I Built My Own App With AI, Here Is What I Learned

Stop overthinking and start building. Here's how I turned a simple idea into a working quiz app in 30 minutes, and what it taught me about the new era of AI experimentation.

Honestly, I've been kicking around AI content for months now, reading about tools, bookmarking articles. Then I had one of those moments.

You know, when you're scrolling through your feed and see yet another person talking about how "anyone can build apps now" and you think, "Sure, but can they really?" 🤨

So I decided to test it myself at a deeper level. Not with some grand, world-changing app idea, but with something simple and practical: a product management skills quiz.

The results surprised me! 🤯 Not because I built something revolutionary, but because of how ridiculously easy the whole process was once I stopped overthinking it.

---

👷‍♀️ Big Shoutout to Builders 👷‍♂️

Karo (Product with Attitude)

·

Karen Spinner

·

Orel Zilberman

·

Mike Watson

These folks here have inspired me enough to show my silly creations in public starting with this post. Thanks for being such an inspiration here at

Substack

!

---

My Experiment 🧪

- What is it? An application that measures your seniority in product management based on 13 questions.

- What is the Toolbox?

-- Perplexity: I utilized Perplexity’s Deep Research to look out for the right questions to do depending on the product management level based on the Cracking PM Career book.

Perplexity was also responsible for generating the scoring and how to weight each question from aspiring PM to VP.

When the results were right for me, I told Perplexity to generate the prompt to just copy and paste into Lovable.

-- Lovable.ai: Took care of all the frontend creation with React and Tailwind CSS for basic but pretty decent mockup, and made sure to store the values in a simple TypeScript.


*The quiz is designed as a standalone assessment tool that doesn't persist user data anywhere - it's purely for immediate feedback and results display.

The 30-Minute Experiment That Changed My Mind

Here's what I did, step by step, because I know you're probably in the same boat I was:

Step 1: I defined the problem in plain English

Instead of diving straight into tools, I spent 15 minutes writing down exactly what I wanted to build:

A quiz that would help product managers assess their skills across different areas: strategy, execution, leadership, and technical knowledge.

Simple enough for the experiment, and useful.

Step 2: I chose Lovable (and stopped second-guessing myself)

After reading and writing about different AI builders, I picked Lovable for one simple reason: it promised to turn plain English into working apps. Perfect for someone who didn't want to get bogged down in technical details.

Not that I don't like to waste hours coding, but not for the sick of this experiment! 😆

If you want to see all the tools available, take a look here.

How To Jump Into The Rapidly AI Prototyping Era

Step 3: I had a conversation with AI (Perplexity)

This is where it got interesting. Instead of trying to write perfect specifications, I literally described my quiz idea like I was talking to a colleague:

"I want to create a product management quiz that tests people on strategy, execution, leadership, people management and product skills (based on Cracking the PM Career book). Each question should have multiple choice answers, and at the end, users should get a score breakdown by category."

Step 4: I watched it build in real-time (Lovable)

What happened next felt like magic. Lovable started generating the app structure, creating the database schema, building the user interface, and implementing the scoring logic. All while I watched in real-time.

You can try it here: https://product-expertise.lovable.app/

💡 Within 30 minutes, I had a working quiz app. Not a prototype. Not a mockup. A functional application that people could actually use.

The Not-So-Smooth Parts

Let me tell you what the articles don't always mention: AI tools have a mind of their own sometimes. 🧠

When I asked Lovable to create my quiz app, it initially built something that looked great but didn't quite work the way I expected. The scoring system was calculating percentages differently than I intended, and the results page was showing categories I hadn't asked for.

At first, I was frustrated. This was supposed to be the "easy" part, right? Perplexity already did the math for you. 😒

But here's what I learned: the key is in how you refine your conversation in Lovable, and most of this tools I guess.

Instead of getting annoyed when Lovable modified something I didn't expect, I started treating it like a collaborative design session. When the scoring didn't work right, I said:

"Actually, I want each category to be weighted equally, and the final score should show both individual category scores and an overall percentage."

Magic happened in the back-and-forth. Each time I clarified what I wanted, Lovable got closer to my vision. But I'm glad I got off to a good start with Perplexity's Deep Research, otherwise instead of 30 minutes it would have been like 2 hours or more. 😅

This taught me something important: these AI tools work best when you're prepared to have multiple rounds of conversation. Don't expect perfection on the first try. Expect to refine, clarify, and iterate.

