Tensorflow's tf.contrib.rnn.NASCell for optimal network architecture

Hi, I saw this cell - tf.contrib.rnn.NASCell being mentioned in the Tensorflow Summit 2017 videos as something that through Reinforcement Learning finds the best/optimal neural network architecture. Has anyone tried it yet? I can’t seem to understand how to use it as my python skills are not great and I don’t see any examples on web.
Cheers.