Can mentions and inline search support slow hunches? 🐢

Hi everyone,
I’m sorry, we had a server error and I only had daily backups active.
The below post by @hanbzu was lost. I restored it from my email notification.


Post from 2. October 2020 by @hanbzu:


Slow hunch is creating an idea slowly, in small bits—opposite to the famous eureka in the bathtub moment. Also referred to as slow burn.

Charles Darwin himself (in his autobiography) said he came up with the idea of natural selection in an eureka moment but then later Howard Gruber studied Darwin’s notes and came to the conclussion that it was more of a slow hunch.

This seems like a good thing to exploit when building effective tools for thought.

We’ve seen this getting popularity with Roam Research, where people describe they use “backlinks” (= mentions) as a way to slowly build content for a page:

  1. Whenever they encounter related content they write a comment and they [[mention]] the target note, event if it’s empty.
  2. They navigate to the note and, out of the collection of backlinks they sketch the content of the note.

What I’m trying to do here is analyse the process.

I see two things that afford this:

  • You can mention inline, whether it’s by typing @ or [[. This affords for disambiguation (the autocomplete interface shows options, you pick one) without losing the train of thought (it’s a micro-interruption because it does not involve a big search). @johannesmutter has talked about this naming it “pull from workspace”. It effectively reduces a linking task to a micro-interruption of the flow you are in:
  • Mentions are visible, so you can see the paths that lead to the place you are in. This is giving a lot of information of that place from different angles, which is a proto-description, or even a complete description, depending on the mentions.

@johannesmutter thanks a lot for restoring it. Also I’d like to thank you for hosting the site. If there’s any way I can help/support you doing that, please tell me.


Just for reference: Under this topic I also mentioned Promnesia, a browser extension by @karlicoss. Because I believe that the basis for slow hunches is being able to build upon small interactions with information in the past.

Here’s what the author @karlicoss (maybe his new profile is also gone in the downtime) replied (restored from email notification):

IMO a very important thing that helps with slow hunches is some form of quick capture/clipping.
It has to be seamless and mechanical, requiring minimal cognitive effort.

When I’m programming, I’m very liberally using todo comments for that. If I have some side thought/improvement idea, I just type //todo: idea and forget about it immediately. I probably have a todo per 10 lines of code on average. Now and then, I can deliberately go through todos, and either delete them if they turned out to be not so relevant, resolve if they are minor, or promote to more specific tasks.

Otherwise, most of the time I’m using org-capture (in my Emacs), or quick capture in Orgzly on Android to log interesting thoughts/ideas I don’t have time to think about at the moment. It’s incredibly quick, just a matter of hotkey, typing your thoughts, maybe a tag; done. Similarly, now and then (e.g. every 3-7 days), I would go through the captured ideas and refile them somewhere more specific.

I have a similar approach when I’m in my browser, I even wrote an extension for that: https://beepb00p.xyz/grasp.html

Similarly, if I see a somewhat interesting link, I can capture it with a single hotkey (possibly add some comment), and it’s immediately in my org-mode capture file, so I can easily find it later, refile, etc. Basically I use that instead of browser bookmarks. This is actually one of the reasons I wrote Promnesia, so I have a proper integration of my browser with my notes.

1 Like

Here Conor White-Sullivan describes a writing process in Roam:

The first part is a way in which we can do slow hunches through mentions (backlinks).