I have a folder of highlights from about sixty books. I have referred back to it maybe four times, and on three of those I could not work out why I had highlighted the sentence.
This is the common case, and it is not a discipline failure. Highlighting is a genuinely poor note-taking technique wearing the costume of a good one: it produces visible output, it feels like engagement, and it requires no thought at the moment it happens — which is precisely the problem, because the thinking is the part that makes a note worth having later.
Why highlights do not work
They record attention, not understanding. A highlight means "this seemed important for four seconds". Four seconds is not enough to know whether it is important, and by the time you find out, you have lost the context that would tell you.
They are the author's words. Which is why they read as inert later: you are looking at a sentence somebody else wrote, with none of the reason you cared about it. The thing you actually wanted to keep — what it connected to, what you disagreed with, what it changed — was never written down.
They scale badly in the wrong direction. A well-read year produces thousands of highlights, and a pile of thousands is not a resource. It is a second library, with worse search than the first one.
The export is the trap. Getting all your highlights into a notes app feels like completing the system. It is completing the collection, which was never the hard part. Nobody has ever been short of material.
The only reading note that works
One question, asked after you finish something: what would I tell someone about this?
Then write the answer, in your own words, in a few sentences. That is the whole method.
It works because it forces the two things highlighting skips. You have to have understood it, since you cannot summarise what you did not follow. And you have to compress it, which is where the actual thinking happens — deciding what mattered is the intellectual work, and a highlighter lets you skip it entirely.
The output is small. A book becomes a paragraph or two, maybe a page for something genuinely dense. That feels like too little to people who are used to fifty highlights per book, and it is worth more, because you will actually read it again.
What to write down while reading
Not nothing — but much less than the highlighting instinct wants.
The page number and one word. Enough to find your way back to something you want to think about, without stopping to think about it now. Reading and note-taking are different modes and doing both at once damages the reading.
Anything you disagreed with. Disagreement is the highest-value thing to capture, because it is where your existing model met a different one. It is also the thing that evaporates fastest — you will not remember, tomorrow, that you thought chapter four was wrong.
Anything that connected to something else you know. "This is the same argument as X." Those connections are the entire reason to read widely, and they are ephemeral.
Quotes, rarely. Save the exact words when the exact words matter — a definition, a formulation you could not improve. That is a handful per book, not fifty.
Cornell, and where the old methods still hold
The formats that survived from an era before software mostly survived for a reason.
Cornell notes split the page into a wide column for notes taken during, a narrow column for questions written after, and a summary at the bottom. The reason it works is the second pass: the cue column forces you to reread and ask what each part was for, which is retrieval practice rather than transcription. It was designed for lectures and it is equally good for a dense chapter.
A summary at the bottom is the part to keep even if you skip everything else. It is the same "what would I tell someone" question, given a fixed place to live.
The mistake people make importing these into software is reproducing the visual layout — two columns, a box — and dropping the discipline, which was never about layout. A plain note with your summary at the top does the same job. Layout is the least transferable part of any method.
Where reading notes should live
Not in their own app, and this is the practical part.
A reading note is only valuable when it turns up next to something else — when you are writing about a topic and the thing you read in March is right there. That happens if the notes are in the same collection as everything else you write, and it does not happen if they are in a dedicated reading app you open when you are being a Reader.
Two mechanics make the difference:
They have to be findable by content. You will not remember which book said the thing. You will remember a word from it. That is a full-text search question, and it is the whole retrieval story for reading notes.
They should link to the things they relate to. A reading note that mentions [[a project you are working on]] shows up on that project's page without you having filed it there. This is the one place the linking habit pays off fastest, because reading notes are exactly the material you cannot predict the future use of.
The honest version
If you are doing academic work — citations, page-accurate references, a bibliography that has to be correct — you want a reference manager, and Zotero is free, open-source and very good at exactly this. Nothing in a general notes app replaces it, and trying to run a dissertation out of a folder of paragraphs is a bad idea.
If you read to think rather than to cite, the method above is most of what there is, and it is unusually low-tech: one question, a few sentences, somewhere searchable.
Cyanote has Cornell and reading-note templates in the / menu, and both are meant as starting shapes rather than a system to adopt. Reading notes sit in the same collection as everything else, searchable by any word in them with ⇧⌘F, linkable with [[ so they surface next to whatever they relate to. It is all one database on your own disk, so a decade of reading notes stays yours regardless of what happens to any app — including this one.