Strip a generative system down far enough and you get something like booleanSequence: one rule, applied per cell, forever. Artist and developer Antonio Pezzella (niceunderground) built it in p5.js, and the rule is a single threshold. Each cell in the grid samples a Perlin noise field that changes over time, and whichever side of a cutoff value it lands on decides which of two opposing characters it shows — plus or minus, zero or one, greater-than or less-than.
That’s the whole system. What makes it worth watching is what a threshold does to a smooth, continuous noise field: it carves it into regions, and because the underlying field keeps drifting, those regions keep reshaping their borders. Pezzella’s own framing draws on cellular automata and Schelling-style segregation models — the same mathematics used to explain how neighborhoods sort themselves by a simple local preference rule, here repurposed as a generative aesthetic rather than a social simulation.
The piece exhibited at TRIP 2026 in Turin on three separate computers, one per character pair, each running its own threshold setting. A looser threshold produces large, slow-moving territories; a tighter one produces a finer, more agitated mix — the same rule, three different temperaments, side by side.
There’s no training data and no AI here, which is worth saying plainly in a field where “generative” increasingly means “model output.” This is Perlin noise, decades old, filtered through a comparison operator — proof that the oldest tools in procedural graphics still have unclaimed territory in them.