transcript
Matthew Oatts
I’ve spent a bunch of time with undergraduate students at Georgia Tech who I’ve started in a mentoring program around. And undergrads in schools these days are just still as eager beaver as they were when I was in school. Just always wanting more and more time and idea. So I see somebody here, Fran, I don’t know that we’ve met. Maybe we have or not.
Matthew Oatts
But I want to be extra people join to make sure that there’s no bad actors joining because we had a little bit of a disruption one time. You did? Yeah. I wonder who unknown is. Do you want me to give it a shot, Brian? Or do you want me to hold off?
🌴 Brian
What do you think? I say go for it.
Matthew Oatts
Okay. Watch out, Fran. If it’s someone who is going to start displaying unacceptable content, we will hype them out as quickly as we can. Hey. Hey. All right. He’s a real person. We didn’t know who unknown was, but you look real and you look like you’re not a bad actor.
Unknown
So we’re good. Thank you. I signed up with a different email address, but then due to Google things, I was logged in. Okay, cool.
Matthew Oatts
I’m Matthew, by the way, obviously. Brian there. And then we got others joining here. Fran, is it Fran or Fran?
Fran Abenza
Fran. Fran. Yeah, whatever. Actually, yeah. Quickly, I must admit that I won’t be able to participate now much. I’m in a hotel there. Like, there is dinner and the father of my girlfriend was like, not too happy about me doing the call at the same time. So I will just, like, quickly drop by here. You know what would be awesome for me? Can I. Can I get to know what is the agenda? Or can I actually watch this later?
Matthew Oatts
Yes, absolutely. So go back to your dinner and spend time with real people and not us virtual people. It’s going to be recorded. You see an AI agent there that’s not only going to record it, but organize the recording for you to be able to watch it later. So, yeah, do that. And yeah, go enjoy time with real humans. And I’ll make sure that you get. There’ll be a Miro board and I’ll even post it in the lean coffee stuff. We’ll make sure you get access to all that information.
Fran Abenza
Thanks a lot. Yeah, if you can later send to me, Brian, via, like the link via maybe Twitter or something, it’d be awesome.
🌴 Brian
Okay. Yeah, I’ll definitely touch base with you about that.
Fran Abenza
Thanks. Otherwise, I’ll ask later. Thanks a lot.
Matthew Oatts
Thanks. Okay. All right, I see Tapmab. I’m going to admit them too. Again. We’re going to, we’re going to wing it here, see who joins.
Tap mab
Hi there.
Matthew Oatts
Hi there. Welcome. Oh, got an echo. All right, let’s go ahead and get started here. I’m going to mute whoever that was. Make a viewer. Oh, there we go. There we go. Sweet. All right, I’ll start my screen. So how many people are new? I think there might be a couple new folks here to the lean coffee. For folks that are new. Are you new to the concept of a lean coffee or just to this one? You heard of Lean Coffee?
Unknown
I’m new to both.
Matthew Oatts
Sweet.
Tap mab
All right, same here, Matthew.
Matthew Oatts
Okay, great. Well, let me show you this. By the way, I’m Matthew Oates. I am a. I’ve been in the orbit of Brian and his awesome plugins for a while, but I decided to lean into the community to help facilitate connection around people, because I’m not going to actually be able to, like, do any coding contribution or anything like that. But what the lean coffee is, is essentially. And I’ll put the link in the chat for folks. What this is, is just a space for us to come together and share ideas and for, you know, see anyone with the link, copy board link to share ideas, have conversation, communication. Maybe pick Brian’s head a little bit as we learn about new things that are either in. That are Smart Connections or maybe Smart Connections adjacent. And so. Yeah, so we do that. The structure of a lean coffee is pretty straightforward. And so I’ll explain that and then let’s do this. What is today? Today is the 18th. Let me get some of these things. Structure of a lean coffee is relatively straightforward. What we do is there’s sort of. There’s a cadence where this is all emergent ideas of what we want to talk about. So, by the way, is it. Is it names wise, so I can refer to people we got. Is it tap?
Tap mab
Yeah, it’s just tap.
Matthew Oatts
Tap. Okay. And then I know it’s. I know your name’s not unknown. Sorry, what was your name again? I forgot that. Catch it. Auden.
Unknown
Okay, cool.
