The baffling part is how disproportionate the initial creation to the physical realization of this is. Like they really just slapped generate on a weird resolution ratio that made the model break down even more, then upscaled it in one of several possible ways depending on the hardware used, and then went to orders of magnitude more effort to actually put it up, seemingly without ever actually looking at it after slapping generate.

Like they could have taken a few more minutes to throw together a collage with some rough sketchwork to img2img it with a high denoise pass (translation: made the robot draw over a crude reference image they made), then spent like an hour editing and doing inpainting passes to fix the worst errors, and gotten something that at least served the purpose of "visual noise to fill space with some vibes" that they were after, and they'd still have spent orders of magnitude less labor creating the design vs actually putting it up.