YT YouTube2Text

YouTube2Text → Transcripts → How I Use Obsidian + Claude Code to Run My Life

How I Use Obsidian + Claude Code to Run My Life — Transcript

by Greg Isenberg · 10,672 words · 1,728 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00This is Obsidian. And Obsidian is this
0:03little tool that people are using as
0:05their second brain. But what's really
0:07cool about it is they're pairing it with
0:09Claude Code and they're getting crazy
0:12results out of it. It's literally a
0:14game-changer. Now, I've been slow to
0:16adopt Obsidian cuz to me it's been a
0:19little daunting to look at. So, I had my
0:20friend Vin and he clearly explains what
0:25Obsidian is, how to use it with Claude
0:27Code, how to set up these commands
0:29[music] that really drive the most out
0:32of Claude and all the LLMs. And it's an
0:35incredible episode, like a really
0:37game-changing episode, because I think
0:39that people who understand how to use
0:41Obsidian and how to use Claude Code
0:43together, they're going to be able to
0:45live happier, healthier, and wealthier
0:48lives. Why? Because it gives incredible
0:51ideas to you on tap. So, I know that the
0:55people that stick around to the end of
0:56this episode, I think that for a lot of
0:59them it's going to [music] absolutely
1:01change how they use AI and it's going to
1:03be a super impactful way because you're
1:05going to get better ideas at the right
1:07time, the right moment, and it's going
1:09to make you happier, healthier, and
1:11wealthier. Enjoy the episode.
1:14>> [music]
1:18[music]
1:21>> I've got my dear friend Vin, also known
1:24as Internet Vin, on the podcast. I
1:26literally begged him to come on. I
1:29>> [laughter]
1:29>> I begged him. I begged this man to come
1:32on and
1:34to teach us a very specific thing. Vin,
1:37by the end of this podcast episode, what
1:39are people going to learn?
1:41I want you to have an understanding of
1:43how you can use Claude Code and Obsidian
1:46as a thinking partner. I want you to
1:48have an understanding of how you can
1:51stop uh having to explain things to
1:53agents over and over again and just pass
1:55specific files in. And I want you to
1:56understand how you can use Obsidian
1:58Claude Code to notice things about the
2:00way you think that you would not have
2:02noticed on your own without these tools.
2:04All right, from your lips to God's ears.
2:07[laughter] Let's get into it.
2:09Okay, so first is like what is Claude
2:13Code?
2:14So, Claude Code is this like agent that
2:17you can use
2:18um in a command line interface. So, it's
2:21just basically this tool you can use
2:23that can control your computer and you
2:26can use it through natural language,
2:27right? So, I can say make a file or make
2:30a file on my desktop that says "Hello
2:34Greg"
2:36in plain text.
2:37Right? And it's going to go and do this.
2:40That's really cool. That's that's
2:42something that's that's new. That wasn't
2:44possible before.
2:46Before this, I had to go to the desktop,
2:48open some text editor, and then like
2:50create that file, right? And now this
2:52file is on my desktop.
2:54So, you can say open the file.
2:56There we go. Hello Greg. That's
2:58crazy.
2:59Right?
3:00Um
3:02now
3:03what's interesting about this is if you
3:05have this agent that can like control
3:06and do things on your computer,
3:08that means that whatever you can
3:10describe to it, like it can start to do.
3:13And so, when you if you describe a
3:15project to it or you get into these long
3:17conversations
3:18uh with an agent,
3:20um it can do more and more complex
3:22things. The more information it has, the
3:24more complex things it can do. But, the
3:25problem is that if you have to let's
3:28say, you know, like I write some super
3:30long description about a particular
3:31project or I have like an hour
3:33conversation with this agent about a
3:34particular project, it's like I don't
3:37want to have to create a new session to
3:39explain that all I don't want to have to
3:41explain that over and over and over
3:42again.
3:43Um a lot of people are using like Claude
3:46uh or ChatGPT on the web and it has
3:48things like memory, but you you can't
3:50like control you don't know what's in
3:52that memory, right? You don't know what
3:53it knows and what it doesn't know.
3:55And so, there needs to be some way of
3:58like, you know, passing information into
4:00these agents that is easier and faster.
4:03And the better information you can give
4:04it and the faster the information you
4:06can give it, the more stuff if it can do
4:08for you and the better the faster you
4:10can delegate to it.
4:12Okay?
4:13So,
4:14now
4:15even if I let's say let's say I had
4:17like, you know, let's say I wrote like a
4:19big project description here, right?
4:21Create a file that describes, you know,
4:24a project um
4:29about
4:31uh to-do list app that
4:34um
4:36is very minimally designed
4:39and reads from
4:41uh all of my calendar
4:43iMessage is and
4:46um
4:47my slack and slack
4:49and interprets it into a task list that
4:53of tasks
4:54that it thinks that I should do. I don't
4:57know. Some some idea, right? So, now
5:00this this is a file that could be on my
5:02desktop.
5:03And what I can do when I use Cloud Code
5:05is I can reference that file and pass it
5:07in whenever I want.
5:08And why that's important is because
5:12it's the context, right? Yes.
5:14>> game is
5:17feeding the beast good context. Yes.
5:21Exactly. And I don't want to have to do
5:23this over and over again. And and and
5:25and when I work on this uh over days, I
5:28I'm not going to remember like what we
5:29talked about, right? So, I want some
5:30kind of file uh that I can like pass in.
5:34Oh, sorry, Greg. One sec.
5:36Yeah, and that's that's sort of the
5:37problem that a lot of people are facing
5:39with Cloud Code is like they're using it
5:41and then they're saying, "Well, it's
5:43it's okay. It's not like game-changing."
5:46And the issue is they don't have they're
5:48not they're not feeding the right
5:49context at the right time. Yes, exactly.
5:53And so here's
5:54So here's like a a project description
5:57that it wrote. And obviously I can pass
5:58this in. And this is like a this is like
6:00a general one that I just created. But
6:02you you can make these like very
6:03complex. You can build them into like
6:05robust files, right? Over time. So we
6:07know that
6:09Cloud Code can create files and it can
6:10repeat and it can um
6:13uh read files, right? So now I can say
6:14let's say I created a new session. So
6:16here's a new session.
6:17And now I can say like
6:19I want to work on this project. And I go
6:23here and it's going to be
6:25um
6:26to do {dash} app. Here it is. Boom. Now
6:29I didn't need to explain the file again.
6:30I need I didn't need to explain the
6:31project again, right?
6:33So it's going to read this file and it's
6:34going to start like
6:36you know, that saved me a lot of time.
6:37Great project before diving in a few
6:39questions to scope the first session. So
6:40that'll continue. So now what is
6:43Obsidian?
6:44Right? Obsidian is this tool that it's
6:48kind of like an interface that sits on
6:51top of um
6:53uh a collection of markdown files,
6:55right? So here like this is reading a
6:57markdown file. How I use Obsidian story
7:00development.
7:01Right?
7:02Um I have uh daily notes.
7:05This is my daily notes. This is a this
7:07is also
7:08um a markdown file. I should do my own
7:10fundamental analysis into
7:13into how things stay pure when they grow
7:14and become more mainstream, right? This
7:16is just a file that I have. I have um
7:20you know, like a file on Greg Eisenberg
7:23that I haven't put any
7:23>> weird. That's weird. Yeah. That's pretty
7:26weird.
