From "I have an idea" to a working app you can test.

Finding your saved place...

Start here

Short reads for when you need them

You do not need to read everything before starting. Follow the path, and open a lesson when you want a little more explanation. Tap a word’s + to see its meaning.

When a lesson mentions Lunchbox, we mean our example app. It's an app for parents of toddlers to help them plan school lunches for the week.

01 · 3 min read

An app is like a house

A house has rooms you can see and pipes and wires you may not see. An app also has visible parts and work happening behind them.

A drawing of a house cut open, labeling the plan, the foundation, the rooms, the look, the hallways, and the plumbing and electrical, like the parts of an app.
The planThe blueprint

Your idea. What are we making, who is it for, and what does it need to do?

The foundation

The work behind the screens: saving information and checking who may use it.

The rooms

The part people see and move through: screens, buttons, forms and lists.

: User InterfaceThe look

The things you see and use, such as buttons, words and pictures. Like the doors, switches and signs in a house.

: User ExperienceThe experience

What it is like to use the whole app. Can you find what you need and finish the job?

NavigationThe hallways

How people move from room to room. The paths that connect everything.

SystemsPlumbing and electrical

The parts that connect services, such as sign-in and payments. An is an agreed way for one program to ask another to do a job.

You can sketch the rooms and choose a look first. Before we build the finished rooms, we need a clear plan and a safe foundation.

02 · 9 min read

How a computer follows your instructions

Code is written instructions for a computer. Different languages and tools can do the same job. You do not have to write the examples below. Read the plain English that goes with each one.

Programming languages are languages

This was one of the things that finally made code less intimidating to me. The languages people speak do not always put words in the same order.

In English, we might say:

I put on my shoes.

Japanese can structure the same thought more like:

I, shoes, put on.

That is simplified, obviously, but the point is that the thought can be the same even when the order of the words is not. Some languages commonly put the verb earlier. Some put it later. The grammar changes. Code started making more sense to me when I realized programming languages can work like that too.

I might give a computer an instruction in the order that sounds natural to me:

When somebody taps this button, change Hello to Hi there.

One programming language might essentially say:

Remember the message. Show the message. Make the button. Now put the instruction for what happens inside the button.

Another might say:

Here is the button. When it gets clicked, go find the message and change it.

The words move around. The punctuation looks different. The order can be different. But they can still be communicating the same basic instruction. The computer is not speaking nonsense.

I just do not know its grammar yet.

Same instruction. Different computer languages.

We would say it like this

When somebody taps this button, change Hello to Hi there.

Hello

When you’d use it

A common choice for websites and web apps. It runs in every web browser. TypeScript is JavaScript with extra checks.

How this language tends to say it

  1. Thing
  2. event
  3. find the target
  4. change it

here's what the code says, in plain English

There is some text called message. There is a button. When the button is clicked, find the message and replace its words with “Hi there.”

Actual code

<p id="message">Hello</p>

<button onclick="changeMessage()">
  Change it
</button>

<script>
  function changeMessage() {
    document.getElementById("message").textContent = "Hi there";
  }
</script>

The highlighted part changes Hello to Hi there.

Translation: what this code is saying in plain English

  • onclickWhen clicked
  • getElementById("message")Find the thing called message
  • textContent = "Hi there"Replace its words with Hi there

The examples are short pieces of an app. Your would add the other files needed to run them.

3 min read

Other languages you’ll hear about

There are many other programming languages. Most of the time, your picks which ones to use. Knowing the names helps you follow along when they come up. Tap a name to learn more.

is a programming language that reads almost like English.

Common for: working with data, AI and automation (little programs that do chores), and the behind-the-scenes part of websites.

Why people pick it: it is easy to read, and people have shared many ready-made tools for it.

You might hear it when: your AI writes a little script to clean up a spreadsheet.

: HyperText Markup LanguageWhat is on a web page

It labels things such as headings, pictures and buttons. For Lunchbox, puts a Save lunch button on the page.

: Cascading Style SheetsHow the page looks

It sets colors, sizes and spacing. For Lunchbox, makes that button green and gives the words room inside it.

What the page does

It can respond to a tap. For Lunchbox, it can send the chosen lunch to be saved.

What you want to buildCommon languages or tools your AI may use
A website for the page parts, CSS for their look, and or for actions.
An iPhone or iPad app is one language used for Apple apps.
An is one language used for Android apps.
One project for iPhone and Android is one tool. It uses or , and the result still needs checks on both kinds of phone.

Ask your which tools fit your app and why. Your job is to explain what you want, try what it makes and point out anything that does not work.

03 · 3 min read

The two architectures

Two plans for the same app: what the person needs, and how the working parts support it.

A diagram showing the foundation files, the technical architecture on one side, the product architecture on the other, both leading to the app in the middle.

means a plan for how things fit together. You do not need to be an engineer to describe what you want.

The person’s experience

For Lunchbox: a parent plans five school lunches, reviews them, and gets one shopping list. Describe the steps, the wording, and what happens if something goes wrong.

The working parts

For Lunchbox: save meals under the right family, let approved family members share the plan, and prevent other families from seeing it.

Your written instructions

Save the agreed plans in short text files. Each file has a job, such as product.md for what you are making or build-order.md for what comes first.

