It may not surprise you to learn that people use Scrivener for exactly this.
I totally agree. The reason I’m splitting notes is because I’ve decided for now to make DT the first home and hub of my gigantic, 10-year-long history of self-help/human potential/self-development project. I chose DT because of (a) replicants and (b) DT Groups can contain literally anything without fuss or translation.
The history contains every person, text, concept, thesis, theory, school, and tradition in the fields of philosophy, religion, Western esotericism, psychology, and then these two weird more recent movements—(a) the late 19-cent “occult revival” in France and England, and then (b) the 20-cent explosion of movements in the US that eventually became “self-help.”
This latter phase is gigantic, complex, and also hilarious. It includes —
- Crowley’s invention of placebo exercises, perfected later by LRH
- W.P. Patrick’s Holiday Magic, which repurposed sales training (!) as therapy, perfected by Erhard (est/Landmark), who combined it with LRH drills
- The translation of Sanskrit, which produced the systematic occultism of Theosophy and treating of religion as metaphor therapy
- Dale Carnegie and N. V. Peale, who saw that communication and assumption improve lives
- Huxley, psychedelics, and human potential movement
So I got ChatGPT to give me a list of 100s of names. I place this text file in DT and explode it into 100s of records (maybe markdown), which I then run my great make-groups-and-tuck script, which replaces them with groups and tucks the original records inside.
Tinderbox is even better at exploding: merely pasting a chunk of text that contains lines starting with TAB will explode into proper outline structure. Great for seeing form. But it’s not a form you can quickly fill by dragging in all filetypes.
DT does allow this, and it has replicants, which are key here since one node might properly belong inside many histories/diagrams/timelines. So I saw DT as a good place to set everything up.
What about export and final publication, where I want to also have a great UX? I’ve looked at —
- Obsidian publish
- BoxPress (an independently developed Tinderbox export system that turns your Tinderbox maps into clickable HTML image maps and exports your whole TBX as a self-contained portable HTML bundle)
- Kumu, probably the greatest option of all, but I doubt that its working-and-thinking environment is a good place for all my stuff. The great thing about BoxPress is that it WAS my working place, and I was exporting from it directly.
The end product will be a website. I want the UX to let the user —
- Click on a node (person, school, concept, model, theory, etc.) to read more, the more containing lots of links
- See that node in the multiple “influence diagrams” in which it appears. E.g, idea A was made famous by person Z, but it is actually a spin on older idea Y, created by persons B, C, and D. And the new spin was shaped mostly by idea E, invented by person F. Incidentally, idea A went on to father ideas G, H, and I.
- See everything in a single gigantic imeline, which you can filter. Remove all philosopher? No problem. See occultists only? No problem. See only topics containing a certain tag? No problem.
- Same for the influence diagrams. You want to see only the people who directly influenced Freud? Here you go. You want to trace influences back to the Stoics? Here you go.
I had planned on using BoxPress, but the Tinderbox developer effectively destroyed export speed around version 7. What used to take 50 seconds to export suddenly actually took over four hours. And as of today (version 10), it’s even slower. I actually kept a log of the new speeds that I carefully measured and recorded, which I planned to send to him along with a plastic bag full of my tears, but I was too hurt to carry it out.
He also made changes with $DisplayExpression, which I used extensively. What used to work and without lag suddenly caused INSANE lag—I mean 7 seconds of beachball for every 10 seconds of real time. When I emailed him about it, he advised me not to use complex display expressions any more. It taxed the CPU too much. (But it didn’t in the version prior!)
After that one-two punch, I had a psychotic break and stopped using Tinderbox completely, which was hellish since I had 10 years of nicely linked notes in there.
So am now using DT to split lists of 100s of names into lists of 100s of groups, which I can then fill with anything and (because of replicants) move anywhere—without any lagging or beachballs.
My progress is slow, and made slower because I keep experimenting with making it work in different programs. Where “it” means both (a) the organizing and note taking, AND (b) the setting up of the final diagram-rich export with a great UX—something like Kumu, ideally.
(BoxPress was so great until the anti-export tariff was added to Tinderbox. Just thinking about the day export died fills me with yellow bile and shuts me down.)
Another once-promising option: Excalibrain. That seemed like a good way to work, and maybe even export. And using YAML to hold all the relationships—that’s like future-proofing my notes. As long as the relations are spelled out in the metadata somewhere, I can write a script to shape it into something the new exporting program can understand.
