DEV Community: Amber

I don't know if people even read my stuff or not.

Amber — Thu, 22 Oct 2026 12:12:50 +0000

If you could build anything, like an app, website, or game, what would you build?

Amber — Tue, 20 Oct 2026 11:23:03 +0000

How to Get Your Lovable App Into the App Store and Google Play (The Easiest Way I Found)

Amber — Tue, 20 Oct 2026 11:18:40 +0000

I'm not a developer. Here's how I turned my Lovable web app into real iOS and Android apps and got both submitted in under a day.

I built my app with Lovable and it works great in the browser. Then people started asking if there was an app they could download, and I realized I had no clue how to get it into the App Store or Google Play.

I'm not technical. I can describe what I want and click the right buttons, and that's about where it ends. This is the route I wish someone had shown me before I burned all that time searching.

Everything else I found was too hard for me

Lovable makes web apps. There's no "export to iPhone" button, so you need something else to get into the stores. When I searched, I kept landing on the same few options.

The first was wrapping the site with Capacitor. Every tutorial assumed I had a Mac with Xcode, plus Android Studio, and already knew what signing certificates were. I didn't get far.

Then there are paid services that put your website inside an app shell. That sounded perfect until I read that Apple rejects a lot of these under Guideline 4.2, which basically says your app has to be more than a repackaged website. I didn't want to pay for something and then get rejected.

The last option was hiring a freelancer. Possible, but expensive for a side project, and I'd need them again every time I wanted to change something.

I went back and forth between these for way too long.

Then I asked Claude

At some point I stopped googling and asked Claude something like "what's the easiest way to get a Lovable app into the App Store and Google Play if I can't code?" One of the tools in the answer was CatDoes. I'd never heard of it, but it said it could take an existing project and rebuild it as a real native app instead of wrapping the website. So I gave it a shot.

What I actually did

First I connected my Lovable project to GitHub. It's in Lovable's project settings, and it pushes your code to a repo. If you've already done this, skip it.

Next, in CatDoes I created a new project, chose import from GitHub, picked my repo, and hit import.

Then I mostly waited. The agent read my code and rebuilt it in React Native, which is what makes it an actual iPhone and Android app and not a website in a box. My Lovable app already used Supabase for logins and data, and the mobile version just kept using the same backend. Same database, same user accounts.

About 30 minutes later the native iOS and Android builds were done.

One tip: try the app on your phone with the Expo Go app before you submit anything. The preview in the browser isn't always exactly what you get on a real phone.

Submitting to the App Store

I never opened Xcode. You need an Apple Developer account, which is $99 a year. CatDoes asks for your Apple details and an Expo token, and it walks you through making the token. Then it submits the app to the App Store for you. You can send it to TestFlight first if you want friends to test it.

Submitting to Google Play

Google Play took one manual step. CatDoes builds the Android file (an .aab), I downloaded it, and I uploaded it in the Google Play Console myself. The developer account costs $25 once.

The Play Console has a lot of screens. CatDoes has a step-by-step Google Play guide that I kept open in another tab while I filled them in.

Heads up: if your Google Play developer account is a new personal one, Google makes you run a closed test with 12 testers for 14 days before you can release to everyone. That's Google's rule, not the tool's, but it's worth knowing before you promise anyone a launch date.

All done in under a day

I still can't believe how much time I spent looking for a way to do this. Once I found one, the whole thing took less than a day, from importing the code to having both apps submitted. Apple's review takes its own time after that, but that part is out of your hands anyway.

What it cost

I bought the Plus plan for $100 a month, which comes with 400 credits. Credits get used as the agent works, so keep an eye on them if you ask for a lot of changes.

You might not need Plus. The $50 Starter plan also publishes to both stores. I went with Plus because it has two-way GitHub sync and code export, and I wanted my code to stay in my own repo.

On top of that you pay the stores directly: $99 a year for Apple and $25 once for Google.

