If I was designing it, my first instinct would be to make a headset that renders fairly arbitrary surfaces over the environment using as much hardwired custom silicon logic as possible, and leaves everything else external. You should be able to achieve pretty incredible energy efficiency that way, using any number of unconventional logic schemes, as well as minimising headset weight. The main question is how much you can pre-calculate without knowing the fine details about how the user is oriented this millisecond.

We’ll see what the R2 looks like. As far as I can tell the R1 is just a bespoke arrangement of more conventional cores.