The interface that imparted information into a human subject was just a standard transcranial magnetic stimulation coil that they futzed around with until for each subject they found an orientation and intensity and frequency that made them see a spot in their visual field when it was activated. They then would report when they saw a spot and slowly decode a binary string of seeing and not-seeing. That's nothing terribly new, what was new was that the on/off state was decided upon by a computer using electrodes to get a rough look at a sitting person's imagery/intent of moving their arms, and they would visualize moving their arms or not in order to binary-encode the message.