Getting started · Claude Code Club

Where to Start: Stage 1, Your First Build This Week

16 minute read

I got laid off from a startup after two years. Before that I'd freelanced for 14 years, and it was my first full-time job. I'd spent weeks building a 60-plus slide plan to fix the company's YouTube channel. Then someone else was put in charge of it. The official reason for letting me go was "not enough motion graphics work." That was true. It still stung, because I didn't feel seen. I knew I could help in a dozen places.

There was no mentor and no roadmap, so I started posting every day. About six months later my YouTube channel was at 56,000 subscribers. I've also never written a line of code by hand. I build with AI anyway.

You're probably in a different spot. You're busy, you're not a coder, and all the AI advice out there doesn't tell you what to do first. That's the problem this guide fixes. It puts eight steps in order, and every step has a prompt you can paste.

What Stage 1 is, and how long it takes

This is Stage 1 of Where to Start, a step-by-step path for people who already have a full plate and want to build something of their own with AI. One stage, one goal. The goal of Stage 1 is a working build that you've used this week. Not a course. Not a pile of tools. One small thing that runs and that you actually use.

You don't need to code. You do need to be able to describe a problem in plain English, the way you would to a new assistant. Claude Code (an AI helper you talk to, which writes and runs the code for you) does the typing. You do the deciding.

Plan on a few evenings. Most steps take 20 to 45 minutes, and Step 6 is just a few minutes a day while you use the thing. Every step below has the same parts: why it matters, what to do, one prompt (the message you send to Claude) that you can copy and paste, the mistake people make most, and how long it takes. The order is the point. Don't skip ahead.

• Evening 1: Steps 1 and 2. Pick the annoyance and set up Claude Code.
• Evening 2: Steps 3 and 4. Write the brief and build version 1.
• Evening 3: Step 5. Fix what breaks until it works.
• The next several days: Step 6. Use it for real.
• One more evening: Steps 7 and 8. Make it yours and pick build number 2.

Step 1: Pick the one annoyance

I'll walk through this guide next to a working mom in the Claude Code Club. She went from first installing Claude to a live app in about six weeks. Her first build started with a chore she already had: planning what her kids would do each school holiday.

Why it matters: the first build decides whether you keep going. Pick something big and you stall. Pick something you don't care about and you quit. The right first build is a small, repeated annoyance in your own week, because you'll notice right away whether it helps.

What to do: look back over the last seven days. Open your calendar, sent email and messages, and scroll. Write down every task that was repetitive, slow, or that you did by hand when you wished you didn't. Messy is fine. Aim for 10 or more. Then paste the list into the prompt below. You can do this in a normal Claude chat, since you haven't set up Claude Code yet.

CopyI run [what you do, for example: a coaching business, a design freelance practice, a small bakery]. I am not a coder and I have very little spare time.

Here is a messy list of things I did by hand last week:
[paste your list here, one per line, rough is fine]

Help me pick ONE thing to build with AI as my first project.
1. Rank every item on my list as a possible first build. Score each one on: how often it happens, how long it takes me, how clear the input and the output are, and how low the risk is if it goes wrong.
2. Tell me which one to pick and why, in 3 sentences.
3. Describe what the first small version would do in plain English, in under 100 words.
4. Name one backup pick in case I do not like the first.

Use plain words. No jargon. If you need to know more about my work before you can rank, ask me up to 3 questions first.

Time: about 30 minutes. Your pick should pass one test: if it vanished tomorrow, would you notice? If yes, you have the right one.

Step 2: Set up Claude Code in about 30 minutes

I've never written a line of code by hand, and I work in the terminal anyway. The working mom's clock started at the same point. She said she first installed Claude about six weeks before her app went live.

Why it matters: this is the step that scares people most, and it's the shortest once you start. You're doing three things only: installing Claude Code, opening a folder, and checking it can make a file.

What to do: use Anthropic's own install instructions, because they're kept current and our copy of any command would go out of date. Plans and prices change too, so check Anthropic's page for what you need to run Claude Code. Our beginner guide walks through the setup, including the terminal (the text window where you type commands to your computer, which looks scary and isn't). Find it here: https://www.claudecodeclub.ai/free-resources/coding-with-ai-beginners-guide

1. Follow the install steps in the beginner guide or on Anthropic's site.
2. Make a new empty folder on your computer. Name it after your project, for example client-intake.
3. Open the terminal inside that folder, type claude, and press Enter.
4. Paste the test prompt below. If a file appears in the folder, you're set up.

CopyCreate a file called hello.txt in this folder that says "Claude Code is working". Then tell me exactly where the file is on my computer, in plain words.

If something fails, copy the exact error message and paste it into a normal Claude chat with this line: "I am not technical. I got this error while installing Claude Code on [Mac or Windows]. Tell me what it means and what to do, one step at a time." This solves most setup problems. If you want Claude to know how you work from day one, our starter files guide covers the five files worth making: https://www.claudecodeclub.ai/free-resources/claude-code-starter-files

Time: about 30 minutes.

