MODULE 2

Your First AI Conversation

Get something real built fast.

Spoons: Medium

Time: 30–60 min

Done when: you’ve had a first Claude conversation and produced a working first draft (even if it’s ugly).

You have a blueprint. You have an idea. Now you're going to hand it to an AI and watch it turn into something real.

This is the part where most people freeze. Not because it's hard. Because it feels like crossing a threshold into territory where they don't belong.

You belong here. Let's go.

Wait, Who's Claude?

You're about to have a conversation with an AI named Claude. Before you do, here's what you need to know about who you're talking to.

The short version: Claude is an AI assistant made by Anthropic, a company focused on AI safety. Claude has been trained on massive amounts of text - including code, documentation, and tutorials. Then Anthropic fine-tuned Claude specifically to be helpful, harmless, and honest. The result: an AI that can write code, explain what it's doing, and actually listen when you say "that's not what I meant."

Why does this matter for you?

Claude was fine-tuned to be patient with beginners. It won't judge you for not knowing what a "function" is. It won't assume you have a computer science degree. It will explain things in plain language if you ask.

Claude was also trained to admit when it's wrong. If it builds something that doesn't work, it won't gaslight you. It will try again. This matters because AI makes mistakes. A lot. The difference is Claude will work with you to fix them instead of insisting it's right.

You're not talking to a search engine. You're not talking to a chatbot with canned responses. You're talking to something that can reason, iterate, and build.

Where We're Doing This

For your first AI conversation, we're using claude.ai - the web version. No installation. No setup. No GitHub account. Just a browser and your blueprint.

Go to claude.ai

Open your browser. Type claude.ai. Hit enter.
Create a free account

Sign up with your email. Verify it. You're in.
Start a new chat

Click "New Chat" or just start typing in the box.

What about fancier tools? There are more advanced ways to work with Claude - like Claude Code (which you'll learn about in Module 5), a command-line tool that runs on your actual computer and creates real files you can deploy. For now, claude.ai gives you something powerful: artifacts. These are live previews that let you see and click on what Claude builds, right in your browser. You'll iterate faster, learn faster, and get a win faster. Trust the process.

One more thing: Claude doesn't remember previous conversations unless you're in the same chat session. Every new conversation starts fresh. This is actually helpful - if things go sideways, you can just start a new chat. Clean slate. No judgment.

I have a claude.ai account
I can start a new chat
I have my blueprint from Module 1 ready

The Opening Move

This is it. The moment you actually talk to an AI about your app.

Here's what you're going to do:

Get your blueprint ready

Open Module 1. Take a photo of your completed blueprint pages, or have the document open.
Open claude.ai

Make sure you're logged in and have a fresh chat open.
Send this message

Copy the script below, paste it, attach your blueprint photo or type out the key details.

Your opening message - copy this:

I want to build an app. I'm not a developer, but I've filled out a blueprint with my idea. I'm going to share it with you.

Please:

1. Read through my blueprint

2. Ask me any clarifying questions before you start building

3. Start with the simplest possible version that works

4. Show me what you're building as you go so I can give feedback

Here's my blueprint:

[ATTACH PHOTO OR PASTE YOUR BLUEPRINT DETAILS]

Why this script works: You're setting expectations. You're telling the AI you're not a developer (so it won't assume knowledge you don't have). You're asking for questions before building (so it doesn't run off in the wrong direction). You're asking for the simplest version (so you don't get overwhelmed). You're asking to see progress (so you stay in control).

After you send this, the AI will respond. It might ask questions. It might start building. Either way, the conversation has begun.

You just did the thing.

What to Expect Back

The AI responded. Now what?

Here's what typically happens in the first few messages:

Scenario 1: It asks clarifying questions

This is good. This means it's trying to understand before it builds. Answer honestly. If you don't know the answer, say "I'm not sure - what would you recommend?" The AI can make suggestions. You just have to react to them.

Scenario 2: It starts building immediately

This is also fine. Watch what it makes. You'll either think "yes, that's it!" or "no, that's not what I meant." Both reactions are useful information. If it's wrong, you'll tell it. That's how this works.

