.YOUvague askGUESSWORKplausible, wrongclear brief + examplesGREAT OUTPUTexactly on targetSame intern.Different briefing.FIG. 01.03 — MENTAL MODELS FOR PROMPTINGPROMPTMASTERY · ILLUSTRATED GUIDE
Foundations of LLMs

Mental Models for Prompting

20 min
📄LESSON NOTES
Start with the diagram, then read the notes

Three mental models that permanently change how you think about prompts:

Mental Model 1: The Brilliant New Hire

Imagine a brilliant person on their first day at your company. Smart, eager to help, but knows absolutely nothing about your situation. You must explain everything from scratch. This combats the #1 prompting mistake: assuming the AI already knows your context.

Mental Model 2: The Specification Document

Engineers write a spec before building anything. Treat every complex prompt like a spec: define inputs, define outputs, list edge cases, write constraints, clarify success criteria. The clearer your spec, the more predictable your result.

Mental Model 3: The Funnel

Picture a funnel. An empty prompt = wide open — thousands of valid responses possible. Each word you add narrows it down. Your job is to funnel the AI toward the exact response you need.

Simple test: Can you imagine 5 very different valid answers to your prompt? If yes, it's too vague. Keep adding constraints until only 1–2 good responses are possible.

Detailed Analysis

The Brilliant New Hire model explains immediately why vague prompts produce vague outputs. If you hired someone brilliant and said "help me with this project" with no further context, you'd get a useless response. Give them background, explain the task clearly, and they'd produce excellent work. The AI is exactly the same.

Use this model to check if you've included:

  • Company or project context: "We're a B2B SaaS company targeting HR teams"
  • Prior decisions: "We've already tried approach X and it failed because Y"
  • Audience info: "This is for non-technical stakeholders"
  • What success looks like: "A good response makes the CFO feel confident about the decision"

The Specification Document model comes from software engineering. A spec answers: What is the input? What should the output look like? What are the edge cases? What must never appear in the output? What happens if data is missing? Thinking of prompts this way forces you to anticipate failure modes before they happen.

The Funnel model explains why being specific is so powerful. Every additional constraint removes possible responses from the pool. Adding "in formal business language" eliminates casual phrasing. Adding "in 3 bullet points" eliminates paragraph-form answers. Each constraint removes unwanted results.

Combine all three: "I'm the brilliant new hire who needs a clear spec for this task, and I'll use the funnel to narrow down to exactly what I want."

Take-Home Points

  • The Brilliant New Hire: provide context as if talking to someone who knows nothing about your situation
  • The Spec Document: define inputs, outputs, edge cases, and success criteria before writing the prompt
  • The Funnel: every constraint narrows the range of possible responses — be specific
  • If 5+ different valid responses are possible, your prompt is underspecified
  • Mental models are more useful than memorized rules — they apply to every new situation

Conclusion

Mental models are the foundation of real expertise. Rather than memorizing a list of prompting rules, these three models give you a thinking framework for every prompt you'll ever write. Before you type anything, ask: Have I given enough context for the Brilliant New Hire? Have I written a clear Spec? Have I narrowed the Funnel enough? When all three pass, your prompt will reliably produce what you need.

📝
Quiz — 3 questions

Check what you picked up, graded instantly.

📊
Save your progress

Pick up where you left off, on any device.

🎓
Certificate of completion

Finish the course and earn a shareable certificate.

That's the end of the free preview

You've finished Module 01. The remaining 23 lessons cover the techniques the rest of the course is built on — chain-of-thought, prompt chaining, RAG, evaluation and agentic patterns. Free, no card needed.