Module 425 min read · Foundations

Prompt Engineering

The quality of what you get from an AI is almost entirely determined by the quality of what you give it. Prompt engineering isn't a technical skill — it's a communication skill. And like all communication skills, the gap between someone who does it well and someone who doesn't is enormous.

Why this matters more than which tool you use

Most people spend their time debating which AI is best. The people who actually get the most out of AI spend their time getting better at prompting. The same model that produces a mediocre answer to a vague question will produce a genuinely useful answer to a well-constructed one.

Here's the same request, written two different ways:

❌ Weak prompt

Tell me about investing.

✓ Strong prompt

I'm 17 years old with $5,000 invested in index funds and individual tech stocks. Explain the concept of compound interest and why starting early matters — use concrete numbers to show what $5,000 growing at 10% annually looks like at age 30, 40, and 65. Keep it under 300 words and avoid financial jargon.

Same model. Completely different outputs. The second prompt tells the AI exactly who is asking, what they want to know, how to illustrate it, how long the response should be, and what language to use. Every one of those constraints makes the output better.

The six elements of a strong prompt

Element 01

Role and context

Tell the AI who it's talking to and what situation you're in. "I'm a high school student writing a college application essay" produces a different response than no context at all. The AI will calibrate vocabulary, depth, and tone based on who it thinks it's talking to.

Element 02

Specific task

State exactly what you want. Not "help me with my essay" but "rewrite this paragraph to be more concise while keeping the central argument intact." Verbs matter: explain, compare, rewrite, summarize, critique, draft, list — each one signals a different kind of output.

Element 03

Format and length

Tell the AI how you want the output structured. "Give me three bullet points" vs. "write three full paragraphs" vs. "write a table comparing these options" — all produce dramatically different results. If you don't specify, the AI guesses, and it often guesses wrong.

Element 04

Examples

Show the AI what good looks like. If you want a specific tone, paste an example of writing in that tone. If you want a specific format, show one instance of it. Examples are often more powerful than descriptions — the AI can extract patterns from what you show it better than from what you tell it.

Element 05

Constraints and exclusions

Tell the AI what NOT to do. "Don't use bullet points." "Avoid technical jargon." "Don't suggest I see a doctor — I need practical information." Constraints are just as important as instructions. Without them, the AI fills in gaps with defaults that may not match what you want.

Element 06

Step-by-step reasoning

For complex problems, ask the AI to think through the problem step by step before giving an answer. This simple instruction — "think through this carefully before answering" or "walk me through your reasoning" — measurably improves accuracy on reasoning tasks. The AI produces better answers when it works through a problem explicitly rather than jumping to a conclusion.

Advanced techniques

Chain of thought prompting

For problems that require reasoning — math, logic, complex decisions — tell the AI to show its work. "Solve this step by step" or "explain your reasoning as you go" produces more accurate results than asking for a direct answer. This works because generating the intermediate steps helps the model arrive at better conclusions.

✓ Chain of thought example

I have $500/month to invest. Should I put it all in an S&P 500 index fund, split it between index funds and individual stocks, or keep it in a high-yield savings account? Think through the tradeoffs of each option step by step, then give me your recommendation with reasoning.

Iterative refinement

Don't expect the first response to be perfect. The best AI users treat it as a conversation — get a first draft, then refine. "Make this more concise." "The third paragraph is too vague — rewrite it with a specific example." "Change the tone from formal to conversational." Each instruction moves the output closer to what you actually want.

Assigning a persona

Asking the AI to take on a specific role can dramatically improve the quality and relevance of the output. "You are an experienced venture capitalist reviewing a pitch deck" produces very different feedback than no framing at all. The persona shapes what knowledge the AI draws on and what perspective it takes.

✓ Persona prompting example

You are a ruthlessly honest college admissions officer at Georgetown University. Review this essay opening and tell me what a real admissions reader would think in the first 10 seconds — and whether they'd keep reading. Be direct, not encouraging.

The "what am I missing" prompt

One of the most underused prompting techniques: after you get an answer, ask "What important considerations did I not ask about?" or "What would push back on this argument?" AI models are often better at identifying gaps and counterarguments when explicitly asked than when trying to give a comprehensive answer upfront.

Common mistakes

Being too vague. "Help me write better" tells the AI nothing. "Help me make this email more direct and cut it from 200 words to 80" is actionable.

Not iterating. If the first response isn't right, refine it. The conversation is the tool — not any single exchange.

Accepting everything without verification. AI is a drafting and thinking tool, not a source of truth. Anything factual that matters should be verified independently.

Not giving enough context. The AI doesn't know who you are, what you're trying to accomplish, or what you've already tried. The more relevant context you provide, the better the output.

The single most important insight

Treat prompting like writing a good brief for a talented contractor. They can't read your mind. They will do exactly what you ask, no more and no less. Give them everything they need to do the job right — and be specific about what "right" looks like.