I've always been a Kubernetes Maximalist, but recently I've been thinking about the secondary effects of API and technical decisions. The argument goes, developers don't need to know how to use Kubernetes. Data scientists shouldn't need to know how to build a container.