Tinderbox Forum

Roam Research - interesting approach to note taking

Completely agree — this thread has been great at unearthing some features of Tinderbox I wasn’t aware of, even as a fairly heavy user over the past few years.

So an interesting aspect, hypertextually (or ‘hypermedially’) is node size and addressability. An interesting learning point for me running aTbRef over some 14+ years is big nodes are not good for allowing clear traversal of the corpus by different route, for different means, whether or not the end point is the same. Smaller nodes make linkage easier.

Think: people from different disciplines with attendant concepts/vernacular differences trying to understand the same generality but needing a different path (‘trail’) through the corpus. To me this is part of what Vannever Bush’s trails were all about.

Englebart’s NLS (oNLine System) had very tightly scoped addressability. But, for all its many impressive firsts, it remained a system used by a small highly trained group of users. Had things gone that way, I doubt the Web would be as open now as it would need a technocratic layer of users who understood its complexity.

The pertinence here of the above, is the amount of info at our fingertips infantilises us. It probably is good if Netflix can suggest a movie I might like for an evenings relaxation. Making suggestions about my work or my research—I’m less sure. Suggestions based on what? Garbage in, garbage out.

Tinderbox has a lot of thought and expertise in its toolbox, and stretching back before the dawn of the Web. Using the features as existing, offer up a lot of insight and I think it was the moment I stopped worrying about how the app didn’t look like how I wanted/thought it should look and just used the features, did I start to make progress. Otherwise, there are lots of 0.¢¢ apps out there.


Based on your experience, what constitutes a smaller node? How does this make the linkage easier?

Not ‘how’, it does, because it makes it easier to link unambiguously to the idea/concept/learning-point. Seriously, try it.

If a node covers 6 things to what does a link refer? Better to put a (primary) concept per note, making inbound links unambiguous in intent.

Although Tinderbox supports inbound links to defined target_$Text anchors, few use this affordance.

A node that doesn’t contain more than one significant concept/learing outcome.

Before anyone notes it, parts of aTbRef currently fail this test. Now I’ve internalised the lesson, I’m trying to improve the resource.

I think were are in agreement when it comes to node size. Thanks for clarifying.

1 Like

A little late to the party but: I’m also along with a revised version of Road Map that follows the text.

I am considering purchasing Tinderbox, and I’ve recently discovered Roam. The back-linking references and unlinked references display is a godsend for me, but Roam is obviously still in beta, and I find it clunky. Definitely cannot be my primary tool, but I’m trying to run all my data through it purely because of the features being described here. FWIW, and I know it’s not worth very much, if Tinderbox had the ability to display this information, I’d buy it in a heartbeat.

I have been doing a huge Civil War-era historical research project, with tons of data, mostly using Scrivener and Evernote. A lot of it is transcribed newspaper articles. It is difficult to organize and track it on Scrivener, which doesn’t really do useful tags and has somewhat clunky searching, which displays little context around the term. An article may mention several people and several key events, and then there is keeping track of its date of publication, where it was published, and the date of the events described. It’s very difficult to keep a separate file for every single person of any prominence in the 1800s, and connect the note to all the other information. Copying and pasting it throughout the document makes it bloated. And, sometimes, I won’t know who “Mr. Smith” is, and only later will realize he is “Caleb B. Smith,” but in the meantime I’ll have totally forgotten about the mention. I regularly review notes from not terribly long ago and am amazed at all the things I failed to pick up on, like an earlier reference of someone I now know is important.

