Aalap Davjekar lives in Goa, India, and works as a creative technologist — not an artist in any institutional sense, not a VJ in any professional one. In May 2026, he wanted a live visual hook for a video he was making, found nothing cheap or pliable enough on the market, and built his own. What came out of that is bad-ASCII: a live visual synthesizer running entirely in the browser, no installation, no downloads. It feeds a webcam's human image and a microphone's audio through several mathematical engines, producing a stream of characters — ASCII, Chinese glyphs, binary, brutalist blocks — that are no longer text. They are particles: deflected by sound, colonized by physics, seeded by the body's edge.

The center of bad-ASCII's claim to 余项之美 is not its visual effect but the logic of how it uses the Gray-Scott reaction-diffusion equations. Gray-Scott describes morphogenesis — the same differential equations produce leopard spots, coral branching, shell patterns, depending on initial conditions. In bad-ASCII, Davjekar uses MediaPipe to isolate the human silhouette in real time and feeds that silhouette's edge as the seed of the simulation. Stand still: mathematical coral and worm-shaped patterns grow from the boundary of your body and spread into the background. Your outline becomes the initial condition of a growth field; your body's geometry becomes the 凿 (chisel) that initiates a reaction that no longer belongs to you. What the equation produces from that chisel is the remainder — growth patterns that cannot be predicted from the already-named input (the segmented human body) and that don't belong to any prior category.

Seen through the SAE chisel-construct cycle: the character systems (ASCII, Chinese) are already-construct, fully named symbolic structures. But in bad-ASCII their semantic content is stripped away. What remains is their shape — the shape as pure pixel carrier, as a grid cell that can be rotated by optical flow and bounced by spring physics. The meaning is chiseled out; what remains is the form. That form is the remainder of the chisel-construct. Similarly: the human body on camera is already-construct, a recognized and segmented object. But when its edge seeds the reaction-diffusion field, it escapes the category of "human subject" and becomes a mathematical seed growing into patterns that belong to no existing art-historical category. The growth is the remainder.

Why look now? Because this structure's logic is still forming. bad-ASCII openly calls itself "still a toy and an experimental playground." It lives outside gallery systems, outside the formal market for VJ software, useful to VJ culture without being claimed by it. It uses machine learning — but to segment bodies, not to generate images, which sets it entirely apart from AI-image discourse. The art world, the tech world, and the music world are all slowly approaching this kind of practice, but none has yet produced a name for it. The moment a naming vocabulary arrives, the remainder calcifies into construct. Right now, the window is still open.

One further detail worth naming: the placement of Chinese characters alongside ASCII as equivalent visual rendering units. This is a specific remainder. Chinese glyphs carry far denser semantic content than Latin letters — they were never designed to be pixels. Using them that way is neither calligraphy, nor typography, nor Orientalist symbol consumption. It is a fourth thing that does not yet have a name, which is precisely why it sits at the edge of the remainder.

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