Other things worth knowing

It has its own backend built in, with a database, logins, file storage, and server functions. I didn't need it since my app already had Supabase, but if you start a new app from scratch you don't have to set up Supabase or Firebase separately.

GitHub import is how I brought my Lovable code over. You can also paste your site's URL, but if your app has pages behind a login, like a dashboard, GitHub works better because the agent reads the actual code instead of only what's public.

Before you submit, it can check your app against the App Store review rules and point out what would probably get rejected. I'd never submitted an app before, so I liked having that.

It also builds websites and connects custom domains, though I haven't used that part yet.

If you want the longer version, CatDoes has a full guide on converting a Lovable app to a mobile app. It goes deeper on URL vs GitHub import and why wrappers get rejected.

My advice

Don't do what I did and spend days reading Capacitor tutorials you don't understand. Connect GitHub, import the project, and put the time you save into your store screenshots and description.

How I made $5,000 in a week selling websites to local businesses

Amber — Wed, 14 Oct 2026 20:15:41 +0000

I made $5,000 in one week building websites for local businesses. I spent close to $900 doing it in September.

I didn't write any code. I didn't run ads and I didn't have a portfolio. I had a list of businesses from Google Maps, my phone, and an AI agent that turns a business name and a few details into a working website in about ten minutes.

I contacted 11 businesses that week, and five of them paid me to build the real site. It's one of the highest-margin things I've ever done. So here's all of it: where I found the businesses, what I said on the phone, what I emailed them, what it cost, and what happens once the site is live.

The short version

Find local businesses on Google Maps with no website, or one that's broken.
Build each of them a mock of their future website before you get in touch. Mine took about 10 minutes each in CatDoes.
Call, tell them you've already built them something, and ask for an email to send the link to.
The ones who like it pay you to finish it, with their own photos and their own domain.
After launch, they either pay you monthly to look after it, or you hand them the account and they update it themselves by chatting with the AI.

Why this works

Plenty of small businesses still have no website. When Clutch surveyed 406 US small business owners in August 2025, 17% didn't have one (Clutch). Then there are the ones whose site is down, years out of date, or unusable on a phone. Every town has a list.

The same survey asked the owners without a website why. The most common answer, at 34%, was that a website isn't relevant to their industry. Only 16% said cost.

That second number changed how I think about the pitch. Most of these owners aren't saying no because they can't afford it. They just can't picture what a website would do for them, and explaining doesn't fix that. Showing them does. So I stopped explaining and built the site first.

People turn down sales pitches all day. Turning down a website that already has your name, your prices and your reviews on it is harder.

Step 1: Find businesses on Google Maps

Open Google Maps and search a type of business plus your town. "Hair salon [town]." "Plumber [town]." Nail salons, barbers, auto repair, cleaners, cafés. Then click through the listings one at a time.

You're looking for any of these:

No website button on the listing at all
A website button that leads to an error page, an expired domain, or a "this domain is for sale" page
A site that loads but falls apart on a phone
A Facebook page or a link-in-bio page doing the job of a website
A copyright year in the footer that's several years old

Then narrow it down to businesses that are obviously doing fine: lots of reviews, recent photos, an owner who answers reviews. They have customers and they have money. What they don't have is a page to send people to.

I'd skip listings with a handful of reviews and no photos. If the owner hasn't touched their Google listing in years, a website probably isn't on their mind either.

Keep a spreadsheet as you go with the business name, phone number, what's wrong with their web presence, the mock link, and where things stand. By the middle of the week you'll need it.

Step 2: Build the mock before you call

This is the step that makes the rest work. A few years ago it would have killed the whole idea, because nobody builds eleven free websites on spec. At ten minutes each, you can.

I used CatDoes, an AI agent that builds websites and apps from a chat. You describe the site, it builds it, and you get a live link you can send to anyone. Each mock took me about 10 minutes, sometimes less. All 11 together came to under two hours.

