채널: Greg Isenberg — 2026-02-26 241 프로그래밍 8e31b4

How I Use Obsidian + Claude Code to Run My Life

채널: Greg Isenberg 링크: https://www.youtube.com/watch?v=6MBq1paspVU

트랜스크립트

[0:00]

[0:00] This is Obsidian. And Obsidian is this [0:03] little tool that people are using as [0:05] their second brain. But what's really [0:07] cool about it is they're pairing it with [0:09] Claw Code and they're getting crazy [0:12] results out of it. It's literally a [0:14] gamecher. Now, I've been slow to adopt [0:17] Obsidian cuz to me, it's been a little [0:19] daunting to look at. So, I had my friend [0:21] Vin and he clearly explains what [0:25] Obsidian is, how to use it with Claude [0:27] Code, how to set up these commands [0:29] [music] that really drive the most out [0:32] of Claude and all the LLMs. And it's an [0:35] incredible episode, like a really [0:37] gamechanging episode because I think [0:39] that people who understand how to use [0:41] Obsidian and how to use Claude Code [0:43] together, they're going to be able to [0:45] live happier, healthier, and wealthier [0:48] lives. Why? because it gives incredible [0:51] ideas to you on tap. So, I know that the [0:55] people that stick around to the end of [0:57] this episode, I think that for a lot of [0:59] them, it's going to [music] absolutely [1:01] change how they use AI and it's going to [1:03] be a super impactful way because you're [1:05] going to get better ideas at the right [1:07] time, the right moment, and it's going [1:09] to make you happier, healthier, and [1:11] wealthier. Enjoy the episode. [music] [1:18] >> [music] [1:21] >> I've got my dear friend Vin, also known [1:24] as Internet Vin, on the podcast. I [1:26] literally begged him to come on. I [1:29] [laughter] [1:29] I begged him. I begged this man to come [1:32] on and to teach us a very specific [1:36] thing. Vin, by the end of this podcast [1:39] episode, what are people going to learn? [1:41] I want you to have an understanding of [1:43] how you can use claude code and obsidian [1:46] as a thinking partner. I want you to [1:48] have an understanding of how you can [1:51] stop uh having to explain things to [1:53] agents over and over again and just pass [1:55] specific files in. And I want you to [1:57] understand how you can use Obsidian and [1:58] Cloud Code to notice things about the [2:01] way you think that you would not have [2:02] noticed on your own without these tools. [2:04] >> All right, from your lips to God's ears, [2:07] [laughter] let's get into it. [2:09] Okay. So, first is like what is cloud [2:13] code? So, claude code is this like agent [2:17] that you can use um in a command line [2:20] interface. So, it's just basically this [2:22] tool you can use that can control your [2:24] computer and you can use it through [2:26] natural language. Right? So, I can say [2:29] make a file or make a file on my desktop [2:33] that says hello Greg in plain text. [2:37] Right. And it's going to go and do this. [2:41] That's really cool. That's that's [2:42] something that's that's new that wasn't [2:44] possible before. Before this, I had to [2:47] go to the desktop, open some text [2:49] editor, and then like create that file, [2:51] right? And now this file is on my [2:53] desktop. So, I can say open the file. [2:56] There we go. Hello, Greg. That's crazy, [3:00] right? Um, now what's interesting about [3:04] this is if you have this agent that can [3:06] like control and do things on your [3:07] computer, that means that whatever you [3:10] can describe to it, like it can start to [3:13] do. And so when you if you describe a [3:16] project to it or you get into these long [3:17] conversations uh with an agent, [3:21] um, it can do more and more complex [3:22] things. The more information it has, the [3:24] more complex things it can do. But the [3:26] problem is that if you have to let's say [3:28] you know like I write some super long [3:30] description about a particular project [3:32] or I have like an hour conversation with [3:34] this agent about a particular project [3:36] it's like I don't want to have to create [3:38] a new session to explain that all I [3:40] don't want to have to explain that over [3:41] and over and over again. Um, a lot of [3:44] people are using like claude uh or [3:46] chatbt on the web and it has things like [3:48] memory but you you can't like control [3:51] you don't know what's in that memory [3:53] right you don't know what it knows and [3:54] what it doesn't know and so there needs [3:56] to be some way of like you know passing [3:59] information into these agents that is [4:01] easier and faster and the better [4:04] information you can give it and the [4:05] faster the information you can give it [4:07] the more stuff if it can do for you and [4:09] the better the faster you can delegate [4:11] to it. Okay. So now even if I let's say [4:16] let's say I had like you know let's say [4:18] I wrote like a big project description [4:20] here, right? create a file that [4:23] describes, you know, a project um [4:29] about [4:31] a to-do list app that um [4:36] is very minimally designed [4:39] and reads from uh all of my calendar [4:43] and my messages and um my Slack and [4:48] Slack and interprets it into a task list [4:52] that of tasks that it thinks that I [4:56] should do. I don't know some some idea, [4:59] right? So now this this is a file that

[5:00]

[5:01] could be on my desktop and what I can do [5:04] when I use cloud code is I can reference [5:06] that file and pass it in whenever I [5:08] want. [5:08] >> And why that's important is because [5:12] >> it's the context, right? [5:14] >> Yes. [5:14] >> The whole game is [5:17] >> feeding the beast good context. Yes, [5:21] exactly. And I don't want to have to do [5:23] this over and over again. And and and [5:25] when I work on this uh over days, I I'm [5:28] not going to remember like what we [5:29] talked about, right? So, I want some [5:31] kind of file uh that I can like pass in. [5:34] Oh, sorry, Greg. One sec. [5:36] >> Yeah. And that's that's sort of the [5:38] problem that a lot of people are facing [5:39] with cloud code is like they're using it [5:41] and then they're saying, well, it's it's [5:43] okay. It's not like gamechanging. Mhm. [5:46] >> And the issue is they don't have they're [5:48] not they're not feeding the right [5:50] context at the right time. [5:51] >> Yes, exactly. And so here's so here's [5:55] like a a project description that it [5:57] wrote and obviously I can pass this in [5:59] and this is like a this is like a [6:00] general one that I just created but you [6:02] you can make these like very complex you [6:04] can build them into like robust files [6:06] right over time. So we know that uh [6:09] cloud code can create files and it can [6:11] repeat and it can um uh read files right [6:13] so now I can say let's say I created a [6:15] new session so here's a new session and [6:18] now I can say like I want to work on [6:22] this project and I go here and it's [6:24] going to be um to-do app here it is boom [6:29] now I didn't need to explain the file [6:30] again I need I didn't need to explain [6:31] the project again right so it's going to [6:34] read this file and it's going to start [6:35] like you know That saved me a lot of [6:37] time. Great project. Before diving in, a [6:39] few questions to scope the first [6:40] session. So that'll continue. So now [6:42] what is Obsidian? Right. Obsidian is [6:46] this tool that it's kind of like an [6:49] interface that sits on top of um uh a [6:53] collection of markdown files, right? So [6:56] here like this is reading a markdown [6:58] file. How I use Obsidian story [7:00] development, right? Um I have uh daily [7:04] notes. This is my daily notes. This is a [7:07] This is also um a markdown file. I [7:10] should do my own fundamental analysis [7:12] into thing into how things stay pure [7:14] when they grow and become more [7:15] mainstream. Right. This is just a file [7:17] that I have. I have um you know like a [7:21] file on Greg Eisenberg that I haven't [7:23] put in. [7:23] >> Oh, that's weird. That's weird. [7:25] >> Yeah, [7:26] >> that's pretty weird. [7:27] >> And so I I make files like notes of [7:29] things that I'm learning from people and [7:31] stuff like that, too. So I have [7:32] different um files for everything, [7:35] right? And the interesting thing that [7:37] makes a vault, which is Obsidian, what [7:39] Obsidian interacts with, this whole [7:40] thing is called a vault. What makes it [7:42] different than a folder is that Obsidian [7:46] is not only interacting with just like a [7:49] you know a folder of files. Um but what [7:52] it does is it also allows you to make [7:54] inter uh to connect relationships [7:57] between files. So I can say today I am [8:00] on um a podcast with Greg Eisenberg. Now [8:05] this file is linked to that Greg [8:08] Eisenberg file. Super interesting. Super [8:10] interesting. And so when people like [8:12] like like people uh there's a lot of [8:14] people who really really like using [8:16] Obsidian and tools like Obsidian because [8:18] of this ability to form inter [8:20] relationships. This is unique to just [8:23] having a folder. a folder on your [8:25] computer cannot show these inter [8:27] relationships. [8:28] And so it gets really interesting when [8:31] you start to um keep making these like [8:34] interreationships over time, right? And [8:37] so what happens here's a little [8:39] visualization. And so in the here, these [8:42] are each one of these circles is a file [8:45] and and and it's it's showing how it's [8:47] like connected to all of these other [8:49] files where I've written things about. [8:51] So here's like personal agent [8:53] infrastructure, [8:55] right? And so I could look, you know, [8:58] and I guess I should also add just kind [9:00] of a comment on this and what was [9:01] difficult about doing this demo, there's [9:03] like so much personal information in [9:05] here because this is like my personal [9:06] thing. So I don't even know like what's [9:08] going to show up on the screen here, [9:10] right? Um, but that's part of doing [9:12] demos like this, which is which are kind [9:13] of weird and interesting. But you can [9:16] see personal agent infrastructure. It [9:17] links to like Aentic AI. There's like a [9:20] link here to Telegram. There's a link [9:22] here to like Toby, [9:24] the founder of Shopify. There's a link [9:25] to like Presence Log, Claudebot, you [9:29] know, and then here's like I I have a [9:31] podcast too called the other stuff. And [9:32] like you'd see I'm obviously doing a lot [9:35] of like thinking about that a lot, [9:37] right? And so I can also, let's say if I [9:42] go to Greg Eisenberg and I go to local [9:44] graph. So, here's like all the times [9:46] I've written about Greg Eisenberg, [9:48] right? Notes on time constraints, how I [9:50] use Obsidian, which is just kind of [9:52] interesting. Um, so if I'm listening to [9:55] >> I love you, too, Vin. I love you, too. [9:57] >> If I'm listening to a show and I'm [9:59] picking up different patterns, I can I

