Why I wrote this vibe coding case study
This was my first iOS app built with AI, and my first time running the full loop of turning an idea into a real product and shipping it to the App Store. I wanted to document not just what I made, but how each step actually worked, what tools I used, and where I got stuck, especially around App Store review.
For me, the real meaning of this app was not the app alone. It was proving that I could complete the loop: vibe coding → design → development → review → launch. I gave myself one hard constraint: finish it in one week.
That time limit forced the product to stay simple. If I had chosen something with accounts, databases, and server sync, the foundation alone would have eaten the week. For a first AI-built app, I strongly recommend choosing something simple and personally interesting: a local-first tool with no backend is the easiest shape for completing the full loop.
The tools I used to build an iOS app with AI
For this vibe coding workflow, these were the main tools:
- Antigravity: the AI agent I used for most of the coding and logic work. It was the core tool in the flow.
- Google Stitch: an AI prototyping tool I used to generate the first interface direction.
- Google AI Studio: where I previewed the interaction and motion after exporting from Stitch.
- Xcode: Apple's official iOS development environment. If you want to ship to the App Store, you cannot avoid it.
- Apple Developer account: the hard requirement for App Store distribution, currently $99/year.
- Visual references: images that helped the AI understand the record-player mood I wanted.
The full process: from AI prototype to App Store launch
Generate the first product prototype in Google Stitch
I started with Google Stitch to turn the rough idea in my head into something visible. This step changes “I want to make this kind of app” from a vague feeling into an interface you can actually inspect. For beginners, the useful part is that you do not need to draw screens or know Figma.
Export to Google AI Studio to preview motion
The Stitch output is mostly static. After exporting it to Google AI Studio, I could preview the interactive version. This mattered because motion tells you whether the product feels right in a way static screens cannot.
Download the project to my computer
Once the prototype felt close enough, I downloaded the full project. This was the key move from “online prototype” to “local development project.”
Open the project in Antigravity
I opened the downloaded folder in Antigravity. From here, the AI agent became the main coding partner, and the project officially entered the vibe coding phase.
Ask AI to turn the project into a shippable iOS app
This was the biggest jump in the whole process. I asked Antigravity directly: “How do I turn this product into a complete app that can be submitted to the App Store?”
Then I followed the steps. It guided me through installing Xcode, setting up an Apple Developer account, changing the project structure, and generating the native iOS project files.
From there, the workflow became a two-track setup: Antigravity handled code and logic; Xcode handled device preview, packaging, signing, and upload. Xcode is what turns the project into an iOS-ready build, signs it, and uploads it through App Store Connect.
This is why building an iOS app with AI still requires Xcode: it is the final mile of shipping.
Use AI to polish the interface with references and exact numbers
I wanted the whole app to feel like a record player. This was the most interesting part of vibe coding: AI + natural language + visual references. I gave the AI record-player images, explained the mood, then iterated in plain language.
For detailed changes, the prompts that worked best were specific: “move this 20px left,” “move it 10px down,” “make this button 1.2x larger,” or “reduce this font size slightly.”
The UI lesson from vibe coding: do not just say “this feels wrong.” Give a direction and a number. That makes each iteration visible.
Submit to App Store review, and get rejected once
Once the code and interface were ready, I uploaded the build through Xcode to App Store Connect and submitted it for review. The first submission failed.
Apple's reason was simple: I had a “login” button in the interface, but the app did not actually have a login feature. It runs locally and stores everything on the user's device. I had added the login entry as a “maybe later” placeholder.
But the App Store Review Guidelines are clear: if a feature is not available now, it should not appear in the app. This is a common beginner mistake.
The fix was simple: remove the login entry, resubmit, and the second review passed.
A few honest reflections after shipping with AI
Looking back, I learned much more than “how to make an app.” The fastest way to learn AI tools is to use them on a real project. Tutorials and prompt collections help, but nothing teaches like making something that actually has to work. I wanted this to become a real usable product, not just another prototype.
That is why the $99 Apple Developer account felt worth it to me. The feeling of seeing your product live in the App Store is completely different from stopping at a demo: friends can download it, people can use it, and that momentum makes you want to build the next thing.
One more honest note: I spent a long time watching AI from the sidelines before actually building with it.
What helped me move was not one viral tutorial. It was a community: the AI builders community. I kept seeing people with similar starting points actually build, launch, and improve their products with AI. Slowly, my thought changed from “this seems hard” to “maybe I can do this too.”
Learning AI is easier with people beside you. Alone, it is easy to get stuck and stop. Around people who keep building and sharing, you borrow their rhythm and keep moving.
Join the AI builders community
If this case study made you think “I want to build my own app too,” you are welcome to join us, whether you are starting from zero or already using AI but stuck somewhere in the process. In the community, you can see real people ship products with AI, solve practical problems, and keep each other going.
thanks for reading this far ✿
Learn AI together, and build products that can actually launch.
Join the AI builders community →