Stop Taking Regular Notes; Use a Zettelkasten Instead

That's a good quote. I've been trying to figure out what a programmable wiki would look like and I always end up with something that looks like a smalltalk VM.

I think it's possible to build a personal memex now. I'm looking at couchdb, lucene, apache tika, deno, and node.js as the initial set of tools to mash together and expose through the browser. CouchDB and Lucene will be used for persisence and searching along with Apache Tika for extracting metadata and doing OCR on images and PDFs. Deno and Node.js will be used for executing code on the server and the client. Deno can be used for sandboxing and exposing each person's knowledge base to programmatic control by other people like a federated search engine where people can share interesting code patterns for acting on knowledge bases and exposing hidden patterns.

Combine with some basic machine learning models and you get a pretty useful personal toolkit for knowledge enhancement.

Also, a quote by another good thinker

> Civilization advances by extending the number of important operations which we can perform without thinking of them. - Alfred North Whitehead

Maybe a description of my notetaking practice will be useful to others. I don't know if this is strictly a "Zettelkasten" or not.

My note box for the last little while is Dercuano: http://canonical.org/~kragen/dercuano. The notes are longer than a single index card (there are 882 notes averaging about 1300 words each), and they are hierarchically organized into sections and subsections. They often include links to other notes or other web pages, and they also have a chronological order and a set of tags. And I frequently grep them. I kept them in Git to keep them synchronized across different computers.

Typically a single note draws from many sources, and I tried to keep their contents in a readable order, which is easier to do on a computer where you can insert stuff in the middle. Some notes are one afternoon's thoughts, while others got added to and reorganized for years. So that's one way I make connections: I add hyperlinks, new data, explanations of how and why the entire rest of a note is wrong, new sources, etc. Another is that the process of writing a note involves bringing together different ideas, since few of the notes are just summaries of things others have written.

I learned a lot writing Dercuano, but in the end I was more frustrated than effective. I have a series of paper notebooks that are much less voluminous, only a few hundred pages a year, which are worse for text and hyperlinks, but much better for drawing, even things like schematics. Computers offer the potential of doing simulation, calculation, logic, optimization, rendering, data analysis, and data integration, the things that Engelbart and Kay envisioned in the 1960s. Yet Dercuano is stuck in the textual world without even transclusion or sketches. A much better medium is possible.

And we have lots of examples: spreadsheets, Jupyter notebooks, observable explanations in JS, diagrams in KSeg or GeoGebra, Observablehq and all of Bostock's previous excellent work. These show us that a better medium is achievable. But incorporating such a model into a note like the ones in Dercuano maybe requires some UI work. I mean the whole revolution of VisiCalc was that you could see the results of your calculation as you changed the calculation, and that doesn't really fit well into editing text files in Emacs.

Things like Darius Bacon's Doe and Halp do go some distance toward that integration for textual output. And R-Markdown integrates statistical analysis and data visualization with text, and Ward's Smallest Federated Wiki incorporates different media types on one page, including bytebeat. And I understand that org-mode SRC blocks offer the possibility of displaying image output inline. Still, none of these seem like they can scale to interactively doing a compass-and-straightedge construction or drawing a schematic for an analog circuit. Maybe I'm dismissing them too readily?

So I'm trying to figure out what this looks like in https://gitlab.com/kragen/derctuo.

I'm confused as to why you find Obsidian not matching most those criteria. I heard about it here last week, and just installed it finally tonight, but as I understand it. It's just a bunch of markdown files in a folder on the local disk. That seems.pretty easy to access/convert, is private, is lightweight, and is future proof (I could write something in Perl to generate the graph in a night most likely).

I don't know anything about the others, since I haven't really used a note system before, so I can't comment on them. Your system seems to violate a few of your stated concerns though (is it cross compatible?), so I'm confused as to why you find it better.

It's not really new, but finally the years of spreading the cargo-cult is paying.

There is now roam, a new fancy tool which does nothing new, but spreads hard with the new generation and connects well with the zettelkasten-terms. And there recently was was a book about smart notetaking, which sold somewhat well with people who seek some solution but don't know much about this space.

> you gain nothing by creating a new linked note under the 'for loop construct' heading. Just do some practice problems instead!

This is when you are learning very basics and not equivalent of literature review.

The proper equivalent would be doing overview of current frameworks. Or learning about bitcoin where you are going through math, design, algorithm, economics considerations and social consideration. Or security where you jump between modeling, attack, defense and tradeoffs.