I posted this on YouTube, however it may be helpful to other. This is my understanding of the kinematic videos on YouTube in particular episode 6:
So basically, you find the inverse of common positions, which I will call posses, that can be put into a sequence. For example, let us say the first sequence goes to J2 pose as it is the closest move to the object we want to interact with. Once this is achieved the arm then moves in that pose to the object as one unit instead of having to jog the x, y, z individually to achieve the task. Am I on the right track?