- [[how to take notes]] # Idea ## Core ideas Bottom-up approach. Don't waste time/effort organizing/managing initially. Just start taking notes however messily. One vault is more than enough. Flat hierarchy is also more than enough. Resist the urge to create multiple vaults for different areas of your life—[[premature optimization is the root of all evil]]. Differentiate between two types of notes: [[fleeting notes]] and [[permanent notes]]. Take notes liberally, especially [[fleeting notes]]. Take modular notes. Most notes should/can be super messy. Only the ones you've thought through and revised multiple times will look good. Embrace the messiness. You shouldn't waste your precious life organizing folders etc. (this note [[premature optimization is the root of all evil]] is an example of an incomplete/messy note). It's highly personal what's fleeting/permanent/modular. It's supposed to be personal. It's your second brain! What you think is a modular and permanent idea might be fleeting/gibberish to me. When taking notes, make it a habit to constantly connect ideas while you type with `[[` connect with other notes or existing concepts in your vault. You'll see this behavior in this note itself. Rely mostly on local graph view instead of global graph. After you get comfortable with the above, then start exploring fancier stuff like templates, tags, properties, YAML headers/metadata (`aliases` is super useful [to give each note other names that help you retrieve them easier later]), community plugins (e.g., AI autocomplete etc.). ## Details The goal of note-taking isn't to capture a static state of the world. It's not at all about capturing (see [[collector's fallacy]]). It's to help you capture ideas in a way that makes it easy for you to work and play with and combine them into new ideas and insights later. It's to help you think better. Your notes should help you think about what your care about, develop those ideas, and come up with new ones. To think with your notes, you have to constantly edit, revise, and link them to other concepts as your understanding evolves (like how writers constantly revise their prose, because the very act of writing helps them think). So your notes should never be static. - I change my note titles/names all the time. - I have two types of notes: [[fleeting notes]] vs [[permanent notes]] Raw ideas are supposed to be messy. Embrace the messiness and take a lot of [[fleeting notes]]. - Many of my notes are just empty placeholders for ideas I want to consider in the future. - Others are long, rambling thoughts that I don't bother distilling/summarizing at the moment (but when I need them in the future, I'll clean them up). Take notes liberally, and in a [[modularity|modular]] way. One note should contain only one distinct idea or concept—you decide what is one "idea" or "concept." - What is important is that the note remains a cohesive building block that can be easily linked to other thoughts/notes. - What is considered one idea/concept will evolve over time as you think/reflect on this idea and related ideas (similar to [[memory chunking]]). Initially, don't bother categorizing, organizing, or tagging notes. This includes not using nested folders or any kind of hierarchy or management system. Resist the urge. - I have >9000 notes in the same folder. I never look at or open this folder. - I have an `attachements` folder for all images/videos/pdfs inside my vault. - I have only one vault for everything (actually, two vaults—the other one is personal journaling, so it's less about knowledge/ideas). Merge and connect [[fleeting notes]], and when new ideas/concepts seem to emerge or when you see clarity in your thinking, transform them into more refined [[permanent notes]]. - Rename fleeting notes into permanent notes. - Create even more fleeting notes and permanent notes (because concepts/constructs/ideas often have to be broken down). ## Example Say I start reading this article by [[Paul Graham]]: [Superlinear Returns](https://paulgraham.com/superlinear.html). I'll start a new note titled `Graham 2023 superlinear returns` (see [[Graham 2023 superlinear returns]]. I always have local graph opened. Immediately, even before reading, I think of exponential growth (an existing note), so I link it immediately at the top. ![[20251226111624.png]] I see a potential modular idea/argument: "returns for performance are superlinear" so I try create a new potential permanent note immediately. Guess what, immediately, as I start typing `[[returns for performance`... a note with that idea immediately popped up (surprise even for me as I write this). I took that note in 2023! ![[20251226111701.png]] ![[20251226111707.png]] Immediately after linking that note, my local graph has expanded because of the ideas/notes I thought about over the years. I immediately can see relationships between this Paul Graham essay and all my other ideas. ![[20251226111732.png]] I continue reading and somewhat clean my notes a little along the way... ![[20251226112008.png]] He mentions "winner take all" and unsurprisingly, I already have a note for that taken in 2020. I link it and edit my note a bit. ![[20251226112339.png]] He claims that fame combines both sources of superlinear returns. I'm like, hmmm.. really? Need to think more about it (before I think it's a real/legit/modular/permanent idea/concept). It's fleeting thought/note, and I capture it as such (but you might think it's a legit/permanent idea—it's subjective/personal!). Important point: I prefix my fleeting note titles with yymmdd_hhmmss (see [[251226_112501 fame consists of exponential growth and winner-take-all effects]]) ![[20251226112559.png]] And so on... If I decide to keep working on this note [[Graham 2023 superlinear returns]], I might start looking at my local graph, see what ideas they connect to, merge different notes or delete old ones when they become irrelevant/merged, rename notes, and keep working on related fleeting notes. When fleeting notes become substantive enough, I rename them and remove the timestamp prefix to signal their transition into permanent notes. And with just a few connected notes, my local graph already looks like this. Lots of ideas... ![[20251226114813.png]] I have >9000 notes in my vault in a flat hierarchy, and I can always find relevant notes because my notes are highly connected. To retrieve old ideas, I do the following: - Use the "quick switcher" command to search for the note name. - If I find related notes but not exactly what I want, I look at the local graph to see if that thing I'm looking for exists elsewhere in the network. - Use the global search feature. ## My own "top-down" system over time Something else might work for you better. But top-down structure evolves over time (since this approach focuses on bottom-up note-taking to let ideas and structure emerge organically over time). Because of the flat hierarchy, I rely on (re)naming conventions and internal links to maintain order. - projects notes have `project - ` prefix - [[project - llm classifier]] (check it out) - [[project - echo chamber perception]] (unpublished so you can't see it) - after a project is done, i archived it by renaming the note - [[archived project - effort training]] (unpublished) - fleeting notes (as noted above) has `yymmdd_hhmmss` prefix (Obsidian actually helps you insert this prefix easily with a keyboard shortcut!) - courses I take have `course - ` prefix - [[course - cognitive modeling academy hamburg]] - conferences have a prefix - [[conference - SPSP2023]] - and so on... you get the idea - structure/organization emerges organically later (when I have lots of notes and see the need to impose some structure structure) rather than being imposed/predetermined at the outset with clear folder hierarchy/tags, etc. If I change my structure/conventions later again, I don't bother rerevisiting old notes to update them; I simply apply the new system going forward and let the old notes exist as they are until I naturally encounter them again. ## Flat hierarchy examples Flat hierarchy with a lot of fleeting notes: ![[20251226114305.png]] A lot of permanent notes: ![[20251226114352.png]] # References - [[how to take notes]]