we started to split stuff into 'layers' so that e.g. deploying apps to kube didnt involve checking or refreshing any of the infrastructure around it