# The Folgezettel Conundrum

• This is a really great article! My (too long) response below details some of my thoughts on the subject, and ends with the surprise twist that I might be coming around to the idea of abandoning sequence IDs in favor of date IDs.

@ethomasv said:
Let’s say we are adding a new note related to 3b, 4c11, 4d, and 7a10. We decided to place it after note 3b, so it becomes 3b1. Although it could be positioned after all of them equally, we will store other connections as references.

How truly equal are these connections? Especially in the moment, I may be more interested in 3b as a connection than 4c11 or 7a10. I don't lose out on those connections, since I can link to them in the note itself. I also will link 3b1 to 3b so that I can note down why I put 3b1 after 3b. What was the connection that lead me to choose 3b over the rest?

The downside of this is the fact that with time we will develop this primary sequence more and more while neglecting the other three. Next time we branch inwardly from 3b1, a new note will be 3b1a and it will be related to the overall context of the 3b sequence.

I don't know if this is a foregone conclusion, especially if notes are written to be atomic. If written correctly, 3b1 does not require me to read 3b to understand it. 3b is one possible context, just as 4c11 was another possible context. In both places, the new note will stand alone. Its location is a decision of a moment and should not, I think, be a defining decision. If there is something else to say about 3b1 in the context of 4c11, then I can create 4c11a that describes that difference, which will link both to 4c11 and 3b1.

If note's location changes its contents, then I need to make those different contexts into different notes. If its location doesn't change its content (i.e. it really is equally related to 4c11 and 3b) then the locational choice is arbitrary, and so long as I add links then I'm not losing out on anything.

Linear arrangement of notes in the form of an array or a list can deceive us into thinking that each note has only one parent - note that is one level above in the sequence.

I think that this is a very important point, and something to strongly keep in mind while making new notes and looking through old notes. When I place a note in a sequence, I am not imbuing that sequence with meaning through the placement alone. That sequence only has meaning through its description in the note contents. If I don't go that extra step of writing out the connection, then my future self won't be able to understand the sequence, and it will collapse to meaninglessness. But, if I'm writing down the reason for each connection and adding a link, then I'm also not losing out on the meaning of the connection to 4c11. The two connections have been described in the new note, and so I can still create the radial branching that you discuss.

This could be difficult to do in a physical ZK, as I would have to flip through cards in the sequence that I place them. In a digital ZK, though, I can add any number of links. I never run out of space. I also don't have to encounter notes in the order that they're sequenced. If a note links to 3b1, I don't have to first find 3 and then 3b and then 3b1. I'm not forced to look through that sequence. I can skip straight to 3b1, and then its link to 3b is spelled out in the note itself, as is its link to 4c11. In practice, the 3b location is not given favoritism in the contents of the note.

At the top of each sequence note, we add a short description of what topic or question the sequence is tackling.

I am definitely coming around to the idea of sequence notes. @sfast showing how easily a note can be created made making sequence notes a more attractive prospect for me. Sequence notes are useful even when using folgezettel, though. If 3->3b -> 3b1 is meaningful in a way that links alone won't fully capture, then a sequence or structure note is important. Similarly, if 4c11 -> 3b1 -> 3b1a has its own meaning, distinct from 3b -> 3b1, then that needs a place in a sequence or structure note. Folgezettel are not useful if these other tools are ignored.

That said, our brains are wired to recognize patterns, and to even find false patterns where there are none. Folgezettel feeds into this, by giving an obvious pattern to follow which your brain might see as more important (since it is more obvious) and thus bias you towards that sequence over the other possible connections.

Date IDs are a way to remove your ability for pattern recognition in this manner, forcing you to explicitly create meaning through the structure and sequence notes and your links.

I think, though, that pattern recognition is an important tool in our arsenal. If I have several notes on optical thermometry, I put them in a sequence: 3b -> 3b1 -> 3b2. I put them in this sequence not because reading 3b before reading 3b1 is important--each note is atomic and can be read as a standalone note. I put 3b1 after 3b because 3b1 has to go somewhere in my slidebox, and this seems like a reasonable place to put it since the two notes are on a similar topic.

I've tagged these notes with #optical-thermometry, so if I'm looking for an optical thermometry technique I can easily search the tag and find them. So, in that way, it doesn't matter where those notes are in my ZK. Even if I didn't tag them, I could search for "optical thermometry" and probably turn them up pretty easily. Once again, it doesn't matter where I've put those notes.