Almost everything you need for a decent first draft is already on their Google listing: name, address, opening hours, phone number, services, reviews. A prompt like this gets you most of the way:

Build a one-page website for [Business Name], a hair salon in [Town]. Put the name and a "Call to book" button for [phone number] at the top. Add a services section with these services and prices: [list]. Add opening hours: [hours]. Add the address with a map, and three customer reviews: [paste]. End with a contact section. Make it look clean and warm, and make sure it works well on a phone. Use placeholder photos. (or Just copy paste the whole page from google map and paste it into catdoes)

When it's done, ask the agent to deploy it to the web. You'll get a link like bellas-salon.catdoes.app that opens on any phone. That link is what you're selling with.

A few things make a mock land better. Use their real name, hours, prices and reviews, because those details are what make an owner stop and look. Keep it to one page; it's a preview, not the finished site. And use placeholder photos and say so. Their real photos are part of the paid job, and you shouldn't be lifting pictures off their listing for a site that might go live.

Step 3: Make the call, and only ask for an email

This is not a sales call. You're not trying to close anything on the phone. You want an email address so the mock can do the selling.

Roughly what I said:

"Hi, is this the owner? My name is [name], I build websites for local businesses around [town]. I was looking at [Business] on Google Maps. You've got great reviews, but I couldn't find a website for you, so I went ahead and built you a sample of what one could look like. There's no charge for it. What's the best email to send it to?"

If their site was broken, swap the middle for "your website didn't load when I tried it on my phone."

Most owners will give you an email just to see what you made. If they ask about price straight away, give them a rough number and say you'd rather they look first.

Keep it honest. Don't say you're calling from Google, and don't hint that something's wrong with their listing. Owners get those calls from shady web design outfits all the time and they hang up fast. "I made you something, can I send it?" is true, and it works better anyway.

Step 4: Send the link while they still remember you

Email right after the call. Short beats clever:

Subject: The website I made for [Business Name]

Hi [Name],

Thanks for taking my call. Here's the sample site I put together for [Business Name]: [link]

It's a draft. The photos are placeholders and the text came from your Google listing. If you like it, I can finish it with your own photos and put it live on your own domain within [X] days.

Happy to change anything you want.

[Your name]
[Phone]

If you don't hear back in a couple of days, follow up once. Then let it go.

Don't like the phone? Walk in

You can run the whole thing in person. Build the mock, walk into the local hair salon on a quiet weekday morning, ask for the owner, and show them their future website on your phone.

Watching an owner scroll through their own salon on a real website is the whole pitch. You barely have to say anything. It's also a lot harder to ignore someone standing at the front desk than an email.

Depending on where you live, walking in may also be the safer option legally. More on that further down.

Step 5: Build the real site

Five of the 11 said yes. From there, the job is finishing something that's already mostly built.

A finished site usually means swapping in their real photos and logo, rewriting whatever text they want changed, and adding any pages beyond the one-pager, like a full service list, a gallery or a booking page. Every one of those changes is a chat message to the agent.

Then their domain. Register it in the client's name so they own it, not you. Connecting it to CatDoes is one DNS record at the registrar, and HTTPS gets set up automatically (CatDoes docs).

One thing to know before you quote: custom domains need the Starter plan or above. The Free and Core plans publish to a .catdoes.app link, which is fine for mocks, but they can't use a custom domain and they put a "Made with CatDoes" badge on the site.

On price, the five sites brought in $5,000 together, so about $1,000 each on average. [FILL: flat price or a range, and whether the domain and setup were included.] I didn't price by the hour. Priced by the hour, I'd have been charging for an afternoon. The owner is paying to finally have a website without dealing with any of it, and that's worth a lot more to them than my afternoon.

Step 6: After launch, they pay you monthly or run it themselves

Every client gets the same two options.

Option one is a monthly fee [FILL: amount], and I handle everything: new prices, holiday hours, new photos, an extra page when they need one.

