About eight months ago, my company's director told us to move to an AI-first development process. We were expected to build projects with Claude Code, with the AI also handling security checks and code reviews. Since then, I've created web projects along with iOS and Android apps almost entirely through AI assistance. I'm an experienced web developer, but I don't have much understanding of mobile development or Kotlin, and I often don't fully understand the code the AI produced. When changes broke things, I usually relied on the AI to repair them, sometimes creating more problems in the process.
So although I can technically say I've built several mobile apps, I'm not confident I could explain their architecture, debug them independently, or pass a conventional mobile development interview. If companies increasingly expect developers to work this way, how should someone in my position rebuild the fundamentals and prepare for technical interviews?
5 Answers
The best approach is probably not to reject AI completely, but to change how you use it. Pick one app and audit it feature by feature: trace the data flow, read the platform documentation, write tests, introduce a controlled bug, and fix it yourself before asking for hints. Ask the tool for explanations or review questions rather than complete implementations. That turns the existing projects into a curriculum instead of leaving them as opaque demos.
For a conventional technical interview, you need to study and practice the material the same way developers always have. Work through the language basics, data structures, debugging, testing, and the framework or platform involved. Spend some sessions coding with only a text editor and documentation, then practice explaining code and solving problems without AI. Building apps, whether manually or with assistance, doesn’t automatically prepare you for interview questions.
You probably learned more than you realize about project structure, workflows, and assembling features, but that isn’t the same as understanding the underlying platform or language. Start working through the mobile fundamentals directly instead of treating the AI as an oracle. Read each generated change, ask it to explain the reasoning, and then reproduce small pieces yourself. If you can’t explain or debug the code without assistance, you don’t really own that part of the application yet.
A workplace that insists on AI-generated code for everything can easily damage long-term skill development. Fast prototypes are one thing, but production software still needs judgment, maintenance, and accountability. Keep using the tools if your job requires them, while deliberately reserving time to hand-code and study the stack you want to be hired for. Otherwise, you may become efficient at producing code you can’t independently evaluate.
AI tools are useful when they fill gaps for someone who already understands the engineering decisions. They are much less useful when they replace the learning process entirely. Employers may appreciate AI workflow experience, but they still need people who can review generated code, recognize security and performance problems, understand system design, and maintain the result months later. A portfolio of apps helps only if you can clearly explain how they work and what trade-offs you made.
That’s the gap I’m worried about. I can follow the broad behavior of the apps, but I still need to build enough hands-on understanding that I can make changes without blindly asking the AI to fix its own mistakes.