[10:00]

[10:00] can reference that back to to Greg. So, [10:03] that's really interesting. But here's [10:05] the thing. The reason why people love [10:07] Obsidian is because of these [10:09] interreationships. the idea that you [10:10] could open a file and then you know I [10:13] just open this file and then I'm like oh [10:14] interesting I mentioned Greg Eisenberg I [10:16] can click that and it goes to that file [10:17] that's interesting right it shows it [10:19] works more to uh it works more like the [10:22] way your brain works your brain connects [10:24] these patterns all the time [10:27] >> here question yeah [10:29] >> so I see why it's interesting [10:32] >> but how does this get me better output [10:36] >> exactly yeah so the next thing is [10:39] obsidian in released this new tool [10:41] called Obsidian CLI. And what that [10:44] allows you to do is it allows you to use [10:47] Cloud Code and it can go and it can read [10:51] all of the files in your Obsidian Vault, [10:53] which is a folder of text files. [10:56] But with the Obsidian CLI, it can give [10:59] Claude code not only those files that it [11:01] can read and access, but it can also [11:04] give Claude code information about the [11:07] inter relationships of those files. So [11:09] you can see so cloud code can see that [11:12] oh this file is connected to this file [11:14] and this file and this file and that [11:16] gets very interesting in terms of what [11:19] claude code can understand about you and [11:21] what claude code can uh understand about [11:24] uh all of the relationships between the [11:26] things that you're working on. It can [11:28] start to surface patterns about what [11:30] you're thinking about that you are not [11:32] seeing for yourself. some idea that you [11:34] might have been writing about for a year [11:35] in this vault. It could be a latent idea [11:38] and it can just immediately say like, [11:39] hey, did you know that you've been [11:41] writing about this same pattern in [11:43] startups or in this particular project [11:45] you're working on in every single note [11:47] you're making across these different [11:49] domains and and in seeing that for the [11:51] first time can be like a huge light bulb [11:53] effect. It can cause like huge [11:55] progressions in your learning and your [11:56] understanding and your point of view on [11:58] the world but also in what you're [12:00] working on. Um, so I've written out I [12:04] wanted to demonstrate how that actually [12:06] works in terms of how I can pass [12:08] information into an agent that would be [12:10] impossible without Obsidian and Cloud [12:11] Code. [12:13] So here's some commands that I have that [12:15] I use. And I don't want you to be afraid [12:17] of like all this stuff. I know this can [12:19] look intense, but um here here's what um [12:23] here's what I've got some commands. And [12:26] this is just terminal that I've created [12:27] and I'm running it in Obsidian. You [12:30] don't need to use this. You can just [12:31] also do this in your own terminal [12:33] session on in whatever tool you want. [12:35] But I put it in obsidian because I want [12:37] to see it all together and I wanted to [12:39] show you the ways in which you can like [12:40] integrate and customize this [12:42] environment. [12:43] So here's a cool thing. So [12:46] context/context [12:48] load full context about my life, work, [12:51] and current state. Reads context files, [12:53] daily notes, and follows back links to [12:55] build a complete picture. So I'll just [12:56] show you that right here. So, like let's [12:58] say I open a new session in in uh in [13:02] Claude just on my desktop. And now it's [13:05] like I'm about to work on something, but [13:06] before I work on it, I can just type [13:08] context demo. Now, it's going to read a [13:10] whole bunch of files about where I'm [13:12] currently at. Done. [13:15] Like, I've already preloaded I've [13:16] already preloaded in all this context [13:18] now. So, you can see it's going to start [13:20] reading all these files. It's reading a [13:22] read me. It's reading context about new, [13:24] which is a media company that I'm [13:25] working on. It's reading about other [13:27] stuff. It's reading my personal the [13:29] other stuff is my show. It's reading a [13:30] personal workflow context. And so I [13:33] don't have to worry about it not knowing [13:35] the key information that I wanted to [13:37] know. I just did that one command and [13:38] now it's going to get all that [13:39] information done. So I can use slash [13:42] today, which is a morning review. It [13:44] pulls calendar, tasks, IME messages, and [13:46] the past week of daily notes in into a [13:48] prioritized plan for the day. Why does [13:50] this matter? Well, okay, sure. You can [13:53] set up an agent and give it access to [13:55] your calendar and your tasks and iMes [13:57] and things like that, but it's miss it's [13:59] that's that doesn't have all of the [14:01] information about what you're thinking [14:02] about and why. If I'm writing daily [14:04] notes about some particular technology [14:06] or project or thing that I'm interested [14:07] in, does my calendar reflect actively [14:11] like does it match the subjects I'm [14:13] actually writing about? If an agent has [14:15] that context, you can it can it can uh [14:18] more effectively give you um information [14:21] about what you should do or not do or [14:23] and it can more effectively make [14:25] decisions on what should be in your [14:26] calendar or not in your calendar. Here's [14:28] another one. Slashclose day [14:29] [clears throat] [14:30] end of day processing extracts action [14:33] items, surfaces vault connections, [14:35] checks confidence markers, needs to be [14:36] updating. So I have a bunch of [14:38] hypotheses that I I think about and I [14:41] give them a confidence rating. This is [14:43] an idea I'm working on. I feel very [14:45] solid about it. Here's another idea I'm [14:46] working on. I'm not sure about it. So, [14:48] these are like um daily operations [14:51] things, but [clears throat] [14:54] this is what I use Obsidian for the [14:56] most, which is thinking tools. I really, [14:59] really, really like working with LLMs as