Will this change with the new inclusion of GPT-5?

I Learned More By Having a Conversation

The experience taught me three important lessons that I wish someone had told me earlier:

➡︎ AI tools work best when you stop trying to be perfect

My first instinct was to create detailed wireframes, write comprehensive specifications, and plan everything out. But that's not how these tools work. They thrive on conversation, iteration, and refinement. The best approach is to start with a rough idea and let the AI help you shape it.

➡︎ The real power is in rapid experimentation

Building my quiz app wasn't about creating the perfect product. It was about testing an idea quickly and cheaply. In the time it would have taken me to write a project brief for a developer, I had a working prototype that people could actually test.

➡︎ You learn more by doing than by planning

All those articles I bookmarked about AI development? They were helpful, but they couldn't replace the experience of actually building something. The moment I started experimenting with Lovable, I understood its capabilities and limitations in a way that no blog post could have taught me.

The Simple Framework That Worked For Me

Based on my experience, here's the approach that works for AI experiments:

1. Start with a problem, not a solution 💯

   Don't begin with "I want to build an app using AI." Instead, start with "I want to solve this specific problem." In my case, it was helping product managers understand their skill gaps.

2. Choose your tool and commit 🧰

   Pick one AI builder (Lovable, Bolt, v0, whatever) and stick with it for your first experiment. Don't spend weeks comparing features. They're all good enough to get you started.

3. Have a conversation, not a specification 💬

   Talk to the AI like you're explaining your idea to a friend. Be conversational. Ask questions. Iterate based on what you see. The tools are designed for natural language, so use it.

4. Ship fast and learn faster 🍔

   Get your experiment in front of real users as soon as possible. By leaving my quiz application exposed in this post, I know I will collect useful interaction from users.

What This Means for Product People

Here's the thing that really struck me: we're in an era where the cost of experimentation has dropped significantly.

Think about what this means for how we work. Instead of writing lengthy requirements documents and waiting for development cycles, we can build and test ideas in real-time. We can create working prototypes to validate concepts before committing resources.

Actually, when I’ve got the app working like I wanted, I asked Lovable to analyze the project and generate the requirements and technical specs for someone to build it:

And it actually foresaw things that may affect in the real development:

Current Limitations & Technical Debt

Data Persistence

- Limitation: Quiz answers stored only in browser memory
- Impact: Data lost on page refresh or browser close
- Solution: Would require backend integration for persistence

Backend Dependencies

- Current State: Purely client-side application
- Newsletter Integration: Redirects to external Substack page
- Analytics: No user behavior tracking implemented

Scalability Considerations

- Quiz Content: Hardcoded questions (no CMS integration)
- User Management: No user accounts or progress tracking
- Analytics: No data collection for improvement insights

For product managers, this is transformational. We can now test our assumptions with actual applications, not just wireframes or presentations. We can validate ideas with users using functional prototypes, not theoretical concepts.

But here's what I think is even more important: this democratizes innovation.

You don't need a computer science degree or years of coding experience to build and test your ideas anymore. If you can describe what you want to build, you can probably build it.

Final Thoughts

The Real Secret is to Get Started

Want to know the real secret to getting started with AI experiments? Stop reading about it and start building.

I spent months consuming content about AI development tools. But until I actually opened Lovable and started building my quiz app, I didn't really understand what was possible.

The barrier isn't technical knowledge or perfect planning. The barrier is taking the first step.

Pick a simple idea. Something you can describe in a few sentences. Open one of these AI builders. Start talking to it about your idea. See what happens.

You'll probably surprise yourself with what you can build!

My product management quiz isn't going to change the world (I know that! 😆). But building it changed how I think about experimentation, prototyping, and the role of AI in product development. And that's worth way more than the 30 minutes I spent building it.

The tools are here. They work. The only question is: what are you going to build first?

Ready to start experimenting? Try building something small today. A simple quiz, a basic calculator, a landing page for an idea you've been thinking about. The goal isn't to create the next unicorn startup. It's to understand what's possible when you stop overthinking and start building.

What's your first AI experiment going to be? Let me know in the comments below! 👇