Following up on the previous post, here is a good way of beginning a project.
Day one: only brainstorm. Write as much as you can in a given period of time (an hour?)
Day two: Begin by reading what you have written. Rewrite some of the sentences or phrases you have written as though they were sentences you would actually put in a finished version of a the paper. When you have done that, then brainstorm some more. Take about an hour and a half the second day.
Day three: Begin by tweaking some of those "finished" sentences. Revise them for style. Connect some of them up into a fragment of a paragraph. Then convert some of the brainstorming of days one and two into serviceable prose. By now you should have an actual paragraph. End by brainstorming again.
Day four: Repeat. At the end of day four, if you are like me, you will have a document of about 1000-2000 words. Some of it will be finished prose, some merely coherent prose, and some informal brainstorming.
Continue until the article is complete. During the last few days everything will be at stage two (serviceable prose) and your entire effort will be devoted to making it finished prose (stage three). When the entire manuscript is in stage 3 (every sentence tweaked, all references complete) then it is a penultimate draft that you might want to show someone else. It's penultimate because it is the best you can do by yourself, without outside critique. (I never show a "rough draft" to anyone else, because that would be asking someone else to do my work for me: improve sentences I could easily improve myself. I never write rough drafts, only smooth ones. That doesn't mean I write perfect drafts, but that I get them as good as I can before showing them to anyone else. If I want earlier feedback I explain my ideas to others, but in very careful language. I might spend more time phrasing a question in email than I would writing a paragraph of the article itself.)
(The only caveat here is that a document will sometimes get too messy, with too many unconnected brainstorming ideas and not enough finished prose; in this case I start a new "cleaned up" version in another document, retaining only finished sentences. I don't throw away the brainstorming, but I remove it from the document to increase efficiency, usually around day 10.)