Option two is that I hand over the CatDoes account and they run the site themselves. At that point they don't need me or any developer. Changing their Saturday hours is a message to the agent, something like "change Saturday hours to 9 to 3." Then they ask it to deploy again and the live site updates. (Edits don't go live until you ask for a redeploy. Tell clients this twice, because they'll forget.)

If they take it over, they pay for the CatDoes plan themselves. Keeping their own domain and losing the badge means Starter, which is $50 a month. Tell them that before they choose, so the bill isn't a surprise.

A tip if you want handovers to be painless: build each client's site in its own CatDoes account from day one. Then the handover is just giving them the login, and you're never untangling one client's project from someone else's.

[FILL: how many of the five took the monthly fee and how many took the account.]

Some people will tell you handing over the account is throwing away recurring income. I don't see it that way. If they want me, they pay me. If they don't need me, I'd rather they own something they can actually run than keep paying me out of habit.

The numbers


Businesses contacted	11
Mocks built	11, about 10 minutes each
Paid for the full site	5
Didn't buy	6
Revenue	$5,000
Costs	about $900
Profit	about $4,100
Margin	about 82%

Where the $900 went: [FILL: the real breakdown, e.g. CatDoes plan and credits, domains, anything else. Readers will ask.]

Five out of eleven is a 45% close rate. I'd love to tell you that holds at a hundred calls, but I have no idea. Eleven is a small number. My guess is the mock did most of the work, since people were saying yes to something they'd already seen.

The parts that weren't easy

Building took under two hours all week. The calls are the real job. [FILL: one real moment from the calls, e.g. a hang-up, a skeptical owner, or the first yes.] If you hate the phone, walk in instead, but you still have to talk to people. There's no version of this that skips that part.

More than half of the businesses didn't buy. Six of eleven said no or never got back to me.

And this was one week. The money is real, but one week doesn't tell you much about the next one. Some weeks will be slower, and I'd plan around a slow one.

Check the rules before you start dialing

Cold calling is regulated, and the rules depend on where you are. This isn't legal advice, just what I'd check first.

In the US, most calls from one business to another are exempt from the FTC's Telemarketing Sales Rule. Since 2024, though, the rule's ban on misrepresentations and false or misleading statements also covers business calls (FTC). Dial by hand. Autodialers and prerecorded messages fall under separate rules, especially when the number is a cell phone, which the number on a small business listing may well be.

You need less than you think

The same agent builds mobile apps too, so a business you've already delivered for is a warm lead for whatever comes next. A salon with regulars might want a booking app. Putting an app on the App Store needs the Starter plan and an Apple Developer account, which is $99 a year.

That's what gets me about this whole thing. You need one or two tools and the ability to describe what you want to an AI agent. You don't need to code. The rare skill turned out to be calling a stranger and saying "I made you something."

Quick answers

Do you need to know how to code?
No. Every mock and every finished site here was built by describing what I wanted to an AI agent.

How long does a mock take?
About 10 minutes in CatDoes, sometimes less. Eleven mocks took me under two hours in total.

How much can you charge?
Five sites brought me $5,000, about $1,000 each on average.

Can the client update the site without you?
Yes. Hand over the CatDoes account and they change things by chatting with the agent, then ask it to deploy. To keep a custom domain they need the Starter plan or above.

Is it legal to cold call businesses?
In the US, most business-to-business sales calls are exempt from the FTC's Telemarketing Sales Rule, but misleading claims on those calls are banned. In Germany, sales calls to businesses need at least presumed consent. Check your local rules, or walk in.

If you try this, tell me how your first week goes in the comments. I'd like to know whether my 5 out of 11 was luck.

CatDoes Explained: How It Works, Key Features, and Top Alternatives

Amber — Tue, 29 Sep 2026 14:51:27 +0000

Building a mobile app used to mean learning React Native or Swift, setting up a backend, and then losing a week to App Store certificates. AI app builders have made the first part a lot easier. Most of them still leave the last part to you.

