One of the project owners here. I've used the capsense arduino library a long time ago and it's decent when you don't need the accuracy. I found that the sensor values were affected by the environment quite a bit e.x. I'd have to recalibrate everything if the project was being used in a room with a carpet which affected the body charge. Maybe you found a way around it?
Re: the sensing method, they're similar in the sense that both the avr and mgc3130 detects a change in capacitance to figure out the position. But the mgc3130 controller can detect capacitance changes in the femtofarad range.
The capacitive electrodes form an electric field around the surface of the pcb. When your hand or finger (conductive surface) is in range, the chip senses the distortion of the e-field (change in capacitance) to track your hand movement and calculate the position.
We created this dev-kit to let Arduino projects work with gestures + touch. It came out of the need for a project which required basic hand gestures to work accurately. We tested with proximity and distance-IR sensors but wasn't satisfied with the accuracy and development time required to make it work. We decided to build a dev-kit that uses e-field distortion to detect gestures accurately. We just launched it hoping it might help other hardware hackers looking to add gestures/touch to their projects.