It's even worse than that. The cubes are generated and overlaid on the image based on textures from the images. In other words, the lighting from the scene on a

It's even worse than that. The cubes are generated and overlaid on the image based on textures from the images. In other words, the lighting from the scene on a real textured cube would look nowhere near as good as what it is shown. Just consider that the cubes in these images leave no shadows or affect ambient occlusion, nor are they self-shadowed. The article and problem is interesting in itself, but the portrayal of how effective this would be in real life is greatly exaggerated.