Matthew Oatts
Auden. Tap Brian myself. So what we’re going to do is, essentially, we’re going to spend some time, because you all have spent the time committed to be here. We get to prioritize what we want to talk about. The idea of a lean coffee is we come up with topics and ideas. We’ll sort of put them in a board of things that we can prioritize. And then as a group, we’ll sort of vote on what’s the most important thing that we want to talk about. That’s what’s important to us. And then we’ll have some time to talk about that thing. And then once we talk about it, We’ll move that off, sort of. If you’re familiar with like a Kanban thing or like progressive tasks stuff, we’ll move that topic off and we’ll, we’ll talk about the next topic. So it allows us to really, as a group, talk about things that we want to talk about and that we prioritize. Brian, did you have anything else you wanted to add before we hop into that cadence?
🌴 Brian
No, all sounds good to me. Thanks for facilitating.
Matthew Oatts
Cool. All right, so I’m going to start a timer here. And Brian, I am sorry, refresh me. Are we doing more technical stuff today or non technical?
🌴 Brian
I say we just do free for all. Whatever anybody has questions about, just put them up there and we’ll do the vote if we have enough questions and see how it goes.
Matthew Oatts
All right, sounds good. I’m going to start a timer for four minutes. You can hop into the mirror board. If you have trouble, you can just let me know. Also, stop sharing my screen, but I’m happy to, actually, I’ll share it for a little bit so people can sort of see what’s going on, and we’ll do that. So if you’re not familiar with Miro, sort of like an infinite whiteboarding tool, so you can select a post it notes. I actually have a topic on mind comparing notebook ln to our usage of Smart Connections and Obsidian. Notebook ln has been like a topic of conversation for colleagues and people in my ecosystem a lot lately, and so I want to pick y’all’s brain around if you’re familiar with it. And how does it relate to Obsidian and how you use Smart Connections?
🌴 Brian
Yeah, that’s a good one.
Matthew Oatts
And then other topics. What’s that thing? Yeah, feel free to just hop in, folks, into stickies if you have topics you wanted to ask about, even obviously if it’s specific to Smart Connections too, that would be useful. I just haven’t been hands on a lot with Smart Connections lately.
Tap mab
Yeah. And Matthew, sorry, apologies. I’m also new to this setup. How do you zoom in?
Matthew Oatts
You should just be able to, if you’re using the notepad, just like Zoom in and out. If that’s not working, there’s this little thing down here. You can zoom in and out. If you’re also looking for someone you could follow, for instance, I could follow Brian right now, and this is Brian’s view. You could click on my face.
Tap mab
You can follow there if it’s actually quite limited because I’m not able to see that I’m on a Mac, so I can’t see the bottom left.
Matthew Oatts
Oh, strange.
Tap mab
So does.
🌴 Brian
Do you have a topic in mind and we can just add it for you?
Tap mab
Yeah, I think it was mostly just around, like, Smart Connections. So I’m definitely new to like, Obsidian and stuff. So I just saw, like, I was like, looking at your video to connect, like, because I just had the idea of just how can I combine like, folders with ChatGPT and how can I go browser? Then I saw that you came across your video and that’s when I came to in terms of, like, Smart Connection. So I just wanted to see more like, the limitations that they are so far, because I was, like, trying to change the folder name with Smart Connections and then I was not able to change the folder name, but I was able to change like, the notes within it as well. And there was also something where you had, like, Smart Connections, experimental. And I just couldn’t get that to work as well on my end. So it’s more just more just to sort of see, just to have a feeling of where they started. And because I kind of see what their vision is. Like, it’s probably almost like you. You’re trying to just like, pull off notes. It’s probably like a, what’s it, like a Javas or Tony Stark kind of thing. And yeah, I can imagine in the future you will be like hands free and just call, like, hey, can you remind me of that note? Can you update that note for me? Or based on this conversation, end up that on Obsidian acting as a second brain. So I think that’s what I kind of got the g side of it, but, yeah, please let me know.
🌴 Brian
Okay, so editing capabilities from within ChatGPT, does that cover, like, the topic that you’re thinking?
Tap mab
Yeah, yeah, that should be fine.
🌴 Brian
Okay. And if you have anything else and you can’t get in there, just let me know and I’ll, I’ll type it up.
Matthew Oatts
There’s one other thing I had.
Tap mab
Another will be a custom, sorry, what was that time? Custom chatbot.
🌴 Brian
Like custom GPTs?
Tap mab
Yeah, custom gpds?
🌴 Brian
Yeah. Okay. Yeah, I’ll make that card custom scGPTs clarified.
