tools

Figma

ChatGPT

Cursor

Figma

ChatGPT

Cursor

TIMELINE

August 2025 - October 2025

Type

IOS Native Mobile App

responsıbılıtıes

Shape Pommie from idea to launch using AI tools for research, design drafts, and faster coding. Build the core features, set up basic tests, and make sure data is handled safely. Prepare the App Store page (keywords, screenshots, notes), run a small beta, and ship. After release, watch installs, ratings, and crashes, talk to users, and keep improving with quick updates.

GOAL

Build Pommie end-to-end—from idea to App Store release—by leveraging AI tools for market research, UX flows, code generation, testing, and content. Ship a production-ready v1 with CI/CD, analytics, crash reporting, and a privacy-first backend, plus an ASO-optimized listing

overview

Trying to Build an AI-Powered App Without Any Swift Experience or Knowledge

Mascot Design

The Pommie character—cute and playful—takes its shape from the classic Pomodoro symbol, softening the discipline of timeboxing with a friendly presence that motivates rather than nags, and I’ve made it the star of the app: it works with you, sleeps with you, and keeps you company through every focus rhythm.

Design Decisions

I kept the design minimal, intuitive, and relentlessly functional—only what you need, exactly where you expect it. A bright orange accent injects energy and joy into otherwise “boring” focus moments, guiding attention without overwhelming the interface. Micro-animations and clear hierarchy reduce cognitive load, so starting, pausing, or reviewing a session is effortless.

Different AI Buddies Along the Way

I started with ChatGPT and shipped fast at the beginning; even though it worked well at first, because it couldn’t fully read and reason over the whole codebase, it began producing errors and unrelated code blocks. To streamline development, I switched to Cursor, which made iterative edits and code awareness much easier.

How can I keep my cool while
arguing with AI agents… especially when they admit I was right?

learnıngs

After countless messages and iterations with ChatGPT and Cursor, I published the app in one month.

Developers are right while saying it works on my machine.

Works on my machine” is real—and I felt it. Simulators and a single iPhone behaved perfectly, then another device exposed permissions, background limits, and tiny layout shifts. Lesson learned: test on multiple devices early, automate a few checks, and treat “it works here” as the start of QA, not the finish line.

Everything should be really consistent when it comes to coding.

Consistency is kindness—to your future self and to anyone who touches the code. Stable naming, predictable file structure, and repeatable patterns (components, state, spacing, tokens) turn chaos into momentum. Once I standardized these in Pommie, bugs dropped and shipping got faster.

Always duplicate your files, as they might be deleted - even 5 times during the journey.

Backups are not optional—they’re survival. I lost files multiple times (yes, five!) due to sync hiccups and refactors, and each duplicate saved hours. Use Git from day one, push to a remote, tag milestones, and keep a zipped export or Time Machine snapshot for extra peace of mind.

It does not mean that it will work on your second build.

Just because it worked once doesn’t mean the second build will behave. Caches, entitlements, provisioning profiles, and dependency versions can quietly shift and break things. I now use a “second-build checklist”: clean the build folder, lock dependency versions, bump the build number, test on a fresh device/simulator, and archive via CI before I celebrate.

Apple has really challenging publishing process.

Apple’s publishing process is no joke. Between App Store Connect forms, screenshots, privacy details, background modes, and guideline compliance, each submission is a mini project. What helped me: a pre-flight checklist (metadata, permissions text, test notes), a reproducible build via TestFlight, and treating rejections as crisp to-do items—not setbacks.