[15:00]

[15:02] a thinking partner. It's my favorite way [15:04] of using LLM. I know people like to use [15:07] uh agents and LLMs to build things, but [15:09] I really like using them to think [15:11] alongside me and build when I feel like, [15:14] you know, I really have a novel way of [15:16] viewing things. So, let's see here. Um, [15:21] so ghost here's here's a command I have. [15:25] It answers a question the way I would. [15:27] It builds a voice profile from the [15:29] vault, writes in that voice, then [15:30] evaluates the fidelity. So, I can just [15:33] say, "What do I think of AI?" And I'm [15:35] going to show you this challenge topic. [15:38] It pressure tests current beliefs using [15:40] the vault's own history, finds [15:42] contradictions, counter evidence, and [15:43] shifts in thinking. Why does that [15:45] matter? Well, if I want to make sure [15:47] that I'm continually developing as a [15:49] human being and as a as a as in my [15:51] skills, I want to make sure that, you [15:54] know, the POV I have isn't overly biased [15:56] or limited. So, this can challenge me. [16:00] Emerge surface ideas the vault implies [16:03] but never states. conclusions from [16:05] scattered premises, unnamed patterns, [16:07] unarticulated directions. This is super [16:09] super useful because a lot of times, you [16:13] know, I I can be stuck just surfacing [16:15] ideas in a in in um in a lot of [16:18] different ways, like for years, and just [16:20] having someone say a simple thing to me [16:22] that just says, "Hey, this is just [16:24] naming the idea. Hey, did you know that [16:26] you keep circling around this pattern?" [16:28] Huge breakthroughs. [16:30] Uh slashdrift. It compares it compares [16:33] my stated attention intentions against [16:36] actual behavior over 30 to 60 days. [16:39] Surfaces what I am avoiding ideas. [16:42] People on this podcast, the listeners [16:43] will probably like this one. Deep 30-day [16:45] vault scan with cross-domain pattern [16:48] techni detection and graph analysis to [16:50] gen generate ideas across all domains. [16:52] This gives me um not just ideas on like [16:55] things I should work on. like it gives [16:57] me ideas for tools and things like this, [16:58] but it also gives me ideas on like films [17:00] I should watch, products I should buy. [17:03] Again, all influenced by like like [17:05] things I'm writing about in my vault. [17:08] Um, [17:09] trace tracks how an idea has evolved [17:12] over time across the vault. So, let's [17:14] see some of this stuff. The trace demo. [17:17] So, I did this one already. And the way [17:19] this would work is I just like create a [17:21] tab here and I could just be like claude [17:25] trace and I had to create demo versions [17:28] of all of these um commands because of [17:30] how much personal information is in my [17:32] vault. But still I don't even know like [17:34] I I I can't even control what is going [17:36] to show up on the screen. And I I have [17:39] just a dumb question. Like all those [17:40] commands that [17:42] >> we saw, [17:43] >> is that commands that you created or is [17:46] that what Obsidian created? [17:48] >> That those are commands that I created. [17:50] >> Mhm. [17:51] >> And um you can create them very easily [17:54] by just asking Cloud Code to create a [17:56] specific command. [17:57] >> Mhm. Um, [17:58] >> we can include I'll include in the show [18:01] notes and in the description [18:03] just like a link where people can click [18:06] to get access to some of Vin's [18:10] uh skills if that's if that's [18:12] >> Yeah, for sure. Totally. Yeah. [18:14] >> So, you can see here what I did was I [18:16] just typed slashtrace demo. Demo is just [18:18] because I'm showing this publicly how I [18:19] use Obsidian. And so what this does is [18:22] it tracks my I u the way I've developed [18:26] and um the way I've navigated this idea [18:29] through my vault. So I have an example [18:31] here. It's really interesting. [18:34] So here I I did it and I and I and I let [18:36] it run. And what it's doing is as it [18:39] says like I'll trace how your [18:41] relationship with obsidian has evolved. [18:42] Let me start by building a vocabulary [18:44] map and searching across a vault. So it [18:47] when you have OB like so now it starts [18:50] going through the vault reading all of [18:52] these different files and it can see all [18:53] of the files that are connected using [18:56] the Obsidian CLI. This is a lot man. [18:58] This is this is something that I would [19:00] never be able to do on my own and to [19:02] read all of these files to know what how [19:05] they're interconnected with each other. [19:06] This is just not possible for me to do [19:08] as a human being. And then it spits this [19:10] out. I now have everything I need to [19:12] construct this trace. Here's the full [19:13] evolution trace how I use Obsidian. [19:16] First appeared January 11th, 2025. Time [19:20] span 13 months. That's in relation to [19:22] this vault specifically, meaning all the [19:24] files in here. [19:26] It says prevaults December 2024. The [19:30] essay how I take notes in the west end [19:32] of Toronto dated December 1st, 2024 [19:34] describes a complete system where [19:36] Obsidian plays no role at all. The [19:38] system is audio dumps via Mac Whisper, [19:40] LLM dialogue loops, Canopia for spatial [19:43] mapping, physical notebooks for [19:44] cognitive friction, Fuseland for [19:46] tracking, arena for fragments. The [19:47] pipeline is capture process [19:49] crystallization. This is the baseline. [19:51] Obsidian is not in the picture. [19:53] Discovery and skepticism, January to May [19:56] 2025. [19:58] The first daily note in the vault, raw

[20:00]

