There is a very interesting thread where we discuss the exact same thing -
TL;dr - AR allows you to receive pointer events for both fingers, and you can use those two points to do your magic. Since we don’t currently provide the screen coordinates, the best way would be to check the distance (in 3d) between the two XR Input points (the two fingers) and use the deltas to scale a model (for example).
I also mention there that it is possible to get the screen coordinates, it just takes a bit of computation that might slow down the AR render loop. If that’s the missing piece I can show you how to project the 3D point to screen space.