But Excalibrain can only show the parents and children of a node, so too bad.
Mermaid is another option. Maybe I can put links inside the nodes and realize my dream that way.
So I am open to any ideas about a way to put 1000s of nodes into a system that would let a website visitor —
- explore and display the nodes in many ways: so that she can see —
- where idea X sits in this influence-diagram
- where idea X sits in that influence-diagram
- idea X’s concept-subordination tree (showing the analytic parents and children of node)
- who invented idea X (and the inputs that shaped it)
- who championed idea X (which is often different)
- and so on;
- explore and display with much control (like the filtering and sorting kind we enjoy with databases)
If anyone has any suggestions, please let me know. At the very least, I’ll put you on the Acknowledgments page and use whatever self-description you send me.
P.S. — I also considered using (and publishing from) Obsidian Canvas, which seemed like the best BoxPress replacement. But now that Mermaid exists, I don’t think manually adjusting every time a new node is added is smart. If we can store all of a node’s relations inside that node—as TB attributes, YAML properties, Notion properties, or DT custom meta data—then it’s all-important relationships are safe and can someday be resurrected as a labelled link-line in a diagram, in some program.
Have you looked at Visualizing SEP?
https://www.visualizingsep.com
I don’t know how it is achieved but it might be worth asking.
Jesus Christ. That captures almost exactly the “influence diagram” part of my dream. Except that mine could show more “generations” (what Obsidian Graph View calls “depth”)—meaning that not only parents and children, but the parents’ ancestors and the children’s descendants up to the user-chosen level.
Actually, I spent a lot of time staring and the Navboxes and Categories portions of Wikipedia. Those things are amazing. Those things supply a ton of relation information.
Imagine if ChatGPT could —
- comb through 500 Wikipedia pages
- see all of its person/concept/school/tradition inputs
- see all of its respective outputs
- and print me a list that I could explode into markdown notes containing YAML sections that hold all the relevant properties.
Then the project would be 90% complete except for rendering it as the half-dozen diagram types I have in mind (influence, concept subordination, timeline, etc.).
Thanks so much for sharing this. The display method is absolutely incredible (except for the fact that making the window wider crops the diagram).
I’m glad it is helpful. You might want to look at this interview with the guy who created the website, to get a bit more insight into it:
Cheers!
Thanks to @mbbntu for the video. I have tried the website but was not aware of this detailed explainer. It’s an impressive project. I have no doubt that it can be very helpful to someone.
I have all SEP articles, up-to-date, in one of my databases. They are always accessed through full-text search. I used to be impressed by the graph view and spent quite some time playing around that in Obsidian, however I have found that graphs do not actually help with my work. Here’s why:
- From a subjective perspective, any connection between nodes falls into one of the following two categories: those I already understand, and those I don’t know yet.
- I don’t need anything about the connections I know. I’m interested in the unexplored connections.
- I have to open and read the relevant part of a linked article in order to understand why it is linked. There is no shortcut (unless you trust AI summaries).
- The graph view is just one of multiple ways to present all linked articles in a single place. Text search does it, in a better way. While clicking a search results brings you directly to the first mention mention, clicking a graph node does not.
If I increase the depth of a graph, it may reveal more connections than I can get from a text search. There is a major drawback, though: getting what you want from a deep, complex graph demands considerable effort. It’s guaranteed that I would be distracted from what I was thinking of.
Distraction is not an issue when my brain is empty in the first place. That means a deep graph is useful for at least two things: an introduction to a subject, and a summary of it. I suppose the Visualizing SEP project is fantastic for beginners looking for an onboarding experience, but less attractive to professionals of the relative fields, whose typical workflows involve few introductions or summaries.
I have since returned to using dedicated TOC notes for outlining. I still like fancy graphs, but I’m unable to make them useful. It should be noted that I don’t consider myself a “visual person”.
Ugh. You’re making me glad I never really clicked with Tinderbox.
Not to be a wet blanket but…
Dreaming is free. This reality would be $$$
Wikipedia is open source. You don’t need ChatGPT to download the relevant pages and the links between them, including the code that generates the Navboxes and Category pages. And once you have that, building your own site map is a pretty standard webcrawler-type task.