A PDF on your desktop is a claim about what was true on the day you downloaded it. No version, no date, no route back. Point at the publisher's page instead and the address survives every redesign, the revision history stays attached, and a reader who wants to check your reasoning can follow it to the top rather than trusting a filename. When a preprint is revised or a result is withdrawn, the sources tell you which of your notes are now resting on something that moved.
Nobody can tell you which corner of a field has gone unread, because nobody is holding the map. Split the sources across strands and the shape of the coverage becomes obvious at a glance: forty entries against two is not a balanced review, it is a hole. It also makes the gap arguable — point a colleague at the thin strand and they can fill it, so coverage improves instead of quietly staying wrong.
Article URLs rot, repositories reorganise, and a link pasted into a document two years back has a fair chance of landing somewhere unrelated now. Add the live page to the Glubler that already carries that source and the entry is replaced wholesale, annotation and all. What you open at the start of a project points at the current record, not at an archived shell of it.
Methods references, datasets and primary literature are three different searches, and bundling them into one Glubler means three times the scrolling to reach any of them. Keep them apart and each question has its own address. A group can nest all of them under a single profile — nothing gets lost, and nobody has to wade through a colleague’s data repositories to reach the theory papers.
Glublers are open unless their owner closes them. Follow the researchers whose judgement you rate and their lists turn up in your feed — a methods index for a study you are scoping, a source group for a field you have just moved into. Take theirs as a starting point and add whichever venues your own question needs; the list grows in both directions.