7:27And so I I make files like notes of
7:29things that I'm learning from people and
7:31stuff like that, too. So I have
7:32different um
7:34files for everything, right? And the
7:36interesting thing that makes a vault,
7:38which is Obsidian what Obsidian
7:39interacts with, this whole thing is
7:40called a vault. What makes it different
7:42than a folder
7:44is that
7:45Obsidian is not only interacting with
7:48just like a, you know, a folder of
7:50files, um, but what it does is it also
7:53allows you to make inter, uh,
7:56to connect relationships between files.
7:58So, I can say, "Today I am on
8:01um, a podcast with
8:03Greg Eisenberg." Now, this file is
8:06linked to that Greg Eisenberg file.
8:08Super interesting. Super interesting.
8:11And so, when people like, like, like
8:13people, uh, there's a lot of people who
8:15really, really like of using Obsidian
8:16and tools like Obsidian because of this
8:19ability to form interrelationships.
8:22This is unique to just having a folder.
8:24A folder on your computer cannot show
8:26these interrelationships.
8:28And so, it gets really interesting when
8:30you start to, um, keep making these like
8:34interrelationships over time,
8:36right? And so, what happens, here's a
8:39little visualization.
8:40And so, in the here, these are Each one
8:42of these circles is a file.
8:45And in and and it's it's showing how
8:47it's like connected to all of these
8:49other files where I've written things
8:50about. So, here's like personal agent
8:53infrastructure.
8:55Right? And so, I could look, you know,
8:58and I guess I should also add just kind
9:00of a comment on this and what was
9:01difficult doing this demo. There's like
9:03so much personal information in here cuz
9:05this is like my personal thing. So, I
9:07don't even know like what's going to
9:08show up on the screen here, right? Um,
9:11but that's part of doing demos like
9:12this, which is which are kind of weird
9:14and interesting.
9:15But, you can see personal agent
9:16infrastructure like it links to like
9:18agentic AI. There's like a link here to
9:20Telegram. There's a link here to like
9:22Toby,
9:24uh, with the founder of Shopify. There's
9:25a link to like presence log, Claude bot,
9:28you know? And then here's like I I have
9:30a podcast too called the other stuff and
9:32like you can see I'm obviously doing a
9:34lot of like
9:36thinking about that a lot, right?
9:38And so
9:40I can also, let's say if I go to Greg
9:42Eisenberg and I go to local graph, so
9:45here's like all the times I've written
9:46about Greg Eisenberg.
9:49Right? Notes on time constraints, how I
9:50use Obsidian, which is just kind of
9:52interesting.
9:53Um so if I'm listening to his
9:55>> too, Vin. I love you, too.
9:57If I'm listening to his show and I'm
9:58picking up different patterns, I can I
10:00can reference that back to to Greg.
10:03So that's really interesting. But here's
10:05the thing.
10:06The reason why people love Obsidian is
10:08because of these interrelationships. The
10:10idea that you could open a file
10:12and then, you know, I just open this
10:13file and then I'm like, "Oh,
10:14interesting. I mentioned Greg
10:15Eisenberg." I can click that and it goes
10:17to that file. That's interesting, right?
10:18It shows it works more to
10:20it works more like the way your brain
10:22works. Your brain connects these
10:24patterns all the time.
10:27Here's the Yeah, yeah.
10:29>> So I see why it's interesting,
10:32but how does this get me better output?
10:36Exactly, yeah.
10:37So the next thing is Obsidian released
10:40this new tool called Obsidian CLI.
10:43And what that allows you to do is it
10:45allows you to use Claude code and it can
10:49go and it can read all of the files in
10:52your Obsidian vault, which is a folder
10:54of text files.
10:56But with the Obsidian CLI, it can give
10:59Claude code not only those files that it
11:01can read and access, but it can also
11:03give Claude code information about the
11:07interrelationships of those files. So
11:09you can see so Claude code can see that,
11:12oh, this file is connected to this file
11:14and this file and this file. Now that
11:16gets very interesting in terms of what
11:19Claude code can understand about you and
11:21what Claude code can understand about uh
11:24all of the relationships between the
11:26things that you're working on. It can
11:28start to surface patterns about what
11:30you're thinking about that you are not
11:31seeing for yourself. Some idea that you
11:34might have been writing about for a year
11:35in this vault, it could be a latent idea
11:38and it can just immediately say like,
11:39"Hey, did you know that you've been
11:41writing about this same pattern in
11:42startups or in this particular project
11:45you're working on in every single note
11:47you're making across these different
11:48domains?" And And And seeing that for
11:50the first time can be like a huge
11:52lightbulb effect. It can cause like huge
11:55progressions in your learning and your
11:56understanding and your point of view on
11:58the world, but also in what you're
12:00working on.
12:01Um so, I've written out
12:03I wanted to demonstrate how that
12:05actually works in terms of how I can
12:07pass information into an agent that
12:09would be impossible without Obsidian and
12:11Claude code.
12:13So, here's some commands that I have
12:15that I use and I don't want you to be
12:16afraid of like all this stuff. I know
12:18this can look intense,
12:20but um here here's what um
12:23here's what I've got, some commands. And
12:26this is just terminal that I've created
12:27and I'm running it in Obsidian. You
12:29don't need to use this. You can just
12:31also do this in your own terminal
12:33session on in whatever tool you want.
12:35But, I put it in Obsidian because I want
12:37to see it all together and I wanted to
12:38show you the ways in which you can like
12:40integrate and customize this
12:42environment.
12:43So,
12:44here's a cool thing. So, context {slash}
12:47context. Load full context about my
12:49life, work, and current state. Reads
12:52context files, daily notes, and follows
12:54backlinks to build a complete picture.
12:55So, I'll just show you that right here.
12:58So, like let's say I open a new session
13:00in in a in Claude just on my desktop.
13:04And now it's like I'm about to work on
13:05something, but before I work on it, I
13:07can just type context demo. Now it's
13:09going to read a whole bunch of files
13:11about where I'm currently at. Done.
13:14Like I've already preloaded I've already
13:16preloaded in all this context now. So,
13:19you can see it's going to start reading
13:20all these files. It's reading a read me.
13:22It's reading context about New, which is
13:24a media company that I'm working on.
13:26It's reading about other stuff. It's
13:27reading my personal The other stuff is
13:29my show. It's reading a personal
13:31workflow context. And so, I don't have
13:33to worry about it not knowing the key
13:35information that I want it to know. I
13:37just did that one command, and now it's
13:38going to get all that information. Done.
13:41So, I can use {slash} today, which is a
13:43morning review. It pulls calendar,
13:44tasks, I messages in the past week of
13:46daily notes in a into a prioritized plan
13:48for the day. Why does this matter?
13:51Well, okay, sure. You can set up an
13:53agent and give it access to your
13:55calendar and your tasks and I messages
13:57and things like that. But, it's miss
13:59it's that's that doesn't have all of the
14:00information about what you're thinking
14:02about and why. If I'm writing daily
14:04notes about some particular technology
14:06or project or thing that I'm interested
14:07in, does my calendar reflect actively
14:10like it like does it match the subjects
14:12I'm actually writing about?
14:14And if an agent has that context, you
14:16can it can it can uh more effectively
14:19give you
14:20um information about what you should do
14:22or not do or it can more effectively
14:24make decisions on what should be in your
14:26calendar and not in your calendar.
14:27Here's another one, {slash} close day.
14:29>> [clears throat]
14:30>> End of day processing extracts action
14:33items, surfaces vault connections,
14:35checks confidence markers, needs to be
14:36updating. So, I have a bunch of
14:38hypotheses that I I think about and I
14:40give them a confidence rating. This is
14:43an idea I'm working on. I feel very
14:45solid about it. Here's another idea I'm
14:46working on. I'm not sure about it. So,
14:48these are like um daily operations
14:51things. But,
14:52>> [clears throat]
14:54>> this is what I use Obsidian for the
14:55most, which is thinking tools.
