1. I have a Daily Note which is my starting point of the day. Aside from some TODOs and some daily reflections, it also contains an image map of things that are currently important. This map gets updated (in a template) once every couple of weeks, depending on where my focus is. Currently it links to a big project for a customer, my running progress, a webdav tool I'm working on, and a book I'm reading.
2. Many of my note templates automatically include a dataview query which spits out a list of related notes. E.g. for People that will be a list of recent meetings and mentions. For Programming related notes there's a list of notes that link to it, like for Java I have ~200 other notes that point at Java, from Method References to Apache Commons libraries.
3. Lots of links. This makes the local Graph View somewhat useful for rediscovery. This includes links to things that don't exist yet (and may never exist). Like I'm probably not going to create a note about Mitochondria, but several notes link to it.
Some more serendipitous discovery might be nice. I've experimented with various Related Notes plugins, most are garbage which find similarities where none exist because they can't tell the note structure apart from its salient content.
First, a disclaimer. I’ve worked with Ankur, and he is a brilliant, super-organized, and methodical engineer.
Personally, I’m on the other end of the spectrum. I’m still up for a good organization, and I do my own simplified version of PARA[1] + Johnny Decimal.[2] However, I’m beginning to tend more and more towards how Steph Ango uses Obsidian[3] while still maintaining simplicity and ways to walk out when needed. I’m either listing a list of files based on keywords, either via the Open command or the Search feature. I, sometimes, use a Universal Search Extension for Obsidian, but I can live without it.
My final thoughts, especially on the UX, are that, eventually, when it becomes commonplace to run a local LLM, I should most likely be just using Spotlight/Alfred or something to open, search, link, and all of that while I just keep adding contents into their own ”boxes” that I have categorized at the gate while filing the documents/content.
I’m not sure if I can even explain what I’m thinking, but the above is a precursor to what I believe will happen soon enough.
Lots of Obsidian talk here. As someone who started with the OG https://zim-wiki.org, I tried Obsidian and never found it compelling enough to switch.
I did, however, with Logseq.
On the surface they're deeply similar -- but Logseq's native handling of outlines -- including collapsing and and expanding on the fly (org-mode does this as well, but you know) is absolutely killer. It's the first one to kind of reliably "free" me from the whole "every note is a page" thing that causes lots of friction.
I think it could go well to solving op's problem as well. Quick expanding and collapsing and easy search helps me to poke around older knowledge in a fun and useful way.
(actually, I didn't "switch," I still keep day-to-day important stuff in zim, but for that storing and exploring old knowledge, I'm now all about logseq).
For me, the goal of PKMs is to store information for later reference.
I know that some people use them to learn, however I've never found this method to be useful for me, so I just stick to spaced-repetition (Anki) to help me learn the relevant facts.
Done this way, I have two "stores" of information. In Obsidian I put information I don't want to memorize or at least not long term, like work related topics (terminology, the domain, etc), project planning, project details, roadmaps, and in Anki I put information I care about learning long term, like facts about X technology or language, vocab, etc.
Point is, i think obsidian is not the right tool for learning. SRS sounds like a better fit. I'd suggest taking the effort to learn how to make good clozes (card type).
The author's wishlist for an Obsidian homepage could be done with a custom plugin. That's the beauty of Obsidian, you don't have to wait, you can extend the software yourself.
The LLMs are quite good at writing one-off Obsidian plugins, in my experience.
> But something crucial is missing from modern PKM apps: they do a poor job of helping me re-engage with information that I’ve already captured but forgotten about.
> If I’m not careful, my PKM apps become black holes where information goes to die.
Well, that's because just writing notes != learning. Obviously the information get forgotten.
I can't believe the author were so obsessed with PKM, but there zero mention of spaced repetition to help them learn and remember information. (Anki, Supermemo, or at least some Obsidian plugin with same function)
For Obsidian, I use hierarchal tags (eg. #python/django/models) and the Omnisearch extension[1] for resurfacing and I don't have any complaints.
For tags, I use it much like a file navigator. In fact, it sits exactly where the files tree is in the default config. But, instead of files, its hierarchal tags.
Recently, I had issues with Obsidian running out of memory and closing while in the foreground on my 6 GB iPhone, and the culprit turned out to be this add-on. Removing it reduced idle RAM usage on the desktop from ~1.1 GB to ~600 MB.
Just a warning for people like me who tend to collect nice-sounding extensions and then forget to actually use them.
I accidentally started using Pinterest to organize infographics on my phone, which led to resurfacing material on a daily basis.
The process of organization with the app on my phone is itself mindless enough that I can focus on the meaning of what I'm looking at.
That's the friction point in a lot of these PKM solutions - the process of using them is set up for Big Important Tasks. The process of using the interface interferes with the flow of thought.
I agree with the author, but the irony is that resurfacing is akin to UI friction. I resurface content in a journal because I have to flip pages to find the content I seek.
I've emulated some of the features the author wants in my obsidian using Dataview, but it's more the human ritual of review that accomplishes this for me.
Indeed, I wonder to what degree these problems would be better solved by a habit of exploring what you earlier wrote, and having categorized it well, and constantly recategorizing. The power of writing is not in the initial draft, but in how one wrestles to understand that content.
Agreed, this is a big problem. I tried a paper zettelkasten to get around the taking notes and losing them problem, but my collection grew too big to maintain on paper. Because of that, I've been building Zettelgarden (https://github.com/Zettelgarden/Zettelgarden) to help and try to solve this problem for myself.
Among other things, it breaks your notes apart into facts and entities, then stores those along with embeddings. This helps surface things you've seen before since it makes and surfaces links based on individual things, instead of the text as a whole.
This article talks about resurfacing from current context - which is handy. A feature i added to my own tools was to also select a random old note and bring it back up. This allows me to re-discover or to cull the information. It a pattern from when I was a bug-wrangler.
Yeah, agreed that some of the features the author cites sound great...but, then i wondered similar to you that even if i *only* bring up some random notes, it would still be better than what i have now (which is nothing!). I imagine bringing back random notes helps for a few reasons:
* Can prune old/outdated/now-irrelevant stuff
* Can learn/re-learn stuff
* Can consider re-writing/updating some the re-surfaced note, maybe because i have better understanding of some concept nowadays as compared to when i originally wrote some note
* Maybe some old ideas spark thoughts that generate new idea(s)
...etc., etc., etc.
You have inspired me to craft a little script to bring up random notes from my collection. Thanks! :-)
Good article. I am solving the similar problems in my product[1], which purpose is more practical and uses hierarchy (like file system) and keywords to organize and link information.
If you want to resurface them, you can show them in hierarchy(show children, or descendants in any level), search or filter by keywords, sort them by created time, URL, order, or other fields.
Resurfacing from current context can be very complicated, you should create some
fields to track your activities, create many routines to handle these information automatically, and create relative UI. These are advanced features based on the current basic features.
I have been using Obsidian for 5 years and I strongly agree with this. My solution is just occasionally clicking random notes in my vault and reading them, 5-7 notes at a time. I do this a couple of times in a month
There's also a core plugin for opening a random note. I believe it adds a command prompt (or whatever they call it) option and a little die icon on the left side.
The first example cited in Obsidian is easily solvable. Stick to consistent terminology for notes covering a particular subject and the back links panel will auto suggest linking all the notes that cover that area together whenever you create new ones.
As for rediscovery of old stuff, the best solution is to get comfortable with the fact that certain notes won't resurface and to let them die. This is actually a good thing. It allows you to keep focus on your interests as they evolve. Even the legendary Niklas Luhmann, whose paper zettelkasten was arguably the inspiration for the current lmk craze had what we called "black holes" in his system—there are several notes he filed that never saw the light of day again. This was by design. These systems are supposed to be dynamic mechanisms for helpings you think, not immemorial retrieval mechanisms. If that's what you want you might as well make every note chronological.
Trying to use a PKM strictly as a "remember anything at any time" is largely missing the point. The goal is to create knowledge by stringing relevant ideas together and, as Luhmann did, to use the system to produce surprises. The best way to do this for obsidian in particular is to use terminology and link liberally whenever some term occurs in a note even if it at first seems completely unrelated—that's what you want, you want the system to surprise you as you navigate h the graph of notes.
Here are some of the things I do in Obsidian:
1. I have a Daily Note which is my starting point of the day. Aside from some TODOs and some daily reflections, it also contains an image map of things that are currently important. This map gets updated (in a template) once every couple of weeks, depending on where my focus is. Currently it links to a big project for a customer, my running progress, a webdav tool I'm working on, and a book I'm reading.
2. Many of my note templates automatically include a dataview query which spits out a list of related notes. E.g. for People that will be a list of recent meetings and mentions. For Programming related notes there's a list of notes that link to it, like for Java I have ~200 other notes that point at Java, from Method References to Apache Commons libraries.
3. Lots of links. This makes the local Graph View somewhat useful for rediscovery. This includes links to things that don't exist yet (and may never exist). Like I'm probably not going to create a note about Mitochondria, but several notes link to it.
Some more serendipitous discovery might be nice. I've experimented with various Related Notes plugins, most are garbage which find similarities where none exist because they can't tell the note structure apart from its salient content.
Would you happen to have these templates on Github or public somewhere?
First, a disclaimer. I’ve worked with Ankur, and he is a brilliant, super-organized, and methodical engineer.
Personally, I’m on the other end of the spectrum. I’m still up for a good organization, and I do my own simplified version of PARA[1] + Johnny Decimal.[2] However, I’m beginning to tend more and more towards how Steph Ango uses Obsidian[3] while still maintaining simplicity and ways to walk out when needed. I’m either listing a list of files based on keywords, either via the Open command or the Search feature. I, sometimes, use a Universal Search Extension for Obsidian, but I can live without it.
My final thoughts, especially on the UX, are that, eventually, when it becomes commonplace to run a local LLM, I should most likely be just using Spotlight/Alfred or something to open, search, link, and all of that while I just keep adding contents into their own ”boxes” that I have categorized at the gate while filing the documents/content.
I’m not sure if I can even explain what I’m thinking, but the above is a precursor to what I believe will happen soon enough.
1. https://fortelabs.com/blog/para/
2. https://johnnydecimal.com
3. https://stephango.com/vault
Lots of Obsidian talk here. As someone who started with the OG https://zim-wiki.org, I tried Obsidian and never found it compelling enough to switch.
I did, however, with Logseq.
On the surface they're deeply similar -- but Logseq's native handling of outlines -- including collapsing and and expanding on the fly (org-mode does this as well, but you know) is absolutely killer. It's the first one to kind of reliably "free" me from the whole "every note is a page" thing that causes lots of friction.
I think it could go well to solving op's problem as well. Quick expanding and collapsing and easy search helps me to poke around older knowledge in a fun and useful way.
(actually, I didn't "switch," I still keep day-to-day important stuff in zim, but for that storing and exploring old knowledge, I'm now all about logseq).
For me, the goal of PKMs is to store information for later reference.
I know that some people use them to learn, however I've never found this method to be useful for me, so I just stick to spaced-repetition (Anki) to help me learn the relevant facts.
Done this way, I have two "stores" of information. In Obsidian I put information I don't want to memorize or at least not long term, like work related topics (terminology, the domain, etc), project planning, project details, roadmaps, and in Anki I put information I care about learning long term, like facts about X technology or language, vocab, etc.
Point is, i think obsidian is not the right tool for learning. SRS sounds like a better fit. I'd suggest taking the effort to learn how to make good clozes (card type).
This makes me think: wouldn't an ideal play for Obsidian, Logseq et al to be a cloud service with some kind of RAG augmentation?
Yet as far as I know, the only thing an Obsidian account really unlocks, is a graph search. Which would seem like a missed opportunity
The author's wishlist for an Obsidian homepage could be done with a custom plugin. That's the beauty of Obsidian, you don't have to wait, you can extend the software yourself.
The LLMs are quite good at writing one-off Obsidian plugins, in my experience.
> But something crucial is missing from modern PKM apps: they do a poor job of helping me re-engage with information that I’ve already captured but forgotten about.
> If I’m not careful, my PKM apps become black holes where information goes to die.
Well, that's because just writing notes != learning. Obviously the information get forgotten.
I can't believe the author were so obsessed with PKM, but there zero mention of spaced repetition to help them learn and remember information. (Anki, Supermemo, or at least some Obsidian plugin with same function)
For Obsidian, I use hierarchal tags (eg. #python/django/models) and the Omnisearch extension[1] for resurfacing and I don't have any complaints.
For tags, I use it much like a file navigator. In fact, it sits exactly where the files tree is in the default config. But, instead of files, its hierarchal tags.
1. https://github.com/scambier/obsidian-omnisearch
Recently, I had issues with Obsidian running out of memory and closing while in the foreground on my 6 GB iPhone, and the culprit turned out to be this add-on. Removing it reduced idle RAM usage on the desktop from ~1.1 GB to ~600 MB.
Just a warning for people like me who tend to collect nice-sounding extensions and then forget to actually use them.
Good to know, thank you.
I accidentally started using Pinterest to organize infographics on my phone, which led to resurfacing material on a daily basis.
The process of organization with the app on my phone is itself mindless enough that I can focus on the meaning of what I'm looking at.
That's the friction point in a lot of these PKM solutions - the process of using them is set up for Big Important Tasks. The process of using the interface interferes with the flow of thought.
I agree with the author, but the irony is that resurfacing is akin to UI friction. I resurface content in a journal because I have to flip pages to find the content I seek.
I've emulated some of the features the author wants in my obsidian using Dataview, but it's more the human ritual of review that accomplishes this for me.
I do agree, there is a good product there.
Indeed, I wonder to what degree these problems would be better solved by a habit of exploring what you earlier wrote, and having categorized it well, and constantly recategorizing. The power of writing is not in the initial draft, but in how one wrestles to understand that content.
Agreed, this is a big problem. I tried a paper zettelkasten to get around the taking notes and losing them problem, but my collection grew too big to maintain on paper. Because of that, I've been building Zettelgarden (https://github.com/Zettelgarden/Zettelgarden) to help and try to solve this problem for myself.
Among other things, it breaks your notes apart into facts and entities, then stores those along with embeddings. This helps surface things you've seen before since it makes and surfaces links based on individual things, instead of the text as a whole.
In case anyone is interested, I wrote a bit about the process here: https://nsavage.substack.com/p/facts-arguments-theses-buildi...
This article talks about resurfacing from current context - which is handy. A feature i added to my own tools was to also select a random old note and bring it back up. This allows me to re-discover or to cull the information. It a pattern from when I was a bug-wrangler.
Yeah, agreed that some of the features the author cites sound great...but, then i wondered similar to you that even if i *only* bring up some random notes, it would still be better than what i have now (which is nothing!). I imagine bringing back random notes helps for a few reasons:
* Can prune old/outdated/now-irrelevant stuff
* Can learn/re-learn stuff
* Can consider re-writing/updating some the re-surfaced note, maybe because i have better understanding of some concept nowadays as compared to when i originally wrote some note
* Maybe some old ideas spark thoughts that generate new idea(s)
...etc., etc., etc.
You have inspired me to craft a little script to bring up random notes from my collection. Thanks! :-)
Good article. I am solving the similar problems in my product[1], which purpose is more practical and uses hierarchy (like file system) and keywords to organize and link information.
If you want to resurface them, you can show them in hierarchy(show children, or descendants in any level), search or filter by keywords, sort them by created time, URL, order, or other fields.
Resurfacing from current context can be very complicated, you should create some fields to track your activities, create many routines to handle these information automatically, and create relative UI. These are advanced features based on the current basic features.
[1] https://metaessen.com/
Linking, filtering, searching; anything that suggests an active presence in my knowledge base works better for me.
The author’s own gap fillers that he mentions sound cool, but his fixation on randomness makes me curious about what the purpose of his system is.
I made thinktype[1], a notes app that uses the same text field for writing and search, so that you discover related information when you write.
[1]: https://thinktype.app
I have been using Obsidian for 5 years and I strongly agree with this. My solution is just occasionally clicking random notes in my vault and reading them, 5-7 notes at a time. I do this a couple of times in a month
There's also a core plugin for opening a random note. I believe it adds a command prompt (or whatever they call it) option and a little die icon on the left side.
The first example cited in Obsidian is easily solvable. Stick to consistent terminology for notes covering a particular subject and the back links panel will auto suggest linking all the notes that cover that area together whenever you create new ones.
As for rediscovery of old stuff, the best solution is to get comfortable with the fact that certain notes won't resurface and to let them die. This is actually a good thing. It allows you to keep focus on your interests as they evolve. Even the legendary Niklas Luhmann, whose paper zettelkasten was arguably the inspiration for the current lmk craze had what we called "black holes" in his system—there are several notes he filed that never saw the light of day again. This was by design. These systems are supposed to be dynamic mechanisms for helpings you think, not immemorial retrieval mechanisms. If that's what you want you might as well make every note chronological.
Trying to use a PKM strictly as a "remember anything at any time" is largely missing the point. The goal is to create knowledge by stringing relevant ideas together and, as Luhmann did, to use the system to produce surprises. The best way to do this for obsidian in particular is to use terminology and link liberally whenever some term occurs in a note even if it at first seems completely unrelated—that's what you want, you want the system to surprise you as you navigate h the graph of notes.