What if I'm not specifically thinking about optical thermometry? I'm looking through my x-many thousand notes for links to a new note that I just created. The pattern recognition portion of my brain snags on a cluster of notes about optical thermometry, and I'm suddenly reminded of a link that my new note has to optical thermometry! Or, this recognition spawns a new note that describes how optical thermometry might be combined with whatever the subject of the new note is. If these optical thermometry notes were split up by date, my ability to be reminded of that (non-obvious) link would only be as great as my ability to pick an optical thermometry note out of a list of seemingly disparate topics.

I think, here, that it is possible to have my cake and eat it, too. Everything that can be done with date IDs can also be done with sequence IDs. Yes, it is very important to not rely on a sequence alone to imbue meaning. If a sequence has meaning, then it needs a sequence or structure note describing that meaning. At the same time, building up clusters of notes can be useful for stumbling across topics that I have forgotten about, or finding non-obvious links. Alone, in a list of disparate topics, these notes may not stand out.

That said, now that I've written this out, I can also see how this might run into an issue of nomination. By clustering my optical thermometry notes, I am defining that as a topic worthy of clustering. The larger a cluster gets, the easier it is to find a place in that cluster for a new note. As you say in your article:

We can’t predict what branch will be important to us in the future. Therefore, we must keep all branches available for further expansion. We decided against categorization by topics, but imposing categorization by favoring certain references is just as bad and limiting.

Which, I don't have an argument against. I think it comes down to how well I believe I can separate the knowledge a note contains from the note's cluster or sequence. How much do I believe that clustering notes like this removes my ability to think of them as discrete entities, divorced from their sequence? Will I lose out on making new connection because the pattern recognition part of my brain doesn't catch how a standalone note could be useful, since that note doesn't stand out in the middle of a cluster of notes? It comes down to how confident one is in their ability to recognize patterns and then divorce the individual bits of knowledge contained inside of that pattern from the pattern itself.

I think this might be pushing me a little bit towards abandoning my sequence IDs... I'm not sure. I'll have to play around with this. You may have converted me.

Looking back over how many words I've written in response to this article, it's a little funny for me to reach the conclusion that folgezettel might actually detract from the use of a ZK, since I started strongly feeling like date IDs are the ones that detract, and most of the above (too many) words are me arguing for the use of folgezettel. Hopefully the thought process written out helps someone. If I do come down on the side of date IDs, it's currently only by a hair's breadth, though maybe working with date IDs for a while will widen that gap.

• @prometheanhindsight thank you for taking the time to write you thoughts here!

How truly equal are these connections?

The idea behind statement that all references are equal is this:
You are able to determine the value and importance of references and compare them only in the moment of interest in the card. The first moment of interest is when you create the card and make initial links and decide where will you place it - you actively decide that that link is of most interest to you at THIS moment. But that importance is not set in stone. xy years from now you will look at that note again, maybe for the first time since you created it, and your interests will be different, your thinking will be different and your reasons for looking at that note will be different. So, how can you be sure that the link you picked as most important before will still be the most important at that moment? You probably can't. That's why my position is to work in a way that won't limit my capacity to objectively compare links and make the best decision every time when I have to decide what reference is of most importance for my current needs. Reference is there to answer to our needs, it is not there to be some ultimate truth. And you can't tell in advance what reference will be used the most.

3b - 3b1 comment

You are making a great point here. And I will start from the back. "...Locational choice is arbitrary, and so long as I add links then I'm not losing out on anything." Yes, this is my reasoning. And this is the reason why I think ID nesting can be confusing. What you described at the beginning of the comment, "If there is something else to say about 3b1 in the context of 4c11, then I can create 4c11a that describes that difference, which will link both to 4c11 and 3b1" describes that you can track sequences without ID nesting, in fact, you can track only one of them effectively with ID nesting.
Whenever you have to branch from another context, you will do that by using regular links (creating 4c11a and linking it to 3b1 is regular linking that you can perform with timestamps as well). You definitely don't need to read 3b to understand 3b1, but adding note 3b1a from 3b1 presupposes that 3b1a belongs to the same sequence as 3b and 3b1. As you said, if you want to add branch from different context related to 3b1, you will use for example 4c11a and link it to 3b1 from outside the 3b sequence, but you won't add a note related to 4c11 as "3b1a".

But, if I'm writing down the reason for each connection and adding a link, then I'm also not losing out on the meaning of the connection to 4c11. The two connections have been described in the new note, and so I can still create the radial branching that you discuss.

This what you described is exactly what I am doing. And once I started to do it I realized that I don't need ID nesting anymore, I can use whatever I want as identifier, as long as it is unique for every note. I tried to argue in the article that sequences are not pointless, but that ID nesting, as Luhmann used it, is not perfect. If you can work with the traditional way and stay objective - it is great - tell us the secret! I find it hard to stay objective when I get into the reading - thinking - writing flow, I feel like I am thinking on auto-pilot, so this was a distraction, a limitation of my possible insights and discoveries. It doesn't mean that everyone share that problem, it's personal imperfection - I fall easily for false patterns as you defined it (I like this definition by the way, will be stored to my ZK)

I've tagged these notes with #optical-thermometry, so if I'm looking for an optical thermometry technique I can easily search the tag and find them.
...
What if I'm not specifically thinking about optical thermometry? I'm looking through my x-many thousand notes for links to a new note that I just created. The pattern recognition portion of my brain snags on a cluster of notes about optical thermometry, and I'm suddenly reminded of a link that my new note has to optical thermometry!

My argument is that you don't need ID nesting in order to achieve this. I think I wasn't clear enough in my article, I am not against folgezettels as a concept, I am against ID nesting as a method of implementation. I basically have sequence card for every keyword - I definitely use folgezettels, just with different implementation. I use keywords much more selectively than tags. Title of the sequence is my keyword. Tag is like a general, broad term in my ZK, but keywords are short, very descriptive, clear and specific.

So, I am processing my notes into zettels, I do keyword search, and I will open a few keywords (i.e. I will open a few sequence notes) and there I have all notes from the sequence, with titles, and I can preview their content (I guess this varies from app to app) and I can find relevant references. Sometimes I will preview a note and I will open it and follow its links and I will find relevant notes through those links. With ID nesting, if I observe 2 sequences for comparison, one note will be nested only in one of them, if it is linked to a note from the second one I would have to go through all notes and through all links to find the one that it is linked to. With sequence cards I can clearly see that one note shows up in multiple sequences. So if I opened keywords relevant to my new note and I find a note that shares a lot of these keywords as well, there is a good chance that these two notes are a good combo, or at least that following the links of this second note will lead to some discovery.

But I have to admit, the best references are ones that just pop in my mind, they are the most unpredictable ones. And that is hard to explain, because it just happens, we cannot define something that our brain does in the background, but it does a better job than my conscious search every time.

Looking back over how many words I've written in response to this article, it's a little funny for me to reach the conclusion that folgezettel might actually detract from the use of a ZK, since I started strongly feeling like date IDs are the ones that detract, and most of the above (too many) words are me arguing for the use of folgezettel.

Timestamps are not my favorite thing. I don't use app that can generate them automatically, so I don't use them. I just use titles. And since I can preview the content of the note with hover, navigation is pretty fast. I still think folgezettels are helpful, but ID nesting is problematic - so you don't have to give up folgezettels, you can just find a way to make them better. I changed the way I use them because I encountered problems with the traditional way, and the solution I picked solves those issues. It doesn't mean I won't find other problems with this current workflow, when I do I will adopt again. Implementation is there to enhance the concept, whenever I find a better one I will adjust my workflow, after all, the goal is to be more efficient, even if it means abandoning traditional approach.

My goal is not to make people feel like they are doing something wrong, I am not an expert myself. My intention is to show how I deal with problems I encounter and how I examine concepts in order to make safe adjustments that will fix issues and help me to get better results with my ZK. So, I don't want you to think that my way is The way, but to use my experience to define what is the source of friction in your workflow and use understanding of core principles and concepts to fix it in a way that won't hurt functionalities, but improve your workflow.

p.s. Sorry for the total lack of short and concise writing!

• @ethomasv said:
You are able to determine the value and importance of references and compare them only in the moment of interest in the card. The first moment of interest is when you create the card and make initial links and decide where will you place it - you actively decide that that link is of most interest to you at THIS moment. But that importance is not set in stone. xy years from now you will look at that note again, maybe for the first time since you created it, and your interests will be different, your thinking will be different and your reasons for looking at that note will be different. So, how can you be sure that the link you picked as most important before will still be the most important at that moment? You probably can't. That's why my position is to work in a way that won't limit my capacity to objectively compare links and make the best decision every time when I have to decide what reference is of most importance for my current needs. Reference is there to answer to our needs, it is not there to be some ultimate truth. And you can't tell in advance what reference will be used the most.

I think this is a great way of putting it. I've spent the morning playing around with building sequence notes for topics; at the bare minimum, this is just trying to highlight the different sequences I have running through my sequence based IDs. My immediate observation is that the note sequences I made even just a couple months ago are difficult for me to fully parse. The branching ID sequences makes this even more complicated to figure out.

This what you described is exactly what I am doing. And once I started to do it I realized that I don't need ID nesting anymore, I can use whatever I want as identifier, as long as it is unique for every note. I tried to argue in the article that sequences are not pointless, but that ID nesting, as Luhmann used it, is not perfect. If you can work with the traditional way and stay objective - it is great - tell us the secret! I find it hard to stay objective when I get into the reading - thinking - writing flow, I feel like I am thinking on auto-pilot, so this was a distraction, a limitation of my possible insights and discoveries. It doesn't mean that everyone share that problem, it's personal imperfection - I fall easily for false patterns as you defined it (I like this definition by the way, will be stored to my ZK)

My argument is that you don't need ID nesting in order to achieve this. I think I wasn't clear enough in my article, I am not against folgezettels as a concept, I am against ID nesting as a method of implementation. I basically have sequence card for every keyword - I definitely use folgezettels, just with different implementation. I use keywords much more selectively than tags. Title of the sequence is my keyword. Tag is like a general, broad term in my ZK, but keywords are short, very descriptive, clear and specific.

I think this is where my main misunderstanding has always been, and you've greatly helped to resolve it.

Sequences are important for tying notes together into a through line, but using IDs to build those sequences cements that sequence in your notes. Even if other sequences can be built using sequence notes, the primary sequence will always be that of the ID numbers.

Giving arbitrary ID numbers that don't branch prevents this cemented note sequence, and so the sequences that exist in your notes are only the sequences that you build through structure or sequence notes. These are easy to define and redefine in future years and as new notes are added.

So, I am processing my notes into zettels, I do keyword search, and I will open a few keywords (i.e. I will open a few sequence notes) and there I have all notes from the sequence, with titles, and I can preview their content (I guess this varies from app to app) and I can find relevant references. Sometimes I will preview a note and I will open it and follow its links and I will find relevant notes through those links. With ID nesting, if I observe 2 sequences for comparison, one note will be nested only in one of them, if it is linked to a note from the second one I would have to go through all notes and through all links to find the one that it is linked to. With sequence cards I can clearly see that one note shows up in multiple sequences. So if I opened keywords relevant to my new note and I find a note that shares a lot of these keywords as well, there is a good chance that these two notes are a good combo, or at least that following the links of this second note will lead to some discovery.

I think that this is a great way of doing it. It feels like a more flexible way of establishing structure, especially when one note is important to multiple sequences. I've been trying this out today, and it's going a lot smoother than I previously thought it might. Usually, the zettel that I'm processing are coming from a series of thoughts that can help define their initial sequence. That makes it easy to place them in a sequence to start (what would previously be done through their nested ID numbers) before searching through the rest of my notes for new connections or other sequences.

So, I don't want you to think that my way is The way, but to use my experience to define what is the source of friction in your workflow and use understanding of core principles and concepts to fix it in a way that won't hurt functionalities, but improve your workflow.

You have helped me a lot! Working through your explanation, and comparing it to my current workflow, helped me to catch some issues that I probably wouldn't have noticed otherwise. Luckily, I don't have so many notes yet that it'll be too much trouble to fix.

p.s. Sorry for the total lack of short and concise writing!

Only if you forgive me for the same!

• This is really interesting. I always thought i will be able to detect the context by the topics/titles automatically. But that may be because i use a physical zettelkasten for only a specific, exam related purpose. For the digital one, maybe time ID's are enough and next to a reference/link you could just add a word implying a sequence, e.g. "=>/ therefore/ following [[link to folgezettel]]. Thus you can see old argumentation-sequences and "normal" links at the same time?

• edited August 2020

@amunicapunica Yes, I like to add tags to my references like "Prompted By" for example. And I do look at both, regular links AND sequences, I compare them all in the moment of interaction.

• edited August 2020

As a user of both The Archive and a physical index card system, I've stumbled across this issue a couple of times...
Whilst using unique indices is great for software where you can continue linking to eternity and it really doesn't matter what the unique ID is, its not sustainable in a physical index card system.

The way I've approached this is kind of like the same way that RoamResearch implements block references. If you aren't familiar with this approach, it essentially shows you the content of a bullet point from any other page, on the page you are currently on. It creates a unique link ID, but shows the actual content you've written. When you update the original page block that you are referencing from elsewhere, the references also display that same update.

In the physical Zettelkasten I sort of copy this block referencing idea byputting one card behind where I think its most relevant at that time - using the articles example, 3b - calling the new one 3b1. I then create a new card behind each of the other branches, labelled as they would be if the card was created there. So behind 4c11 I'd create one called 4c11a, behind 4d I create one called 4d1 and behind 7a10 one called 7a10a. But on the 4c11a, 4d1 and 7a10a cards, I write just the title, and the link to 3b1. This way I have some context about what the original note/thought is about in each of the branches, and more importantly I can still continue the other branches without losing the train of thought.
Most importantly though in my opinion, it also means I can physically see when an idea is big enough that I may be able to actually do something with it.
I think this is something that you just don't get when using unique ID's such as dates in a piece of software (ANY piece of software other than those that have the graph view like Roam/Obsidian/etc).
If the context then changes in each of the branches, I can just pull the index card out, and write a more accurate description/summary that still doesn't break the train of thought, more accurately fits the branch that that physical card lives in, but is still related to the other cards if I need it to be.
I can go and update The Archive version at the same time nice and easily.

The benefit of doing it this way, is that the moment I see just a title and a link on a card, I know its related to another branch, so I then have the context in my head that the larger ideas may separate at some point - and that point could well be right now...

The one caveat I have here is that my Zettelkasten isn't that big currently, so it could be working well for me at the moment but may fall to pieces later when I've built it out to a larger size.

Does anyone see any issues with this other than the time/effort of creating duplicate cards (which doesn't happen THAT often, and even when it does, you aren't writing a tonne of content - just adding a link).

A note here on Roam...
I think RoamResearch is a great tool, but I feel like its extreme overkill for a Zettelkasten (although you can happily implement one here). There is definitely a place for The Archive living alongside Roam/vice versa, hence why I use both systems.
But I use Roam in a "second brain" type of way, by creating notes from books, articles, digital Bullet Journal, task management, reminders, etc, and The Archive to create order out of the chaos thats stored in Roam to help extract the best of whats there. Keeping them separate (at least for the time being), helps me keep my head clear (a lot goes on up there), but doesn't create a bomb site in my Zettelkasten.
Again, if anyones got thoughts around this workflow, happy to hear them over on my intro post

• Yes, I understand you approach - I was doing something similar in my physical zettelkasten at one point, but it was too much re-writing for my taste. If it works for you, you keep doing it, you will see yourself if it will survive scaling. Any aspect of your Zettelkasten that is not good will crumble as the number of your notes grows. It sound dramatic, but it is a good thing, it will help you to solidify your workflow.

• Any aspect of your Zettelkasten that is not good will crumble as the number of your notes grows.

Very good point! Sometimes, I feel that many people see adding techniques and methods as the way to go. In my experience, the less techniques and methods you use the better your system will sync with you.

I am a Zettler

• @sfast said:

Any aspect of your Zettelkasten that is not good will crumble as the number of your notes grows.

Very good point! Sometimes, I feel that many people see adding techniques and methods as the way to go. In my experience, the less techniques and methods you use the better your system will sync with you.

"Keep It Simple Stupid"

• @ethomasv said: Timestamps are not my favorite thing. I don't use app that can generate them automatically, so I don't use them. I just use titles.

Hello Eva, I enjoyed your two articles. I always use a unique identifier (UID) before the title in a filename. There's multiple times I've changed the title of a note both inside the note and in the filename, the only thing that remains fixed is the UID. My filenames are a short as possible and I use few words and sometimes generic titles. Inside the note I have a long and descriptive title. Whenever I change filename or title none of the links in other notes suffer because I keep the UID and that's what I use for searching.

• @sfast definitely! But sometimes we need to add everything at the beginning and simplify later. It is hard to grasp downsides when starting out (I struggled with it)

@Splattack I am glad you enjoyed reading, and I agree with you regarding the UIDs. I didn't use them at first because I was building my digital ZK in a program that will change all links automatically if I change just the title in one place, so I didn't see the purpose. But I decided to add them since maybe in the future I will change the app and I want to have my links saved and stored safely