I Built an Entire AI Product with Zero Coding Experience — Here's What Happened

j-river — Fri, 03 Apr 2026 09:09:00 +0000

I have zero frontend development experience. I don't know React, never touched Next.js, and CSS makes my head spin.

But last month, I built and launched a real paid product — Pet Sticker Maker — entirely using AI coding assistants. It's an app that turns your pet photos into adorable Funko Pop style animated GIF stickers.

In this post, I'll share my honest experience: what worked, what didn't, and whether AI-assisted coding is actually ready for non-developers to ship real products.

What I Built

Pet Sticker Maker does one thing really well:

Upload a photo of your cat or dog
AI generates 4 unique animated stickers in Funko Pop cartoon style
Download MP4 videos to share anywhere

It's live at petsticker.shop, processes real payments via Stripe, and runs on Cloudflare's global edge network.

My Background (Or Lack Thereof)

Let me be honest about where I started:

❌ No frontend development experience
❌ Never used React or Next.js
❌ No knowledge of CSS frameworks
✅ Basic understanding of what APIs are
✅ Willingness to learn and iterate

I'm currently in an "AI Coding Bootcamp", where the premise is: can someone with no coding background use AI tools to build and launch a real product in 30 days?

Spoiler: Yes. But it wasn't easy.

The Tech Stack (Chosen by AI, Not Me)

I didn't choose any of these technologies — my AI assistant did:

Layer	Technology
Framework	Next.js 15 (App Router)
AI Image Generation	Volcengine Doubao API
AI Video Generation	Volcengine Doubao Video API
Authentication	NextAuth.js + Google OAuth
Payments	Stripe Checkout
Hosting	Cloudflare Workers
Styling	Tailwind CSS

The final codebase is ~5,800 lines of TypeScript/React. I understood maybe 10% of it when it was written. I understand about 40% now.

How I Actually Worked with AI

The Pattern

My workflow looked like this:

Me: "I want users to upload a photo and see a preview"
AI: *writes the upload component, API route, and file handling*
Me: "The image preview looks too small"
AI: *adjusts the CSS and layout*
Me: "Now connect it to the Doubao API for sticker generation"
AI: *writes the API integration with error handling and polling*

I never wrote a single line of code myself. But I had to:

Describe what I wanted clearly — This is harder than it sounds
Review the output visually — Does it look right? Does it work?
Debug when things broke — Copy error messages, paste to AI, apply fixes
Make product decisions — AI doesn't know what your users want. You do.

What AI Was Great At

✅ Scaffolding — "Create a Next.js app with upload, payment, and result pages" → done in minutes
✅ API integration — Connecting to Doubao, Stripe, Google Auth → mostly correct on first try
✅ UI iteration — "Make the button bigger", "Add a gradient", "Animate this" → instant results
✅ Debugging — Pasting error logs → accurate fixes 80% of the time
✅ Deployment — Walked me through the entire Cloudflare Workers setup

What AI Struggled With

❌ Deprecated libraries — It suggested @cloudflare/next-on-pages which is deprecated and breaks with Next.js 15. I had to discover this through error logs and find @opennextjs/cloudflare myself.
❌ API quality differences — AI initially recommended MiniMax for image generation, but their subject_reference only works for humans, not pets. I had to test and pivot to Volcengine Doubao.
❌ Complex state management — When the polling logic for AI generation got complicated, AI sometimes introduced bugs that took multiple iterations to fix.
❌ Production edge cases — Things that worked locally but failed on Cloudflare Workers (async_hooks issues, file path differences).

The Moment It Actually Worked

I'll never forget the first time I uploaded a photo of my cat and saw four animated Funko Pop stickers staring back at me. They were actually cute. They actually looked like my cat.

That's when it hit me — I built this. Me. The guy who can't write a div.

The Numbers So Far

⏱️ Time to MVP: ~2 weeks (from zero to working product)
💰 Total cost: ~$50 (domain + Cloudflare + API testing credits)
📦 Codebase: ~5,800 lines of TypeScript/React
🤖 Lines I wrote myself: 0
🧠 Percentage I actually understand: ~40% and growing

Honest Takeaways

1. AI Coding is Not "No Code"

You're not clicking buttons on a canvas. You're directing an AI to write production code. You need to:

Read error messages (even if you don't fully understand them)
Understand the general architecture (even at a high level)
Make product decisions that AI can't make for you
Test relentlessly

2. The "AI Coding" Learning Curve is Real

It took me about a week to learn how to effectively communicate with AI coding assistants. Clear instructions, specific feedback, and iterative refinement are skills that develop over time.

3. You Still Need to Think Like a Product Person

AI can write any feature you describe. But which features should you build? What's the right user flow? How should pricing work? These are product decisions that require human judgment.

4. Deployment Will Hurt

Local development? Smooth. Deploying to production? Expect 2-3 days of debugging edge runtime issues, environment variables, and platform-specific quirks. This was the hardest part for me.

5. It's Exponential

The more I built, the more I understood. The more I understood, the better my AI prompts became. By week 3, I was shipping features in hours that took days in week.

What's Next

I'm now working on:

🐰 Support for more animals (rabbits, hamsters, birds)
📱 WhatsApp sticker pack export
🎨 Custom sticker text and emojis
🐾 Batch processing for pet shelters

And I'm building all of it with AI, one prompt at a time.

Try It

🐱 petsticker.shop — Upload your pet photo, get 4 animated Funko Pop stickers for $2.99.

If you're a non-developer wondering whether AI coding tools are ready for real products — my answer is yes, with caveats. You won't write code, but you will need to think, test, and iterate.

This post was also written with AI assistance. The product, however, is very real.

Forem: j-river