Scenario 3: It gives you a wall of text or code

Don't panic. Don't try to understand everything. Scroll to the bottom and look for either: a preview you can click, or a question it's asking you. Focus only on that. The rest is the AI's problem, not yours.

Scenario 4: It does something completely wrong

Congratulations, you found a miscommunication. This is normal. This is not failure. Say "That's not what I meant. Let me clarify..." and try again. The AI doesn't have feelings. It won't be offended. Just redirect it.

The golden rule: You are the boss. The AI is the builder. You say what you want. It figures out how.

The 5 Magic Phrases

You don't need to know how to code. You need to know how to communicate. These five phrases will get you through 90% of AI conversations about code.

1. When it's not what you meant:

That's not quite what I had in mind. What I actually want is [describe it differently].

2. When you're overwhelmed:

I'm lost. Can you explain what you just did in plain English, like I've never seen code before?

3. When you want to start over:

Let's take a different approach. Forget what we've built so far and start fresh with [new direction].

4. When it's close but not right:

This is 80% there. The part that's not working for me is [specific thing]. Can you change just that?

5. When you don't know what you want:

I'm not sure what I want here. Can you show me 2-3 different options and I'll pick the one that feels right?

Permission slip: You are allowed to say "I don't know." You are allowed to say "Show me options." You are allowed to say "Start over." You are allowed to change your mind. You are allowed to close the tab and come back tomorrow. This is your project. There are no wrong answers, only iterations.

When It's Not Working

Sometimes the conversation goes sideways. Here's how to diagnose and fix it.

Symptom: The AI keeps building the wrong thing
Fix: Your blueprint might be unclear. Go back to Module 1, tighten up your one-sentence pitch, and start a new conversation.

Symptom: You can't see what it's building
Fix: Say: "Can you show me a preview I can actually click on and interact with?" In claude.ai, this creates an artifact - a live preview in your browser.

Symptom: It's asking questions you can't answer
Fix: Say: "I don't know enough to answer that. What would you recommend for a simple first version?" Let it make the technical decisions.

Symptom: You're frustrated and want to quit
Fix: Close the tab. This is allowed. Come back tomorrow with fresh eyes. The conversation will still be there, or you can start a new one. Nothing is lost except frustration.

Symptom: The AI mentions GitHub, repos, or deployment
Fix: Say: "Let's stay in artifacts for now. I'll worry about saving and deploying the code in a later step." That's Module 5 stuff. Not now.

The nuclear option: If everything is broken and you hate it: start a completely new conversation. Copy your blueprint text, open a fresh chat, paste the opening script from page 3, and try again. Sometimes a fresh start is faster than fixing a tangled conversation.

Pre-Flight Checklist

Before, during, and after your first conversation. Check these off as you go.

Before you start

I have my Module 1 blueprint completed (or mostly completed)
I have a claude.ai account
I have 30-60 minutes of relatively uninterrupted time
I have a drink and my brain is reasonably online

During the conversation

I sent the opening script
I answered any clarifying questions (or said "you decide")
I saw something - a preview, an artifact, a working thing
I gave feedback at least once ("yes that's it" or "no, change X")
I used at least one of the 5 magic phrases

After the conversation

I have something that works (even if it's ugly or incomplete)
I know at least one thing I want to change next time
I didn't mass delete everything in a shame spiral
I gave myself credit for doing a hard thing

Parking Lot

Stuff that came up during your conversation that you want to remember, ask about later, or figure out in another module. Get it out of your head and onto paper.

Questions I still have: _______________________________________________

Things the AI said that I didn't understand: _______________________________________________

Ideas that came up that I want to add later: _______________________________________________

Module 2 Complete

You talked to an AI about your app.

You saw something come to life.

You gave feedback and watched it change.

That's building software. You just did it.

What's next - Module 3: Module 3 is called "WTF Is Happening" and it's going to give you just enough technical vocabulary to understand what the AI is doing - without making you learn to code. You'll learn what repos, commits, and deployments actually mean, in human terms. Not so you can do them yourself, but so you're not lost when the AI mentions them.

Take a break. You earned it.