Building an AI App Solo: What I Actually Learned

I'm not a developer. I'm a technical consultant — strong business logic, solid understanding of architectures, and zero code mastery in the strict sense. 18 months ago, building a mobile app with a cloud backend felt completely out of reach. Today, Mental Loadless runs on Azure, React Native, and Expo. Here's what I learned along the way.

AI as co-developer: what actually changes

You hear a lot of "AI will replace developers." That misses the point. What AI really changed for me is the barrier to entry. Code I could never have written alone, I can now iterate, debug, and understand with an AI assistant in real time.

But there's an important nuance: AI doesn't replace business logic. It doesn't know what your user actually wants. It doesn't understand your industry's regulatory constraints. It doesn't know your overall architecture better than you do. What it does very well is translate your specifications into code — as long as your specifications are precise.

The skill that helped me most: knowing how to break complex problems into clear logical steps. That's exactly what a good consultant does. And it's exactly what AI needs to be effective.

What takes 10x longer than you think

The code is the fastest part. What takes time is everything else:

• Cloud infrastructure: Azure Functions, Communication Services, DNS, CORS, App Registration… each service has its own quirks. I spent more time on configuration than on business logic.
• The App Store: Apple's requirements change, review timelines are unpredictable, rejections can feel arbitrary. Plan for 3x more time than you think.
• Content: copy, SEO, support pages, legal notices, privacy policies. Invisible but mandatory.
• Distribution: having a product is useless if nobody finds it. GEO, press, Product Hunt — that's a full-time job on its own.

The moment it almost stopped

About halfway through the project, I realized my initial architecture wouldn't scale. The backend I'd built worked locally but would cause serious problems in production. I had two options: rebuild from scratch, or patch.

I took a week to refactor the core. It was the longest and hardest decision of the project — and the best I made. Technical shortcuts always come due, and they always come due at the worst moment.

What I'd do exactly the same

Validate before coding. Before writing a line, I spent time understanding the real problem. Mental load isn't a lack of organization — it's an asymmetric coordination problem. That clarity guided every product decision I made.

Choose mainstream stack. React Native, Next.js, Azure. Well-documented technologies with large communities and excellent AI support. Not the time to be original on the stack when you're building solo.

Iterate fast and show early. I showed prototypes to real families within the first weeks. The feedback fundamentally changed the product — and that's a good thing.

My advice if you want to do the same

If you're a consultant, project manager, or product manager — if you have the logic but not the code — this is the best moment in history to build a tech product. AI tools are powerful, infrastructure costs are low, and the domain knowledge you've built up is valuable.

Start small, document everything, and don't try to be a developer. Be what you are: someone who understands problems better than most.