CatDoes is one of the tools trying to handle the whole thing. You describe the app you want, and an AI agent writes the code, sets up the backend, and publishes it to the web, the App Store, or Google Play.

In this guide you'll learn what CatDoes is, how it works, its main features, what it costs, and where it falls short. At the end there are six alternatives worth a look if it doesn't fit what you're building.

What is CatDoes?

CatDoes is an AI agent that builds mobile apps and websites from plain English. You tell it what you want, it asks a few questions, and then it builds it. That includes the screens, the database, user login, and the deploy.

Mobile apps are built with React Native and Expo, so one project runs on both iPhone and Android. Websites are built with Vite or Next.js. These are normal, widely used frameworks, so you're not stuck with a proprietary format if you want to take the code somewhere else later.

The main difference from AI coding tools like Cursor is who does the work. With Cursor you're the developer and the AI helps. With CatDoes the agent is the developer. It runs on its own computer in the cloud, installs packages, runs the app, fixes its own errors, and keeps working even if you close the browser tab.

It's aimed at founders, small businesses, and anyone who has an app idea but doesn't want to learn mobile development to ship it. It also works for developers who just want to skip the setup and the store publishing work.

How CatDoes works

1. Describe what you want

You start with a prompt, something like "a habit tracker where users log in and see their streaks." Before it writes any code, the agent asks follow-up questions to fill in what you left out. If you'd rather see a plan first, you can ask it to lay out what it's going to do without building anything yet.

2. Watch it build and test it on your phone

While the agent works, a live preview updates next to the chat. For mobile apps you can scan a QR code with the Expo Go app and run it on your own phone. It's worth doing early, because the browser preview doesn't always match how the app behaves on a real device.

3. Keep chatting to make changes

From there it's a conversation. Ask for a new screen, a different layout, or a bug fix, and the agent makes the change. Each new chat saves a checkpoint you can roll back to. When a chat gets long, the "Compact and Save" button starts a fresh one, so the agent doesn't lose track and burn extra credits.

4. Publish it

Websites go live at your-project.catdoes.app, and you can connect your own domain. For iOS, you add an Expo token and your Apple Developer credentials, the build takes 10 to 30 minutes, and then the agent submits it to the App Store. You can send it to TestFlight first or release it publicly. Android builds the same way, but you download the finished file and upload it to Google Play yourself.

Either way, the store fees are on you: $99 a year for Apple and a one-time $25 for Google.

Core features of CatDoes

Mobile apps and websites. One agent builds both. Apps are React Native with Expo, and websites are Vite or Next.js.
An agent that works in the cloud. The Compose agent runs on its own cloud computer and can split big jobs across subagents. It keeps going after you close the tab.
Built-in backend. Every plan, including the free one, comes with CatDoes Cloud: a database, user auth, file storage, edge functions, and realtime updates, hosted in the US or the EU.
App Store and Google Play builds. The agent runs the builds for both stores and submits iOS apps to the App Store for you.
App Store review check. Before you submit, it simulates Apple and Google's review and points out what would probably get rejected.
Website to app. Paste the URL of any public website and it rebuilds it as a real mobile app instead of just wrapping the site.
Checkpoints. Every chat saves a restore point, so a bad change is easy to undo.
Custom domains. Connect your own domain to the websites you publish.
GitHub import and code export. You can import an existing repo on the Core plan and up. Two-way GitHub sync and full code export come with Plus.
Watch. A built-in error tracker that catches crashes in your live app. You can then ask the agent to fix what it found.

Benefits of using CatDoes

1. You skip the App Store setup

Signing, certificates, and build tools are where a lot of people give up on their first app. CatDoes runs the builds and handles the iOS submission, so your part is mostly entering your credentials.