14:58I really really really like working with
15:01LLMs as a thinking partner.
15:03That's my favorite way of using LLMs. I
15:06know people like to use uh agents and
15:08LLMs to build things, but I really like
15:10using them to think alongside me
15:13and build when I feel like, you know, I
15:15really have a novel way of viewing
15:16things. So,
15:18let's see here.
15:20Um
15:21so, ghost
15:23is a command I have. It answers a
15:25question the way I would. It builds a
15:27voice profile from the vault, writes in
15:29that voice, then evaluates the fidelity.
15:32So, I can just say, "What do I think of
15:34AI?" And I'm going to show you this.
15:36Challenge topic. It pressure tests
15:39current beliefs using the vault's own
15:40history, finds contradictions, counter
15:43evidence, and shifts in thinking. Why
15:45does that matter? Well, if I want to
15:47make sure that I'm continually
15:48developing as a human being and as a as
15:50a as in my skills, I want to make sure
15:53that, you know, the POV I have isn't
15:55overly biased or limited. So, this can
15:58challenge me.
15:59Emerge. Surface ideas the vault implies
16:03but never states, conclusions from
16:05scattered premises, unnamed patterns,
16:06unarticulated directions. This is super
16:09super useful because
16:12a lot of times, you know, I I can be
16:14stuck just surfacing ideas in a in in a
16:17in a lot of different ways like for
16:18years.
16:20And just having someone say a simple
16:21thing to me that just says, "Hey, this
16:24is just naming the idea. Hey, did you
16:25know that you keep circling around this
16:27pattern?"
16:28Huge breakthroughs.
16:30Uh slash drift. It compares It compares
16:33my stated attention intentions against
16:36actual behavior over 30 to 60 days.
16:39Surfaces what I am avoiding.
16:41Ideas. People on this podcast, the
16:43listeners will probably like this one.
16:44Deep 30-day vault scan with cross-domain
16:47pattern
16:48detection and graph analysis to generate
16:50ideas across all domains. This gives me
16:53um not just ideas on like things I
16:55should work on. Like it gives me ideas
16:57for tools and things like this, but it
16:58also gives me ideas on like films I
17:00should watch, products I should buy.
17:03Again, all influenced by like like
17:05things I'm writing about in my vault.
17:07Um
17:09Trace. Tracks how an idea has evolved
17:12over time across the vault. So, let's
17:13see some of this stuff.
17:15The trace demo.
17:17So, I did this already. And the way this
17:19would work is I just like create a tab
17:21here and I could just be like Claude
17:25trace.
17:26And I had to create demo versions of all
17:28of these um commands because of how much
17:31personal information is in my vault, but
17:33still I don't even know like I I I can't
17:35even control what is going to show up on
17:37the screen. And I I have just a dumb
17:39question like all those commands that we
17:42saw is that commands that you created or
17:46is that what Obsidian created? That
17:48those are commands that I created. Mhm.
17:51And um
17:52you can create them very easily by just
17:54asking Claude code to create a specific
17:56command. Mhm. Um We can include I'll
18:00include in the show notes and in the
18:01description
18:03just like a link where people can click
18:06to get access to some of Vin's
18:09uh
18:10skills if that's if that's all
18:12>> Yeah, for sure. Totally. Yeah.
18:14So, you can see here what I did was I
18:15just typed {slash} trace demo. Demo is
18:17just because I'm showing this publicly
18:19how I use Obsidian.
18:21And so, what this does is it tracks my I
18:25uh
18:25the way I've developed in um the way
18:27I've navigated this idea through my
18:29vault.
18:30So, I have an example here. It's really
18:32interesting.
18:33So, here I I did it and I and I and I
18:36let it run and what it's doing is as it
18:39says like I'll trace how your
18:40relationship with Obsidian has evolved.
18:42Let me start by building a vocabulary
18:44map and searching across a vault.
18:46So, it when you have
18:48like So, now it starts going through the
18:50vault, reading all of these different
18:52files and it can see all of the files
18:54that are connected using the Obsidian
18:56CLI. This is a lot, man. This is
18:59this is something that I would never be
19:00able to do on my own.
19:02And to read all of these files, to know
19:04what how they're interconnected with
19:06each other, this is just not possible
19:07for me to do as a human being.
19:09And then it spits of out. I now have
19:11everything I need to construct this
19:12trace. Here's the full evolution. Trace,
19:15how I use Obsidian.
19:16First appeared January 11th, 2025. Time
19:20span, 13 months. That's in relation to
19:22this vault specifically, meaning all the
19:24files in here.
19:26It says,
19:27pre-vaults, December 2024. The essay,
19:30how I take notes in the West End of
19:32Toronto, dated December 1st, 2024,
19:34describes a complete system where
19:36Obsidian plays no role at all. The
19:38system is audio dumps via a Mac Whisper,
19:40LLM loops, Kanopy for spatial mapping,
19:43physical notebooks for cognitive
19:44friction, Fusion for tracking, Arena for
19:47fragments. The pipeline is capture,
19:48process, crystallization. This is the
19:50baseline. Obsidian is not in the
19:52picture. Discovery and skepticism,
19:55January to May 2025.
19:58The first daily note in the vault, raw
20:00excitement mixed with uncertainty. I
20:02could probably drop transcriptions in
20:03here, too, as a way of storing them. The
20:06current theory is that in terms of
20:08note-taking, bidirectional linking is
20:09not that useful, but I don't know.
20:12Then,
20:14it says, the chosen tool.
20:17Um originally, I was backlinking It
20:19says, um here,
20:21how I use Obsidian note, a pivotal
20:23realization about backlinking.
20:25Originally, I was backlinking to general
20:26terms, podcast or physical fitness or
20:28filmmaking. I'm realizing it's not the
20:30most useful way to use Obsidian. The
20:32most important thing to do is to create
20:34notes for each of my patterns, theories,
20:36projects, or perspectives and get them
20:38documented out of my head, and then link
20:40to those notes. So, it's just It's just
20:42like pulling from
20:45things I've written, and it's forming
20:47this historical uh this history of this
20:50concept, and I could do this with
20:52anything. I can do this with like
20:53startups, I can do this with a
20:55particular project, with foods, my
20:57relationships, like a hobby, anything,
21:00right? And then phase four, January
21:022026, a month of explosive building.
21:06Everything still require And then like
21:08everything still requires me to actively
21:10prompt and manage each section session.
21:12The next unlocked is figuring out how to
21:14get agents task to run automatically.
21:16The friction is no longer with Obsidian
21:17itself, but with the boundary between
21:19the vault and agent execution.
21:21So you can see I'm really pushing
21:22myself.
21:23Right?
21:24And
21:26it's cool. This is a This is a very
21:28useful thing for me to understand how my
21:32um use of this tool is evolving. And
21:35it's just I think it's just absurd that
21:37I can just be making notes and then
21:41about all of these different things
21:42through my life. Like even as a parent
21:44I I can I can like reflect on the
21:47different things I'm learning. I just
21:48think this is insane that
21:51and uh a computer can have this much
21:53information about me and surface these
21:55patterns. I would not be able to do this
21:57on my own and this fast.
21:59And what a great tool it is for me and
22:01like like I can just like now write in
22:03here and I'm you know, as I'm thinking
22:05about things and it gives me ideas,
22:07right? About um my life and the projects
22:11that I'm working on.
22:12So I can say
22:13you know, it's interesting the way that
22:16my relationship with Obsidian has
22:20evolved over time.
