separation of concerns, easier testability, easier mocking..., the thing is especially in more complex applications the code that generates/validates the data and the code that displays them are usually written by two different people.
nowadays once the json schema design is settled, they can work in parallel, each of them can test their parts without needing the other and the merges can be simpler, because the parts do work more or less stand-alone.
nowadays once the json schema design is settled, they can work in parallel, each of them can test their parts without needing the other and the merges can be simpler, because the parts do work more or less stand-alone.