Xoogler from a few years ago. I'm not saying they have nothing cutting edge, just that the rest of the industry eventually reached parity with a lot of their everyday platforms. If you come in knowing HDFS and Cassandra and Spark and k8s, moving to CFS and BigTable and Flume and Borg involves learning a whole lot of minor differences that don't make you much more capable but screw up code reuse.

I was pleased to see Cloud Spanner hit the market, and trying to nudge the industry towards hermetic builds with Bazel.