[20:00] excitement mixed with uncertainty. I [20:02] could probably drop transcriptions in [20:04] here, too, as a way of storing them. The [20:06] current theory is that in terms of [20:08] note-taking birectional linking is not [20:10] that useful but I don't know then [20:14] it says the chosen tool [20:17] um originally I was back link it says um [20:20] here how I use obsidian note a pivotal [20:23] realization about backlinking originally [20:25] I was backlinking to general terms [20:27] podcast or physical fitness or film [20:29] making I'm realizing it's not the most [20:31] useful way to use obsidian the most [20:33] important thing to do is to create notes [20:34] for each of my patterns theories, [20:36] projects or perspectives and get them [20:38] documented and out of my head and then [20:40] link to those notes. So, it's just it's [20:42] just like pulling from things I've [20:46] written and it's forming this historical [20:49] this history of this concept. And I [20:51] could do this with anything. I can do [20:53] this with like startups. I can do this [20:55] with a particular project, with foods, [20:57] my relationships, like a hobby, [21:00] anything, right? And then phase four, [21:02] January 2026, a month of explosive [21:05] building, [21:06] everything still require and then like [21:08] everything still requires me to actively [21:10] prompt and manage each section session. [21:12] The next unlocked is figuring out how to [21:14] get agents task to run automatically. [21:16] The friction is no longer with obsidian [21:17] itself, but with the boundary between [21:19] the vault and agent execution. So you [21:21] can see I'm really pushing myself, [21:23] right? And it's cool. This is this is a [21:28] very useful thing for me to understand [21:31] how my um use of this tool is evolving. [21:35] And it's just I think it's just absurd [21:37] that I can just be making notes and then [21:41] about all of these different things [21:42] through my life. Like even as a parent I [21:45] I can I can like reflect on the [21:47] different things I'm learning. I just [21:48] think this is insane that an a computer [21:52] can have this much information about me [21:54] and surface these patterns. I would not [21:56] be able to do this on my own and this [21:58] fast. And what a great tool it is for me [22:01] like like I can just like now write in [22:03] here and I'm you know as I'm thinking [22:05] about things and it gives me ideas right [22:08] about um my life and the projects that [22:11] I'm working on. So I can say you know [22:14] it's interesting the way that my [22:17] relationship with Obsidian has evolved [22:20] over time. [22:23] It makes me think a lot about [22:28] the way in which my relationship with [22:31] computers has evolved over time since I [22:34] was a kid to now. It's interesting [22:39] how these things just happen and [22:42] compound over time and we don't really [22:45] realize it. So, it's just like a note, [22:48] right? It's an idea. So that's an [22:50] example of something. [22:52] >> So I think a part of getting good at [22:55] Obsidian sounds like refle, you know, [22:58] inserting reflection into your everyday [23:02] life because a lot of people, [23:04] >> you know, we're moving from meeting to [23:06] meeting. We're busy. We're parents. Um, [23:09] you know, we grow up and we of course [23:12] write things down in notebooks and stuff [23:14] like that, but I feel like as we get [23:15] older, we actually write and reflect [23:18] less and less. [23:19] >> Yes. [23:20] >> Uh, you know, h how do you [23:24] how have you been able to insert [23:26] reflection into into your life? [23:29] >> Yeah. I think um for me it's really [23:32] about there's there's two reasons that I [23:34] think reflection is interesting and [23:36] making notes a lot are is interesting. [23:38] One is that it's great to be able to [23:41] look back on them. Like for me now, [23:43] obviously I can use an agent, but to for [23:45] me to go back and see these notes and [23:47] realizing that, oh, like I'm a person [23:49] that's continually changing. My skill is [23:51] continually changing. Projects are [23:53] continually evolving, it's it's just [23:55] it's an amazing part of life to be able [23:57] to reflect on how things are changing [23:59] over time and how you are changing over [24:01] time and how the world is changing over [24:02] time. But the other thing is that [24:04] there's like a functional reason too. [24:07] The reason I like to make notes a lot is [24:08] because that's how I generate ideas. [24:10] When I like get when I sit down on my [24:12] computer and I write things down, that's [24:13] where ideas come. For example, this [24:15] thing I just wrote here, it's just a [24:16] quick note. I'm just doing it in real [24:17] time. I'm just making it up right now. [24:20] But um by writing it out, I feel like I [24:23] internalize it more. And I like having [24:25] good ideas. I like progressing. So [24:28] because I like having good ideas and I [24:29] like progressing, writing is how I do [24:32] that. And so I think um to you know if [24:35] you want to cultivate like a writing as [24:38] a habit I think first you have to uh [24:40] connect it with the idea that this is [24:42] how you progress and this is how you [24:44] generate ideas and this is how you uh [24:46] have form original ways of thinking. Um [24:49] the other thing I'll say is that writing [24:53] right now is a big way of how you [24:55] delegate things to agents. That's like a [24:58] whole new that's a whole new aspect of

[25:00]

[25:00] it. So if you can develop a writing [25:01] habit, [25:03] you have a lot more context that you can [25:05] pass over to an agent which then um [25:08] dramatically I think increases the [25:10] amount of like things you can delegate [25:11] and the amount of things you can build. [25:13] I hope that was a good answer. [25:16] >> How does how [snorts] does this relate [25:18] to open claw? Because if you think of [25:20] open claw, it's essentially, [25:25] you know, at the best case an extension [25:28] of you. Yeah, [25:29] >> that could go and do things you know uh [25:33] independently [25:35] slashbased on your guidance. So how can [25:39] you use commands obsidian [25:42] [clears throat] and open claw and [25:44] reflection harmoniously? [25:47] >> Yeah. So I think if you look at like [25:49] here's an example of one command that I [25:52] do. It's just like a schedule command. [25:54] Um, and so what I asked this thing to do [25:58] is I said schedule I said can I take a [26:00] meeting with Greg Eisenberg today at at [26:02] Feb 20 at 2 p.m. Right? And what this [26:05] does is of course it can look at my [26:07] calendar and stuff like that but it's [26:08] also going to look through my daily [26:10] notes. It's going to look through what I [26:12] care about and then it's going to give [26:14] me some perspective. So it says your day [26:17] is stacked. You're already recording on [26:19] Greg's podcast this morning followed by [26:21] a team lunch outing and meeting with [26:22] Peter and Vince. Your Feb 17 note show [26:25] the Greg episode has been top of mind. [26:26] The vault has a dedicated Greg Eisenberg [26:28] note. No, not a two. So the [26:30] recommendation is no, not a 2pm, but you [26:32] might not need a separate meeting at [26:33] all. Yeah, that's val that's actually [26:35] the correct answer. How does that relate [26:37] to OpenClaw? Well, OpenClaw is this like [26:41] autonomous agent that can go and do [26:43] things if you uh set it up to do that. [26:46] Like it can do things without you having [26:48] to prompt it all the time. it can just [26:49] go and make decisions and build things [26:50] for you um on your own. So now what [26:55] OpenClaw can do is in the same way I [26:57] just did this command, OpenClaw can do [26:59] this on its own as well. And it can go [27:02] and read my vault, find connections and [27:04] then make decisions uh on behalf of me [27:07] with like a deeper understanding of me. [27:08] And now instead of like managing an [27:10] agent or talking to another human about [27:13] working on something, I just focus on [27:14] managing this vault. This is like the [27:16] new source. I just continually try and [27:19] make it so that this vault has all of [27:21] the information needed so that I can [27:23] delegate to an agent and um the agent [27:26] can just pull from this vault source and [27:28] make decisions. And if it's not making [27:30] the right decisions, I'm changing [27:31] something on the vault, I'm not [27:33] necessarily working with the agent [27:34] specifically. [27:35] >> Um [27:38] kind of kind of um that's my speculation [27:41] on that subject. I think it's very [27:42] interesting. Yeah, I think uh one thing [27:45] that worries me a little bit about it is [27:47] if Obsidian [clears throat] [27:48] is really your second brain, [27:51] >> giving openclaw access to your second [27:53] brain [27:55] >> is [27:56] scary. [27:57] >> Yeah, scary. And I would say that is the [28:01] um [28:03] fundamental um [28:06] the the weird element of this technology [28:09] I would say and I have purposely given [28:13] um Obsidian I mean sorry a claude code [28:16] or any agent access to a lot of [28:18] information. I've purposely done that [28:20] because my um relationship with this is [28:23] I want to understand what these things [28:25] are and I want to understand what [28:27] they're revealing about, you know, how [28:29] our relationship with computers is [28:30] changing. But um it's weird. It's like [28:34] you have to really think about how much [28:36] information you're sharing with these [28:39] agents and whether that's the right [28:41] decision or not the right decision. And [28:44] I think um it's going to be very [28:46] interesting to see how privacy as a [28:47] concept evolves and changes and what we [28:50] fight for or don't fight for in like the [28:52] in the future of our society and our [28:55] world. Even with every one of these [28:56] commands, I had to create a new version [28:59] of them, a demo version, so that I [29:02] wouldn't reveal too much personal [29:04] information while I'm on screen on this [29:05] podcast. And even then, it's like a it's [29:08] a tossup. You know, I could type the [29:09] demo version, but who knows what's going [29:11] to be shown on screen, you know? [29:14] What other commands do you want to show? [29:17] >> Um, so [29:20] there's connect which is allow it allows [29:23] me to take two domains and connect um [29:26] them using the vaults link graph. So I [29:28] can just say I did one here [29:33] and I just asked it to uh connect [29:38] film making and world building. And so [29:40] it goes through and it reads all of [29:42] these different files and then it can [29:44] start to say okay let's let's connect [29:46] these two concepts. So notes in film [29:49] making uh notes in filmmak's [29:52] neighborhood. So I was like 35 film [29:54] watch list my first meeting with Toby [29:57] notes in the world building neighborhood

[30:00]

