Find a reason to experiment.
Find something big and fun to give you a reason to learn what the tools can actually do. Pick something that makes you curious enough to keep playing.
Does Human Creativity Still Matter in the Era of AI?
This all started with a simple thought: "You know what would be cool?" I wanted to capture the feeling of great musical improvisation: musicians finding something new together, groovy visuals vibing to the music, and something a little different every time. So I built a band and lightshow with agents.
As I played with my crazy machine, I learned new things about what was possible. I made a lot of music that wasn't quite right. And even though lots of my results were cool, I realized that I couldn't get what I was chasing without bringing myself.
This is a talk about play, taste, and the importance of your human creativity in the age of AI.
Human Creativity in the Age of AI · AI Rising · October 20, 2026
Find something big and fun to give you a reason to learn what the tools can actually do. Pick something that makes you curious enough to keep playing.
Catalog and save useful generations so you can build on them. Bad generations are useful too - they give you something to learn from.
The musicians follow changing chords because their programs share a musical context. Models write their programs to reference contextual variables, and Tripulator keeps them current. Your team’s card wall, product plan or database schema all serve this same purpose: helping independent contributors pull in the same direction.
The same prompt produces different results through Hydra, p5, and GLSL. AI makes experimentation cheap. But choosing what to keep from the results still takes your judgment.
Sometimes one element of a scene needs specific changes. Sometimes the whole band needs a feeling: “Make it feel like driving fast.” "Vibe" prompts can be an interesting way of unsticking your results from the average. See what different levels of context can add to the abilities of your agents.
The model matters. The same prompt fired to two different models can yield dramatically different results. Maybe my own guardrails were getting in the way? This experiment changed my idea of what the instrument could do. Keep playing, and push the limits of what you've built. If you think you already know what's possible, you're probably wrong.
When I wanted something specific, I had to bring it myself. A riff and a bassline I wrote on my guitar, translated into Strudel. A generative solo recut from my own playing in real time, different every time. The music I was hearing in my head all along, finally real. AI can build the application. It took something human to make it matter.
Tripulator is a browser-based Rube Goldberg machine for generating music and visuals. Load a scene, change a part, give a musician some direction, and see where it takes you. Turn your own agents loose and change the whole thing. Try different models and capture your best results. Make something cool and tell me about it!
An hour wasn't long enough to share everything interesting about this project. Some things I plan to write about:
Get the making-of posts and future experiments in your inbox.
This subscribes you to my posts via Substack, including the Tripulator series. The articles will also be available on my blog.