Linking is the process of creating pointers between two pieces of separate information. In the case of this zetttelkasten, the information is separated by posts and comments.

Problems with Linking

  • Links can be distracting
  • Link is not helpful because the information it leads to is either of low suprisal or useless
  • Can be low suprisal by sending you relationships you already know, context you already know, and useless information.

Good Links Are

  • High in Suprisal
  • Provide you with Needed Context
  • Show you useful information
  • Show you connections you weren’t aware of (suprisal)

Zettelkasten and Linking

I think we can distinguish two aspects of linking, which you allude to:

  • Selection - Which links to show?
  • Context - How much surrounding content to show?

The selection question is interesting, as it will vary depending on the activity. If you are trying to find relevant connections, you will want that surprisal connection. If you are trying to synthesize a coherent narrative, you may want a more structured representation of the connection.

I think jump links (traditional web links) are generally inferior to transclusion, where the destination content is embedded into the existing context. How the transcluded content is embedded depends on the structure of the representation (graph, outliner, etc).


The difficulty in the selection process is that relevant connections are highly individualistic, with individuals being you/others or you/future you. A link to a term can be helpful when you are first learning a subject matter, but it becomes useless or a distraction after you’ve mastered it.

I can see how it would be useful for quickly navigating the structure of a model you are actively building, even if it is a structure that is already well established in your head.

I think transclusions have been gaining traction. I don’t know if Wikipedia official has changed, but the Chrome add on I use to browse Wikipedia (Wikiand) started implementing popovers on links in the past year or two. I also use popovers within my notes (Obsidian). Too bad they don’t have popovers on Discourse.

This all being just preliminary thinking on linking for me. I haven’t really settled on a linking philosophy yet.

Agreed. I tend to focus more on structure and synthesis rather than raw connectivity. I’ll leave selection to search.

Yes, actively building a model is my prime pattern of note-taking behavior.

That’s the right direction for sure. I want navigating a thought graph to be continuous, more akin to navigating a physical space then jumping from page to page. Part of the power of digital is discontinuity though, so there has to be a way to allow discrete transitions while maintaining as much continuity as possible in the user experience.

I do a lot of sensemaking on the go, on my phone, so popovers aren’t as useful for me.