I co-taught a class called “Computational Textiles” this semester…

I co-taught a class called “Computational Textiles” this semester with @tkttran. It was our first time teaching together, and we jumped right in—exploring the interplay of weaving and computation.

When an article about a Boltzmann machine simulator popped up in my feed, I knew I had to try weaving it. But how?

All semester, I’ve been trying to figure out how to do procedural weaving. Imagine an evolving, dynamic draft where each row depends on the previous ones.

We tried a Fibonacci pattern. Each value was derived from the ones before it, and it felt like a promising direction—beautifully facilitated by @tkttran. I didn't have any more inspiration until now.

But why this?

The Boltzmann machine is an early generative network. It learns by mimicking how energy flows in physical systems. In the simulator, you can see visible “neurons” on the left and “hidden neurons” on the right. Their connections/weights drive the output: the warp threads move to indicate lifting of the warp threads and selection of the weft color. The result is a material record of the machine’s learning process.

But who would warp a loom like that?