Others on the page here (specifically @Dpthomas87’s A, B, C) have done a great job at outlining their methods which I’m generally following. So I’ll focus a bit more on the mechanics.
I rely pretty heavily on Hypothes.is for most of my note taking, highlights, and annotations. This works whether a paper is online or as a pdf I read online or store locally and annotate there.
Then I use RSS to pipe my data from Hypothes.is into a text file in OneDrive for my Obsidian vault using IFTTT.com. I know that a few are writing code for the Hypothes.is API to port data directly into Roam Research presently; I hope others might do it for Obsidian as well.)
Often at the end of the day or end of the week, I’ll go through my drafts folder everything is in to review things, do some light formatting and add links, tags, or other meta data and links to related ideas.
Using Hypothes.is helps me get material into the system pretty quickly without a lot of transcription (which doesn’t help my memory or retention). And the end of the day or end of week review helps reinforce things as well as help to surface other connections.
I’m hoping that as more people use Hypothesis for social annotation, the cross conversations will also be a source of more helpful cross-linking of ideas and thought.
I prefer to keep my notes as atomic as I can.
For some smaller self-contained things like lectures, I may keep a handful of notes together rather than splitting them apart, but they may be linked to larger structures like longer courses or topics of study.
If an article only has one or two annotations I’ll keep them together in the same note, but books more often have dozens or hundreds of notes which I keep in separate files.
For those who don’t have a clear idea of what or why they’re doing this, I highly recommend reading [[Sönke Ahrens]]’ book Smart Notes.
I do have a handful of templates for books, articles, and zettels to help in prompting me to fill in appropriate meta data for various notes more quickly. For this I’m using the built-in Templates plug-in and then ctrl-shift-T to choose a specific template as necessary.
Often I’ll use Hypothes.is and tag things as #WantToRead to quickly bookmark things into my vault for later thought, reading, or processing.
For online videos and lectures, I’ll often dump YouTube URLs into https://docdrop.org/, which then gives a side by side transcript for more easily jumping around as well as annotating directly from the transcript if I choose.
I prefer to use [[links]] over #tags for connecting information. Most of the tags I use tend to be for organizational or more personal purposes like #WantToRead which I later delete when done.
When I run across interesting questions or topics that would make good papers or areas of future research I’ll use a tag like #OpenQuestion, so when I’m bored I can look at a list of what I might like to work on next.