I think you could think of "infrastructure as code" as he described it as a superset of "infrastructure as data". Both have the benefit of being able to be repr

I think you could think of "infrastructure as code" as he described it as a superset of "infrastructure as data". Both have the benefit of being able to be reproducibly checked into a repo. Declarative systems like Kubernetes/"infrastructure as data" just go even further in de-emphasizing the state of the servers and make it harder to get yourself into unreproducible situations.