22:23It makes me think a lot about
22:28the way in which my relationship with
22:30computers has evolved
22:33over time since I was a kid
22:36to now.
22:37It's interesting
22:39how these things just happen and
22:42compound over time
22:44and we don't really realize it.
22:47So it's just like a note, right? It's an
22:48idea. So that's an example of something.
22:52So I think a part of getting good at
22:55Obsidian sounds like reflect you know,
22:58inserting reflection into your everyday
23:02life cuz a lot of people,
23:04you know, we're moving from meeting to
23:05meeting, we're busy, we're parents, um
23:09you know,
23:10we grow up and we of course write things
23:13down in notebooks and stuff like that,
23:14but I feel like as we get older, we
23:16actually write and reflect less and
23:18less. Yes. Uh
23:20you know, how do you
23:24how have you been able to insert
23:26reflection into into your life?
23:29Yeah, I think um
23:31for me, it's really about There's
23:33There's two reasons that I think
23:34reflection is interesting in making
23:36notes a lot of is interesting. One
23:39is that it's great to be able to look
23:41back on them. Like for me now, obviously
23:43I can use an agent, but to for me to go
23:45back and see these notes and realizing
23:48that oh, like I'm a person that's
23:49continually changing, my skill is
23:51continually changing, projects are
23:52continually evolving.
23:54It's It's just a It's an amazing part of
23:56life to be able to reflect on how things
23:58are changing over time and how you are
24:00changing over time and how the world is
24:02changing over time.
24:03But the other thing is that there's like
24:04a functional reason, too.
24:06The reason I like to make notes a lot is
24:08cuz that's how I generate ideas. When I
24:10like get When I sit down at my computer
24:12and I write things down, that's where
24:13ideas come. For example, this thing I
24:15just wrote here, it's just a quick note.
24:16I'm just doing it in real time. I'm just
24:18making it up right now.
24:19But
24:21um by writing it out, I feel like I
24:23internalize it more and I like having
24:25good ideas. I like progressing. So
24:28because I like having good ideas and I
24:29like progressing,
24:31writing is how I do that. And so I think
24:34um
24:34to you know, if you want to cultivate
24:37like a writing as a habit, I think first
24:39you have to uh connect it with the idea
24:41that this is how you progress and this
24:43is how you generate ideas and this is
24:45how you have
24:46form original ways of thinking.
24:49Um the other thing I'll say
24:51is that
24:52writing right now is a big way of how
24:55how delegate things to agents.
24:57That's like a whole new That's a whole
24:59new aspect of it. So, if you can develop
25:01a writing habit,
25:03you have a lot more context that you can
25:05pass over to an agent, which then um
25:08dramatically, I think, increases the
25:09amount of like things you can delegate
25:11and the amount of things you can build.
25:13I hope that was a good
25:15answer. How does
25:17How does [snorts] this relate to Open
25:19Claw? Cuz if you think of Open Claw,
25:22it's essentially
25:25you know, at the best case, an extension
25:28of you Yeah.
25:29>> that could go and do things, you know,
25:32uh independently
25:35{slash} based on your guidance. So, how
25:38can you use Commands, Obsidian,
25:42>> [clears throat]
25:42>> and Open Claw, and Reflection
25:45harmoniously?
25:47Yeah, so I think if you look at like
25:49here's an example of one command that I
25:52do. It's just like a schedule command.
25:54Um
25:55And so,
25:56what I asked this thing to do is I said,
25:58"Schedule I said, "Can I take a meeting
26:00with Greg Eisenberg today at at Feb 20th
26:022:00 p.m.?" Right?
26:05And what this does is, of course, it can
26:07look at my calendar and stuff like that,
26:08but it's also going to look through my
26:10daily notes. It's going to look through
26:11what I care about. And then it's going
26:13to give me some perspective. So, it
26:15says,
26:16"Your day is stacked. You're already
26:18recording on Greg's podcast this
26:20morning, followed by a team lunch
26:21outing, and meeting with Peter and
26:23Vince. Your Feb 17th notes show the
26:25episode has been top of mind. The Vault
26:26has a dedicated Greg Eisenberg note.
26:29No, not a two So, the recommendation is
26:31no, not at 2:00 p.m., but you might not
26:32need a separate meeting at all." Yeah,
26:34that's that's actually the correct
26:35answer. How does that relate to Open
26:37Claw?
26:38Well,
26:39Open Claw is just like autonomous agent
26:42that can go and do things if you uh set
26:45it up to do that. Like it can do things
26:46without you having to prompt it all the
26:48time. It can just go and make decisions
26:50and build things for you um,
26:52on your own.
26:54So now what Open Claw can do is in the
26:56same way I just did this command, Open
26:58Claw can do this on its own as well.
27:01And it can go and read my vaults, find
27:03connections, and then make decisions
27:05uh, on behalf of me with like a deeper
27:07understanding of me. And now instead of
27:09like managing an agent or talking to
27:12another human about working on
27:13something, I just focus on managing this
27:15vault. This is like the new source.
27:18I just continually try and make it so
27:20that this vault has all of the
27:21information needed so that I can
27:23delegate to an agent and um, the agent
27:26can just pull from this vault source and
27:28make decisions. And if it's not making
27:30the right decisions, I'm changing
27:31something on the vault. I'm not
27:32necessarily working with the agent
27:34specifically.
27:35Um,
27:37kind of kind of um,
27:39that's my speculation on that subject. I
27:42think it's very interesting. Yeah, I
27:44think uh, one thing that worries me a
27:46little bit about it is if Obsidian
27:48[clears throat]
27:48is really your second brain,
27:51giving Open Claw access to your second
27:53brain
27:55is
27:56scary. Yeah, scary and I would say that
28:00is the
28:01um,
28:03fundamental um,
28:06the the weird element of this
28:08technology, I would say. And I have
28:11purposely given um,
28:14Obsidian, I mean sorry, the Claude code
28:16or any agent access to a lot of
28:18information. I've purposely done that
28:20because my um,
28:22relationship with this is I want to
28:23understand what these things are and I
28:26want to understand what they're
28:27revealing about, you know, how our
28:29relationship with computers is changing.
28:31But um, it's weird. It it's like you
28:34have to really think about how much
28:36information you're sharing with these
28:39agents and whether that's the right
28:41decision or not the right decision.
28:43And I think um,
28:45it's going to be very interesting to see
28:46how privacy as a concept evolves and
28:48changes and what we fight for or don't
28:51fight for and like the in the future
28:53of our society and our world. Even with
28:56every one of these commands, I had to
28:58create a new version of them,
29:00a demo version,
29:02so that it wouldn't reveal too much
29:03personal information while I'm on screen
29:05on this podcast. And even then, it's
29:07like a it's a toss-up. You know, I could
29:09type the demo version, but who knows
29:10what's going to be shown on screen, you
29:12know?
29:14What other commands do you want to show?
29:17Um
29:18so
29:20there's
29:21connect, which is allow it allows me to
29:23take two domains and connect um them
29:26using the Vault's link graph.
29:28So, I can just say
29:30I did one here.
29:33And I just asked it to
29:35uh
29:36connect
29:38filmmaking and worldbuilding. And so, it
29:40goes through and it reads all of these
29:42different files.
29:43And then, it can start to say, "Okay,
29:45let's let's connect these two concepts."
