We'll give an update on all that's happened in TensorFlow 2 since launch, covering recent improvements and launches. In usability, TensorFlow now supports tf.numpy, a NumPy-compatible op subset that works interchangeably with TensorFlow ops. We're working on new research-focused NLP and image libraries. For performance, we'll cover runtime and compiler updates that let TensorFlow run fast on all kinds of hardware, including NVIDIA Ampere and TF-32. We'll also highlight TensorFlow Cloud that lets you train on large-scale cloud setups with just a few lines of code, and tools to let you profile it. Finally, in the TensorFlow ecosystem, we will discuss extension libraries around important fairness and privacy, as well as check in with TensorFlow Extended, Recommenders, TF Lite, and TensorFlow.js. Whether you're a current TensorFlow user or new to it, you'll get an accessible view into the depth and breadth of the TensorFlow ecosystem.