Debating the usefulness of atomic notes: a novel, pragmatic, obsidian-based approach to PKM strategies

Hi all,

in the last months I’ve come to question whether atomic notes are actually a rational choice in the obsidian environment. I didn’t find a lot of debate about their effectiveness; on the contrary, they’re usually proposed as a clever solution.

This opinion actually stems from my personal experience with evergreenish notes, which led my past self to create hundreds of notes in a increasingly specific manner, to the point where there was little text involved apart from the 2-lines long titles.

This is not necessarily everyone else’s experience (yet), but I feel that this problem is intertwined with the impossible definition of “atomic note”, and eventually everybody would run in the unresolved atomic question.

So, the question is, why atomic notes in the first place?


As far as I know, this is a decades old solution originating in the Zettelkasten method, to help us create meaningful connections in our note base.

It is still a pre-digital era solution though, and while Obsidian lacks an outliner or block reference capability (which circumvent the need for atomic notes, although in an unelegant manner), it is nontheless capable of doing lots of things paper based zettels can’t.

…Backlinks being the main one.

As most of you know, thanks to backlinks, a file is pretty much a tag with its own page where we can write stuff. This actually means that when we create a link, the reciprocal backlink already points with perfect precision to where the concept related to that page is, that being, to what you’re writing right now. It doesn’t actually matter if it sits or not in an atomic note.

The problem atomic notes try to address lays at the opposite end of the connection, being that we can only directly link with the degree of precision allowed by how subdivided are our notes.

:dart: e.g.
If I have a single note about the whole anatomy, I can’t direct link to “kidney” if needed, however if I link something else from the kidney part of the anatomy page, the reciprocal backlink will point to the precise argument by virtue of its position)

So I started fragmenting more and more my notes, but the reality is there is virtually no end to how far you can go, and it’s difficult to think ahead to what degree of precision you will need in a particular topic.

But, what if we didn’t try to shape the note around an ill defined element like an atomic concept? What if we turn to something simpler, like the number of characters or likewise metrics?

A note, in the end, is just a container. I can pour my knowledge in whatever kind of note, it is not going to change because of that, and as long as I put a meaningful label on every jar I will find it when I need it.

In this setup, instead of declarative titles which are supposed to capture what the note says in one line, I limited myself to describe what is inside (e.g. social impact of HIV). This wuold even result in different topics sharing the same note (e.g. social impact of HIV + developement of mRNA vaccines)

You might think it’s a mess, but as long you stay within the character limit, you can link to that note from wherever else, and when you read it you know what you’re referencing, because it has a decent resolution (meaning my notes are enough subdivided to allow precision linking) and it takes almost no time to read it.

This is possible in the long term because obsidian allows to merge and split notes too. Although not everybody does it the same way, so I will leave my method here.

  1. if a note title declares a scope that overlap with other notes (e.g. epidemiology of car crash related trauma and incidence of car crash injury) you straight up merge those notes.

  2. if a note grows larger than the character count (or whatever metric you’re using) you split part of that note in a sub-note, which means to leave a link behind. This still allows that note to be referenced by the old links to the parent note.

  3. if a note with 2 or more subnotes does not have any content of its own left, you can delete it from the vault with this workaround: rename it to sub-note 1]]; [[sub-note 2. This will effectively destroy the links to the parent note, leaving behind a list of direct links to the subnotes. (I would also throw in a tag to help remember your future self that this action was automated and some link may need to be eliminated to increase accuracy).

This whole process might become more frictionless in future, if it’s possible to get Note Composer to auto-update links [1], [2]

The ability to split and merge effectively allows a note to grow over time in this kind of cycle, which ultimately leads to “mytosis” and terminates the note itself

In opposition to atomic notes, this highlights the difference between the container and the knowledge itself.

A direct consequence on how I organized my notes also needs to be noted: they are now categorized in “pages” which have a short title and acts as entry point for a topic via the backlink pane, and sort of work like tags, and “notes” which have a long title as discussed above and actually contain informations.

For all pourposes each page has the potential to become a little MoC, although all the relevant information is already avaliable indipendently of how much energy I put into organizing it. As a result MoCs are slowly being namely reconverted to pages of broad topic, which they are.


From my personal sperimentation, chopping up some class notes in paragraphs of defined lenght and quickly describing what they are about is a lot faster than individually trying to define how to group the topics in atomic notes (which is always an approximation anyway, but takes time to be constructed in an acceptable one).

