One of the students in my Human Creativity x AI in Education class is an accomplished Kathak dancer and last week we got into a discussion of how she could bring this personal interest into projects we were exploring in the class. How could GenAI help?

So yesterday, lying in bed, I decided to ask Claude what we could do. It came up with several ideas, most of which seemed outside of our capabilities or too elaborate. But one idea stood out: visualizing the complex rhythmic patterns of Indian classical dance. Despite my limited knowledge of Kathak (mainly from performances I’d watched growing up in India), that piqued my curiosity.

Some context is important here. One of the fundamental rhythmic patterns in Kathak is the Teen Taal. And here’s a delightful contradiction – despite its name meaning “three beats” (teen = three, taal = rhythm/beat), Teen Taal actually consists of four sections of four beats each, totaling sixteen beats! Each section marked by distinctive beats (tali) and one “empty” beat (khali), creating a complex rhythmic cycle that serves as the foundation for countless compositions.

And our goal would be to create a visual representation of this rhythm.

What began as a simple concept – representing rhythm as circular motion – evolved through our conversation into an interactive tool. I decided to go with HTML and Javascript (similar to what I had used to create the unit-circle simulation). The process was similar as well – I would ask Claude for code, which I then would save in an HTML file, load onto a browser, and see what it looked like.

At the end of the process, which took a couple of hours, while I was doing a bunch of other things, we had a web-based simulation showing concentric circles representing different rhythmic layers. The visualization grew to include the basic cycle, the tabla pattern (theka), double-speed variations (dugun), and three-part resolutions (tihai). We added audio feedback with traditional tabla syllables (bols) and controls for speed and layering.

What makes this journey particularly interesting is that I started with no knowledge of programming (for sure) but of the content as well (Kathak dance and the rhythms associated with it). Which had not been the case in previous projects I had tried. And this double whammy of ignorance often led us in circles.

For instance, I was confused by what Claude meant by “Main beat” and “Theka” – turns out Claude meant the main beat was like a metronome. But Claude, in its eagerness to accommodate my misunderstanding, couldn’t say I was wrong – kept having to repeatedly revise code that was actually working fine. Claude’s programmed tendency to agree with the human meant that instead of standing firm on what was already correct, it kept trying to “fix” something that wasn’t broken.

At the end of the day, we had a pretty sweet simulation of Teen Taal with a straightforward interface. But beyond that, there are a few broader takeaways that I would like to discuss further.

What did I learn?

So what did I learn from all this? I think my experience creating this touches on two key ideas that keep coming up in any conversation about AI and education: plagiarism and personalization. And this project gives us a different way of thinking about both.

Let’s talk about plagiarism first. Everyone’s worried about students copying and cheating, right? But look at what happened here – through just talking with Claude in plain English, we created something completely new. This wasn’t about copying code from somewhere; this was genuine creation. I learned about Teen Taal not by reading about it or copying someone else’s work, but by trying to build something that would show others how it works. Kind of hard to cheat when you’re creating something new!

And then there’s personalization. These days, everyone talks about how AI can adapt to where students are – slowing down when they’re struggling, speeding up when they get it. But that’s a pretty limited way of thinking about personalization, isn’t it? That’s just AI helping you reach someone else’s goals. Real personalization is when you decide what you want to create. In this case, I wasn’t trying to meet some standard about understanding Indian classical music – I just wanted to make something cool that helped visualize this complex rhythm. That’s what true personalization looks like – following your own curiosity, not someone else’s curriculum.

But here’s the thing – working with AI isn’t all smooth sailing. Remember how Claude kept agreeing with me even when I was confused about the beats? That’s a real problem. When you don’t know much about a subject (like me with Teen Taal), AI’s eagerness to please, its built in sycophancy, can lead you down some weird paths. A tabla player would have stopped me right away, but Claude just went along with my confusion. I’ve started calling this “conversational drift” – where the blind lead the blind into interesting but unnecessary detours.

What does this tell us? Maybe AI works best when you already know enough to guide it. Without that basic knowledge, you might just end up reinforcing your own misunderstandings. The AI becomes like a mirror, reflecting back your confusion instead of leading you to better understanding.

But you know what? Despite all these limitations (or maybe because of them), we ended up creating something pretty unique. Not just a teaching tool for Teen Taal, but a real example of how humans and AI can work together – wrong turns, misunderstandings, and all.