We got picowatts of power so far, which is similar to preliminary results at Prof. Daniel Sheehan's lab at U.C. San Diego with their devices. The result is quite repeatable. So far the device is unoptimized so we expect a factor of 2 or 4 more power from an optimized device. Update: we have kept the device shorted for a month now with no further reduction in voltage, and we have tested on an antistatic mat: after removing the shorted condition and allowing a day for equilibration it always goes back to about .02 millivolts again.