I Use a FTDI Friend for programming this; it plugs directly into the same connector as the Duet cable; and allows in-circuit programming and serial debugging.

The ATMega runs the Optiboot bootloader; this is the standard loader used in the Uno etc.

During development I found I had to set the FTDI to 5v mode to get the reset pin triggering correctly to force programming mode.  This is not a problem for the 3.3v level converters; they still pass the serial signal properly. 

There are ways of using Arduino Uno, Deci etc boards to simulate the FTDI friend by removing or disabling their CPU and then connecting the RX, TX and RST lines from the board to the relevant lines on the FTDI connector. (note: the CTS line is not needed/used/connected)