What Roam and Bear do is allow me to quickly paste the articles and put hashtags in front of every name in articles where the person is actually importantly featured (for example, in Scrivener, if I search “Grant” for references to Ulysses S. Grant, it will bring up all the times the verb “grant” is used), and also before key terms. This process takes a fraction of the time and tracks every name in advance, and each note by multiple topics. But Bear and Scrivener only tell me where something is mentioned, and my documents are usually long. Roam shows me the exact context, all on one page. And with the unlinked references, it does it without me even having to go in and hashtag the names. Transclusion is also helpful since I use the same article in many narratives. I offer this as an example of why someone would want the ability to do these things, and also because I’m trying to figure out if Tinderbox will help me get more organized. It definitely would make it easier to sort the data, but I’d have to spend a lot of time labeling everything by its attributes first, I think. Still would beat trying to have a hierarchy of 10,000 Scrivenings.

If you search a bit on the forums you will find various discussions of the new Ziplink feature in Tinderbox. You will also find some recent discussion about importing Scrivener projects into Tinderbox. At the risk of stating the obvious, you could set up a trial Tinderbox file and see how you like working with the application. Not everyone takes to it.

My approach to working with a large amount of source material, such as documents and transcriptions, was to use DEVONthink. I’ve stopped using that program nowadays, but it might suit your needs. You might also profit from looking at the discussions at www.zettelkasten.de where there is a lot of discussion of methodology in “knowledge work”.

Best of luck,

UIs can be seductive, but it’s detail that counts. There’s a trade-off to auto-[whatever], that is two fold:

  • it doesn’t warn you about what it can’t/won’t find (but may be highly pertinent).
  • it encourages us to thing what ‘computer knows best’

Accepting the above, my move has been away from seductive “don’t-make-me-think” features. Also, if the app is really doing all the hard work, what intelligence am I applying to the process.

One conceptual/methodological issue is whether you want simply provide large runs of text and the computer guesses the rest for you. Or do you want to understand the content with more precision? The latter is best done by creating metadata that help acts as both an index and scaffolding to your understanding.

Inline hash-tags aren’t very precise. Tagging #Grant might tease Grant from grant. But the latter might be a misspelling. What if there are two Grants and both are relevant. Quick-and-dirty doesn’t do nuance well.

I’m not sure you’ll get a meaningful answer in comparing Roam and Tinderbox as they strike me as doing different things in different ways. Tinderbox works best using small notes and storing big references (pictures, large source texts) externally (be it in something like DEVONthink or just Finder). Tinderbox linking makes it easy to add notes about notes—glosses, linked to but not written on top of the text.

Tinderbox doesn’t use transclusion internally but it certainly does for export. The Preview feature (originally intended for the latter) means you can use transclusion internally too (albeit only when reading in preview).

Overlooked in all the above is using all these notes and links. Tinderbox shines here as it is easy to add structure as it emerges. Rather than tag soup, you can use attributes to tease out strands in the research. Attribute values can be exploded out into new notes. Metadata can be used to generate (other) unseen linkages, etc…

I guess the bottom line is if you want to understand your work use Tinderbox. If you want the computer to do the understanding and you just accept and use the output, other tools are available. :slight_smile:

I second what Mark says above about how Tinderbox probably won’t do what you’re looking for in terms of automated connections. Your description here, though, reminded me of this thread from the DevonThink forum, where a user did something similar, which you might find useful or maybe offers nothing more than what you already can get with Bear and/or Roam.

Than you everyone for the responses and recs, which I will definitely check out. I only recently learned about all of these options–can’t believe how “hard” they were to find–other than Evernote, which I’ve had for years, I never heard them mentioned. Collecting my data in Apple pages was not fun, but typing the articles by hand over years means I have a decent grasp on what I have, so I’m not just blindly throwing data into a program. It’s just too much to remember the fine details of. I agree it is too much to fully understand, but this isn’t like a focused dissertation project, more like storing my information, putting the pieces together on certain things and writing about them. The project isn’t really about a traditionally coherent understanding, since many articles contradict each other, and that’s a topic I write about a lot, debunking accepted narratives. But yes, misspellings are a problem, and some people’s names were actually spelled multiple ways. But I do think I could benefit from learning the features of Tinderbox and teasing the topics out over time, as Mark said. I can just combine it with other programs for the storage and searching. I’m going to end up making a lot of it public (currently working on a TiddlyWiki) for the benefit of other researchers, because there’s just not a market to try and publish all of it in books, and I have to organize it better for that purpose. So I still may get Tinderbox at some point. Scrivener is running really slow under the weight of the information. Thanks for again for the tips!