Matthew Oatts
All right, so now what we can do as again, the group of people that have invested time in here, we get to have the authority to pick what we want to talk about. So do this. I’m gonna start a vote and I’m gonna give everybody three votes. Just stickies. Don’t need this one. You should see a joining to. There and then we’ll start that voting. And so you’ll see a thing called join voting and you can join there and that will allow you to essentially vote on topics that we’re going to talk about. All right, looks like 404 have voted. All right, so we got editing capabilities within ChatGPT, and then comparing notebook LM and then custom Smart Connection GPT is clarified. So let’s stack rank those over here. We’ll do this one is going to be the first topic and we’ll talk about these two down here. So, Brian, I think you’re probably the best person to kick off this part of the conversation. I’ll go down below and start taking notes as best I see fit, but yeah, yeah.
🌴 Brian
So as far as the editing capabilities in ChatGPT are concerned, so this is going to be something that’s specific to the Smart Connect app and not the Smart Connections Obsidian plugin. So in order to use the ChatGPT with your notes, you have to have the Smart Connect app. And those editing capabilities are of course going to be limited by the progress on that application. The current state of the application is there’s this notes version one GPT, and I’d say the large majority of that code for that GPT is actually quite old. So it was one of my very first proof of concepts and prototypes that I built and I. It was really cool. It was like a really nice start. But I quickly started discovering a lot of limitations using the way I was going about the logic in that notes version one GPT. That brings us to notes version two GPT. The way the notes version two GPT works much more dependent upon the structures that you’ll see when you’re using Smart Connections. The Obsidian plugin, for example. In the Obsidian plugin, it breaks down your notes into what I’m calling blocks. So, like, another common term for that would be chunks. But I use the word blocks because chunking, it sounds a little messier than what blocks are supposed to be. Blocks are supposed to be these logical segments of content. When I imagine a chunk, I just imagine chopping up a note like every hundred words or something like that, while a block is this logical segment of the note. And when it comes to markdown, like, which is the syntax used by Obsidian, what that looks like is the headings in your note essentially become part of a file path. And like, how does that matter for editing capabilities? How that matters is when ChatGPT is looking at your notes, instead of just seeing this like, big block of text, it sees these segments, and each of those segments has what I call a key. But that key is very similar to like if you had a file path to a file. So then when the ChatGPT says, okay, I want to update this note, there’s this sentence, and it’s in a heading that is midway through the note. It’s able to say, here’s a path, or here’s the key directly to this smallest segment of the note, and then it can just rewrite that small segment of the note, rather than having to rewrite the entire note in order to change a piece. That might be minuscule compared to the length of the note. So that’s the primary paradigm shift between the notes version one GPT and the notes version two is that the notes version two is much more like the editing capabilities are baked. In to how the Smart Environment breaks down notes in order to make sense of them.
Matthew Oatts
So to repeat that sort, I’m trying to attempt at doing some visuals here for you, Brian, so we can, for the visual, visual learners out there that end up watching this. Yes. Perfect. Am I close in? Like, it’s not actually. Let’s see this.
🌴 Brian
I mean, I definitely see where you’re going with the visuals here.
Matthew Oatts
How off am I?
🌴 Brian
No, I mean, I think that it’s like, definitely going in the right direction. So I don’t know if anybody here is familiar with these, like, coding AI’s like cursor, but the way cursor works is it literally goes through the entire file and reproduces the entire file with the exception. It’ll add placeholders, and then it’s similarly parsing out these placeholders in order to create a final output for editing your code. What’s going on in Smart Connections is just, I would say it takes it to, like, a lower level of granularity, and I think that that is going to add, you know, like. Like not just flexibility when it comes to editing, but this whole, like, breaking things into blocks idea. It also enables things like a change history. So in your Obsidian, like, if you’re using Obsidian sync, or if you’re familiar, like once again with code where they have like GitHub or the Git program, which helps you track change history, the idea behind blocks is that you can have this more granular level, more granular level of history. And while that has direct implications for editing in the long term, when we have a history of these edits, we can also take that and use it as context of what does a successful change look like. I want to improve my note. I have this big long note. It’s a first draft, and I want to make progress on that note. Well, there’s probably a lot of similar segments compared to other notes. For example, I like to have a heading for why does this note exist? And another heading for next actions. What are the next tasks? I need to do to move this note forward and I have these like structured sections. So what I would like to see happen, you know, as like the next step is being able to track change history based on those sections. So then when the GPT is attempting to help me make progress, it can see what did progress look like in the past, you know, so I guess that’s like, that’s a lot of my thinking behind, you know, the architecture that goes into this. Does anybody have any more questions? You know, that can kind of hopefully keep me on the rails with what you guys are interested in hearing about?
Matthew Oatts
One of the things that I’m curious about is relative to like, I think a lot of folks are always interested in timing of availability of these things and we talked. I know I missed last one, but like stability of it and things like that. In terms of like the different, like pragmatically leveraging notes v two versus notes v one is like, they’re a different link that we should. Is this like the experimental one? Is that what you mean by experimental over here and like those types of things?
🌴 Brian
Yeah. So notes V two is currently labeled as experimental. I think I’ll probably be dropping that label once the next version comes out. So I’ve been, you know, that’s been like my primary focus of the latest update I’ve been working on, which hasn’t shipped yet. But the upcoming update, there will be basically more, the features of the version two will be a lot more comparable to what’s available in version one. Right now in version two it lacks things like the alignment note, which I just re added an alignment action. And the alignment action has a little bit of extra, a little extra capabilities compared to the version one. But you know, like I didn’t do too much there because as far as alignment goes, that’s, that’s a whole sub project that I’m working on. You know, so like I wanted to bring some level of alignment like, you know, some like level of control into that version too. So I did what was pretty much like a copy and paste of the version one with the addition of a couple extras. But you know, long term as far as that alignment goes. So if you’re familiar with the Smart Connect app and setting an alignment note, basically it’s like another set of custom instructions that you can give ChatGPT to help you work through your notes. But where I see that. Transitioning into is going to be more of like an alignment folder. So while you’ll still be able to add some static custom instructions, I think what this morphs into is having an entire folder of content that while you’re interacting with the GPT, it’s just adding notes in the background, very similar to how the ChatGPT memory works. So I’m still working out the details. How are we going to. And this brings up a major principle behind all the software, which is how do I complement what OpenAI and ChatGPT and these other platforms are doing? Because I don’t want to spend my time just reproducing what they’re doing. We’re working on something that they’re going to release as a feature in the near term, or even long term for that matter. So what does this alignment folder, what would that look like if it complemented what already exists in the form of ChatGPT memory? And that’s something I’m still working through the details on. But getting back to the difference between v one and v two. V one, like I said, it’s very much legacy code. I see that you have the application brought up there, Matt. So just so you know, to get to the link for version two, you should be able to. Okay, so where it says no speed to experimental, is that clickable? That’s not clickable. Yeah. Okay. That is actually a mistake on my part. There was supposed to be a link there that would just open up the GPT. So I’m making a note to myself here, added backlink to notes v two GPT. And that’s a result of I’ve been working through the user interface architecture. Basically I just like moving fast. So the first iteration of the UIHe was not a very sustainable architecture. So I’ve recently just been rewriting all of this UI to be something that’s going to be more sustainable in the long run. And of course I’m breaking things and losing things in the process.
Matthew Oatts
Cool. Is there a way to find the notes v two thing? One of the things that I’ve recently been struggling with. I know this is a little bit of sidebar. I’m not wanting to hijack this, but in terms of if I wanted to, actually, there’s so many different ones, I don’t know which one to use. Yeah, like, Smart Connections chat. Smart Connections note, it looks like the chat with your Obsidian notes one is like 300 plus. So is that where you’re saying the actual proper links to find out which one is going to be. Like, I could click it from here to know the right one to go to.
🌴 Brian
Yeah. So in the previous version, there was a little, like, you know, like that, like where it said notes experimental. That was clickable, you know, so I just. That just got lost in the. In the update, you know, the last update, which pushed a bunch of new UI stuff, you know, like, as far as, like, where you would find that link. Like, I could give you the link right now, but I’ll say you’ll just want to wait until the next version where the link will be there anyway, because, so, like, I’ve been, you know, like, I’ve been working on that notes version two, except for the. The actual version two, like. Like, where that link would take you is not updated yet. I’ve had to work using one of these custom GPTs, which is also, you know, like, so I know there was a question about the custom GPTs in here. You know, basically I’ve been. I basically just copied that experimental notes GPT into a custom GPT and I started updating it from there, my local version. Version. And that will all be moved over to the publicly available notes v two GPT upon the next release of the software,
Matthew Oatts
which before we move on to this next topic, I want to check in with folks, and I also want to clarify something that took me a while when I first started to get involved in Smart Connections. And what is the difference between Smart Connections and Smart Connect app and all those types of things for those watching the recording, or both of you guys that are on this call, and Brian, correct me if I explain this incorrectly, but there is Smart Connections in Obsidian, which is essentially a plugin within Obsidian, and that’s where you can have, within an Obsidian panel, chats using your open AI, whatever key you have, whatever LLM you connect it to chats within Obsidian with relative to your notes, and it does embeddings. You’ll see it open up with the plugin and it’ll do all the embeddings there. Separate from that, there is the smart. I don’t know how you brand it, but either the Smart Connections with ChatGPT, or I guess that’s probably smart. Is it smart? Do you call it Smart Connections with chat
🌴 Brian
as far as the app is concerned? I’m just calling it Smart Connect right now, but I do apologize. Yeah, the naming is a bit of a mess, yes.
