In about two months, in late 2026, we built eight apps of our own: KADRE, La Grada, Underbuy, No Takeout, Readybell, Dawnsy, Clubkey and Stikki. One person, Dani Gil, with AI doing most of the code. Some are heading to the stores this month; others are still being finished.
This is what that taught us, for anyone thinking about building an app now.
What AI does, and what it doesn’t
AI is fast at the parts that used to fill the calendar:
- Writing screens, forms, database tables and the code that connects them
- Writing tests and fixing what they catch
- Translating the interface (our apps ship in English, and some in Spanish)
- Drafting store descriptions, privacy text and release notes
It doesn’t replace the decisions:
- What the app is for. One job, for one kind of person. An app with two jobs is slower to build and harder to explain.
- What to leave out. AI will happily build everything you ask for. Cutting is still a human skill.
- Taste. Whether a screen feels right, whether the copy sounds like a person, whether the empty state helps.
- Rules. Store guidelines, privacy law and, for La Grada, children’s data. Those need someone who reads the rules, not a guess.
The stack we use
Every app starts from the same template, so day one is never a blank page:
| Part | What we use | Why |
|---|---|---|
| App | Expo (React Native) | One codebase for iOS and Android |
| Website | Next.js | Landing page, web version, legal pages |
| Backend | Supabase in the EU | Sign-in, database, storage, row-level security |
| Payments | RevenueCat | Subscriptions on both stores from one place |
| Analytics and errors | PostHog (EU) and Sentry | See what people use and what breaks |
| Builds | EAS, run locally | No cloud build queue and no build fees |
| Code | Claude Code | Writes and changes the code with us |
The template already has sign-in (try first, save the account later with an email code or Apple), in-app account deletion, payments, push notifications, analytics that wait for consent, English and Spanish, legal pages and the store build setup, all tested. A new app adds what makes it different, not the plumbing.
A realistic timeline
For a focused app with sign-in, a database and 10 to 20 screens, this is what our builds looked like:
- Days 1 to 2: the idea on paper. Who it’s for, the one job, the five screens that matter. We design the website’s phone mockups first, then build the app to match them.
- Days 3 to 10: build. Screens, data, sign-in, the core loop. Daily test builds on a real phone.
- Days 10 to 15: polish. Empty states, errors, onboarding, real copy, icons, store screenshots.
- Days 15 to 20: store prep. Privacy labels, demo account, review notes, submission.
- Then: review. Apple usually replies in days. Google Play can take longer for a new account (see below).
Two to four weeks to a first submission is realistic for one focused app. It is not realistic for an app that tries to do five things.
Store rules that block launches
Most rejections are predictable. Check these before you submit:
- Account deletion inside the app. If users can create an account, Apple requires a way to start deleting it from within the app, not only by email.
- Minimum functionality. Apple rejects apps that are a website in a wrapper or don’t do enough to be useful as an app.
- Too similar to your other apps. Apple’s spam rule targets many near-identical apps from one developer. If you build several apps from one template, make each clearly its own product.
- Sign-in options. If you offer Google or another third-party sign-in, Apple asks for an equally private option too, such as Sign in with Apple.
- A demo account that works. Reviewers need to get past sign-in. Give them an account with sample data and a two-line note on what to try.
- Google Play closed testing. New personal developer accounts must run a closed test with at least 12 testers for 14 days in a row before they can publish. Start it early.
What it costs to run
Building is only part of the bill:
- Store fees: Apple charges $99 a year for the developer program. Google charges a one-time $25.
- Commission: under Apple’s Small Business Program, developers earning up to $1 million a year pay 15% on paid apps and subscriptions instead of 30%.
- Backend: Supabase and similar services have free tiers that cover a launch. Paid plans start when you have real users.
- Your time after launch: reviews, support, updates for new OS versions. Plan a few hours a week.
Where to start
Write the app in five lines: who it’s for, the one job, how they find it, why they come back, and how it makes money. If you can’t fill all five, the app isn’t ready to build yet, with or without AI.
If you can, send us the five lines. Apps start from $15,000, or a fee plus equity for the right idea.
Questions
Can AI really build a whole app?
AI writes most of the code, tests and store text. It doesn't decide what the app is for, who it serves or what to leave out, and it can't judge whether a screen feels right. One experienced person making those calls, with AI doing the typing, is what makes weeks possible.
How long does it take to build an app with AI?
For a focused app with sign-in, a database and 10 to 20 screens, our builds took two to four weeks to a first store submission. Store review and fixes add days to weeks. A new Google Play personal account adds at least 14 days of closed testing before you can publish.
Do I need separate code for iOS and Android?
No. We build with Expo (React Native), so one codebase ships to both stores. The website and the app's server run on Next.js next to it, in the same repository.
What gets apps rejected most often?
Missing in-app account deletion, apps that feel like a website in a wrapper, broken sign-in for the reviewer, and apps too similar to others from the same developer. Give the reviewer a working demo account and a short note on what to test.
Who owns the code?
On client projects, you do: the repository, the store accounts and the backend are yours. On a fee plus equity deal, ownership is set out in the contract.