Goal: Build an app for visualize and navigate your Swift & Xcode codebase.
Role: iOS Engineer, UX/UI Designer
Software: Swift, SwiftUI, Apple Foundation Model (LLM)
Outcome:
1. Leverage Apple's Foundational Model framework.
2. Create a privacy-first fitness tool that processes everything on-device
3. Explore real-time biomechanical analysis through camera sensors.
Duration: January 2026 - current
AI Softwares: Codex, Anti-gravity, Claude Code. For disclaimer, I used AI mainly to help me architect the overall application.
For transparency, I primarily used AI to help architect the overall structure of the application. I did not build the product by blindly prompting step by step. With Swift’s evolving syntax and the Observable framework updates, that approach would quickly lead to inconsistencies and technical debt.
My workflow was structured and intentional:
1. I first designed and mapped out the architecture myself to fully understand how the system should function.
2. I then used LLMs to review my architecture and implementation decisions.
3. If an alternative approach was suggested, I asked the model to explain why it would be better so I could evaluate and understand the tradeoffs.
This ensured that I maintained a deep understanding of how the code works rather than outsourcing the thinking process. Building a complex system requires a strong grasp of the core foundations. While LLMs can generate functional code, it’s very easy to lose clarity as the builder if you rely on them passively.
For transparency, I primarily used AI to help architect the overall structure of the application. I did not build the product by blindly prompting step by step. With Swift’s evolving syntax and the Observable framework updates, that approach would quickly lead to inconsistencies and technical debt.
My workflow was structured and intentional:
1. I first designed and mapped out the architecture myself to fully understand how the system should function.
2. I then used LLMs to review my architecture and implementation decisions.
3. If an alternative approach was suggested, I asked the model to explain why it would be better so I could evaluate and understand the tradeoffs.
This ensured that I maintained a deep understanding of how the code works rather than outsourcing the thinking process. Building a complex system requires a strong grasp of the core foundations. While LLMs can generate functional code, it’s very easy to lose clarity as the builder if you rely on them passively.