Google Glass is probably a good example of halfway failures. They did well on implementation of some sort of screen in front of the eye among other things, but

Google Glass is probably a good example of halfway failures. They did well on implementation of some sort of screen in front of the eye among other things, but in the end, they stopped short of AR in favor of just a screen sitting in front of the eyeball. To your point: it'd be impossible to achieve AR without some major breakthroughs in eye-tracking _or_ general 3D image rendering in front of the eyes. This is what hololens had to achieve before AR made sense, and just to be able to ship it in spite of the hardware constraints, they were stuck with a small viewport in Gen 1. The only real compromise with implementing AR is FOV coverage. If you're going to cover the entire FOV, you should either get creative (rendering low-res outside what hits the fovea of the eye, which would be possible only with eye-tracking) or use a small viewport. Hololens did it right. It seems (judging from the hardware) that Magic Leap is trying to avoid compromises of any sort, so they're probably doing it wrong.