Emotion Studies

2026

The first experiment with Xylem, using its underlying model Telomere 1.0, explored prompting it with concrete things like "flower". This second experiment hands it something abstract instead. I prompted Xylem with different emotions, among them anger, disgust, fear, love, hate, happiness, sadness, and surprise. I wanted to see how a model trained only on my own archive reaches for an idea that has no fixed shape.

Once again there was a colour collapse. The model can't hold natural, in-between colours, so every tone snaps to a harsh extreme of magenta, green, black or white. Those four are the corners of the colour space digital images are built from, which makes them the emptiest place in real photography and the one thing a broken model falls back on.

Every image created in Xylem starts as pure noise, a field of random static. Over 100 steps the model clears the noise away, nudging the mess closer to the prompt, in this case the word "flower", until something is left. What emerges is a flower of sorts, still deep in colour collapse. The model doesn't draw. It surfaces an image from static, nothing entering from outside.