Step 3: Describe it like a client brief

Here's the whole skill, in my words: "It's less about code and more about being willing to try. Know what you want and ask for it. Then make it better." I used the same skill for years as an art director and motion designer. Her finished app asks a few quick questions, then gives a day-by-day plan.

Why it matters: Claude builds what you describe. A vague request gets a vague tool. A clear brief, the kind you would hand a good freelancer, gets something close to what you pictured on the first try. This step is where most of the quality comes from.

What to do: fill in the prompt below. Talk it out if typing is a drag. You can speak into your phone, paste the rough text, and let Claude tidy it. The prompt makes Claude ask you questions before it builds, which catches the gaps while they're cheap to fix. If you want to see this done by talking instead of typing, we have a video on it, right below the prompt.

CopyI want to build a small tool for myself. Here is the brief.

What it is: [one sentence, for example: a page where I type in a new client's answers and get a one-page summary]
Who uses it: [just me / me and my team / my clients]
What goes in: [what you will give it, for example: a client's name and their answers to 8 questions]
What comes out: [what you want back, for example: a clean one-page summary I can print or save]
What "done" looks like: [how you will know it works, for example: I can finish a summary in under 2 minutes]
What it must NOT do: [for example: send anything to anyone, cost money, need a login]

Before you build anything:
1. Ask me up to 5 questions about anything unclear. Ask them one at a time.
2. Then write a plan in plain English, as a short list of what you will build, in what order.
3. Wait for me to say OK before you start.

Step 3 by talking instead of typing: the Ramble Session.

Time: about 20 minutes, including answering Claude's questions.

Step 4: Build version 1

She was about a week into using Claude when she started building her planner. That's early. Version 1 doesn't need you to know much. It only needs to exist.

Why it matters: version 1 means the first rough working copy. Its job isn't to be good. Its job is to exist, so that you can react to something real instead of imagining.

What to do: once you've approved the plan from Step 3, tell Claude to build only the must-haves. Keep it on your own computer, with no accounts and no paid services. That keeps it simple and safe. Claude will ask permission before it creates files or runs commands. Read what it asks and say yes when it makes sense. If it heads the wrong way, press Esc to stop it.

CopyThe plan looks good. Build version 1 now.

Rules:
- Build only the must-have parts of the plan. Leave out anything nice-to-have and list those at the end.
- Keep it simple. Make it something I can open on my own computer, with no accounts, no logins and no paid services.
- Put every file in this folder.
- When you finish, tell me in plain words: (1) how to open it, (2) three things to try, and (3) what you chose NOT to build.

Then try the three things. Use real information from your work, not made-up examples. Real input shows you the real problems.

Time: 30 to 45 minutes, mostly Claude working while you watch.

Step 5: Fix what breaks

She told the Club her app is a bit buggy. It went live anyway. I always say troubleshooting is about 90% of the job. Try it, look at it, try again.

Why it matters: something won't work the first time. That's normal. It doesn't mean you did it wrong. Fixing things is the main job, and it's easier than it sounds, because you're only describing what you see.

What to do: tell Claude three things: what you did, what you expected, and what happened instead. If an error message appears, paste it in word for word. If it looks wrong rather than broken, take a screenshot and give Claude the file name. Fix one problem at a time, and test after each fix.

CopySomething is not working right.

What I did: [the exact steps, for example: I typed 8 answers and clicked Make Summary]
What I expected: [for example: a one-page summary]
What happened instead: [for example: the page went blank]
Any error message: [paste it exactly, or write "none"]

Please:
1. Find the cause and explain it to me in two sentences, in plain words.
2. Fix it. Change only what is needed. Do not rewrite everything.
3. Tell me how to check that it is fixed.

If a fix makes things worse, say so: "That made it worse. Undo that last change and try a different way." Claude can step back. You aren't stuck with a bad change.

Time: 15 to 45 minutes, depending on how many problems show up. Two or three is common.

Step 6: Use it for a week

Her friends tested the app and told her it works outside her own country. She'd mostly tested it on data from home. Then people she didn't know started using it, and she said some of them sent detailed feedback. That's what real use does.

Why it matters: a tool that only works in a demo isn't a build. Real use shows you what's actually missing. Most people skip this step and go straight to adding features, which means they add the wrong ones.

What to do: use it for your real work for about a week. Each time you use it, jot a line in a notes file or on your phone. You're looking for four things.

• Where you avoided it and did the job by hand anyway.
• What annoyed you, even a little.
• What you wished it did.
• What worked so well you didn't think about it.

At the end of the week, paste your notes into the prompt below. It turns scattered notes into a short, ranked to-do list.

CopyI used the tool I built for a week. Here are my raw notes:
[paste your notes, rough is fine]

Sort them into four groups: broken, annoying, missing, and working well.
Then give me the top 3 changes to make next, ranked by how much time each would save me. For each one, say in one sentence what the change is and how hard it is (small, medium, or big).
Do not change anything yet. Just give me the list.

