Finally, Your Mobile App
in a Weekend

A week ago I wrote about opening your vault to agents: put the value you provide in one place, and give every visitor a door that fits them – the website, the Captain on the page, MCP for Claude and ChatGPT, API keys and a CLI for agents running on a server.

I didn’t know it yet, but that was the groundwork for this weekend. A phone app is just another visitor who can’t walk up to the counter. It lives outside the building. It needs a door.

Pirate Skills is built on Next.js, where the front end and the back end sit very close together. A page on the website can simply ask the business logic for what it needs. A phone can’t – it has to call in from outside. So I used the weekend to re-architect: we exposed the functions the app needs through a protected API, and the website and the app now ask the same core. Same rules, same data. A chat you start on your laptop is there on your phone.

The real work for me wasn’t code. It was deciding which parts of pirateskills.com belong in an app, and what its cornerstone is. I started with the no-brainers.

Listing the upcoming events and registering for them was the very first version. Then came the chat agent, the CaptAIn – the companion for all the knowledge we share at Pirate Skills. We always planned to make the Captain the heart of the website and never quite did. On a phone it finally fits as the centerpiece.

It’s like having Ben and the knowledge of the Pirate Skills community in your pocket.

The biggest decision came before any code. I had a big discussion and watched a lot of YouTube videos, deciding mostly between two routes.

At Maple Apps we went fully native, with our own backend. That was also how the three developers split the work: one on the backend, one on iPhone, one on Android. And AI really makes it possible these days to go fully native. If I were building for the iPhone only, I would have considered it.

But Pirate Skills is built on Next.js. Our code already lives in one repository that holds all our apps – developers call it a monorepo. So we could simply add the mobile app to it: the iPhone app, the Android app, and a web version of the app on top. They share one language, TypeScript, one design system, and they all talk to the same API. That had many advantages I wanted to keep, so I went with React Native plus Expo.

One thing to know if you’re weighing the same choice: React Native is not a website in a frame. It draws real native screens. You just write them once. Expo is the toolkit around it that builds, signs and updates the app for both stores.

I had my kids over that weekend, so the time was short. I mostly had Friday afternoon to Saturday afternoon, before they came. Then I spent the time with them, and after I brought them to bed I sat down again. That was it. I didn’t want more, because I wanted to focus on my kids.

So I didn’t sit at the keyboard building. One Claude Code session acted as the control tower: it built nothing itself, kept a live board of the whole project, handed work to other sessions – one for the screens, one for building and releasing the app, one for the API – and brought me only the decisions that were really mine.

The control tower had one rule, and it earned it: never pass on another session’s report without checking it. It once found 24 tasks on our Linear board marked done within the same second – before the work behind them existed. Agents are fast. They also report things as finished that aren’t.

The part I enjoyed most surprised me. I thought screen designs were a thing of the past – the old way, where a product manager writes requirements, a designer draws them in Figma and a developer builds them. I thought it was way better to go straight to implementation. Then I saw how easy and fluent it is to design the screens first with Claude Design, inside Claude Code. I was absolutely convinced this is the future for me. And I used the chance to give the whole Pirate Skills brand more of a product look, and less of a typical website.

Then the part I used to dread: the Apple certificate, signing and provisioning stuff. I’ll use a politer word than I did on the weekend. All of it was handled by agents. Our company account at Apple is still in review, so they found the route that works before Apple says yes: a build signed with my personal Apple ID. It only lasts seven days, and push notifications had to come out of it, because those need the paid account. What I had to do myself fits in one line: plug my iPhone into the Mac, tap Trust, and add my Apple ID in Xcode.

Not everything went smoothly. One night the whole app opened to a black screen – no crash, no error. And every iPhone build failed for a reason nobody would guess: the folder the project lived in had a space in its name. The fix was moving the whole project. After that, the build went through on the first try.

Seeing it in a browser or a simulator is one thing. An installed app on your own phone is something else. The moment it was truly installed on my iPhone, I was in the car with my kids. Christina was driving, we were on the way to a birthday party, and I finally had it. Not a website with an icon. Real React Native, with our icon on my home screen.

All handled by these agents through Claude Code in two days. That was absolutely mind-blowing to me.

For my kids it meant something too. I’m teaching them to vibe code apps, and I wanted to take them along on my own first one. And the phone found what the simulator hid: within an hour of the chat screens shipping, I found two bugs on my own iPhone that no simulator run had shown.

To be honest about where we are: the app runs on my phone, but you can’t download it yet. The app stores don’t review anything on a weekend, and a few steps are still ahead.

I’ll keep you posted as each step lands.

In the age of AI you can literally build anything. So the biggest question isn’t whether you can build an app anymore. It’s should you spend the token budget on it? If the answer is yes, this is the order I’d follow.

Hand this to your coding agent. It won’t build anything until you’ve talked it through together:

Want to hold ours? Members test the Pirate Skills app at Vibe Coding Cologne on 7 October, and it’s meant to be public in both stores by the one on 4 November at the latest. If you’d rather build your own with a crew around you, that’s what the Pirate Forge is for; the next Build cohort starts on 30 September. And if React Native still sounds like another language, start with the AI Compass on 23 September.