Back to Blog
Case Study
October 5, 202410 min read

From Idea to App Store: The BookMatcher Journey

How I built an AI-powered book recommendation platform in 3 months, from initial concept to launch. Includes technical decisions, pivots, and lessons learned.

Next.jsAISupabaseProduct Development

The BookMatcher Journey


BookMatcher started as a weekend project and grew into a full platform. Here's the story of how it came together.


The Problem


I read a lot, but finding my next book was always a chore. Goodreads recommendations felt stale. Amazon's suggestions were too commercial. I wanted something that understood *why* I liked certain books, not just *what* I read.


Initial Prototype (Week 1-2)


The first version was embarrassingly simple:

  • A Next.js app with a single text input
  • Users describe what they want to read
  • OpenAI's API suggests books based on the description

  • Surprisingly, this worked well enough to validate the concept. User feedback was enthusiastic.


    Building the Real Thing (Week 3-12)


    Tech Stack Decisions


  • **Next.js** - Already familiar, great for SEO (book pages should be indexable)
  • **Supabase** - PostgreSQL with auth, real-time, and edge functions built-in
  • **OpenAI API** - For the recommendation engine
  • **Tailwind CSS** - Fast iteration on UI

  • The Recommendation Engine


    The interesting part. Rather than just prompting GPT with "recommend books like X", I built a more sophisticated system:


  • **Preference extraction**: When users rate books, we extract *why* they liked/disliked them
  • **Taste profile**: These preferences build up a "reading personality"
  • **Contextual matching**: Recommendations consider mood, time available, recent reads

  • This approach produces much better results than simple collaborative filtering.


    Pivots Along the Way


  • **Dropped social features** - Users wanted recommendations, not another social network
  • **Added "reading moods"** - Quick filters like "something light" or "make me think"
  • **Simplified onboarding** - Original was too long; now it's 3 books rated and you're in

  • Launch and Beyond


    Launched on Product Hunt, reached #4 for the day. Current users: ~2,000 monthly actives.


    The biggest lesson: ship early, iterate based on real usage. My "perfect" v1 plans would have taken 6 months and missed what users actually wanted.

    Enjoyed this post?

    I write about iOS development, architecture patterns, and building products. Get in touch if you'd like to work together.