Of course I would try my best to minimize fragmentation of topics, but you can’t respect both rules at once. The note is not supposed to last forever anyway, and it will eventually split as soon as a topic gains enough weight to become a note of its own.

The raw notes are revised via the backlink pane from the various pages at a later time, identifying possible merges and adding connections between notes.


In my view there are at least two straightforward argument against this workflow:

  • It is possible to argue you could do the same thing with atomic notes, as long as you keep an eye on the character count, to which I respond:

    1. focusing only on the character count keeps you from creating one-liner notes before it’s actually necessary, which in my experience always defaults to efficiency loss.
    2. you would still need to drop the declarative titles, which are part of atomic notetaking, in order to efficiently merge notes, which is kind of mutilating the original intent of atomic notes in the first place.
    3. If implementing this workflow, the value of defining ahead of time a topic is greatly diminished.
  • Another possible objection might be in regard of the degree of automation in creating new entries into the permanent stash of notes. Most people here have a multi-step workflow to filter what they want to keep in order to maximize insight.

    • This however is not desiderable in some use cases, possibly a consistent minority, especially when dealing with large quantities of trusted informations, such as when reading from a textbook.

In the end this is personal knowledge management and it is not my goal to change anyone’s mind; however the lack of proposed alternatives prompted me to make a post about it and ask you all: How would this apply to your way of using obsidian? Which role did atomic notes play until now for you? Could this improve your experience?

I tried to come up with something readable, but as you might have guessed I’m not a native speaker. Sorry in advance.


Atomic notes are an old answer for improving the meaningfulness of connections between notes. Obsidian-based notetaking gives us enough tools to change this paradigma resulting in both improved precision and faster workflow. A graphic example of this kind of approach


This is based on a (common) misconception that the zettelkasten is for “making connections” or “gaining insight.” A zettelkasten is first and foremost a writing tool meant to create an external, notes environment riddled with trains of thought that can be pulled from and expanded on in written content. The entire point of the short-ish, single-idea note is because it’s easier to connect a single idea to many other single-ideas, then it is to connect a complex idea to other complex ideas. If you’re objective isn’t to become a publishing machine (Schmidt), then there’s no real need for atomic notes or the zettelkasten. If your interest is primarily in “making connections” or examining ideas, etc, then atomic notes really can be set aside.

1 Like

Thank you for your reply. The sentence you quoted is a bit overcompressed by trying to convey in a few lines the message of the whole garbled stuff above, but it was not my intent to discuss the zettelkasten method in particular. In fact, the line of reasoning I tried to push here is based on the idea that atomic notes are not actually limited to the zettelkasten use case you describe here.

They are usually applied to a larger subset of systems and vaults (evergreen notes probably being the biggest example) because it is difficult to find any alternative at all, even if someone does not intend to create something zettelkasten-like in the first place, probably in light of the misconception you talk about.

This is because, at the end of the day, whatever it is you’re doing with your notes, you can put your information about something only in one file; this is a root problem everybody is going to run into. The moment you realize something should go in two files at the same time, you either copy & paste, which is obviously not the correct solution, start transcluding (which is rarely sustainable on the long term) or end up with atomic notes, which incidentally are a great solution to solve this particular problem, but come at a price.

From this point of view, setting aside atomic notes, as you suggested, can’t possibly mean reverting back to longform notes; I wouldn’t need a computer program to do that and it would solve 0 problems; we need an alternative to that.

I would further argue that the amount of people who end up “atomizing” their notes in any contest other than creating a zettelkasten for writing purposes is not a small minority of users, as, as you pointed out, far from everybody here is trying to maximize publication output. The idea I present here is mainly directed at them, as it’s closer to the use case I work with.


With modern computer systems, and the ability to set keywords and links, longer notes are not the “problem” they might have been years ago because you can use search to find any specific phrase. Strict File based systems make lots of users feel good, and that’s fine, but really not necessary. A loose filing system for “neatness” is much more useable.
Its not the 1950’s now, we are doing things digitally. Much of our “systems” are based on how Libraries organize knowledge. They used to have stacks of file cards to find their content. Families used to have a Huge bookcase with an Encyclopedia to reference. Now Libraries no longer have reference cards, but use this digital thing called a computer :):slight_smile: and Encycopedia’s are now available on line :):):slight_smile: I wish I had this when I was in school. Granted the library and encyclopedic references were vetted and assumed to be accurate, now its up to the users to vet internet info, but its still a LOT easier to learn almost anything, and research is much easier. Our note making best practices need to be updated to current technology. Its a fascinating time and I thoroughly enjoy the ability to discuss this with all, I always learn something new.


