Module 630 min · Hands-on project

Your First AI Project

Everything in this course has been building toward this. You understand what AI is, how it works, what tools exist, how to prompt effectively, and what the real risks are. Now you're going to build something with it — a custom AI assistant that solves a real problem in your life.

What you're building

A custom AI assistant — powered by Claude or ChatGPT — that has a specific job, a defined personality, clear rules for how it behaves, and genuine usefulness in your daily life.

This is not a toy. A well-built custom assistant can save you hours every week, help you think through decisions more clearly, or help you get better at something you're trying to learn. The quality of what you build is entirely determined by the quality of your system prompt — the instructions that define how your assistant behaves.

Why this project specifically

Building a custom assistant puts every skill from this course to use at once. You have to understand how AI models work to know what's possible. You have to understand prompt engineering to make it work well. You have to think about the ethical implications of what you're building. And you have to apply it to something real — which is where genuine learning happens.

Step 1: Choose your assistant type

Pick one that solves a real problem you actually have. The more specific the problem, the more useful the assistant.

📚 Study Coach

Best for: Students who want smarter study sessions

Helps you understand material, not just memorize it. Quizzes you, identifies gaps, explains concepts multiple ways, and pushes back when your answers are vague.

✍️ Writing Coach

Best for: Anyone who writes and wants to improve

Reviews your writing and gives direct, honest feedback. Identifies weak sentences, vague arguments, and structural problems — without sugarcoating. Asks you what you're trying to say before telling you how to say it better.

💡 Decision Advisor

Best for: People who want to think through decisions more rigorously

Helps you think through important decisions systematically. Identifies assumptions, asks what you might be missing, steelmans the opposing view, and refuses to just validate what you already think.

📈 Finance Tracker

Best for: Anyone building financial knowledge and habits

Explains financial concepts clearly, helps analyze investment decisions, calculates compound growth scenarios, and challenges you to think long-term rather than react to short-term noise.

🎯 Goal Accountability Partner

Best for: People with big goals who need structured thinking

Tracks your goals and progress, asks hard questions about why you're falling short, helps you identify obstacles and real solutions, and doesn't accept vague answers or excuses.

Step 2: Write your system prompt

The system prompt is the set of instructions that defines everything about your assistant. It runs invisibly in the background of every conversation, shaping every response. Here's the structure to follow:

Define the role and purpose

What is this assistant? What is its one job? Be specific. "You are a writing coach" is weak. "You are a direct, honest writing coach who helps ambitious students write more clearly and persuasively" is strong.

Set the personality and tone

How should it communicate? Direct and challenging? Warm and encouraging? Socratic — asking questions rather than giving answers? Define this explicitly or the AI will default to generic helpful-assistant mode.

Write the behavioral rules

What should it always do? What should it never do? "Always ask what I've already tried before giving advice." "Never just validate what I said without pushing back." "If my answer is vague, ask me to be more specific before continuing."

Give it context about you

What does your assistant need to know to be useful? Your goals, your level of knowledge, your situation. The more relevant context you give it, the more tailored its responses will be.

Define what success looks like

What outcome are you trying to achieve through this assistant? Being explicit about this helps the AI stay focused on what actually matters rather than drifting into generic helpfulness.

Step 3: An example system prompt

Here's a complete, well-written system prompt for a Study Coach assistant. Read it carefully — notice how specific every element is:

Example System Prompt — Study Coach

You are a rigorous, Socratic study coach helping a high-achieving high school junior in Louisiana master challenging academic material. Your job is not to give answers — it is to build genuine understanding. Your personality: Direct, demanding, and encouraging in equal measure. You take the student's goals seriously and refuse to let them settle for surface-level understanding. You ask hard questions. You push back when answers are vague. You celebrate real insight. Your rules: — Never just give the answer. Always ask the student to attempt it first. — When the student gives a vague or incomplete answer, ask "Can you be more specific?" or "What do you mean by that?" before moving on. — After the student answers correctly, follow up with "Why?" or "What would change this answer?" to deepen understanding. — If the student says "I don't know," respond with "Take a guess — what do you think it might be?" — End every session by asking: "What's the one thing from today you want to make sure you remember?" Context about this student: Strong work ethic, ambitious long-term goals, tends to understand concepts quickly but sometimes moves on before truly internalizing them. Preparing for AP courses and standardized tests. The goal: Build the kind of deep, transferable understanding that shows up on tests, in college, and in life — not just the ability to recognize the right answer on a multiple choice question.

Step 4: Build and test it

Open Claude at claude.ai or ChatGPT at chat.openai.com. Start a new conversation. For Claude, you can paste your system prompt directly into the first message and tell it to behave as that assistant going forward. For ChatGPT Plus users, you can create a custom GPT and paste the system prompt into the instructions field.

Then test it rigorously. Give it inputs you'd actually use. See where it drifts from what you wanted. Go back and update your system prompt. Iterate until it behaves consistently the way you designed.

The iteration mindset

Your first version will not be perfect. That's not a failure — it's the process. Every time the assistant doesn't behave the way you wanted, ask yourself: what would I add to the system prompt to prevent this? Then add it and test again. The skill of building good AI systems is mostly the skill of writing clear, specific instructions and iterating on them.

Step 5: Use it for real

The final step is the most important one: actually use what you built. Have real study sessions with your study coach. Work through real decisions with your decision advisor. Review real writing with your writing coach.

This is where you'll discover what works and what doesn't. You'll find places where your prompt was too vague. You'll find behaviors you didn't anticipate. You'll refine it based on real use rather than imagined use. And you'll end up with something genuinely useful — an AI assistant built around your specific needs, that didn't exist before you built it.

What to do with it

Share it. Show it to someone who could use a similar tool. Document what you built and how you built it — the thinking behind the system prompt, what you iterated on and why. This kind of documented, thoughtful work is exactly what college admissions officers, internship managers, and future employers want to see evidence of: real initiative, real building, real learning.