2. The backend is already there

Login, a database, and file storage are ready from the first prompt. You don't need a separate Firebase or Supabase account to get an app that actually saves data.

3. One project for iPhone and Android

Because apps are built with React Native, the same code runs on both platforms. You're not maintaining two separate apps.

4. You can test on a real phone right away

Expo Go puts the app on your phone while it's still being built, long before it goes anywhere near a store. You find the weird mobile bugs early.

5. Fewer surprises in app review

The review check catches common rejection reasons before Apple does. That can save you a round of fixing, resubmitting, and waiting again.

6. You're not locked in

The code is standard React Native, Vite, or Next.js. On Plus and up you can sync it to GitHub or export it and keep building somewhere else.

Limitations to know about

CatDoes isn't perfect, and a few things are worth knowing before you sign up.

App Store and Google Play publishing starts on the Starter plan. The free plan and Core only publish to the web, so if the stores are the goal, Starter is your real starting price.

Google Play isn't fully hands-off. iOS submission is handled by the agent, but for Android you download the build and upload it through the Play Console yourself.

You still need your own accounts: an Expo account and an Apple Developer account for iOS, plus a Google Play developer account for Android. The agent helps you set them up, but it's not zero setup.

How much does CatDoes cost?

There's a free plan with 25 one-time credits that publishes to the web only. Paid plans are monthly, and each one comes with a set number of credits. Core is $20 for 50 credits, Starter is $50 for 200, Plus is $100 for 400, Pro is $200 for 900, and Max is $399 for 2,000.

The pricing page shows yearly prices by default, so Starter shows up there as $42 a month. The prices above are the monthly ones as of late September 2026.

CatDoes alternatives

If CatDoes isn't the right fit, here are 13 alternatives worth checking out.

Conclusion

AI can write app code pretty well now. What still separates these tools is everything after the code: the backend, the builds, app review, and getting into the stores.

That's where CatDoes puts its effort. It gives you a backend on every plan, runs your builds, checks for review problems, and submits iOS apps for you. The trade-offs are that store publishing starts at $50 a month, credit costs are hard to predict, and Android still needs a manual upload.

If you only need a website or a dashboard, Lovable or Bolt.new may be simpler and cheaper. If you want a real mobile app in the App Store and Google Play without learning the whole publishing process, CatDoes is a good place to start.

CatDoes vs Rork, what's your fav vibecoding tool?

Amber — Thu, 03 Sep 2026 19:07:51 +0000

I Finally Built My Own App, and I Still Can’t Write a Line of Code

Amber — Thu, 03 Sep 2026 19:07:02 +0000

For about three years I had a note on my phone titled “app ideas.” Most of it was nonsense, the kind of thing I’d type at midnight and not understand by morning. But one idea kept coming back, and it was almost embarrassing in how simple it was. I wanted one clean place to keep the recipes I actually cook, and a small website I could send to friends when they asked me for “that pasta thing” for the hundredth time.

That was the whole dream. A little app and a little site. And for three years I did absolutely nothing about it, because every time I looked into how, the answer was some version of “learn to code,” or “hire someone,” or “spend a weekend fighting with a tool that assumes you already know what a database is.”

I am not a technical person. I want to be honest about that up front, because I think a lot of these stories quietly skip past it. I do not know what a framework is. I have never used a terminal on purpose. When something on my computer breaks, my strategy is to restart it and hope. So this is not a “look how easy it is when you already sort of know your way around” story. This is the other one.
The part where I almost didn’t

I had poked at the no-code tools before. A couple of years ago I spent a long, joyless Sunday inside one of the drag-and-drop builders, the kind where you snap boxes together to make a screen. On paper it should have worked. In practice it felt like homework for a class I never signed up for. There were workflows and data types and little red error icons, and by the afternoon I had built roughly one screen that did nothing, and I closed the laptop and made an actual dinner instead.

