The current problem I have with AR implementations is they require the user to assist in a relatively slow scanning process of the immediate environment before

The current problem I have with AR implementations is they require the user to assist in a relatively slow scanning process of the immediate environment before recognizing geometry. This has to be done every time you use AR. The geometry recognition isn't all that great, can't handle occlusion, eats battery, etc. Lots of potential, but even with ARKit as the current industry leader, it's still pretty rough. We need LIDAR for phones or something.