Matthew Oatts
So there’s the Smart Connections Obsidian plugin, which is essentially the thing that allows you to have an Obsidian plugin and use your key to actually interact with nodes within Obsidian. Then there is a ChatGPT tool, which is what we were just showing, which gives you the ability within ChatGPT to have a conversation with ChatGPT and directly engage and edit your notes. That’s the thing that requires the Smart Connect app, which is like my understanding, is pretty much like a local server that helps connect the ChatGPT like actions infrastructure and that sort of architecture with your actual notes saved on your local computer. So if you’re confused at all between, wait, what is this app thing? And you’re just a Smart Connections with Obsidian plugin user, they’re sort of two different but related products that have similar roadmaps.
🌴 Brian
So I think I can add a little bit there. Like everything you said is 100% correct. I think I can add a little context about why are there multiple apps or plugins here. So the underlying concept that both of these pieces of software implement is what I’m calling a Smart Environment. And a Smart Environment is essentially a wrapper for your local data that enables you to control how AI interacts with your data. So as far as the Obsidian plugin goes, it’s essentially a window into your Smart Environment. Where the Smart Connections name comes from is that there’s this Smart Connections view, which takes the data from within your Smart Environment and says, hey, based on what you’re currently looking at, this is probably most relevant using AI embeddings. So that’s where the Smart Connections name came from. But really what that plugin is doing, it’s just tapping into your local Smart Environment. Then we have the Smart Connect app, which very similarly connects into your local Smart Environment. How that works, how they’re connecting into the same thing, is that they’re both using the same underlying settings. There’s this Smart Environment configuration file. So when you’re changing the settings, whether in the app or in the Obsidian plugin, they’re both updating the same. Smart Environment configuration file, and that defines what data is accessible. And then in the Smart Connect app, it also defines what actions are available. So in the Obsidian plugin, it’s essentially read only. While the Smart Connect app brings in these extra editing features and the Smart Environment, the design is to give you control of what the AI is able to read, what it is able to edit. I hope that added a little bit of context around what these two pieces of software do. Something I left out is the Smart Chat. The Smart Chat was an additional feature that came into the Smart Connections Obsidian plugin. It’s something I’m working on a major rewrite for right now to bring it up to speed with a lot of the other architecture decisions I’ve made since initially releasing that chat. At the end of that, the Smart Chat will actually become its own plugin. Then this way we have the Smart Connections plugin. It’s specifically for showing you what connections exist to your current content. Then there will be the Smart Chat plugin, which enables you to chat with your notes. Both of these share all the same underlying internals, so it won’t be reloading the same things. It’s the way the software just designed is to reuse those same internals. And if you’re interested in the code, all of these, pretty much all of the code is available in the JSBrainss repository. On my GitHub. I break down all of the pieces into these reusable components. They get reused. The things in Obsidian get reused in the application, the Smart Connect app. I expect that the Smart Chat will be available hopefully by the end of the year in the Smart Connect app. And what that will allow you to do is to work having the same capabilities as ChatGPT gives you, but using APIs, maybe you’re using anthropic or Google’s API or even local models to utilize the actions and capabilities from the Smart Connect app. And you know, That also includes bringing the capabilities from the Smart Connect app into the Smart Chat plugin for Obsidian.
