Music Quiz

I created a music quiz as an experiment for a presentation I was doing on Ai. I needed a demonstration and thought it could be fun with some interaction.

The quiz app I made uses openai's API to generate a list of songs based on a prompt. The user can then players take turns guessing a song by listening to a clip.

The first part of the process was just playing around with ideas on paper, but it crystallized very quickly. I wanted to create a simple game loop that you would then repeat until all the songs had been heard and guessed.

I am currently using nextjs, and wanted to see the capabilities of Cursor and Claude. It turns out that it is pretty good at some things, and horrible at other things.

First of all, I'm a Vim user. Before that, I used VSCode. So it was only a little painful to go back to that environment. Luckily, Cursor has has vim mode, so I could still navigate at lightning speed. There were some hickups however.

I'm not sure if it's just me, but the undo and redo in Cursor seems a little broken. I took a while to get used to the new way of coding with only prompting, but whe I had to go back in history, it was painful. It often failed. I nearly tossed my laptop out the window once or twice.

The promting part was fun as hell. I wanted to push the envelope on how lazy I could be. Turns out, I can be pretty lazy. Whenever there was an error on screen, or the build didn't work, I would just copy paste the message and throw it back into the machine. While waiting Claude to get back to me, I could proceed to reading anohter article online or get myself a cup of coffee. This is the life!

Well, not really. Even though it one-shotted a Spotify API action file, it sometimes failed, what to me were very simple tasks. When I knew it would be complicated, I would describe in detail what I needed it to do. You get back what you put in most of the time. So when I was lazy, I would make errors.

However powerful it felt to have the machine do most of the work, it probably took longer than I had I done it myself. The conclusion here is obvious and boring. A little bit from column A, and a little bit from column B.

We got there in the end. And the result what better than I had hoped for. But there take away is what most people are realizing right now. Ai is a powerful tool. But if you decide to use your hammer on a screw.. You know the rest.

I don't think we've hit the wall just yet on prompting. I look forward to see the evolution of this type of workflow as we crash into 2025.

Check it out: https://musikviz.netlify.app/