29:48So, notes in filmmaking's uh
29:51notes in filmmaking's neighborhood. So,
29:53it's like 35 film watch list, my first
29:56meeting with Toby, notes in the
29:58worldbuilding neighborhood,
30:00um
30:01the worldbuilding essay, um new as a
30:04media company. So, these are different
30:05things I'm thinking about. So, bridge
30:07one, the interview portal and the
30:09constructed world
30:10in filmmaking. If I notice something
30:12specific and ask the question about it,
30:14it would open a portal into a person's
30:16internal world, which is often a vast
30:18universe of concepts and beliefs and
30:20visions. So, worldbuilding essay.
30:22I want my blog to show you what I value,
30:25what I believe, what I worry about. Like
30:27a tomb from ancient Egypt, I want my
30:28blog to be a place that you dig up and
30:30examine long after I'm gone. These are
30:32things that I've written and I can start
30:34to see how these ideas connect together.
30:36Bridge two, always-on documentary equals
30:39continuous worldbuilding.
30:42Always-on documentary is a creative
30:43strategy where companies continuously
30:45narrativize their characters, pursuits,
30:47conflicts, and visions through
30:48documentary.
30:49So, these are like things that I'm
30:50writing about, and it's showing me the
30:52ways in which these are connected.
30:55Um I think this can get
30:57very interesting depending on the kinds
31:00of things that you're willing to connect
31:01together. You could get probably pretty
31:03crazy with it as well, depending on what
31:05you're writing about in your vault. I
31:06could connect like shawarma and startups
31:08if I wanted to, for example, and see the
31:10kind of connections that are coming
31:12between these things.
31:13Um again, really interesting because all
31:16of this is happening super quickly, and
31:18I don't need to explain any of this to
31:20an LLM. I can just type something like
31:22slash connect filmmaking world building.
31:26The the exa- A lot of the examples
31:28you're using is personal reflection.
31:31Yes.
31:31>> you think about
31:33uh
31:34you know,
31:36for example, note-taking in meetings?
31:38Like maybe you have like a granola or
31:41Gemini notes taking,
31:43you know,
31:44uh
31:45taking notes or and and and sort of
31:47putting it into Obsidian or
31:51um and and by the way, when I say notes,
31:53those could be meetings that you're not
31:55even in. They could just be like,
31:57you know, Tommy met with Vince and they
31:59had this meeting and I want to put it in
32:01here. Yes. Um how do you think about
32:04that?
32:05So, it's a really good question. So, I
32:07think you can use these vaults however
32:08you want to use them in terms of like um
32:11you could put any text you want in here.
32:12If you want to put your granola meeting
32:14transcripts in here, you can put them in
32:15here. And you have to just make sure,
32:17you know,
32:18maybe you're doing something like this,
32:19right? So, you're just like meetings,
32:21and then you're like, okay, these are
32:22this is, you know, project one. And then
32:25every time you do a meeting,
32:27um you take your granola notes and you
32:29just put them in here, right? So, you're
32:30like meeting Greg Eisenberg, Venn
32:34plus Venn, and then,
32:36you know,
32:37that file is created now.
32:39And you can just drop your tran- you can
32:41just drop your meeting notes in here.
32:42And now that's in the vault, and then
32:43you can pass that into the agent, or the
32:44agent will discover it, right?
32:46Especially if you start tagging like,
32:47"Oh, I'm going to tag this back to, you
32:49know, like my podcast or something."
32:51There, now it's connected. So now the
32:53the the agent it has like more context,
32:56and now it knows that tran- this
32:57transcript is related to this other
32:59file. Great.
33:01Um I think that's up to you. I think the
33:02way that you the the amount of
33:04information that you put in here is up
33:06to how you want to use the vault, and
33:07how you want to delegate to things to
33:09agents. And maybe you even want to
33:10create different vaults for different
33:11purposes. For me, per- for me, I use
33:15LLMs and agents as a way to
33:18increase my own level of understanding
33:20of subjects. So I use it for a lot of
33:22reflection and things like that. So I
33:24don't want an agent to write into the
33:27files. Like I could easily get it to do
33:29that. Like I could just say, like even
33:31here, I've I've asked it to write a
33:33description of some commands that I can
33:35talk about today. But I don't I don't I
33:37don't want it to make a file to do this,
33:39because I want to control all the files
33:40in my Obsidian vault. Because I I always
33:43want it to pull from what I think about
33:45things, right? Not what it thinks about
33:47things. And if it starts making its own
33:49files in this vault, then I don't know,
33:51like is like when it's finding these
33:52patterns, is it finding patterns about
33:54things it's written, or is it finding
33:56patterns about things I've written?
33:58So I create a rule for myself, which is
34:01like a strict separation between these
34:03things. I only want it to write things
34:05on the side here, and then I will take
34:07that and and and and write what I think
34:11um should be included.
34:13Right? Yeah. Yeah, go ahead. No, I was
34:16just going to say like,
34:18you know,
34:20I could see the power of just using it
34:22for your own reflections.
34:25I can also see the power of,
34:28you know, AI is really good at,
34:30you know,
34:32going out on the internet, finding
34:34information based on trends and stuff
34:36like that, distilling it in a way that
34:37you want, and having that being put into
34:41your world is also interesting.
34:44>> Yes.
34:45Totally.
34:46I think that and also like let's say if
34:48you asked um the Obsidian if you asked
34:51the
34:52um Cloud Code to go through your
34:53Obsidian file and and generate ideas,
34:57which
34:58um you know, ideas for tools that you
34:59should build. Well, then you can just
35:01like say, "Okay, cool. If that's If I
35:02have an idea for a tool that I should
35:04build, just generate a description of
35:06that, and then just build the tool."
35:09Exactly.
35:11>> Yeah.
35:12I want to show this one
35:15so that it's like less on reflection.
35:18Like for example, like, you know, I'm a
35:21I built this thing called
35:22ideabrowser.com and every single day we
35:24give this validated startup idea. Like
35:27someone theoretically can go and
35:30you know, grab that information, put it
35:32in an Obsidian vault, and then based on
35:35that
35:36basically, you know, help them build the
35:38actual thing, right? Totally.
35:41Yes.
35:42So, I want to show you this because I
35:44think it'll it'll really make it
35:46it'll take it out of the realm of
35:47reflection and into the realm of
35:48building. But, the only issue is this
35:50takes a bit, so.
35:53That's okay.
35:54Yeah. So, the other thing is that with
35:56all of these like with these commands,
35:58another pattern that I'm noticing is
36:00that they take a bit because it's
36:02reading so many files, and I would say
36:04that's a big difference between using
36:06Obsidian and
36:08like or using Cloud Code with access to
36:09this Obsidian vault that I'm noticing is
36:12all of my
36:13requests are taking way longer, and it's
36:15just because it's reading so much more.
36:17So, like look at this one.
36:19So, this is ideas demo. So, I'll run a
36:21comprehensive ideas generation. Let me
36:24start by gathering vault structure and
36:25context in parallel. And then if you
36:27look, it's really interesting to see
36:28what it's doing.
36:30Right? So,
36:31it's like Obsidian orphans.
36:34Right? So, it's it's it's it's like I
36:37guess orphans are like files that are on
36:39their own, not connected to things,
36:40right? So, that's interesting that it
36:41knows that. Obsidian dead ends, Obsidian
36:44resolved, Obsidian tag counts. So, it's
36:46just trying to it's trying to figure out
36:48like some connection between all these
36:49things.
36:50Um and then it says, "Okay, daily read."
36:53So, it's reading my daily notes. Then it
36:55found this file called new context,
36:57which is new is this media company I'm
36:59working on. Then it's like read file the
37:01other stuff context that that's podcast.
37:04The other thing I would say to you guys
37:05is I do manage
37:08um I write uh I create context files for
37:10projects that are pretty extensive. I'll
37:13show you.