[30:00] um the worldbuing essay um new as a [30:04] media company. So these are different [30:05] things I'm thinking about. So bridge one [30:08] the interview portal and the constructed [30:10] world in film making if I notice [30:12] something specific and asked a question [30:14] about it it would open a portal into a [30:16] person's internal world which is often a [30:18] vast universe of concepts and beliefs [30:20] and visions. So worldbuilding essay I [30:23] want my blog to show you what I value [30:25] what I believe what I worry about like a [30:27] tomb from ancient Egypt. I want my blog [30:29] to be a place that you dig up and [30:30] examine long after I'm gone. These are [30:33] things that I've written and I can start [30:34] to see how these ideas connect together. [30:37] Bridge to always on documentary equals [30:39] continuous world building. [30:42] Always on documentary is a creative [30:44] strategy where companies continuously [30:45] narrativize their characters, pursuits, [30:47] conflicts, and visions through [30:48] documentary. So these are like things [30:50] that I'm writing about and it's showing [30:52] me the ways in which these are [30:53] connected. Um, I think this can get very [30:58] interesting depending on the kinds of [31:00] things that you're willing to connect [31:01] together. You could get probably pretty [31:03] crazy with it as well, depending on what [31:05] you're writing about in your vault. I [31:06] could connect like shawarma and startups [31:08] if I wanted to, for example, and see the [31:11] kind of connections that are coming [31:12] between these things. Um, again, really [31:15] interesting because all of this is [31:17] happening super quickly and I don't need [31:19] to explain any of this to an LLM. I can [31:21] just type something like slashconnect [31:24] filmm worldbuilding. [31:26] >> The the a lot of the examples you're [31:28] using is personal reflection. Yes. Yes. [31:31] >> How do you think about [31:34] uh you know for example note-taking in [31:37] meetings like maybe you have like a [31:40] granola or Gemini notes taking you know [31:44] uh taking notes or and and and sort of [31:47] putting it into Obsidian or um and by [31:52] the way when I say notes those could be [31:54] meetings that you're not even in. They [31:56] could just be like, [31:57] >> you know, Tommy met with Vince and they [32:00] had this meeting and I want to put it in [32:01] here. [32:02] >> Yes. [32:03] >> Um, how do you think about that? [32:05] >> So, it's a really good question. So, I [32:07] think you can use these vaults however [32:09] you want to use them in terms of like um [32:11] you could put any text you want in here. [32:13] If you want to put your granola meeting [32:14] transcripts in here, you can put them in [32:16] here. And you have to just make sure you [32:17] know maybe you're doing something like [32:19] this, right? So, you're just like [32:20] meetings and then you're like, "Okay, [32:22] these are this is, you know, project [32:24] one." And then every time you do a [32:26] meeting, um, you take your granola notes [32:29] and you just put them in here, right? So [32:30] you're like meeting Greg Eisenberg Vin [32:34] plus VIN and then you know that file is [32:38] created now and you can just drop your [32:40] trans you can just drop your meeting [32:41] notes in here and now that's in the [32:43] vault and then you can pass that into [32:44] the agent or the agent will discover it, [32:46] right? Especially if you start tagging [32:47] like oh I'm going to tag this back to [32:50] you know like my podcast or something [32:51] there. Now it's connected. So now the [32:53] the the agent it has like more context [32:56] and now it knows that trans this [32:57] transcript is related to this other [32:59] file. Great. Um I think that's up to [33:02] you. I think the way that you the the [33:04] amount of information that you put in [33:05] here is up to how you want to use the [33:07] vault and how you want to delegate to [33:09] things to agents and maybe you even want [33:10] to create different vaults for different [33:12] purposes. For me per for me I use LLM [33:16] and agents as a way to increase my own [33:20] level of understanding of subjects. So I [33:21] use it for a lot of reflection and [33:23] things like that. So I don't want an [33:25] agent to write into the files. Like I [33:28] could easily get it to do that. Like I [33:30] could just say like even here I' I've [33:32] asked it to write a description of some [33:34] commands that I can talk about today. [33:36] But I don't I don't I don't want it to [33:37] make a file to do this because I want to [33:39] control all the files in my Obsidian [33:41] vault because I I always want it to pull [33:44] from what I think about things, right? [33:46] Not what it thinks about things. And if [33:48] it starts making its own files in this [33:50] vault, then I don't know like is like [33:52] when it's finding these patterns, is it [33:53] finding patterns about things it's [33:55] written or is it finding patterns about [33:57] things I've written? So I create a rule [34:00] for myself which is like a strict [34:02] separation between these things. I only [34:04] want it to write things on the side here [34:06] and then I will take that and and and [34:08] and write what I think um should be [34:11] included. [34:13] >> Right. Some [34:15] >> Yeah. Go ahead. No, I was just going to [34:17] say like, [34:18] you know, I could see the power of just [34:22] using it for your own reflections. [34:25] >> I can also see the power of, [34:28] you know, AI is really good at, you [34:31] know, going out on the internet, finding [34:34] information based on trends and stuff [34:36] like that. Distilling it in a way that [34:38] you want and having that being put into [34:41] your world is also interesting. Yes, [34:45] totally. I think that and also like [34:48] let's say if you asked um the Obsidian [34:51] if you asked the um cloud code to go [34:53] through your Obsidian file and generate [34:55] ideas [34:57] which um you know ideas for tools that [34:59] you should build well then you can just

[35:00]