Matthew Oatts
I’m going to see if I can repeat back because I found that useful to understand the context there. Brian, am I still sharing my screen? Yeah. All right, so if you look at, like, we see how close I am here, and then I do this, this, this, this got your actual grabs one, right? So is this close in the sense that the, if I repeat back what I was just hearing, you have your Obsidian markdown files, which are essentially middle files here. And if you start to build an architecture that sort of can access those files and create a Smart Environment, sort of wrapping all of those files, these are still like those Obsidian files here. And then if you still have the. Obviously the UI with Obsidian, you can just use the UI to access your notes, and that’s no plugin necessary. But when you add this idea of a Smart Environment, and then you have your Smart Connect app, you have the ability to, and you don’t necessarily even have to have the Smart Connect app. There’s an optionality where you just like, the orange thing is, you just go straight to the Smart Environment. You don’t even need to Smart Connect path. But the idea is folks that have encountered this ecosystem, maybe simply encountered it because they saw Smart Connections as an Obsidian plugin. I bet you that’s probably the most common entry point into what the heck we’re talking about. And so they download the Smart Connection subsidian plugin underneath the covers. All the subsidian plugin is leveraging a wrapper around their notes that allows you to get some interesting visual insights into how some of your notes are related, and that’s how you’re handling all your embeddings and stuff here. And then you also have this cool little tool where you can have a Smart Chat leveraging the same kind of stuff, and you can go to the Smart Environment stuff that you’ve built underneath the covers. And so I would say probably the vast majority of people that interact with anything Smart Connection ecosystem related, just download the plugin and there. They’re doing that and they get their little list of how closely related these connections are. And that’s neat. And then at some point, they encounter this idea of, oh, wait a second, there’s a way I can actually leverage ChatGPT to update and change my notes, edit notes, and so what that is, well, that’s a sort of a different part of the architecture. That’s part of the architecture that one definitely requires this Smart Connect app. So you have to essentially have this app, that app is running on your local environment, because that’s the thing that you need to have access to actually leverage ChatGPT’s actions architecture to be able to interact with different things. You can interact with notes there. That’s what gives you the ability to have a ChatGPT window, to do something with your notes, to update it. Let me see if I can find a decent example of this that I did the other day. So, so one of the things that I do, and I had to update it, and it’s sort of not working, but it’s besides the point. I. I like to sometimes exactly find in my Obsidian without sharing all of my secrets. Give me 1 second here. Not that I don’t have a lot of secrets. All right, so I have a implementation of this architecture. And by the way, is this useful like this? I’m sort of going off on a tint, like, hijacking the conversation, but I’m seeing some nods that this is useful to at least talk about.
🌴 Brian
So.
Matthew Oatts
Yeah, okay, cool. One of the things that I do, and I’m okay with doing this. Yeah, okay, cool. So here’s my Obsidian vault. I’m gonna have to be super confident that I don’t give away any, like, private stuff. But one of the things that I have done is I like to use Smart Connections, and this is me actually testing something. Let me see. I don’t know if I can have. Let me go back to some older variables real fast. Doesn’t work. I like to. Let me see if I can actually find this article. Here’s like a local consulting firm. I’ve broken it, but here’s the idea. Imagine you encounter a blog post where it’s just a blog post, an interesting blog post. What I like to do is have a Smart Connections conversation in ChatGPT. To have ChatGPT essentially summarize that article and then generate a note for me that creates a note in my vault, my Obsidian vault, that has a summary of the article that maybe also starts with some backlinks that would be useful to create in the future and then give me the source file so I can sort of have it sort. Be an archivist of things that I do that improves pocket like use over like pocket or other sort of bookmarking tools. So I have set up Smart Connections, and again, this is not the blue, this is now getting into the green world. I’ve set up Smart Connections. I have the Smart Connect app running, and so I have a conversation in ChatGPT, and then I give it instructions on what I want it to do. So summarize the article. If it’s over 500 words, reduce it down to 75% of its length. If it’s over, or if it’s under 500 words, just sort of keep it the same. If it’s over, have it be 50% of the lengths, or shorten it. Include some important topics, like extract out from this some insight that you might have as to what are important key topics from the article. Retain the URL that I give you, create a new note or the link, and then put it in a particular folder. So what this is essentially doing is turning ChatGPT with Smart Connections into an archivist of things that I encounter, and it puts it in a specific folder. So for instance, this article, and this is what that’s done, is essentially, hey, here’s this thing that I found. It’s highlighting the article and it’s giving me links so that if, for instance, technical debt happened to also be a note that I had in my vault, that could do that. But I tend to do this for, and I actually did this one AI overhyped. Here is something that somebody published on LinkedIn. It summarized it, I added some own notes myself. And so I essentially have, I’m leveraging Smart Connections to automatically build notes, and we can try to see if this actually works here. On another note, like, I really like this large language model here. Here’s a good article that I encountered the other day. It was just a good explainer of large language models and how they work and that type of thing. So if I do this, the idea here is like, notice that why tech modernization fails is the most recent article. I’m just going to give it another URL. I don’t know if it’s going to work or not because I haven’t been messing with this for a bit, but I’m essentially giving ChatGPT this. It’s going to go read and find the article. And this is a real pressure test, Brian, to see if it’s working out. So it’s starting the action. I could probably open my console if I need to see it working. And then, yeah, I don’t know if it’s going to work or not. Oh, it’s done. There we go. Created a large language article. I sort of covered it up. It’s sort of. I missed the drama. But here it is largely language articles explained. So now here’s the article. It’s got that. It summarized it. And now if I wanted to build connections between this, I could. So that’s for folks that are new. That’s this thing that is different than the Smart Connections.
