Skip to content
Making What I've Built My Own

18 Oct 2026 · 3 min read

Making What I've Built My Own

Doing a lot wasn't the same as knowing it deeply

Weekly

✅ Liked — What I enjoyed

I've been listening to a lot of Clazziquai lately. The more I listened, the more I wanted to make something about them, so I made and posted a Reel.

I also shared my content goals for the week with Layered Club. There are several things I want to make next. I'd like to put together some music for autumn and spend more time talking about an album I love. I'm also thinking about making something around five of my favorite scenes from About Time.

When I share music or scenes I like, I want to talk about why I like them, too. Including the memories I associate with them could make room for more of my own story. I'm also working on a draft about wanting to do too many things.

I was also drawn to an app that lets you leave a song at a place. What would it be like to hear that song again when I return? It would be nice if the music brought back the memories from that time, too. I'd like to connect music and travel, two things I love, in this way.

💡 Learned — What I learned

While building music search, I thought about which songs would make a search result useful.

Music for playing FIFA and background music for a tech review come with different expectations. In a tech review, the music needs to stay out of the way of the person speaking. That led me to think about where people intend to use the music when I look at search results.

I also compared using information from AI analysis of tracks with using the metadata we already have. Along with checking the results, I thought about whether the cost would remain manageable as more tracks were added. It mattered that the search could keep running, as well as work well once.

🧩 Lacked — What I want to look into further

I want to make the criteria for comparing search results more specific. I'd like to try more queries that people might actually use. I plan to look at which searches return the songs I expect and which ones don't, one by one.

I'm also curious about how other music services approach this kind of search. I'd like to compare different explanations and results, then find ideas I could apply to the search I'm building.

🌱 Longed for — What I'll try next week

To revisit how the music search works, I went through explanations of embeddings, vector search, and RAG, including how they differ. Next, I want to study them alongside the code I built. I plan to find where the concepts I read about are used in the actual code.

I'd also like to write small TypeScript examples and tests. I want to check how things work by trying them myself and write about the process of solving a problem.

I want to keep making new things. I also want to spend time understanding what I've already built. Next week, I'll choose one feature and explain, in my own words, what problem I was trying to solve and why I chose that approach.

📝 Lined — A video that stayed with me

A video by Han Hye-jin stayed with me. Watching it made me want to create something about dreams. It also got me thinking about my own dreams again.