10-Year iOS Dev Builds First App Without Coding
An iOS developer with 10 years of experience built their first personal app, AffirmCount, during parental leave entirely without writing code.

Stock photo for illustration only, not from the actual event
- 10-year iOS veteran builds first personal app with zero manual code
- Chooses to verify output results instead of reviewing generated code
- Solves Korean particle complexity using similarity scoring thresholds
Building a software application typically demands intensive programming skills, but one iOS developer with a decade of industry experience took a radically different route while on parental leave. They successfully created their very first personal application without ever typing a single line of code themselves, embracing a fully AI-driven development approach.
The developer acknowledged that after spending ten years working primarily with WebView, they lacked the ability to determine whether generated SwiftUI code was correct or flawed. Consequently, they established a strict rule from the very beginning: they would never read a single line of code because they knew they could not verify AI-generated syntax.
Instead of reviewing code, the creator decided to verify the execution results instead. For instance, in their AffirmCount app dealing with the Korean language, grammatical particles like 나 and 내 change based on preceding words, causing strict exact-matching checks to fail even when user inputs and saved affirmations matched in meaning.

Stock photo for illustration only, not from the actual event
Initially, the developer instructed the AI to hardcode Korean particle handling, but edge cases quickly exploded in volume. That prompted a strategic shift: any similarity score above 85 to 90 percent would be accepted, moving away from exact matching toward a flexible scoring mechanism that required zero manual code.
This approach of orchestrating software without writing manual code highlights an evolving shift in engineering roles. Developers are transitioning from writing low-level syntax to acting as high-level orchestrators who define prompts, test system boundaries, and evaluate final outcomes. This methodology suggests future software creation may rely heavily on managing collaborating AI agents rather than manual implementation.
Ultimately, checking the oldest logs on their MacBook revealed not human-typed instructions, but a conversation log exchanged between autonomous AI agents. The very first prompt the developer actually typed into the repository was simply assigning the AI as the orchestrator, with future roadmap plans involving bringing on-device AI integration to the AffirmCount app available on the App Store and Google Play.
Source: Dev.to
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment