"Smart Linking" in DevonThink

In “How have folks been using PARA” topic, I read that someone actually relying on “link” (not like a function, but like an idea where you build your system from bottom instead of doing the top level design).

I think it’s a brilliant idea, but there are so many details left, for example, I’d like to know how should we do the “link”, is it smart group? tagging? replicant? To me it really sounds like a Zettelkasten thing.

Is there anyone who do not use nested groups and rely on other approaches?

1 Like

A reference to or quote of the original post might clarify this but usually linking in DEVONthink means either Wiki linking (see Preferences > WikiLinks) or item linking (see e.g. Edit > Copy Item Link).

Smart groups, tags and replicants are not links. By “links” I mean connections between one file and another, or one piece of text and another, or sometimes from a file or piece of text to a group in DEVONthink, that are made intentionally. (DEVONthink does allow “automatic” linking, which has its uses, but I find intentional links more useful.) The way the link is made depends on the program. It may be a “wikilink” or something else.

The screenshot below shows a small part of a Tinderbox file with several links between different notes. The point of this, as I’m sure you can see, is to build a network connecting topics, ideas, concepts, and so forth. It is indeed a Zettelkasten type of method.

The developer of Tinderbox, Mark Bernstein, has written about the idea of “incremental formalisation”, which basically means allowing a structure to appear as a result of working with material, rather than trying to define a structure to begin with, and then trying to fit the material into it. I find this idea of allowing the structure to emerge as one works to be very attractive, though I doubt it would work for everybody.


I think emergent structure assumes that you don’t know in advance what the best structure should be. In my experience, that means it works well for smaller clusters of information, not as well for larger clusters.

So if I’m writing an article about two-dimensional semiconductors, I may not know going in that I’ll have clusters of information about different materials, clusters about manufacturing methods, and clusters about device physics. So I can throw all of those in a big bucket and let structure emerge. But there are still going to be vast piles of material – all of human knowledge that isn’t 2-D semiconductors – that don’t go in the bucket, and I can use pre-existing categories to get those things out of my way so I can focus on the topic at hand.

Disclaimer: I’ve played with Tinderbox a bit, but it is not currently part of my toolset.

1 Like

You have made me wonder if it is a characteristic of the way that my mind works that I tend not to see hard boundaries between things, or if it is a characteristic of the area that I work in these days. Or perhaps it is a bit of both. Categorical thinking does seem to be a feature of Western styles of thinking, however (see https://web.mit.edu/6.969/www/readings/culture_thought.pdf) so perhaps I have some non-European cultural influence lurking inside me …

1 Like

Depending on context, a “boundary” could be either hard or fuzzy.

A Tinderbox map supports the idea of notes being “near” each other. At a sufficient degree of “nearness,” they could be seen as “inside” a boundary region.

If you are scoping a 1000 word article, probably the region will be pretty small. If you are scoping a book, probably larger. If you are designing a curriculum, even larger.

Many of my boundaries are “fuzzy”. I’m an ecologist - everything is related to everything else and the world is complex. I do rely on topic folders/groups, because it reflects the way I like to navigate information. Files get put in a top level folder that’s “broad” in theme, and only get added to more specific subfolders when that is possible.

However, I do find myself relying on replicants more and more nowadays. (It’s not a conscious choice, but I’ve been observing where I need information to surface and using replicants to ensure that happens.)

Here is an example to take us out of the hypothetical:

Wolves prey on deer and reduce deer-related traffic collisions in the U.S. This paper and my notes live primarily in my group on U.S. wolf studies. However, this research is also important for deer, so a replicant is in my deer group. Because I am interested in the societal implications of this, a replicant is in a group I maintain about conservation and society. Finally, it also has interesting consequences for carbon cycling, so a replicant also appears in a group about carbon and the role of predators.

At some point, I might write notes specific to the topics of those groups and remove the replicants, but I’m not in any rush and my group structure and replicants mean things are usually where I expect them to be.

I could do all this with linking, or even just one giant group of everything. But I like groups. My post is to demonstrate how you can reflect fuzzy boundaries even within a group structure with DT.


I think a lot of psychologists would echo that. I could well imagine finding a paper in which semi-conductors are related to some aspect of human behaviour.

I used to do something similar, and I certainly use groups, folders, containers, etc. I’m not sure why I stopped using replicants. It was too long ago. But in general the speed at which I organise material has not kept up with the speed at which I collect it … I use search functions rather a lot :slight_smile:

1 Like

haha, i hear that! I’m trying now to do “housekeeping” at least every couple of weeks so that things are mostly where I expect them to be (I don’t like leaving things in database inboxes until I’m ready to deal with them, because often that doesn’t happen for months). In reality though, I will often spend a few mins tidying up the part of the database I am in when I am looking for something, and in that way the system mostly ticks over (I think it’s similar to how looking for a piece of paper on your desk often results in you tidying up what’s on your desk. For those of us who don’t tidy up every day we might as well accept that this is when tidying is done.)

1 Like

And more and more of this “everything” is being added. :joy: I have given up. I mean the tidying up. Often I wouldn’t even know how to tidy up. So a lot of documents have accumulated in the same folder. But I organize myself with metadata, which I always assign immediately. It’s quick and quite efficient. This creates overlapping universes of knowledge that I can bring together and keep track of to some extent using the search function. If I understand correctly, that’s how @mbbntu does it too. But at the end I know that I will always be unable to find some of the information I have collected. That’s ok :slightly_smiling_face:

My neuromorphic computing group is wild…


I think I got some inspiration now.

I’m gonna do some experiment. First, I will stop using tags and links. Because, TBH, in real working environment, we don’t really have that much time to tide up. Instead, I’m now gonna rely on metadata, groups and replicants. I will give up most top level design thing.

This is my new structure. I now have “workspace” and “Records” as the two database for work. I will group files naturally, adding metadata when a file is created and use smart groups to search for time-sensitive files. I will also try to explore replicants. Hope this will improve my workflow.

1 Like

I don’t think I know anybody who works in an unreal environment!

You might like to read around a bit and find out how others have chosen to work. One example is Kourosh Dini’s blog:

And another is the Zettelkasten site (which you might already know). It has been around for ten years now:

Good luck with the work.

That depends heavily on your description of work. If you see your work as a set of distinct medium and small-scale projects, using only groups can be an effective way to minimize the need to reorganize material.

If you see your work as one single large, evolving project, it is likely worth the time to organize with links and tags. I spend around 20 minutes every night to review and organize the day’s work. Sometimes inspirations come in the process. Most of the time it helps me sleep better knowing that another day has been wasted.

1 Like

I find tags very efficient. While labels and other metadata can only be used once per document, tags have the advantage that a document can have several of them. This makes it easier to cleverly “link” (bundle) documents that are stored in different locations (because nobody has tidied them up).

If you (like me) prefer typing to clicking, DT offers a fairly simple way of combining text and tags in the search field.

Of course, everyone already knows that … just in case someone has overlooked it :slightly_smiling_face:

1 Like

Instead of tag:tag1 tag:tag2 you could also use tag:tag1;tag2

1 Like

Interesting, I didn’t know that. However, I don’t need it because I expand “tag:” with t, space, space. But if someone writes “tag” every time, that’s good to know. :slightly_smiling_face:

I get the feeling that software marketing people live in a kind of unreal environment. When you watch or read them describe how it works or all the great things it will do for you, it never seems grounded in reality.

I’ve been researching Project Manager software and the interconnected realms of PKMs and Note taking software. The unwarranted certainty that those sites express for how their software will solve all your problems is tiresome and offputting.

One thing that attracted me to DevonThink was the rationally stated functions of the app with little added hyperbole about it curing all ills.
I grabbed the free trial and now use it every day in my day job and personal work. But it’s not, and the purveyors of the software state this same thing in this forum regularly, it’s not a be all to end all.

This is also why I am giving LogSeq a try right now. The people who were talking about it were clear about its limitations and uses. There’s a part of my job that isn’t covered by DevonThink and Logseq seems like a good option to try to cover that.


Indeed, that’s our stance. DEVONthink can do more than one thing well. But it’s not made as – nor will it ever be – a replacement for all the apps you use. It will never be a photo editor or a chat client, for example.

We encourage people to use apps that fit their needs and their (subjective) tastes because we know the core of what DEVONthink is and can do is what holds peoples’ commitment or brings them back in. :slight_smile: