But this is why libraries are written. As you use each schema (using json-ld, or turtle, or whatever), wrap it with a library, and it becomes as easy to work with as anything. And you're adding to the critical mass of a read-write "web." Programmers need to work with the data and schemas occasionally, everyone else, forever. You've exchanged YAGNI with NIH. Unless you just expect to gain a critical mass, or don't care about participating in that web.
It really doesn't become as easy to work with. You still have to wrangle around the RDF "everything is a triple and attributes are URLs" concept, and even if you hid all that away you'd be attempting to gain something as simple as this design is. We just need a way to share & publish objects which link to each other. We don't need the graph model.