[35:01] like say okay cool if that's if I have [35:03] an idea for a tool that I should build [35:04] just generate a description of that and [35:07] then just build the tool. So [35:09] >> exactly. [35:11] >> Yeah. I want to show this one so that [35:15] it's like less on reflection. [35:18] Like for example like you know I'm a I [35:22] built this thing called ideab [35:23] browser.com and every single day we give [35:25] this validated startup idea like someone [35:28] theoretically can go and [35:30] >> you know grab that information put it in [35:32] an obsidian vault and then based on that [35:36] basic you know help them build the [35:38] actual thing right [35:40] >> totally yes so I want to show you this [35:43] because I think it'll it'll really make [35:45] it it'll take it out of the realm of [35:47] reflection ction and into the realm of [35:48] building. But the only issue is this [35:50] takes a bit. So [35:54] >> yeah, so the other thing is that with [35:56] all of these like with these commands, [35:58] another pattern that I'm noticing is [36:00] that they take a bit because it's [36:02] reading so many files. And I would say [36:04] that's a big difference between using [36:06] Obsidian and like or using cloud code [36:09] with access to this Obsidian vault that [36:11] I'm noticing is all of my requests are [36:14] taking way longer and it's just because [36:16] it's reading so much more. So like look [36:18] at this one. So this is ideas demo. So [36:21] I'll run a comprehensive ideas [36:23] generation. Let me start by gathering [36:25] vault structure and context in parallel. [36:26] And then if you look it's really [36:28] interesting to see what it's doing, [36:30] right? So it's like Obsidian orphans, [36:34] right? So it's it's it's it's like I [36:37] guess orphans are like files that are on [36:39] their own, not connected to things, [36:40] right? So that's interesting that it [36:42] knows that. Obsidian dead ends, Obsidian [36:44] Resolved, Obsidian tag counts. So it's [36:46] just trying to it's trying to figure out [36:48] like some connection between all these [36:50] things. Um and then it says, okay, daily [36:53] read. So it's reading my daily notes. [36:55] Then it found this file called new [36:56] context, which is new is this media [36:58] company I'm working on. Then it's like [37:00] read file the other stuff context that [37:02] that's podcast. The other thing I would [37:04] say to you guys is I do manage [37:08] um I write uh I create context files for [37:10] projects that are pretty extensive. I'll [37:13] show you I I was I didn't know if I was [37:16] going to show this because it's very [37:17] personal but like for the other stuff. [37:19] Look at this. So other stuff working [37:21] context. What shifted recently [37:24] frontloading profile by traveling to San [37:26] Francisco, New York City to record [37:27] guests. This is super personal stuff, [37:29] but what is the other stuff? The format, [37:31] core beliefs of the show. Research is [37:34] the foundation. Solid. The best [37:36] conversations feel like discovery solid. [37:39] You know, here's the team that's working [37:41] on it. And so what happens is this again [37:45] very personal, but this is um context [37:48] that it just pulled in. So now it knows [37:50] who's working on my podcast, what are [37:53] the what are the recent hypothesis [37:55] hypotheses I'm exploring. Um, and it [37:58] just got that information. That's just [37:59] like one of the things it did. Personal [38:01] workflow context. Super personal file, [38:05] but it shows like um, you know, like uh, [38:09] what my daily schedules is like, you [38:11] know, things that I have to do like in [38:13] my personal life. So it's like pulling [38:14] that how I like to work, how I don't [38:17] like to work. Personal agent [38:19] infrastructure is another thing. Let's [38:22] see what happens if I pull this up. [38:25] This is a project in which I want to [38:26] take a step towards increasing my [38:28] personal infrastructure workflow, [38:29] delegation, however you want to describe [38:31] it with agents, understanding what it [38:32] means to delegate to agents more and [38:34] more implementation approach. So this is [38:37] like what I'm writing about the file [38:40] about how I'm thinking about using [38:42] agents personally. And again, that's an [38:44] example of one of the files it's [38:46] reading. That's just one of them. So you [38:48] saw like the other stuff, the personal [38:49] workflow, and it's factoring that all in [38:52] to this task I've asked it to do, which [38:54] is generate ideas for me. Gathering data [38:56] from your daily notes, calendar, and [38:58] vault structure. This takes a moment [38:59] since it's pulling from multiple [39:01] sources. Again, one of the um things [39:04] with this is that it's just going [39:06] through a lot of information, man. A lot [39:08] of information, so it takes longer. You [39:10] know, it's already been going for five [39:12] minutes, right? And so that's that's [39:14] something I'm noticing. But for me, [39:17] that's what I want. I want that. I want [39:20] I want a response from LLMs that is very [39:23] very contextual uh to the things that [39:25] I'm writing about. And I think a lot and [39:28] I think that's how me and an agent can [39:30] work best together where I just focus on [39:32] continually um noting that my the the [39:37] where I'm currently at in terms of the [39:38] projects I'm working on and what my [39:40] understanding is and what I find [39:41] interesting. I want to maintain that and [39:43] make it as current and as deep as [39:45] possible. So whenever I'm talking to an [39:47] agent, it has the best representation at [39:49] all times of who I am in that moment [39:51] when I ask the agent for something. [39:53] >> Yeah, I mean that's the goal, right? You [39:56] that's the question that we all should [39:58] be asking of ourselves, which is

[40:00]

[40:00] >> does the agent have the most upto-date [40:03] information on the projects, on my [40:05] preferences and my dreams and my hopes [40:07] and my goals? Um because you're it's [40:12] only as good as the up-to-date version [40:14] of that. Correct. [40:15] >> Yes, 100%. The quality the the quality [40:19] of information that the agent has [40:21] entirely determines what it can do for [40:23] you. Right. If it doesn't know a lot [40:25] about you, it's not going to be able to [40:27] do a lot for you. But if it knows a lot, [40:28] then it can it can it can do things for [40:31] you that I I think like even some of [40:34] your in some it's kind of weird to say, [40:36] but I mean like that you don't even know [40:37] about yourself in ways. [40:40] >> I mean it makes sense, right? Cuz [40:42] ultimately what this is doing like to [40:45] distill it to its core, it's connecting [40:47] the dots. [40:48] >> Yes. [40:48] >> Obsidian and Obsidian and Claude code [40:51] here are are connecting the dots. Now, [40:54] it's actually quite difficult as a as a [40:57] business owner or just as a personal in [41:00] our personal lives to connect the dots. [41:01] Like why do people in a lot of ways go [41:03] to you know coaches, therapists? [41:06] >> Yes. [41:06] >> Um you know if you go to a therapist and [41:10] you know you have someone who is you're [41:14] doing most of the talking, right? Think [41:16] about it. You're doing a lot of the [41:17] reflection and the therapist and coach [41:19] is sort of guiding you. That's that's [41:21] what this is doing in a lot of ways. And [41:24] I'm not saying by the way that don't go [41:25] to your therapist, you know, just [41:28] >> but I'm but my point is it helps it [41:31] helps you helps you uncover what are [41:34] their dots and how you can connect them. [41:37] >> Yeah, absolutely. And for me, yeah, it's [41:40] just it's just really exciting. Um, and [41:43] yeah, it's just it's just a crazy time [41:45] with computers. So, let's look at this. [41:47] This thing finished. So, this is an idea [41:48] generation report. vault relationship [41:51] exploration. This is this is pretty [41:53] extensive, right? Like to get an idea [41:57] report [41:59] like I I think this is this is really [42:02] going to show how we can move you move [42:04] from reflection to something actionable. [42:08] So [42:09] structural highlights. So again, this is [42:12] just Obsidian stuff, right? Orphans [42:14] worth noting. There's some defense [42:16] technology stuff here. just a theme [42:18] that's growing in Canada. Massive [42:21] intellectual investment sitting in [42:23] isolation. Also, orphan agentic [42:24] software. So, orphan just means these [42:26] are files I haven't really linked. [42:27] Random notes I just wrote once or [42:29] something. Um, unresolved links that [42:31] reveal latent interests. [42:34] Uh, hidden relationships. Again, all [42:36] reflection stuff. Fine. What's working? [42:39] Obsidian cloud code as a combined system [42:41] is working for me. This is producing [42:43] genuine breakthroughs in thinking and [42:45] output day per domain structure. when [42:47] enforced. This is basically I started u [42:50] splitting my schedule where each day has [42:52] a specific focus. This is cool and this [42:55] is very true. The Greg Eisenberg episode [42:57] as a forcing function. It's compressing [42:59] months of thinking about Obsidian and [43:01] agents into a clear thesis with demos. [43:03] Very true. It, you know, coming on the [43:06] show and doing this forced me to um [43:09] synthesize everything I knew and present [43:10] it. [43:12] But here's where we're going to the [43:14] actionable stuff. Tools to build [43:18] The slashgraduate slash command daily [43:20] note idea extractor based on daily notes [43:23] are full of idea sorry daily notes are [43:25] full of idea tags and interesting [43:27] thinking that never gets developed. The [43:29] vault has nine idea tags but hundreds of [43:30] undiscovered insights. Build a command [43:33] that scans recent daily notes, [43:35] identifies ideas tagged or not, and [43:38] prompts you to decide create a [43:40] standalone note, add to an existing [43:42] file, or dismiss. This turns a daily [43:45] node stream into a structured idea [43:46] pipeline. Obsidian vault for new. It [43:49] says I just have to manage and set up a [43:51] central Obsidian vault for new. What [43:53] that means is in the same way that I'm [43:56] creating this vault and it has all my [43:58] ideas and my patterns and everything [43:59] like that. Um why would I not create one [44:02] for my team where they like as a team we [44:05] can go and ask this vault questions and [44:07] we can all uh contribute to it. Here we [44:10] go. Tools to start using. What is this? [44:13] Typed for all external documents. [44:17] Uh, interesting. A time blocking act [44:19] that enforces day. A time blocking app [44:21] that enforces day per domain. Meaning [44:23] since I'm trying to focus on one thing [44:25] each day, one aspect of my life, it's [44:28] saying why not create a time blocking [44:29] app that forces you to do that. [44:31] Interesting [44:33] systems to implement. [44:35] One sentence in Obsidian agent handles [44:38] arrests. This is literally the demo 3 [44:40] version of the Greg Eisener. [44:42] Eisenberg prep, you're already imagining [44:44] it. The next step is to actually [44:45] building it. Start small. Write schedule [44:48] a call with person about topic this week [44:49] in a daily note and have Otis or [44:51] Claudebot or OpenClaw pick it up and [44:54] handle it. So, it's saying maybe you can [44:56] delegate right from the note itself is [44:58] how I'm interpreting that. Super

