Brazil r/s V2 Field Guide
Hey, that's neat 
I've seen this done in an old raytracer where you basically force the renderer to use a different Z value for a given surface; it did require it to always trace all surfaces, instead of stopping on the first hit (presuming opaque), of course.
Other methods are compositing in post; that way even shadows and reflections can be fooled and the 'impossible' object will be impossible in all three.
What's funny is that I'm more impressed by the top image. I'm used to trickery in renders, not in viewports.
well as I understand it, given that just camera rays are fooled here, it's actually modeled to be this way. Essentially you could take a proper cube lattice, give it enough vertices, and then move the 'behind' vertices in front while keeping them at the same X, Y location in the viewing plane. If rotated, I think you'd see some bendy geometry - but that's just my guess
What? No depth of field?
My first thought was that I wanted to see it animated
Commenting by HaloScan
