Many years ago, I worked with the NYU Music Experience Design Lab to create a browser-based music theory tool called the aQWERTYon. It has two purposes: to turn the computer keyboard into a playable instrument, and to visualize chords and scales in an intuitive way. I am very proud of this tool, it was a lot of hard work, and I use it constantly when I’m teaching music theory. However, in the years since we stopped working on it, I have thought of some features I would want to add and refinements I would want to make. However, I am not much of a programmer and felt too overwhelmed by the idea of going into the code myself.
But then I thought, maybe this is my excuse to try vibe coding. So I told Claude to replicate the aQWERTYon. It made an HTML file that I uploaded to my web host, and there it was, a fully functioning knockoff of the app. Then I started adding missing features, and making some of the changes I had always wanted. Here’s the result, try it yourself.
This was disturbingly easy to do. As of this writing, I have spent a total of, I don’t know, half on hour on this? I’m using the free version of Claude and my own web hosting. Claude exports HTML files that I am just FPTing into my main web site directory.
Let me explain what I mean when I say I’m not much of a programmer. I did not go to grad school with the goal of becoming an academic. I wanted to learn to design and develop software for something music-related. I thought, maybe I can go work at Ableton or something. I went in knowing some HTML and general computer science basics. In the NYU Music Technology masters program, I tried to learn JavaScript, C, Pascal and Max/MSP. I mostly failed. I could kind of do Max, Pascal and JavaScript with a lot of handholding, but C was too much for me.
My masters thesis project was the Groove Pizza. I did all the design and a little of the code, but I had heavy support from my friend Chris Jacoby (he helped me with Max in exchange for improvisation lessons.) My version was a Max executable that could run for about a minute and a half before crashing. It was enough to do my thesis presentation, but it wasn’t a usable tool. But then my advisor, Alex Ruthmann, got a grant from the Thelonious Monk Institute of Jazz to turn the Groove Pizza into a legitimate web app.
I wrote zero percent of this code, and didn’t design the graphics either; we had a whole team of ace developers and designers working on it. The Groove Pizza has been used by many millions of people in over a hundred countries around the world, many of whom are classroom teachers, and helping to bring it into being remains one of my proudest achievements in life.
The Groove Pizza put me on the map as a music academic, but it didn’t launch my music software career. I discovered that I was more interested in the teaching and learning of music than I was in the technical side. I started getting offered teaching jobs, and I found it to be fun and energizing in a way that programming profoundly was not. So, for the last ten years I have been teaching classes, writing and making music, but not writing any code.
This year, however, I did start on a new software project that I’m not talking publicly about yet. I met the lead developer because she was a longtime fan of the Groove Pizza. She has been vibe coding our demos, and I have been amazed at how fast she is able to turn them around, and how richly functional they are. She is an expert developer, but so far she has been working alone, so clearly AI has been a massive force multiplier for her.
I am on record as an AI hater, but that’s in the context of the humanities: writing creatively and analytically, making music and art. I am not surprised to find that AI is so much better at writing code than it is at writing English or music. Code is highly structured and modular in a way that English and music aren’t. I imagine that it’s also easier to train the models, because code either works or it doesn’t, there isn’t all this squishy subjectivity to deal with. I don’t know anything at all about Claude Code’s training data. Does it come from open-source projects or is it stolen intellectual property? I should probably find out, but I haven’t.
Remaking the aQWERTYon is not a completely fair test. I had this already-existing, fully functional web app for Claude to copy. I haven’t tried vibe coding something completely from scratch because I haven’t thought of anything I needed or wanted to make. My next vibe coding project will be to update the Groove Pizza, I have some new functionality I want to try for it.
This experience has not changed my conviction that we should keep students away from generative AI in the humanities. I have no idea how to teach computer science; maybe AI is helpful for that, maybe it’s harmful. My instinct is that people should learn how to code the hard and slow way, but I tried and failed at that, so what do I know? Having this experience made me sympathetic to my tech friends’ conviction that AI is going to transform the world. It has transformed their world already! There is no comparison at all between my fumbling with JavaScript and my fumbling with Claude Code. The question is whether AI is as good at other things as it as at coding. I have never read a piece of AI prose that I wanted to read again, and AI music is relentlessly terrible. So I guess the question is, how much of the world is like code, and how much of the world is like prose or music?