After some requests, I have uploaded my (really short) analysis of Google’s TensorFlow to arXiv: https://arxiv.org/abs/1611.08903.
It is really just a small seminar paper, the main finding is that while using any Machine Learning framework is generally a good idea, TensorFlow has a really good chance of sticking around due to its already widespread usage within Google and research coupled with a growing community.