37:14I I was I didn't know if I was going to
37:16show this cuz it's very personal, but
37:17like for the other stuff,
37:19look at this. So, other stuff working
37:20context, what shifted recently,
37:24front-loading profile by traveling to
37:25San Francisco and New York City to
37:26record guests. This is super personal
37:28stuff.
37:29But what is the other stuff? The format,
37:31core beliefs of the show, research is
37:34the foundation solid, the best
37:36conversations feel like discovery solid.
37:39You know, here's the team that's working
37:41on it. And so, what happens is this,
37:44again, very personal, but this is um
37:47context that it's just pulled in. So,
37:49now it knows who's working on my
37:51podcast, what are the what are the
37:54recent hypotheses This is hypotheses I'm
37:56exploring. Um
37:58and it just got that information. That's
37:59just like one of the things it did.
38:01Personal workflow context, super
38:03personal file, but it shows like um
38:07you know, like
38:08uh what my daily schedules is like, you
38:11know, things that I have to do like in
38:12my personal life. So, it's like pulling
38:14that, how I like to work, how I don't
38:17like to work.
38:18Personal agent infrastructure
38:20is another thing. Let's see what happens
38:22if I pull this up.
38:25This is a project in which I want to
38:26take a step towards increasing my
38:27personal infrastructure workflow
38:29delegation, however you want to describe
38:31it with agents, understanding what it
38:32means to delegate to agents more and
38:34more. Implementation approach.
38:36So, this is like what I'm writing about.
38:40The file about how I'm thinking about
38:41using agents personally.
38:43The And again, that's an example of one
38:45of the files it's reading. That's just
38:47one of them. So, you saw like the other
38:48stuff, the personal workflow, and it's
38:50factoring that all in to this task I've
38:53asked it to do, which is generate ideas
38:55for me.
38:56Gathering data from your daily notes,
38:57calendar, and vault structure. This
38:59takes a moment since it's pulling from
39:00multiple sources. Again, one of the
39:04um things with this is that it's just
39:05going through a lot of information, man.
39:07A lot of information. So, it takes
39:09longer.
39:10You know, it's already been going for 5
39:12minutes.
39:13Right? And so, that's that's something
39:15I'm noticing, but
39:17for me, that's what I want. I want that.
39:19I want I want a response from LLMs that
39:22is very, very contextual
39:24uh to the things that I'm writing about,
39:26and I think a lot, and I think that's
39:28how me and an agent can work best
39:30together, where I just focus on
39:32continually um
39:34noting that my the the where I'm
39:37currently at in terms of the projects
39:39I'm working on and what my understanding
39:40is and what I find interesting. I want
39:42to maintain that and make it as current
39:44and as deep as possible. So, whenever
39:46I'm talking to an agent, it has the best
39:48representation at all times of who I am
39:50in that moment when I ask the agent for
39:52something. Yeah, I mean, that's the
39:54goal, right? You That's the question
39:56that we all should be asking of
39:59ourselves, which is Yes. Does the agent
40:01have the most up-to-date information on
40:04the projects, on my preferences,
40:06and my dreams, and my hopes, and my
40:08goals,
40:09um because you're it's only as good as
40:13the up-to-date version of that, correct?
40:15Yes, 100%. The quality the the quality
40:19of information that the agent has
40:21entirely determines what it can do for
40:23you.
40:24Right? If it doesn't know a lot about
40:25you, it's not going to be able to do a
40:27lot for you, but if it knows a lot, then
40:29it can it can
40:30it can do things for you that
40:33I I think like even some of your in some
40:34it's kind of weird to say but I mean
40:36like that you don't even know about
40:37yourself in ways.
40:40I mean it makes sense, right? Cuz
40:42ultimately what this is doing
40:45like to distill it to its core, it's
40:46connecting the dots. Yes.
40:49>> Obsidian and Obsidian and Claude code
40:51here are are connecting the dots.
40:53Now,
40:54it's actually quite difficult as a as a
40:56business owner or just as a personal in
41:00our personal lives to connect the dots.
41:01Like why do people in a lot of ways go
41:03to
41:04you know, coaches, therapists? Yes. Um
41:07you know, if you go to a therapist and
41:10you know, you have someone
41:12who is
41:14you're doing most of the talking, right?
41:15Think about it. You're doing a lot of
41:17the reflection and the therapist and
41:18coach is sort of guiding you. Totally.
41:21That's what this is doing in a lot of
41:23ways. And I'm not saying by the way that
41:25don't go to your therapist, you know,
41:27just
41:28but I'm but my point is it helps it
41:31helps you helps you uncover
41:34what are there dots and how you can
41:36connect them.
41:37Yeah, absolutely.
41:39And for me, yeah, it's it's just really
41:41exciting. Um and yeah, it's just it's
41:43just a crazy time with computers. So
41:46let's look at this. This thing finished.
41:47So this is an idea generation report.
41:50Vault relationship exploration. This is
41:52this is pretty ex
41:54tensive, right? Like
41:56to get an idea report.
42:00Like I I think this is this is really
42:02going to show how we can move you move
42:04from reflection to something actionable.
42:08So,
42:09structural highlights. So again, this is
42:12just Obsidian stuff, right? Orphans
42:14worth noting. There's some defense
42:16technology stuff here. Just a a theme
42:18that's growing in Canada.
42:20Massive intellectual investment sitting
42:22in isolation. Also, orphan agentic
42:24software. So, orphan just means these
42:26are files I haven't really linked.
42:27Random notes I just wrote once or
42:28something.
42:30Um unresolved links that reveal latent
42:32interests.
42:34Uh hidden relationships. Again, all
42:36reflection stuff. Fine. What's working?
42:39Obsidian cloud code as a combined system
42:41is working for me. This is producing
42:43genuine breakthroughs in thinking and
42:44output.
42:46Day per domain structure when enforced.
42:48This is basically I started um splitting
42:50my schedule where each day has a
42:52specific focus.
42:54This is cool and this is very true. The
42:56Greg Eisenberg episode as a forcing
42:58function.
42:59It's compressing months of thinking
43:00about Obsidian and agents into a clear
43:02thesis with demos. Very true.
43:04It you know, coming on the show and
43:06doing this forced me to um synthesize
43:09everything that I knew and present it.
43:12But, here's where we're going to the
43:14actionable stuff. Tools to build.
43:18The {slash} graduate {slash} command.
43:20Daily note idea extractor. Based on
43:22daily notes are full of idea
43:24Sorry, daily notes are full of idea tags
43:26and interesting thinking that never gets
43:28developed. The vault has nine idea tags,
43:30but hundreds of undiscovered insights.
43:32Build a command that scans recent daily
43:35notes, identifies ideas tagged or not,
43:37and prompts you to decide
43:39create a standalone note, add to an
43:41existing file, or dismiss.
43:44This turns a daily note stream into a
43:45structured idea pipeline. Obsidian vault
43:48for new.
43:49It says, I just have to manage and set
43:51up a central Obsidian vault for new.
43:53What that means is in the same way that
43:56I'm creating this a vault and it has all
43:58my ideas and my patterns and everything
43:59like that, um why would I not create one
44:02for my team?
44:03Where they like as a team we can go and
44:05ask this vault questions and we can all
44:07uh contribute to it.
44:09Here we go. Tools to start using.
44:13What is this? Types for all external
44:15documents.
44:17Uh interesting. A time blocking app that
44:19enforces day A time blocking app that
44:21enforces day per domain, meaning since
44:23I'm trying to focus on one thing each
44:25day, one aspect of my life, it's saying
44:28why not create a time blocking app that
44:29forces you to do that.