[45:00]

[45:00] interesting. [45:01] >> Yeah, I just inline inline delegation [45:04] like [45:04] >> maybe that's even like a new UX pattern. [45:07] I don't even know, right? That you could [45:08] build into like these different tools. [45:10] Um, subjects to investigate. Christopher [45:14] Alexander's pattern language applied to [45:16] digital spaces. Interesting. Um, Black [45:20] Mountain College as a model for the [45:22] stadium. Stadium is a physical space we [45:23] have in Toronto. Authorless media as a [45:25] concept. Uh, how Shenzen's hardware [45:28] ecosystem actually works. Things to [45:31] write and publish. That'd be useful. [45:34] Context architecture essay. The computer [45:36] as a place. Software will become [45:38] fashion. What Toronto theory actually [45:42] is. Editorial thinking zen conversations [45:45] to have. This is interesting. These are [45:48] real people. Aaron stadium workshop host [45:50] about becoming an anchor of technical [45:52] programming. This is a space we have in [45:53] Toronto. Trun another person program [45:57] about making program the flagship [45:58] series. Steph ango obsidian CEO about [46:00] the vault as a place. Um and so this is [46:04] like [46:06] yeah this is crazy. It's it's suggesting [46:08] people I should meet, right? Top five [46:11] high impact do now. Build the graduate [46:14] command or do a manual weekly idea [46:16] review. Um, [46:18] >> this is crazy, dude. This is actually [46:21] crazy. Like, and the fact that it's in [46:25] plain text and it just not be it's not [46:28] there's no images. It It doesn't make it [46:30] easy to read, but I kind of like it [46:32] because it's it's like dressed down. You [46:35] know what I mean? [46:36] >> Yeah. I mean, I like that aesthetic cuz [46:38] I'm a nerd, but you could just say, you [46:40] know, you could you could just say, [46:41] obviously, you could just say, can you [46:42] turn this into a beautiful readable HTML [46:47] file that is on my desktop? This is hard [46:50] to read and it's going to do that. [46:52] >> Mhm. [46:53] >> Right. So, I mean, if you don't that [46:54] says if you don't like that, just do it [46:56] however you want to do it. You know what [46:57] I mean? I I like it like this in this [46:59] kind of like this aesthetic. Um, but [47:02] yeah, that's how you can move out of [47:04] reflection. And of course, you know, of [47:06] course, we could also say like here, you [47:08] know, if we don't if we don't want to do [47:10] this, uh, we could also say like it it [47:12] recommended that /graduate command. So I [47:15] could just say build the slashgraduate [47:17] command. [47:18] >> Mhm. [sighs] [47:20] >> Right. Which is interesting. [47:22] >> And is that how you started building a [47:24] lot of your commands? Like it it sort of [47:27] it started suggesting it and you're like [47:29] just go build it. [47:30] >> Well, I started off actually like [47:32] building them myself. Like I'm like, oh, [47:34] trying to think about commands myself. [47:35] But then yeah, I said I started asking [47:37] the agent like, well, wait a second. [47:40] >> What commands do you think would be [47:41] interesting? And I just and this might [47:43] be useful. Another thing that I like to [47:45] do is I like to move to like higher [47:48] levels of abstraction when I'm using an [47:50] an LLM. What I mean by that is I could [47:53] say like, oh, make a command that um [47:57] tells me what I should focus on each [47:59] day. And that's that's like um a command [48:02] that I thought of. But when I when I [48:04] another thing you can do is you can step [48:06] back and I can say based on my obsidian [48:09] vault and what you know about me form an [48:13] understanding of what you think my level [48:15] of understand like where you think my [48:18] skill level is in terms of of a person [48:20] and the projects that I'm working on and [48:23] based on that suggest the kinds of [48:25] commands I should use that would take me [48:27] from the level I'm at to a higher level. [48:30] Right? get like, you know, get it to [48:33] suggest the commands for me instead of [48:35] me suggesting the commands and I could [48:36] pick between them. So, look at this. [48:38] This is this is it was this was the [48:40] agent's idea based on what it read in my [48:42] vault based on the notes I'm taking. [48:44] Right? So, let's see what this is. Daily [48:45] note idea extractor ideas, insights, and [48:48] original thinking accumulate in detail [48:50] in daily notes, but rarely graduate into [48:52] standalone notes where it can where they [48:55] can compound through back links. This [48:56] command scans recent daily notes, [48:58] surfaces the best candidates, and helps [49:00] decide what to promote into like an idea [49:02] or something, right? Sure. So, this is [49:05] how it's going to work. It scans all the [49:07] recent daily notes. It cross references [49:09] with the existing vault. It presents [49:11] candidates. It graduates selected ideas. [49:15] If uh if creating a new standalone note, [49:17] create the note in the vault route. [49:19] Write the note as a mini essay or [49:21] working document that captures the core [49:24] claim or question context from the daily [49:26] note where it originated connections to [49:28] other vaults notes as backlinks. Now all [49:31] this stuff like it captures the core [49:32] claimer question. You might be looking [49:35] at this and think like okay this is just [49:36] the text that the agent generated and it [49:39] is but also it hits differently for me [49:43] because I know like I'm writing a lot [49:46] about these things. I know. I know. Like [49:48] even like the mini essay thing. These [49:49] are words that mean specific things to [49:51] me, which is just so it's just so crazy. [49:54] It's very contextual. I know what it's [49:56] talking about because I spend a lot of [49:57] time in this tool and I spend a lot of [49:59] time writing. So yeah, created and it's

[50:00]

