Etienne Jacob, known online as bleuje, is a Paris-based software engineer who makes generative animation in his spare time. Since 2017 he has been writing perfectly looping GIFs in Processing, later moving to openFrameworks and GPU compute shaders. Around 2024 he built a tool called Variograph: a real-time, interactive generative image instrument that can be operated with a gamepad and used as a gallery installation. It has been exhibited at least five times in live settings and is downloadable on itch.io.

The core algorithm is simple and deep. Variograph maintains a list of functions called "variations," each taking a 2D coordinate as input and returning another 2D coordinate. These variations come from the fractal flame algorithm — hundreds of different positional transformation rules, most with randomizable parameters. The key operation is composing these functions end to end. Then millions of particles, starting from a uniformly filled 2D square, pass through the entire function chain and are plotted on screen as near-transparent dots. The image is not "drawn" — it is the trail sediment of particles passing through the transformation chain. Each pixel's brightness is the cumulative count of particles that landed there.

This is where the remainder occurs. In the SAE framework, each variation function is a chisel: it takes a position and outputs another, and the gap between them — the geometric displacement — is the residue of that single chisel stroke. But Variograph's point is not the remainder of a single chisel. It is the cascading remainder of composed chisel-constructs. When you chain five or six variations together, each step's remainder becomes the next step's input. Residue compounds, folds, deforms. What emerges on screen is not the "result" of any single function but the aggregate sediment of the entire chain. Nobody — including the operator — can reverse-engineer why the image looks the way it does from the parameters alone. This is the signature of the remainder: it is a byproduct of process, not a product of intention.

Even more telling is Variograph's resistance to mastery. Jacob himself writes in the documentation: "Getting interesting results is not really easy the more you become used to the tool." This is not modesty — it is a structural description. When the operator develops intuition for certain variation combinations — when those combinations become "already-construct" — the remainder dries up. You must keep introducing variations you do not yet understand, keep re-randomizing parameters, keep returning yourself to uncertainty. The tool itself is a remainder-production apparatus, but it demands that the operator continuously dismantle their own constructed experience. This is the chisel-construct cycle made concrete at the human-machine interface: once a construct solidifies, it must be broken by a new chisel, or the image degrades into style.

Seeing Variograph now matters more than seeing it later, because its categorical home has not yet hardened. It is not fractal art — fractal art has its own already-constructed aesthetics (Julia sets, Mandelbrot renders, systematic parameter-space exploration), and Variograph borrows from fractal flames only the vocabulary of variation functions, not the goal of exploring parameter space. It is not traditional generative art — it has a gamepad, live interaction, the operator's body as part of the image. It is not mathematical visualization — its use of mathematical structure is aesthetic, not explanatory. It is something not yet named: a real-time variation synthesizer, a visual instrument that patches mathematical functions like a modular synth patches oscillators. The art world is slowly approaching this territory — "real-time generative installation" appears more frequently in exhibition contexts — but Variograph's specificity is that its remainder is mathematical, not pixel-based. Its raw material is function composition itself.

Han Qin (秦汉) · @hqinjarsy · SAE Judgment & Aesthetics