These files use , a simple way to write text with headings and lists. Here you are writing your app’s , not app code.

Your is the agreed version. If Lunchbox defines a week as Monday through Friday, every builder follows that definition until you change it.

You will save both guides and fill them in with your in the Foundation phase. Read and correct the files before asking it to build.

04 · 3 min read

When the AI gets it wrong

You direct the work. When an says “done,” open the app and check what it actually does.

Start with what you expected and what happened. For Lunchbox: “I chose five lunches and saved. After I reopened the page, the plan was empty.” Write the steps and exact . Remove passwords and private details before sharing.

Pause and describeName the problem

Try the task again. Say which screen you used, what you tapped and what happened. Correct anything the misunderstood.

Ask it to show whyCheck the evidence

For a , ask the to show which part failed and what check will catch it. For a factual claim, open the source link and check that it says what the AI claims.

Ask your other to review itGet another view

Give it the same files, instructions and test results. Ask what the first missed. Two matching answers can still be wrong.

Try the task yourselfSee whether the fix works

In Lunchbox, choose five lunches, save, close and reopen. Check the same meals and shopping list are there. Also check that another test family cannot open your private plan.

Keep changes small. Save a version you can return to, then send it to . Your tools call those two steps and . Try new changes in a separate working version, called a . Check them before adding them to the main version, called a . Ask whether sending or adding the changes will also put them online.

If the same problem keeps returning, stop adding fixes. Ask how to return to the last working version without losing saved information. Get a fresh review of the problem.

Copy the reusable prompt
05 · 2 min read

Keep a short note for your next chat

A new chat may not know what you already decided. A short note helps you continue without starting over.

Long chats can use more of your tool's limit and become harder to follow. Tools handle old messages in different ways. A short helps you keep the decisions that matter.

Try to keep one chat to one job: one phase, one feature, one problem. Start a fresh chat when you move to something else.

When a chat gets long, or you're starting something new, write a instead of scrolling back to find things.

A is a short note to your next chat: what's done, what's next, and anything it needs to know to pick up exactly where you left off. Paste it in before you ask for anything new.

A new chat may not have the details from your old chat. Keep your foundation files in a project or shared space your can read. Ask it to name the files it can see before it starts. If your tool has no project space, attach the files or paste the short .

Save the conversation somewhere you can search later, like a Google Doc. It matters most for your long build chats: the decisions, the reasoning, the exact wording you landed on. Searching a doc for "the button color we picked" beats scrolling through months of chat history.

06 · 3 min read

Which AI for which job

Different AIs are good at different jobs, and the names change every few months. Learn the four jobs once. Swap in the newest names whenever you need to.

Start with the tools you already have. I recommend ChatGPT and Claude so you can get two perspectives. You do not need a different for every job.

JobWhat to ask for
PlannerHelp me turn my Lunchbox idea into a clear first version. Ask questions before filling gaps. When the plan is ready, write the prompt for my builder.
BuilderUse a coding tool that can change the app files and run checks. Give it the prompts your planner writes. Cursor is optional.
ReviewerGive a second the plan, actual changes and test results. Ask what is wrong or missing.
Quick helperExplain this word or in plain English and tell me the next small step.

Your planner is your planning partner. Pick one and stay with it. Use it to brainstorm, plan it all out and make your instruction files. Then it writes every prompt for your builder, including changes you want later. Before it writes one, it repeats back what it understood in 3 sentences or less, so there's no misunderstanding.

names, available features and plan limits change. Check the tool’s current options when you choose it. Paying more is useful only if it solves a limit you actually have.

Two AIs agreeing is not proof. Open the app and check that the promised task works.

07 · 3 min read

Build from anywhere

Some tools keep your work online so you can reach it from another device. Check what your chosen tool saves before switching.

The means computers you reach over the internet. Work saved there can be opened from another device if your tool and account allow it. Work saved only on your own device will not appear elsewhere by itself.

Your code on An online copy you can check

Ask your to save a version and send it to . Open the files there before switching devices. Only work that reached GitHub is in that copy.

Use the same account in each toolSame account, any device

Use the same sign-in on your phone and your laptop. Check what each tool actually . In this Playbook, signed-in checkmarks and Your app can sync. Brand notes currently stay in this ; use their option before moving devices.

A An AI helper running online

Some building tools can keep working online after you close your laptop. Check that your tool supports this. Give it one small job and ask for a draft in a separate working version, called a . Review the draft before publishing.

If you have only a little time, choose a small job such as fixing a typo. Come back and check the result before using it.

Good jobs for a short window: fix a typo, change some wording, write a first draft of a page, or add one small feature you've already described in your . Save big decisions and anything involving money, logins or private data for when you're at a real keyboard.

Check the changes before adding them to your main version. Your makes a draft. You decide what to keep.

Codex and Claude Code are tools that can help change app files and run checks. Choose a version you can open on your device. Check its current access and costs in the official instructions. A free chat account may not include work done by an building tool. See https://developers.openai.com/codex and https://code.claude.com/docs/en/overview.

Before switching tools, save your work and send it to . Open the files there to check that the new version arrived, including your app’s instructions. Use the same GitHub account on each device. Never put passwords or secret keys into the chat.

Less guessing Fewer mistakes A working app
a kinder way to build