You point to things by looking at it, which is very precise according to people who have tried it, like MKBHD. Clicking and dragging is done by finger pinching. So it works quite similar to a touchscreen or a mouse.
Yes apparently it is surprisingly precise. There might be an uncanny valley when the system is perhaps a little too good at predicting what your next move is. Are the goggles monitoring your facial expressions, looking not just for eye movement but also wrinkled noses, scrunched-up foreheads, etc. ? Will it be able to function as (for example) a lie detector ?