44:31Interesting.
44:33Systems to implement.
44:35One sentence in Obsidian, agent handles
44:38the rest. This is literally the demo
44:39three version of the Greg Eisenberg
44:42Eisenberg prep. You're already imagining
44:44it. The next step is to actually
44:45building it. Start small. Write schedule
44:48a call with person about topic this week
44:49in a daily note and have Otis or Claude
44:52bot or open claw pick it up and handle
44:54it. So it's saying maybe you can
44:56delegate right from the note itself is
44:58how I'm interpreting that. Super
44:59interesting. Yeah. I just inline inline
45:03delegation. Like
45:04>> Yeah. Maybe that's even like a new UX
45:06pattern. I don't even know, right? That
45:08you could build into like these
45:09different tools.
45:10Um
45:12Subjects to investigate.
45:14Christopher Alexander's pattern language
45:16applied to digital spaces. Interesting.
45:19Um Black Mountain College as a model for
45:21the stadium. Stadium is a physical space
45:23we have in Toronto. Authorless media as
45:25a concept. Uh how Shenzhen's hardware
45:28ecosystem actually works. Things to
45:31write and publish.
45:32That'd be useful.
45:34Context architecture essay. The computer
45:36as a place. Software will become
45:38fashion.
45:40What Toronto theory actually is.
45:42Editorial thinking zine.
45:45Conversations to have. This is
45:47interesting. These are real people.
45:48Aaron, stadium workshop host about
45:50becoming an anchor of technical
45:52programming. This is a space we have in
45:53Toronto. Tarun, kind of another person,
45:56program about making program the
45:58flagship series. Steph Ango, Obsidian
46:00CEO, about the vault as a place.
46:02Um
46:03And so this is like
46:05Yeah, this is crazy. It's It's
46:07suggesting people I should meet. Right?
46:10Top five
46:11high impact do now. Build the graduate
46:13command or do a manual weekly idea
46:15review.
46:17Um
46:18This is crazy, dude.
46:20This is actually crazy.
46:22Like
46:23And the fact that it's in plain text and
46:26it's just not be It's not There's no
46:28images. It It doesn't doesn't make it
46:30easy to read, but I kind of like it cuz
46:33it's
46:34it's like dressed down, you know what I
46:35mean? Yeah, I mean, I like that
46:37aesthetic cuz I'm a nerd, but you could
46:39just say, you know, you could you could
46:40just say, obviously you could just say,
46:42"Can you turn this
46:44into a beautiful readable HTML file that
46:48is on my desktop. This is hard to read."
46:51And it's going to do that. Mhm. Right?
46:53So I mean, if you don't That says If you
46:54don't like that, just do it however you
46:56want to do it, you know what I mean? I I
46:58like it like this and this kind of like
47:00this aesthetic. Um
47:02but yeah.
47:03That's how you can move out of
47:04reflection. And of course, you know, of
47:06course we could also say like here,
47:08you know, if we don't if we don't want
47:09to do this, uh we could also say like it
47:12it recommended that {slash} graduate
47:14command. So I could just say, "Build the
47:16{slash} graduate command."
47:18Mhm.
47:19>> [sighs]
47:20>> Right? Which is interesting.
47:22And is that how you started building a
47:24lot of your commands? Like it it sort of
47:27it started suggesting and you're like,
47:29"Just go build it." Well, I started off
47:31actually like building them myself. Like
47:33I'm like, "Oh, trying to think about
47:35commands myself." But then yeah, I said
47:36I started asking the agent like, "Well,
47:38wait a second.
47:39What commands do you think would be
47:41interesting?" And I just And this might
47:43be useful. Another thing that I'd like
47:44to do is I like to
47:46move to like higher levels of
47:48abstraction when I'm using an an LLM.
47:51What I mean by that is I could say like,
47:53"Oh, make a command that um
47:57tells me what should focus on each day.
47:59And that's
48:00that's like um
48:02a command that I thought of.
48:03But uh when I when I Another thing you
48:05can do is you can step back and I can
48:06say
48:08based on my Obsidian Vault and what you
48:11know about me, form an understanding of
48:14what you think my level of understand
48:16like
48:17where you think my skill level is in
48:19terms of a person and the projects that
48:21I'm working on.
48:22And based on that, suggest the kinds of
48:25commands I should use that would take me
48:27from the level I'm at to a higher level.
48:30Right? Get like, you know, get it to
48:32suggest the commands for me instead of
48:35me suggesting the commands and I could
48:36pick between them. So look at this. This
48:38is This is It was This was the agent's
48:40idea based on what it read in my vault,
48:42based on the notes I'm taking, right? So
48:44let's see what this is. Daily note idea
48:46extractor.
48:47Ideas, insights, and original thinking
48:49accumulate in detail in daily notes, but
48:51rarely graduate into stand-alone notes
48:53where it can where they can compound
48:55through backlinks. This command scans
48:57recent daily notes, surfaces the best
48:59candidates, and helps decide what to
49:01promote into like an idea or something,
49:02right? Sure.
49:04So this is how it's going to work.
49:06It scans all the recent daily notes, it
49:08cross-references with the existing
49:10vault, it presents candidates, uh it
49:13graduates selected ideas. If uh if
49:15creating a new stand-alone note, create
49:17the note in the vault root.
49:19Write the note as a mini essay or
49:21working document that captures the core
49:24claim or question, context from the
49:25daily note where it originated,
49:27connections to other vault's notes as
49:29backlinks. Now,
49:30all this stuff, like it captures the
49:32core claim or question,
49:34you might be looking at this and think
49:35like, okay, this is just a text that the
49:37agent generated, and it is,
49:40but also it hits differently for me
49:43because I know
49:44like I'm writing a lot about these
49:46things. I know I know like even like the
49:48mini essay thing. These are words that
49:50mean specific things to me.
49:52Which is just so it's just so crazy.
49:54It's very contextual. I know what it's
49:56talking about because I spend a lot of
49:57time in in this tool and I spend a lot
49:59of time writing. So, yeah, create it.
50:02And it's going to create that command.
50:05Right? And it's like
50:07that's that's nuts because I'm just
50:09going about making notes and I have this
50:11parallel agent that is looking at my
50:13notes and giving me ideas on how I can
50:15improve my workflow, improve my life.
50:17And then and then not only can it just
50:18suggest it, it could just build the
50:19thing.
50:21And it's done.
50:22And we have it right here, /graduate. I
50:24could just hit it.
50:26And it's going to run.
50:27It's crazy.
50:29If if I'm OpenAI or Anthropic, I'm
50:33buying Obsidian.
50:35>> [laughter]
50:37[gasps]
50:38>> Right? Cuz it's the missing link. Yeah,
50:40it's nuts. It's the missing link. Why
50:42the fact that there are people like
50:44you've sold me on this, by the way. I
50:46have
50:47I've downloaded Obsidian. I think it's a
50:49free tool, right?
50:50>> Yeah, it is, yeah.
50:51>> It's open-sourced.
50:53I've downloaded it, but I have not
50:57created my vault because I wanted you,
51:00Vin, to take me [laughter] through. I
51:02knew that this was going to be great. I
51:04knew that I would go through this. This
51:06actually exceeded my expectations. Like
51:08this is
51:10the the fact
51:12like
51:13it makes no sense. It makes no sense.
