I did this once using graph paper. It worked perfectly.

That is not feasible, especially when it covers to the type of QR codes shown in the example, because humans can’t read QR codes, which means we can’t warp the data points in the QR just barely enough so that a reader could still pick it up because it’s within the error threshold - we don’t know what the threshold is. We could do something black and white with exact accuracy, like by using graph paper, but that wouldn’t be the same as there’s no warping of the code points, so what a human could do would at best always be a worse illusion than what the AI could do, because the AI can know what limit it can warp the QR too while keeping it readable, while we can’t.

That’s why I used this as an example - this is something only a machine can feasibly do with any practicality, because we don’t have machine vision so as to calculate where error thresholds would lie if trying to reproduce by hand. I suppose if you’re really, really good with math, you might be able to replicate something like what I posted, such as the red panda - but at that point, is using math to draw art? If so, then AI would be considered art too. If not, then 3D rendering isn’t art either.

The AI goes much further then what you are thinking of as stylized. The amount of skill needed is I imagine very very high to make something like the picture below while keeping it scannable.