Time: 5 minutes a day for a week, plus 10 minutes to sort your notes.

Step 7: Make it yours

After my layoff I posted every day, sometimes two or three times a day. One series became my signature: "5 ChatGPT prompts for side hustles," "for passive income," "to learn anything." I became the ChatGPT prompts guy. That's what making it yours looks like. One thing, repeated, until it's recognizably yours.

Why it matters: version 1 works. Now it should feel like it belongs to you. A bit of polish, in your name, colors and wording, is the difference between a rough experiment and something you're proud to use and show.

What to do: pick the top item from your Step 6 list, plus one polish pass. Do both in one go, but no more than that. First ask Claude to save a copy of the working version, so you can always go back.

CopyBefore changing anything, save a copy of the current working version into a folder called backup-v1.

Then make two changes:
1. Polish: make it look and feel like it belongs to my business. My business name is [name]. My colors are [colors, or "pick calm, professional ones"]. My tone is [for example: warm and plain]. Use my wording, not tech wording.
2. New feature: [the top item from your Step 6 list, in one sentence]

After you finish, tell me what changed and how to check each change works. If anything I could lose is at risk, tell me first.

Time: 30 to 45 minutes.

Step 8: Pick build #2

The working mom didn't stop at one build. She said she's pursuing four or five different ideas. Her next challenge is using her Claude skills on a more corporate-focused project. My version was posting every day after the layoff, with no roadmap.

Why it matters: one build proves you can. Two builds make it a habit. The second is faster and better because you now know the steps. This is where building stops being a project and becomes how you work.

What to do: go back to Step 1, with a better list. You now know what Claude can do and how long things take, so you'll spot buildable things you could not see before. Put a 30-minute slot in your calendar each week for the next build. Then ask Claude to write down what you made, so it remembers the next time.

CopyI just finished my first build. Help me start the next one.

1. Write a short summary (5 lines max) of what I built, how it works, and where the files are. Save it in a file called ABOUT.md in this folder.
2. Based on what you learned about my work while we built this, suggest 5 more things I could build, from the quickest and simplest to the biggest.
3. Recommend which one to build next, and say why in two sentences.

Use plain words. No jargon.

Time: about 20 minutes to plan, plus a weekly slot.

A worked example: a working mom in the Club builds a school-holiday planner

This is a real story from a Claude Code Club member, told without her name. She's a working mom. About six weeks after first installing Claude, she had a school-holiday and weekend planner live. Here's how her path lines up with the eight steps. She didn't share every step, so here's what she did share.

• Step 1: she picked a problem from home. She wanted a better way to plan what her kids do each school holiday.
• Step 2: she first installed Claude about six weeks before the app went live.
• Step 3: the finished app asks a few quick questions and gives a day-by-day plan for school holidays and weekends.
• Step 4: she started building about a week into using Claude.
• Step 5: she said the app is a bit buggy, and she put it live anyway.
• Step 6: friends tested it and said it works outside her own country. Then people she didn't know started using it and sent detailed feedback.
• Step 8: she said she's pursuing four or five different ideas, and her next challenge is a more corporate-focused project.

Two honest notes. Her six weeks is longer than the few evenings in this guide, and she went further by putting the app online for other people. That's Stage 2 territory, and your timeline will be your own. What she didn't do is wait for it to be perfect. She started from a problem at home, kept it small, shared it, and moved on to the next idea.

The one-page checklist

Copy this part. It's the whole sequence on one page.

1. Pick the one annoyance. List last week's chores, rank them, choose one. (30 min)
2. Set up Claude Code. Install it, open a folder, run the hello.txt test. (30 min)
3. Describe it like a client brief. What goes in, what comes out, what done looks like. Let Claude ask questions first. (20 min)
4. Build version 1. Must-haves only, on your own computer. Try it with real information. (30 to 45 min)
5. Fix what breaks. Say what you did, what you expected, and what happened. One problem at a time. (15 to 45 min)
6. Use it for a week. Jot notes. Sort them into broken, annoying, missing, working well. (5 min a day)
7. Make it yours. Save a backup, polish the look, add one feature. (30 to 45 min)
8. Pick build #2. Write down what you made and book a weekly build slot. (20 min)

You're done with Stage 1 when you've used something you built, for real, on at least three different days.

Stage 2 preview: make it save you hours

Stage 1 gets you a working build. Stage 2 is about making that build work harder for you. Here is what it covers.

• Connect your build to the tools you already use, like your email, calendar, and spreadsheets.
• Make it run on its own, so you don't have to start it each time.
• Put it online, so your team or your clients can use it too.
• Turn what you learn into rules Claude remembers, so every new build starts faster.

If you want to keep going with other people who are building the same way, the Claude Code Club is $9/month. Members work through each stage together and share what they make. You can find it here: https://www.skool.com/claudecodeclub/about

If you'd rather go at your own pace, that works too. Finish Stage 1 first. One working build is worth more than ten ideas.

claudecodeclub.ai