I’ve been a developer for nearly 20 years. I built and launched my latest app using 100% "vibe coding" with AI agents. Here is what I learned.
Hey everyone,
I have nearly 20 years of software development experience, mostly focusing on js for a long time. Over the years, whenever freelance projects popped up, I’ve also built mobile apps with Swift and written backends in php etc. I’d like to think I know what it actually takes to ship a finished product, it’s not just about writing code; it’s understanding business requirements, adhering to clean code standards, using tools like SonarQube, and monitoring or debugging live production environments.
However, since AI emerged, many of these traditional steps have started to feel redundant. I personally believe programming languages as we know them will eventually become obsolete and forgotten. AI will likely move toward something much closer to machine language.
For a while now, I wanted to build a fully working product at maximum speedpurely through "vibe coding" using AI agents, with zero self-imposed limitations or strict rules. The exact nature of the app didn't matter much to me; my primary goal was to experience end-to-end how efficient vibe coding actually is. To get to end-users as fast as possible, I targeted Google Play and the App Store using Flutter, aiming to have a fully working app live in 2 weeks.
I let AI agents handle the market research, competitor analysis, and product ideation. The AI suggested a home inspection app: users create a property, add areas, scan rooms using AI to generate a checklist, and export it as a PDF. The AI defined the scope, design, architecture, monetization, and literally wrote 100% of the code. My role was strictly human Qa testing flows and checking for security or financial vulnerabilities. True to the goal, the app went live on the App Store in just 2 weeks, and the Google Play release is right around the corner.
App Store link: https://apps.apple.com/ng/app/habiproof/id6797038152
The tools I used: ChatGPT (Sol 5.6), Claude Opus for planning and advisory, and Sonnet for code generation. I also used Gemini like a scratchpad for quick research, essentially replacing standard google searches.
I’m sure there are plenty of bugs to fix and areas to polish, but I intentionally ignored them to stick to the MVP scope. If needed, I might release updates later based on user feedback.
What really got me thinking is this: A few years ago, as a developer, all I needed was a laptop to build something from scratch. But after AI, I feel less like a pure builder and more like a consumer—or at least a creator with heavy dependencies.
I'll probably launch another app next week, maybe a variation of this one. The web has felt like a digital junkyard for a while now, and app stores are heading in the same direction. Sure, there are treasures hidden in junkyards, but I feel we’re slowly reaching the end of the traditional WWW and smartphone era as we know it.
What are your thoughts on building apps this way? How do you maintain long-term motivation in this new setup?