So when a friend mentioned CatDoes, my first reaction was a very tired “sure.” I assumed it would be the same thing wearing a different outfit. I only tried it because it was late, I was procrastinating on something else, and typing a sentence felt like less effort than not typing one.

The thing that got me to keep going was that it did not ask me to learn anything first. There was no tutorial to sit through, no concepts to absorb before I was allowed to begin. It just asked me what I wanted to make. So I told it, in plain English, the way I’d describe it to a friend. Something like: an app where I can save recipes with a photo, tag them by how long they take, and mark the ones I’ve actually made.

Then it started building it.
What it was actually like

I want to describe this part carefully, because it’s the bit that genuinely surprised me.

It didn’t hand me a blank canvas and a hundred settings. It went and made a first version, and then I looked at it and told it what was wrong. “The photo should be bigger.” “Can the tags be at the top instead.” “I don’t like this color, make it warmer.” I was talking to it the same way I’d talk to a friend who happened to be very fast at this and never got annoyed with me. And it just did the things.

That changed the whole feeling of it. With the older tool I’d tried, I was the one who had to know how to build the thing. Here I was the one deciding what the thing should be, and the building happened on the other side. Those are very different jobs, and it turns out I am only qualified for one of them.

There were moments where it felt a little magical and I don’t fully trust that word, so let me be specific about one. I asked it to let me search my recipes by ingredient, which in my head was a “maybe next year” feature, the kind of thing I assumed real developers cried over. It added it. I typed “lemon” and my three lemon recipes showed up. I actually said “oh” out loud to an empty room.
The website happened almost by accident

I only set out to make the app. But once the recipes existed, sending people a screenshot felt silly, so I asked whether I could also have a simple website that showed the same recipes. I half expected to be told that was a separate project, or a different tool, or a different price of admission.

It just built the site too. Same recipes, laid out for a browser, with a real link I could send. A little while later I connected my own domain name to it, which is a sentence I could not have written a year ago and would not have understood if you’d read it to me.

That was the moment the whole thing tipped from “neat toy” into “wait, this is real.” I sent the link to my mum. She texted back a photo of the pasta. The loop closed.
What I want to be honest about

This is the part I’d want if I were reading this instead of writing it, so here it is without the shine.

It was not instant, and it was not effortless. It was fast in a way that genuinely startled me, but I still had to know what I wanted, and I still had to look at each version and decide what was off. If you don’t have a clear picture in your head of the thing you’re making, no amount of clever software will invent taste for you. It’s very good at building. It cannot want something on your behalf.

There were a couple of moments where a change I asked for did something to another part I liked, and I had to go back and say “no, keep the old version of that.” So it wasn’t a perfectly straight line. It was more like editing than typing. You write a messy first draft by describing it, then you shape it by reacting to what comes back.

And I’d gently push back on anyone who says tools like this mean nobody needs to think anymore. If anything I thought more, just about the right things. I spent my attention on what the app should feel like to use, instead of on why a button wouldn’t turn blue.
What actually changed

The app is not going to be in any headlines. It’s a recipe app that maybe forty people will ever open, and most of them are related to me. By any serious measure it is small.

But I made it. The thing that lived in a note for three years exists now, and I use it every week, and there’s a website with my name on the domain. For someone who had quietly filed “building your own app” under “things other people get to do,” that shift is hard to overstate.

I think that’s the part worth sitting with. It isn’t really about recipes, or about any one tool. It’s that the distance between “I wish this existed” and “this exists” got dramatically shorter for people like me, the ones who were never going to learn to code and had made a kind of peace with that. The peace turned out to be unnecessary.

So if you have your own version of that note on your phone, the one full of ideas you assume aren’t for you to build, I’d just say try describing one of them out loud to something that can build it. Worst case you lose an evening, which I was going to lose to scrolling anyway. Best case you say “oh” out loud to an empty room, and then you text your mum a link.

If you end up making something, I genuinely want to see it. Reply and show me.