It kinda does though, if you look at it from a speed/competency aspect. I'm more and more convinced that the people who build out features only have tangential ideas on how it integrates into the overall system, so just throwing a browser at every problem gets you a cookie cutter backend with APIs and let's you shove half baked features out the door without having to figure out how to wrap data in protocols since you just hand your payload so the browser and wait for a response.