51:16The the fact like that if you are if you
51:19are use if you are serious about
51:23using LLMs to take your ideas and put in
51:26and and get the most out of them, if
51:28you're serious about building, you know,
51:30what people are calling a personal OS,
51:33uh
51:34and you are not using a centralized
51:36note-taking tool like this Yeah. that
51:39uses markdown as the foundation, then
51:43you are not using LLMs properly. Yeah,
51:47or at least not at the limit.
51:50Yeah. Yeah, exactly. You're not getting
51:51the most out of it. Yeah. You're not
51:53getting the most out of it. So, I think
51:55what's what's difficult about this is
51:57that it require It does require a lot of
51:59time. Mhm. And to actually set it up
52:03properly, it requires
52:05uh like
52:08Yeah, I mean it takes a lot of time and
52:10and and the UI is in an is is so
52:12daunting in the sense that it's a blank
52:14canvas. And it's not like, "Hey, you
52:17should like write your preferences over
52:18here." or you know, you kind of just
52:20have to
52:21you have to
52:24come up with these ideas yourselves.
52:26Yes.
52:27But
52:29that's still so amazing, right? Because
52:31I mean, even when we work with other
52:32humans, we have to find a way to explain
52:35things to them. And I just think it's so
52:37cool that
52:38now we can work with these agents, and
52:40we still have to explain things to them,
52:42but we only need to explain them once.
52:44Because once we get it down on into a
52:46file, we can always reference that file,
52:47that explanation of a project or a
52:49preference or anything, and it's always
52:51there, and you can pass it in. Yeah. A
52:53file is like essentially perfect a
52:56perfect memory. Yes.
52:58>> Right? Human beings have memories, like
53:00we recall things. Yes.
53:02>> But there's tons of studies that show
53:04that what we remember in fact is
53:06completely different than reality. For
53:08example,
53:10when we went and got our that haircut in
53:12Mississauga,
53:13>> [laughter]
53:14>> I could have thought that I had the best
53:16haircut.
53:18You know, that's what my memory reme-
53:20remembers. It was a great haircut. But
53:22who knows? It could have been the worst
53:24haircut I had ever gotten.
53:25>> [laughter]
53:27>> Now,
53:29Obsidian or whatever tool you end up
53:32using,
53:33like
53:34you know, if I had written Yes. Like it
53:36it the memory, the file, the markdown
53:39file is perfect. So, that when I link
53:42that or I I recall it, it is going to
53:44give me a perfect
53:46uh da- you know, data point. And the
53:48other thing about, you know, these files
53:51is that you hope, well, they're not
53:53biased, basically. They're as biased as
53:54the human being is
53:56in terms of writing the reflections.
53:58Yes. At that moment in time. Yes.
54:03It's crazy, man. And
54:07yeah, it's just crazy. And there's all
54:09There's all of these different aspects
54:10to it. There's the privacy of it, and
54:12what that means.
54:14There's the power of it, the fact that
54:16now you can just work with these
54:17computers in natural language, and just
54:19delegate to them.
54:20Um
54:21There's the fact that there's people
54:22like me that are that are using these
54:24tools and and trying to figure out how
54:25to delegate stuff to agents in this way.
54:28There's people that are like me that are
54:29even more hardcore in in different ways,
54:31and and pushing them. And I just think
54:33it's such a crazy time to be alive,
54:36because I think we are potentially
54:39watching a fundamental shift in the
54:42human relationship to computers. And
54:44it's just [snorts] I just I'm just
54:45really happy to be alive while this is
54:47happening, and I'm curious like how is
54:49this all going to unravel? Well, what's
54:51what's cool about this is
54:5399.99%
54:55of people are not going to spend the
54:58time
55:00to actually set up something like this,
55:03and make it a part of their daily lives.
55:05And
55:06the alpha, so to speak, is in terms of
55:10leading a more productive, happier,
55:13healthier,
55:15better, more money-making career is in
55:18in in using something like this with an
55:21LLM. I think so.
55:23I'm not saying download Obsidian today,
55:25and I have no affiliation or whatever
55:27with them. But I'm saying like
55:30pick a It sounds like what we should all
55:32be doing, and I'm I'm talk- I'm giving
55:34myself this this advice, is like
55:38there's no excuse anymore for me not to
55:40be writing down and reflecting Yes.
55:43>> into markdown files. Yes. In a world
55:45that where LLMs use markdown files as
55:48the oxygen. Yes.
55:51Like people think tokens are the oxygen.
55:54Yes. But they're not.
55:56Yes. The markdown files are the
55:57memories. Like think about what a human
56:00being is.
56:01Yes. You know?
56:04Is a human being
56:06the the energy of a human being or is it
56:09the memories of it, you know, what we
56:11recall, you know? Yes.
56:13>> that's like a philosophical question and
56:15maybe it's a a bit a you know,
56:18a bit of both, but it's
56:20you know, I think that
56:23there's something really really
56:24fascinating about MD files as
56:28an underrated about them in order to
56:31to have a true
56:34computer experience in in today's day
56:36and age. Yeah, there's definitely
56:38something going on here. Some
56:40fundamental shift. Yeah.
56:42>> Yeah. And like I have bad my word I you
56:44know, I'm learning in real time, right?
56:46Like and I I don't have the right
56:47vocabulary to even explain this. Yes.
56:50And neither do I, man. Neither do I. I
56:52I'm trying to I'm trying to figure out
56:54in real time. That's that's why I think
56:56like I know I show something and and for
56:58me I I'll do something or I'll see
57:00something.
57:01And my friends are like they kind of
57:02laugh because I'll just be sitting at my
57:03computer just tripping out and I think
57:06it's because I really like computers and
57:09and I cannot believe that this is
57:11possible. I cannot believe that I can
57:14just be making notes on my computer like
57:16I have been since I was a kid
57:19and then all of a sudden this agent can
57:21scan through it and build things because
57:23of it and and like connect patterns that
57:26I could never see. It's nuts, man.
57:28It's nuts and
57:30and at the root of it, you're right, is
57:32just a collection of interrelated
57:34markdown files. Yeah. Cool, man. I
57:37appreciate you. I don't know if you can
57:39see my mind, but my mind is blown right
57:41now. Thank god.
57:42>> Yeah.
57:43>> [laughter]
57:43>> Yeah. I wanted to do right by you. I
57:45also just like
57:46I say this every time, man, but I'm just
57:48going to keep always saying it to you
57:50all the time. I really, really, really,
57:52really appreciate everything that you
57:53do.
57:54I think your pattern recognition and
57:56your pattern matching is like
57:59like really underrated. I think there's
58:00a lot of things that you do that I don't
58:02think it's like difficult to see if
58:04you're not really paying attention. I I
58:05just want to say like thank you for for
58:07everything that you do. You're always
58:09putting on like new voices on your show.
58:11I see it. I really appreciate it. And
58:13it's just been It's an honor to know you
58:15and yeah, just thanks for the
58:16opportunity. Thanks for everything, man.
58:18I appreciate you, Vin. You're a legend.
58:20I'll include links for where to follow
58:24criminally underfollowed internet Vin
58:27on on X, on his YouTube show
58:31uh podcast in the show notes and
58:34description. You can go and check him
58:36out there.
58:37Uh people, please play with some of
58:40these tools. Uh and let me know what you
58:43think. Let Vin know what you think.
58:45Please.
58:46>> And uh
58:47Vin
58:49I will beg you to come back on the show
58:51>> [laughter]
58:51>> another time.
58:52And I hope you come back on again. For
58:54sure, man. Thank you. Thank you.

About this transcript

This page contains the full transcript of How I Use Obsidian + Claude Code to Run My Life by Greg Isenberg, generated from the public captions YouTube serves with the video. The transcript has 10,672 words across 1,728 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.