🌴 Brian
Anyway, yeah, thanks for sharing that example. It’s, you know, I’m constantly like, you know, working to figure out, like, how can this thing actually be valuable, you know, because, like, I don’t want it to just be a novelty, you know, like, what, what is a workflow that is going to actually produce, like, value rather than just being, oh, wow, look, it can, you know, read back my notes to me and, you know, whatever. So it’s like, it’s really cool to see how you’re actually making this useful for you.
Matthew Oatts
Yeah. The value for me is it’s an aggregate. These are all articles that I’ve encountered since I started doing this that are just things that I just don’t have the time to actually go figure out. But I get summaries of, that’s funny that, that actually I came in with, like, an actual person, but I get summaries of different articles. And then what it does is when I have to craft, because I’m in consulting, so when I have to craft advisement material around certain things, particularly if it’s actually flagged a note that already had connections, I could start building, like, literally Smart Connections between articles that I’ve read, that I didn’t actually read, that this thing parsed for me so that I can get more value for my clients out of things that I’m multiplying the amount of connections that I can make for my clients when I give them advisement around stuff without actually having to do the work myself.
🌴 Brian
Yeah, awesome.
Matthew Oatts
Yeah, cool. Anyway, that was a little side quest for us there, so pause, because I’ve been talking a lot.
🌴 Brian
Let me just, I think I can add this in 30 seconds or less. So just to get a little bit more of an idea about why is all of this necessary? So what this is replacing is a lot of what would be happening behind the scenes if you were to use. And this might hopefully tap into another one of the topics a little bit, like open or not open, but notebook lm. So when you upload your content to notebook LM, it’s essentially creating a Smart Environment and it’s providing all of these different functions to interact with. That content that you uploaded to their servers, except for the differences, is all of that is on some third party server somewhere. And what I’m working on is building something that you can be confident in. The data that you’re sharing with AI is done, so on your terms. And you don’t have to just be thinking that, oh, all of my proprietary information is now sitting on some server somewhere just waiting to be accessed by the next hacker or whatever, or even the company themselves, trying to utilize your data for their own profits.
Matthew Oatts
Yeah. And I think that’s the thing for me, too, that just drew me to this, like, for instance. And I’m going to do it because I don’t actually particularly care, but I’m going to let me create something and I’m going to take some of my own proprietary content from my consulting firm around some things. So I will, oh, I just did this. And I’m just, I’m using this example. Does it sort of add some context to what you’re saying? So I just did a presentation to some folks at Georgia Tech. So technically, in theory, I’ve already given it away, but it’s a presentation around what is consulting and perspectives around. So this is technically my proprietary information around points of view around, and it’s not important. So it’s not that big a deal. But imagine if it was client proprietary information or, or something. The idea here is if I were to drop that content into this and then I were to maybe pick some other things from my vault of things, maybe there’s an interesting HBR article. So where else do I put stuff in here? Sources? Do I add more? So I’m going to add like a Harvard Business review article. So now I’m starting, when people use AI, you’re starting to now breach like a little bit of that ethics boundaries as to whether or not I really should be uploading a Harvard Business review article to a thing if I don’t know actually where it’s going to go. And then here’s an interesting article that I reference a lot about the autonomous enterprise of the future. So the idea of what I’m doing right now, real time in notebook lnzenhe, is if I’m understanding oop, I keep not doing that to make your point. I’ve just had, I’ve essentially given away, technically the rights to full control. I’m essentially having to trust Google that they’re not going to do anything with these and protect this information. To be able to have this set of sources, that’s going to give me something. So if I wanted to say, please prepare a TED talk on the sources. It’s essentially having to do all this stuff, but it owns that information. So Brian not owns, but they have access to the information. So Brian is essentially your vision for Smart Connections about. I already had that stuff on my own local environment. Keep it on my local environment. I know what I’m doing with it and I’m just applying everything else between what I just did, you know, all of the features that are in notebook lm, you’re looking to have everything, but in a saved in a safer space that you have more control over.
