Reevaluating my apps and finding what I need

Spent a lot of time recently kind of reevaluating my apps. What do I really need? What can I minimize? But not trying to go too far. You know sometimes when you get in these minimalism modes you go too far and you get rid of stuff that's actually useful. And in the past, I would have kind of just jumped into whatever note looked shiny and interesting at the moment. But what I've actually been doing for the past couple weeks is just thinking about it and giving the problem time. I say problem in the, you know, the sense of a mathematical problem not as in the more common usage of problem. It's something for me to puzzle over. Something for me to, like I said, think about. To go in depth on.

A lot of this is driven by artificial intelligence and the realization that certain paradigms of apps may no longer be necessary. And the more that AI improves, the less and less those things are gonna be necessary. So like complicated folder structures and complicated workflows within apps. Probably we won't need those at all. Right now, semantic search in the apps that have AI-enhanced semantic search is pretty amazing. I think it's like I'd say 80 to 90 percent accurate. So you can only imagine that that's going to improve. But I mean right now it's enough to consider the fact that you can pretty much find anything you want. So I don't worry about those things anymore.

So what I switched the question to was what do I need or what do I want that isn't about finding something. Or in what ways can I find apps that will help me find things that I don't know I need or that I don't know that I want. You know, like an example would be you're looking at a file. Maybe a piece of writing that you're working on. And in the sidebar, things are suggested to you that are related or useful to what you're doing right then. Some apps do this okay. The only one that I found that actually does this really well is if you get the third-party plug-in Smart Connections for Obsidian. That works really well. Even though the local models work really well, but if you plug it into OpenAI, it's eerie how good it is. And it pulls up things that are semantically related, not just matching keywords like I said jogging.

So here's a file that says jogging in it. And I tested this actually; I went through, like I said, I've been putting a lot of thought into this. I actually tested this. I went and I had ChatGPT write me a short description of the novel A Passage to India. Then I asked it to write the description of a book with similar themes. And then I asked it to write an atomic note about something that would be thematically connected to A Passage to India but not specifically about A Passage to India. And if I remember correctly, the note that it wrote was about cultural differences and the way that communication between cultures can be stunted because of the way that they don't understand each other. Which thematically is what happens in A Passage to India but that's not even what you would say is the plot of the movie. But it is that it's thematically similar to what happens in the movie. I mean, I said movie twice. It's also been made into a movie, so that's where my confusion is coming from. And the other one, the book, I can't remember if it gave me—oh, I think it was The Boy Who Would Be King or The Man Who Would Be King. It was a Rudyard Kipling book that took place in Pakistan, but it is thematically very similar to A Passage to India.

So I put those three things into multiple apps and saw how it did it. Like I put it in Reflect Notes. It's okay. I put it in Sublime. Theoretically, Sublime did well, but Sublime isn't about putting the most relevant result to the top. It's kind of about shuffling and just kind of showing you random stuff that's related. So in some ways, it's not a fair test for Sublime because it's asking it to do something that it's not doing. But if you scroll far enough, you will find those notes suggested as related to each other. They just get buried amongst everything else. And then I put it in Mem. I tried Mem's new 2.0. Wow, that one was disappointing. Granted, still in beta to be fair.

And I think that's it. I think these are the only apps that do this right now. I know that Capacities is working on it. To be honest, my faith in Capacities has dwindled a little bit. Just, yeah, I think it'll work okay, but it's not gonna work to the level that I want. And I was actually resistant to doing third-party plugins with Obsidian. I just feel like the more you add to an app like that, the more clunky it gets and the more messy it gets. But man, that plug-in just works. Well, it nailed if I went to A Passage to India, the other two things were like within the top three suggestions. If I went to one of the other ones, same thing. So they were always connected to each other.

And this makes things very interesting because you can actually create a pretty powerful Zettelkasten. Fairly traditional. A Zettelkasten, and I say fairly because in order for it to be traditional, there's certain things about analog cards that you can't really replicate digitally. But as close as you can get in a digital format, you can do some pretty amazing things using that Smart Connectors plug-in because you have your implicit connections. Things that you've yourself have connected, like I want this note to be connected to this note, you know, like an author to a book or a note about colonialism connected to A Passage to India. But then in addition, now you have suggested semantic connections as well. So now that the possibility when you go to that note—because the idea, of course, was Zettelkasten is it's not about retrieval. It's about kind of getting lost and finding ideas you've forgotten about or seeing them in different contexts.

So when you go to a note that's supposed to be a jumping-off point, go to a card that's supposed to be a jumping-off point, you're supposed to jump to the next one and jump over here and jump over there and see where you end up. So if you can imagine doing that digitally with Obsidian and going to a note, you've got your connections you've already made. Maybe those are good places to jump to next; maybe they aren't. But now you also have the suggestions of all these other places that you can jump that you know are semantically related with a score telling you how semantically related they are, which is pretty powerful. In addition, there are settings in that plug-in where you can tell it do not suggest notes that are similar to this one if they are already linked in the file, which is one of the things that other ones do not do. And that's really powerful because then in some ways you can begin to use the sidebar as kind of a Zettelkasten inbox. Because all you're gonna see on the side is stuff that's not connected yet.

So as you begin to explore those in those, "Okay, how is this connected? What do I feel about this?" And then you maybe you do like it. Then you link it in the note. It's gonna come out of the sidebar. But now it's linked in the actual note. That's pretty cool. And this just discovered all this stuff because I wasn't busy trying to make a system for myself. I was just exploring the capabilities of the systems while also thinking about the possible ways that I need and want to use apps and also rethinking the way that we use apps within the context of AI. Now there's a lot of other thoughts I have about all of that. But yeah, I think that's enough for now.