What works very well with the device is coarse movements, especially relative hand movements. What doesn't work so well is finer gestures (1/100th of a millimeter motions of all 10 fingers? ha).
This app is a perfect example: https://airspace.leapmotion.com/apps/cyber-science-motion
You can use your hands, kept flat, to spin around / zoom a 3-D rendering of a human skull. You can also point at specific elements on the skull. Both of these coarse gestures work great, and the experience is incredible.
However, the app also unfortunately has a "click" gesture to pick apart elements of the skull - you click by spreading out your thumb and then folding it back in. Works terribly, as this fine gesture is detected maybe 50% of the time. It should've simply been left out.
I showed this app to my dad, who's a doctor, and he was blown away. He was visibly excited about the potential for a device he can use to spin around CT and MRI scans in the operating room without having to touch a mouse/joystick - currently he has a person doing this for him to keep things sterile, and this can sometimes be frustrating.
The leap, at least in its current incarnation, reminds me a lot of Google glass. Both Google/LEAP and their proponents say the devices are going to change the world. Maybe, maybe not. Neither device works perfectly like what you see in the heavily edited demo videos. But both can be invaluable in certain specialized fields, today, as long as folks are realistic about what can be done with them.