[50:02] going to create that command, [50:05] right? And it's like that's that's nuts [50:08] because I'm just going about making [50:10] notes and I have this parallel agent [50:12] that is looking at my notes and giving [50:14] me ideas on how I can improve my [50:16] workflow, improve my life. And then and [50:18] then not only can it just suggest it, it [50:19] just build the thing and it's done. And [50:22] we have it right here. I could just hit [50:25] it and it's going to run. That's crazy. [50:29] >> If if I'm Open AI or anthropic, I'm [50:33] buying Obsidian. [laughter] [50:37] >> [gasps] [50:38] >> Right. Cuz it's the missing link. [50:40] >> Yeah, it's nuts. [50:40] >> It's the missing link. Why? The fact [50:42] that there are people like you've sold [50:45] me on this, by the way. I have [50:47] >> I've downloaded Obsidian. I think it's a [50:49] free tool, right? [50:50] >> It is. [50:51] >> It's open sourced. I've downloaded it, [50:54] but I have not [50:57] created my vault because I wanted [51:00] >> you, Vin, to take [laughter] me through. [51:02] I knew that this was going to be great. [51:04] I knew that I would go through this. [51:06] This actually exceeded my expectations. [51:08] Like this is the the fact like it makes [51:13] no sense to me. It [51:15] >> makes no sense. [51:16] >> The the fact like that if you are if you [51:19] are using if you are serious about [51:23] using LLMs to take your ideas and put [51:26] and and and get the most out of them. If [51:28] you're serious about building, you know, [51:30] what people are calling a personal OS, [51:33] uh, and you are not using a centralized [51:36] note-taking tool like this, [51:38] >> Yeah. [51:39] >> that uses markdown as the foundation, [51:43] then you are not using LLMs properly. [51:47] >> Yeah. Or at least not at the limit. [51:50] >> Yeah. Yeah. Exactly. You're not getting [51:51] the most out of it. [51:52] >> Yeah. [51:53] >> You're not getting the most out of it. [51:54] So I think what's what's difficult about [51:56] this is that it require it does require [51:59] a lot of time. [52:00] >> Mhm. [52:01] >> And to actually set it up properly. It [52:04] requires [52:05] uh like [52:08] Yeah. I mean, it takes a lot of time and [52:10] and and the UI isn't is is so daunting [52:13] in the sense that it's a blank canvas [52:15] and it's not like, hey, you should like [52:17] write your preferences over here or, you [52:19] know, you kind of just have to you have [52:22] to [52:24] come up with these ideas yourselves. [52:26] >> Yes. But [52:29] that's still so amazing, right? Because [52:31] I mean, even when we work with other [52:33] humans, we have to find a way to explain [52:35] things to them. And I just think it's so [52:37] cool that now we can work with these [52:40] agents and we still have to explain [52:41] things to them, but we only need to [52:43] explain them once because once we get it [52:45] down on into a file, we can always [52:47] reference that file. That explanation of [52:48] a project or a preference or anything [52:50] and it's always there and you can pass [52:52] it in. [52:53] >> Yeah. A file is like essentially perfect [52:56] a perfect memory. [52:57] >> Yes. [52:58] >> Right. Human beings have memories like [53:00] we recall things. [53:02] >> Yes. But there's tons of studies that [53:03] show that what we remember in fact is [53:06] completely different than reality. For [53:09] example, when we went and got our that [53:12] haircut in Missaga, [laughter] [53:14] I could have thought that I had the best [53:16] haircut. [53:19] You know, that's what my memory [53:20] remembers. It was a great haircut, but [53:22] who knows? It could have been the worst [53:24] haircut I had ever gotten. [laughter] [53:27] Now, [53:29] Obsidian or whatever tool you end up [53:32] using like you know if I had written [53:36] >> yes [53:36] >> like it it the memory the file the [53:39] markdown file is perfect so that when I [53:41] link that or I I recall it it is going [53:44] to give me a perfect [53:46] >> uh you know data point and the other [53:49] thing about you know these files is that [53:51] you hope well they're not biased [53:53] basically they're as biased as the human [53:55] being is in terms of writing the [53:58] reflections. [53:59] >> Yes. [53:59] >> At that moment in time. [54:01] >> Yes. [54:03] It's crazy, man. And [54:07] >> yeah, it's just crazy. And there's all [54:09] there's all of these different aspects [54:10] to it. There's the privacy of it and [54:12] what that means. There's the power of [54:15] it, the fact that now you can just work [54:17] with these computers in natural language [54:18] and just delegate to them. Um, there's a [54:21] fact that there's people like me that [54:23] are that are using these tools and [54:25] trying to figure out how to delegate [54:26] stuff to agents in this way. There's [54:28] people that are like me that are even [54:29] more hardcore in in different ways and [54:31] and pushing them. And I just think it's [54:33] such a crazy time to be alive because I [54:37] think we are potentially watching a [54:40] fundamental shift in the human [54:42] relationship to computers. And it's [54:44] [snorts] just I just I'm just really [54:46] happy to be alive while this is [54:47] happening. And I'm curious like how's [54:49] this all going to unravel? [54:50] >> Well, what's what's cool about this is [54:53] 99.99% [54:55] of people are not going to spend the [54:58] time to actually set up something like

[55:00]

[55:02] this and make it a part of their daily [55:05] lives. And the alpha, so to speak, is in [55:10] terms of leading a more productive, [55:12] happier, healthier, better, more [55:16] money-making career is in in in using [55:20] something like this with an LLM, [55:22] >> I think. So, [55:24] >> I'm not saying download Obsidian today, [55:26] and I have no affiliation or whatever [55:27] with them, but I'm saying like pick a It [55:31] sounds like what we should all be doing. [55:33] And I'm I'm talking I'm giving myself [55:34] this this advice is like [55:38] there's no excuse anymore for me not to [55:40] be writing down and reflecting [55:42] >> Yes. [55:43] >> into markdown files. [55:44] >> Yes. [55:45] >> In a world that where LLMs use markdown [55:47] files as the oxygen. [55:49] >> Yes. [55:51] Like people think tokens are the oxygen. [55:54] >> Yes. [55:54] >> But they're not. [55:56] >> Yes. [55:56] >> The markdown fires are the memories. [55:58] Like think about what a human being is. [56:01] >> Yes. [56:02] >> You know, [56:04] >> is a human being the the energy of a [56:08] human being or is it the memories of you [56:11] know what we recall? you know, [56:13] >> I mean, that's like a philosophical [56:14] question and maybe it's a a bit, you [56:16] know, a bit of both, but it's, [56:20] >> you know, I think that there's something [56:23] really really fascinating about MD files [56:27] as [56:28] an underrated about them in order to to [56:31] have a true [56:34] computer experience in in today's day [56:36] and age. [56:37] >> Yeah, there's definitely something going [56:39] on here. Some fundamental shift. Yeah, [56:41] >> it's awesome. [56:42] >> Yeah. And like I have bad my word, you [56:45] know, I'm learning in real time, right? [56:46] Like and I I don't have the right [56:47] vocabulary to even explain this. [56:49] >> Yes. And neither do I, man. Neither do [56:52] I. I'm trying to I'm trying to figure it [56:54] out in real time. That's that's why I [56:56] think like I know I show something and [56:57] and for me I'll do something or or I'll [57:00] see something and my friends are like [57:02] they kind of laugh because I'll just be [57:03] sitting at my computer just tripping [57:05] out. And I think it's because I really [57:08] like computers and and I cannot believe [57:11] that this is possible. I cannot believe [57:13] that I can just be making notes on my [57:16] computer like I have been since I was a [57:17] kid and then all of a sudden this agent [57:21] can scan through it and build things [57:23] because of it and and like connect [57:25] patterns that I could never see. It's [57:27] nuts, man. It's nuts. And and at the [57:31] root of it, you're right, is just a [57:32] collection of interrelated markdown [57:34] files. [57:35] >> Yeah. Cool, man. [57:37] >> I appreciate you. I don't know if you [57:39] can see my mind, but my mind is blown [57:41] right now. [57:42] >> Thank God. [57:43] >> Yeah. [laughter] [57:43] >> Yeah. [57:44] >> I wanted to do right by you. I also just [57:46] like [57:46] >> I say this every time, man, but I'm just [57:49] going to keep always saying it to you [57:50] all the time. I really, really, really, [57:52] really appreciate everything that you [57:54] do. I think your pattern recognition and [57:56] your pattern matching is like like [57:59] really underrated. I think there's a lot [58:01] of things that you do that I don't think [58:02] it's like difficult to see if you're not [58:04] really paying attention. I I just want [58:06] to say like thank you for uh for [58:07] everything that you do. You're always [58:09] putting on like new voices on your show. [58:11] I see it. I really appreciate it and [58:13] it's just been it's an honor to know you [58:15] and yeah, just thanks for the [58:16] opportunity. Thanks for everything, man. [58:18] >> I appreciate you, Vin. You're a legend. [58:20] I'll include links for where to follow [58:24] criminally underfollowed internet vin on [58:28] on X on his YouTube show uh podcast in [58:33] the show notes and description. You can [58:35] go and check him out there. Uh people, [58:39] please play with some of these tools uh [58:42] and let me know what you think. Let Vin [58:44] know what you think, [58:46] >> please. [58:46] >> And uh Vin, I will beg you to come back [58:50] on the show [laughter] another time. And [58:53] I hope you come back on again. [58:54] >> For sure, man. Thank you.

관련 노트

• [[260323_moltbook]]
• [[260321_moltbook]]
• [[260322_moltbook]]
• [[260323_reddit]]
• [[260322_reddit]]
• [[260226_llm_wwwyoutubecom_bed57d]]