(a machine learning phd student here) 10 hours a week reading seems about right for me, although often is a lot more or less than that. Usually most of my readi

(a machine learning phd student here) 10 hours a week reading seems about right for me, although often is a lot more or less than that. Usually most of my reading come from: google scholar searches on a topic of interest, following the references cited in a paper I'm trying to study, following the google scholar list of papers that cited an interesting paper I'm trying to gauge the influence of, and (a few times a year) some 10 to 15 papers that are freshly released from one of the top international conferences in machine learning or natural language processing that had interesting titles or abstracts. A very small amount of reading comes from my advisor or colleagues. I do follow some arxiv feeds about research in ML, but it's very rare to find good papers there that I would not find by some of those other means. I'm not sure how much of this translates to software engineering, but I find it helpful to keep a loosely organized reading list with papers and topics I should browse or read more carefully. The important thing, I guess, is bootstrapping an initial reading list for you to expand. Search for top conferences (ACM, IEEE, etc) on SE and browse the titles/abstracts of papers. Also look for seminal papers on something that interests you, and see who cites it, find interesting abstract/titles, etc. Then find out where were these papers published and search for more papers there. Maybe keep a note of the good publishing spaces in the areas that interest you and check up on them once a year or so. I've right now got around 20 papers/books open in my notebook, only 1 was got from arxiv, around 10 from conferences in the area and the rest from google scholar or references. However, a note of warning: most papers I've on software engineering and horribly dull, uninteresting, and unscientific. Take a look, though, at experimental software engineering conferences, because they seem to be producing very interesting things and slightly improving the face of the field.