Atomic notes as a methodology, primarily in academic research, go back over a century; the zettelkasten is simply one example.

The definition is not impossible, and it’s a very functional system and way of thinking.

You have confounded it with a digital implementation, and one associated with a distorted perception of Luhmann’s zettelkasten method which is further distorted by being applied to a range of uses it was not designed for (eg student learning) and even further by artificial and automatic linking systems. The zettelkasten method is an aid to thinking, and writing, but not necessarily to learning or memorising.

The idea of atomic notes is independent of whether they are all contained in one document or file or in many. What is important is that they embody a single concept or a single piece of evidence. Social impact of HIV is neither a single concept nor a single piece of evidence; it’s a topic or a set of other ideas that may be related to the topic. It reflects the way many subjects are taught, but does not reflect the evidence or the methodologies of the research that are the base of that teaching.

If you are arguing that atomic notes and zettelkasten are inappropriate approaches for students who are being taught, then I would agree with you - the teaching is pre-digested and structured with links already made; splitting it up risks losing the relationships the teaching was intended to highlight. The approaches are more suitable for those who write the courses, and even more suitable for those doing the fundamental research.


Thank you for your reply! In my experience search is far from an ideal tool to quickly get the information you need, both online and inside personal notes.

I would love to digress on the former, but it’s sufficient to say search engines didn’t really open science to anyone. In fact, a well curated review or meta-analysis is still something worth of its own article, and there are a bunch of costly, widely used paid services in various fields (e.g. UpToDate or WestLaw) that basically provide notes and selected literature / case law to their subscribers.

On the other hand, it is common opinion that, if my professional depth had to largely be determined by guidelines provided by a single private actor, I might as well be replaced by a trained monkey, which would both be less grumpy and cost far less bananas.


Talking about our notes on Obsidian: at this level, I appreciate your open mind, but I’m not entirely sure a case can be made for longform notes:

If we had to consider the search tool by itself, there are a number of problems I encountered in my past experience:

  1. it is not an automatic process, you kinda have to type in your query and get your result each time for each specific concept, then review all of the results to find what you’re looking for, and if you need to do the same thing tomorrow you have to repeat the entire process.
  2. A lot of stuff has two or more names across different sources. Sometimes in different pages of the same source, which is a major cause of headache. If you get unlucky and encounter the same concept from two points of view, let’s say, 6 months apart, there’s a decent chance you’re not gonna remember, use two different terms, and might take a good while to realize, and even more pain to repair.
  3. The search tool is not “conclusive” enough. There might or might not be something left. The more certainty you need, the more broad the search terms, the more results you’re gonna need to spend time reviewing.

So, at the end of the day, I am gonna need links, because they provide a stable and immediately apparent relationship between two concepts. If links are to be used, we also need files (well you could use ^blocks but isn’t that more work?)

If we need files we also need a way to organize them and work with them, which this post is about. Now, I’m not saying this is the only method; probably most of the problems with search could be resolved (maybe tags can help?) and someone will eventually find a way, but it’s not as seamless as just opening the search tool. I’m just trying to say: hey, links are fine, but atomic notes aren’t made for this and you might want to consider an alternative.

Thank you for your reply. Starting from the end: yes, the student use case is certainly a majority and my main argument is against using atomic notes at that level, although I disagree with the way you expressed:

since this is far from my reality of being a student. Even if spelling word for word what a professor says was enough to pass exams (it isn’t) it wouldn’t be anything that reading a book over and over wouldn’t get me anyway.

As we move across different subjects and we encounter the same concepts over and over again, it’s natural to make connections across what we already learned and what we are learning right now.

And maybe I’m overusing the term “making connections” which probably sounds really poetical and detached from reality, but being able to retrieve in half a second your work on a specific cytokine over the course of 2 years, being able to get a look at which pathogenetic mechanisms are involved with it, and being able to integrate new informations in the same place, instead of writing stuff from scratch, is immensely helpful, and not something a teacher or textbook will provide for you.

But I get that this is a totally different kind of “connection” from what a researcher does, I have no experience in research and, again, I’m not trying to appeal to use cases I know nothing about.

However, this sparks my curiosity, as I would object that there is no concept I can think of that could not be further subdivided into two related concepts. Being this your definition of atomic note, would you provide me with an example? Sorry if I try to digress here but it’s part of why I rebut atomic notes in my use case.

Yes, that’s exactly the idea and why I used that example in opposition to atomic notes, my whole thesis being - there’s not such thing as an atomic concept, only reasonable approximations of it, and here’s how you could do without. - It is also what I’m trying to debate here so different points of view are more than welcome.

Yes, ok, but as I said in the reply above:

so the problem is still here. The “confusion” stems from the fact that people with different use cases are being provided with the same solution (again, evergreen notes tries to do this, but in general you could run a search for “atomic” in the obsidian discord and find plenty of examples).

Let’s say atomic notes are not designed to learn (which, I say again, it’s not a niche use case of what we try to do with obsidian, in this community), then how are we supposed to split and organize our files? To my knowledge atomic notes are the only solution widely discussed for ALL use cases, not only research implementations. (apart from the apps with outliner capability, which have their own shortcomings).

@Dat - I absolutely love the conversation you’ve started here, and I’ve reached the same bifurcation in my own system: atomic notes are starting to become titles only, and while this makes them easier to link, they lose meaning / context.

I’ve not solved for it, but this internal debate has pushed me to try other apps to see how they solve for it, and here are the gaps that I hope at some point the Obsidian team or the plugin community will solve:

  1. Robust blocks / headings embed: the fact that atomic notes end up being title only is a result of Obsidian’s lowest unit being the note. Yes, you can link to a note’s blocks or headings, but a small change in those headings or blocks breaks those links irreparably. I don’t know exactly how doable this is, however, given Obsidian architecture built on markdown files.
  2. Blocks drag and drop: I find that referencing a block from another note interrupts my flow; intuitively, I would love to to just select a block from the origin note, drag it to the destination note to create an embed (with modifier keys doing things like copying or moving / referencing). The drag & drop plugin would be phenomenal for this (and thanks @artem for developing it!), but it’s not reliable (at least for me).
  3. Blocks / Headings Queries: I wished there was an easy way to query blocks (e.g. blocks with a particular tag, like you can do for tasks today with the tasks plugin) or headings (for instance, my thoughts about my son are on random daily notes under a heading with his name; each one of those thoughts is atomic, but if I could build a note with a query collecting the content of all those notes, where I could then drag and drop single blocks out of a query - see 2 above - that would be phenomenal! It is possible with other apps, but I love Obsidian too much).
1 Like

You can query blocks with a particular tag already. Officially, tags apply to the whole note. But in practice, search by default shows each match and its context, and you can set the view to expand the context to the whole block.

1 Like

I love this comment. I think you and @Dat have correctly identified a fundamental tension in Obsidian’s design: the basic unit is the Markdown file, so anything more granular—headers, blocks, sentences—is not going to be as seamless of an experience. Linking to headers or searching for blocks is fragile; these are treated as parts of files, not entities in themselves.

This makes sense when considering Obsidian’s philosophy, but it comes with its tradeoffs. I’d love to see improvements to the within-file experiences.

The “contextual backlinks” (the block in which a page was linked) is a use case that I hacked together a Dataview panel for: Show "contextual backlinks" in Obsidian using Dataview and CustomJS · GitHub if you’re interested.

(Another manifestation of this: the backlinks pane indiscriminately links every mention. There’s no way to exclude, e.g., a mention that you don’t need to link because it’s linked earlier in the block. There’s no way to exclude incorrect suggestions that come from name collisions.)

  • Different minds have different abilities to mentally juggle different chunk size and volume of information. I know one writer in our Obsidian space who writes his missives (some longform) in one huge note, using nothing more than sections and only the core search plugin. Many others spend an inordinate (to me) amount of time fastidiously organizing (folders, tags, and atomizing but, interestingly, very little use of internal links).
  • Of the several hundred folks I know that use Obsidian, only a small number use blocks; an example, I see the same small group of us pouring over the forum posts regarding this area:)
  • Canvas has brought this subject to the forefront for me. I use canvas to toss stuff onto to ponder. For quickly rendered output content for class examples, online posts, and other correspondence I don’t use internal links; just select and drag content onto the canvas, then delete when done. For long form or longer pondered topics, I do bring in internal links; I need them to jog my memory or inspire new thought.
  • I doubt the necessity for strict atomic notes; Luhmann didn’t have our modern tools. Because internal links are not standardized across the PKM realm, they’re not fully implemented or cross compatible. However, given the minds behind the development effort, I am confident we’ll get there.
1 Like

