I copy/paste what I think are the "atoms" of the thing to learn. And i build a small graph of interconnected .md files in Obsidian.
To me, it is very important to break the medium that brings the information into a graph that suits you the most. (call that process my knowledge ETL, and the result .md files my personal knowledge graph :)
PS: the funny thing is that you indeed refactor your graph when you come back to it with more knowledge. You usually do not remove things, but reorganize it in a better way. Very interesting process...