While my answer is specific to CS, might be general enough. After writing it, I noticed that I'm mostly focused on the writing, the technical parts were never an actual problem for me...
Personally, I try to decrease my involvement with Ph.D. students over time. The objective is for them to be independent researchers by the end of it after all.
First years usually require a lot of supervision and hand-holding. Often they don't really know what they want to do, or how, or how to properly write it. I found that putting them to help as "middle authors" of more senior PhDs is a good idea so they get a sense of the process.
When they start writing "by themselves", I usually do not directly edit the text, but rather comment on the PDF explaining why I want stuff changed. Each review takes forever, but at least they should be able to extrapolate the rules and develop a "sense" for it...
By the third year, I fully expect them to "lead" a paper, but I still keep a close watch, albeit with less interference... I see most of what they do but intervene only when unavoidable. Most are starting to develop their own writing style there.
4+ should know enough by then to deal with their own issues and help the "lower" students, so they require minimal supervision if any. And they have a dissertation to finish. Except for a few "dissertation specific" details, I never had to touch the dissertation, especially when it is in the "collection of papers" format.
Of course, that is just the general lines. I had students leading papers as undergrads and hand holding on the fifth year of their Ph.D.s, each person is different, stuff happens, and the advisor needs to be on top of it to "personalize" the supervision.