Early in this thread, there was a call for

seamless, bidirectional links, and the ability to immediately view and access these in a note

I want to come back to this, though it might be too soon.

I’ve written before about why I think bidirectional links are less useful than directed links, but this thread did show that they have advantages for some brainstorming and composition tasks.

Tinderbox 8,6 is an attempt to capture the advantages of seamless, bidirectional links while retaining directed links. My sense is that this works out fairly well — better than I expected. Agree?

1 Like

Agreed! The bidirectional links (backlinks) are extremely useful and much appreciated.

It would also be very useful if there was a solution to the second part of the ask:

Find all other notes that mention this concept by simply clicking on the link I just created, without having to create an Agent or searching for it. Navigating to that concept’s page will reveal not only the note’s name, but the text in which it is embedded. Importantly, the concept I navigated to will also reveal links (and the text in which they are embedded) to other text that include this concept.

Teasing apart UI and hypertextual/network function, is the issue here the ‘effort’ of getting back to the note we just left, or something deeper. It reads to me as mainly the former and ISTM that in the short term letting the Go Back link following the last link, as opposed to the last basic link, would solve some of the ‘difficulty’ of going back to the previous note.

† That basic link bias feels like an inherited Storyspace behaviour. IOW, it does need to be that way in Tinderbox.

Having pairs of one-way links will make for some really messy maps, noting that maps seem to be the most favoured view.

I agree. The new link pane makes all links to and from a note easily visible and click navigable. This addresses the only “backlink” functionality I personally need. I also like that it’s been done without eliminating the advantages that single-direction links offer.


I agree. But one area which is, I think, due to be revisited is the way we draw links. More might be done with link routing, even in complicated maps.


Have to say, loving the new 8.6 upgrade, and impressed at the quick turnaround for adding these features!

Sure, and I’m not a nay-sayer to ‘reverse’ links. Tinderbox already offers—to the experienced, at least—ways to modify or hide link types.

It would be useful, perhaps if all links between same notes were drawn with a single line and arrowhead(s) indicating direction(s) used. Conversely, I don’t know the degree of usage built off current behaviour where links with the same source/target aren’t aggregated to one ‘line’ on the map.

At this point, my feeling is for aggregation as the other approach (separate line per link) is essentially a drawing app task (IOW something better done outside Tinderbox). Making links back into items of structural significance, rather than as drawing affordances, also gets Tinderbox closer to its hypertextual roots and enhances the existing feature set.

I set my default links display to “linear” precisely because Tinderbox displays these double links as a single link with an arrow on each end. So this option eliminates a lot of noise from busy maps. My experience is that non-linear links draw so as to create two distinct link lines on the map, one in each direction, often curving in different directions towards their target.


Yes, this has been a fantastic addition to Tinderbox, which really enhances its application as a Zettelkasten (which concept itself appears to be gaining increasing momentum at the moment among knowledge workers.)

I was wondering whether the functionality couldn’t be further augmented by the addition of link contextualisation. This is something Roam does very well, allowing one to see the surrounding sentence/paragraph within which the link was contained, from the note that has been linked to. I like Andy Matuschak’s thoughts on the matter (can’t provide link as a new user I assume):

“Systems which display backlinks to a node permit a new behavior: you can define a new node extensionally (rather than intensionally) by simply linking to it from many other nodes—even before it has any content […] This effect requires contextual backlinks: a simple list of backlinks won’t implicitly define a node very effectively.”

Even providing a bit more context when hovering over a link in the new links pane would be really beneficial for the connection-forming and sense-making processes.

1 Like