Principle of Atomicity - File-based Zettelkasten vs. Roam Research

Adapted from my original post at [forum.zettelkasten.de](https://forum.zettelkasten.de/discussion/1072/principle-of-atomicity-file-based-zettelkasten-vs-roam-research#latest)

I’ve been using a zettelkasten (Sublime Zettel) for more than 2 years. It has over 2000 notes in it covering diverse topics from neuroscience to cooking to clinical medicine. It’s worked well, and as it enlarges, I get more value from it. But when new and interesting tools come along, I like to check them out. I know Roam Research has gotten a lot of attention around here and around similar corners of the web. So I started playing with it a few weeks ago.

I think one of the most striking and important differences between Roam and Markdown-based systems is that in Roam, the atomicity is even more granular. The fundamental unit is the bullet or line of text. Every bullet has a unique ID that allows it to be linked to anything else.

In Markdown-based ZKs, like mine, or those that use the Archive, the zettel is a markdown file, which usually has several lines of text in it (and sometimes a lot of text). This is helpful in that it consolidates related sentences but it doesn’t allow the level of granularity that Roam does. Sometimes I really do only want to link to a single sentence and not an entire note. Roam allows that. A markdown-ZK could allow this if one made zettel-files line by line (one line per zettel), but this becomes really tedious, mostly because you have to explicitly title each zettel file.

In contrast, in Roam, you can just open up a Daily Note and start writing. There is no cognitive overhead in having to title a zettel. You can basically do a brain dump and if you didn’t feel like it, you could leave that brain dump right there and never move it or properly place those lines into their own pages (groups of related bullets).

But, then, as Sasha Fast talks, you can build infinite hierarchical “Structure Notes” or whatever you want to call them elsewhere.

So all this to say, I think one of the most enabling features of Roam is the size of the atom can flexibly go from a single line up to whole blocks of hierarchical bullets, without the cognitive overhead of having to create distinctly titled pages.

The other things like the back-linking and all that are cool too, but I think the fact that bullets are the unit of operation is perhaps the most powerful difference. I don’t want to jump onto the Roam Wagon, since it’s proprietary and I like my system, but I’m wondering if I’m losing something by sticking to the file-based ZK.

I would only make 2 observations and let others who perhaps have something more comprehensive to say:

  1. I believe Roam’s most powerful feature (second most to the bullet UUID) is full transclusion across the board, and also when implemented, full transclusion across graphs (databases).
  2. Roam is also built on markdown - so not really proprietary in the usual sense. What is unique (in other words not quite markdown) is exportable as a JSON file, which can then be stitched with some code to make-up for what is missing in the markdown files. I export my .md files into DEVONthink and can use 80% of them without much of an issue.
1 Like

@Calhistorian Thank you for your thoughts. I agree, the full transclusion is a powerful feature and distinct from many other tools. I think it removes the need to find titles for every singular piece of information, and you can see the information in-place in some outline rather than having to move away to new note to see the content that is inside.
However, there are some that feel that titling a note is an important act in itself: https://notes.andymatuschak.org/Evergreen_note_titles_are_like_APIs

In what ways does full transclusion help you with your knowledge work?

To your second point, it’s true that the raw content of Roam can be exported, but not everyone who uses it will be able to create custom code to reproduce its features. So most users are getting locked in. Indeed, already with Roam going to a subscription model, I’ve lost what I put in there. I could pay to get it back, but I didn’t do enough to warrant that. But this loss of access validates my initial concerns. I think it’s a valuable tool that is worth paying for some, but many people (myself included) want full ownership of the content.

Gladly respond how transclusion helps my workflow. Though I don’t have the time now.

To your 2nd point (and mine) you are right that it is outside of many users abilities. Mine as well frankly, but there are already tools out there to download and solve that problem for free.

Also, if you were on the Beta program before being charged (or you cancel after joining the sub) your data and access to that data is theoretically forever. This is the stated policy. So if you have lost access for some reason shoot them a tweet or email.

These are reasons why Obsidian is so powerful too. Of which I use for publishing purposes.