I’ve been wrestling with a similar dilemma, with similar reasoning. Sometimes it makes sense to make a freestanding brief note. But it often doesn’t make any sense to me to atomize a longer note purely for the sake of atomizing it, when I can already find all the information in it due to the digital nature of the vault. It seems to me a little like trying to recreated a card catalog in a library catalog – making numerous individual entries for the same piece of content – when you could have all the same information on a single book page and it’s just as searchable.

You can use block references in Obsidian, though. They’re not as easy to use or feature-rich as linking to a heading, which makes that a better option in most cases, but they’re there and I’ve used them under certain circumstances.

I find it useful sometimes to break up notes if I’ve just been piling stuff into a single note — for example, pulling specific “how-to” notes out of the note for an app so I don’t have to dig thru the note to find it. And I definitely prefer normal links to block links, so that can be a factor (tho I do use blocks for some things). But at this point I’m less likely to pile things into a common note like that than I used to be.

If we could link to blocks of text to within a file, and backlinks autoupdate not just on renaming the file, but also when that block is cut and pasted into another file, as Obsidian currently does with filenames, then this dilemma goes away.

Because then we get to decouple a “note” from a “file”. A file just becomes another layer of organization, one level below folders.

So from the conceptual POV of networking information, this:


# Idea1

Idea 1 connects to [[File2#Idea2]]


# Idea2

Idea 2 connects to [[File1#Idea1]]

is the same as this:


# Synthesis

## Idea3
Idea 3 connects to [[#Idea2]]

## Idea4
Idea 4 connects to [[File2#Idea2]]

of anything in between (e.g., single-concept note files linking to particular blocks in large aggregating note files.

There is no compromise in precision, and it comes does to personal preference for metaorganization (i.e., non-semantic organization for book-keeping rather than for note-keeping, if that makes sense). The preferred “packaging” of the ideas can vary depending on the content. Large files (not notes!) for many small related ideas (=“notes”), and small single-note files for for developed (not necessarily more complex!) ideas, e.g. including references etc. And, of course, change freely through the evolution of the knowledge base, as we splice and dice and collate and revise ideas, moving some into aggregating files and splitting other into individual files.

We can sort of implement this already using lists, due to the ability tag individual list items with metadata etc., and of course, Obsidian block reference syntax. But it is a little clumsy.

So we can have our atomic notes AND use files as thin wrappers.

Another advantage of the above approach (once supported better by Obsidian) is solving the tension between optimizing organization for capture vs. optimizing for retrieval. When capturing, there is less friction if I can just open one document and dump my thoughts there, to be refactored later. Right now, the refactoring — whether into atomic notes or larger chunks — is a necessary evil / busywork that has to be done very quickly or otherwise the ideas cannot be linked usefully (due to the lack of precision issue). With stable, robust, and dynamically-updating block-level or line-level linking, we can start linking our information right away in the big capture dump. Then refactor as we go along as needed as part of the flow, rather than switching to clerical mode until we can get going.

One use-case for a multi-“note”/idea file that confers semantic benefit is the ability to organize notes into a sequence and hierarchy. Now, I suppose that the spirit of some note-taking philosophies decry these and a lot of what happens in “smart notes” is to actually break information out of sequential/hierarchical packaging! And definitely that is something that we do not want to stop doing! Also I love the Zettlekasten idea of supporting multiple parallel hierarchies (through tags, for e.g., we can place a single idea in: #topic/science/geology, #topic/science/statistics, #status/pending/review``). But at the end of the pipeline, when synthesizing more complex ideas, it is useful to organize these atoms into molecules and polymers.


Is there any indication from Obsidian devs (who are brilliant and have built an amazing tool) that block linking will be advanced beyond its current state?

New Obsidian user here, just a couple of days. Has anyone considered modeling their vault in a graph database like neo4j and using NLP and link prediction techniques find more relationships between their notes? I did see a plugin where someone was dumping the files to a graph, but then moved to Cytoscape.

Anyway, I’m really enjoying this app!

Thought about? Yes. A future project is implementing NLP, but vague thoughts and limited skill and time → inertia (so far).