Papers are good for communicating the big picture: what is being done, why certain choices were made, and how things are supposed to work. Writing them is usefu

Papers are good for communicating the big picture: what is being done, why certain choices were made, and how things are supposed to work. Writing them is useful, because it forces you to think about the big picture. It's common that people who have spent months or years working on a project suddenly understand some aspects of it better when they try to communicate it to other people. Code is good for communicating details, but very few people actually read research code. Those who do usually just check a few details, because it's hard to justify spending weeks or months studying a single project. It doesn't help that the code was probably written before anyone understood what was the exact problem they were trying to solve and what would be a good way of solving it.