🌴 Brian
Maybe not everything. I think that is 100%. Like one of the beneficial side effects of, of the whole system. But I think it actually, you know, will go a step further to the point where eventually the, the amount of data necessary to maximize AI’s ability to help you as an individual, I think that that amount of data is actually going to cross a threshold where it no longer becomes possible, or at least practical for Google to provide these services, because right now it’s essentially zero cost. You’re not paying Google to use notebook Lm. I think that we’re going to hit a threshold where having the data locally, so there’s like, what you can imagine as like a pre processing layer. So, like, the AI does its processing, like the AI we’re all used to interacting with. But there could be this other layer which I imagine is going to look more like a reinforcement learning system than the large language models. That’s like a specific type of AI network. And a reinforcement learning system is separate. I think that it’s going to become increasingly difficult for a company to manage the necessary data all on servers because there just ends up being so much processing required that it just, like, it starts to make more sense to just have that happen locally instead of like, you know, like, like, basically what Google does is they put the information that’s not accessed a lot into what they would call, like, cold storage, storage that takes a while to read from. So, like, while in theory they could easily store, you know, tons and tons of data, but it then it becomes an issue of, like, how quickly can that data be accessed? And then now all of a sudden, like, if you’re relying on Google for that, you ask the AI a question and it literally could take minutes in order to process all that information. While if it’s on your local computer without even getting any fancy hardware, you have access to all of this RAM. And RAM allows you to access data at, like, I just happened to stumble across a diagram, like, in the past day or so, where it was like, basically millions of pieces of data can be accessed in the time it takes to access one piece of data. When you’re comparing hard disk storage versus ramp, so basically you’re taking advantage of having that. Easily accessible memory to augment and improve the AI. And, you know, while with the tools like that are currently available in Smart Connections and Smart Connect, you know, we’re just talking very rudimentary retrieval and some editing. While that doesn’t necessarily, like, hit this threshold, you know, by itself, you know, like, I’m really still working on the foundations of what I see as a lot more of like, intelligent types of processing being layered into this Smart Environment, you know, so I’ve mentioned reinforcement learning, and that’s like, that’s one of my main focuses is like, what is that like? Because that’ll be like one of the next layers that gets added in, you know, like, what does that look like? And that I think will, you know, it’ll probably exponentially increase the amount of processing even happening, like, in the Smart Environment, you know, and when it comes to having that all happening on servers, like, it’s. It’s much easier to exponentially increase your processing locally than it is when you have millions of people trying to access via the same servers. It’s just like, it’s a different calculation. And I think it’s one that even I’d be surprised if a company like Google doesn’t start offloading some of this processing in the future via local applications.
Matthew Oatts
Well, I know we took a bunch of side quests today with the lean coffee, but hopefully it was some sort of like useful engagement. I don’t know that we. I know a lot of times people will join the lean coffee hoping it become. Hoping it can be like a little bit of a troubleshooting forum. And we tend to not go that direction just because we want to talk more about the bigger picture concepts. And also it’s intentional from a community perspective that we do as much of the troubleshooting as we can out in the open. So we can even enlist people other than Brian to solve the problem, or we can all encounter solutions to things that we probably else other people are encountering. So we tend to sort of do that a little bit more where we don’t do a lot of the troubleshooting, these calls. So if we didn’t get to some of those things, we can. We have a pretty vibrant community that’ll help engage in that. Are there any other sort of questions or things we want to hit on?
Unknown
I’d just like to thank you for the meeting. It was interesting to learn a bit more in an area that’s still a bit diffused to me. So I came, as you mentioned, Matthew, to Smart Connections via Obsidian. So I’ve been listening and learning.
Matthew Oatts
So thank you very much.
🌴 Brian
Cool. Thanks for joining.
Matthew Oatts
Yeah, I know we might not have gotten to all your questions either, but.
Tap mab
Yeah, that’s so good. Just sort of, it’s good to sort of get a framework and just sort of pick your brains. Hearing Brian speak from his perspective and your perspective as well, Matthew, as well. Yeah, I think it’s. Yeah, I’m more than happy to join. I think I’m just going to have to read the forum as well, because I just saw it today. So I’ll probably just go in depth, having understanding the forum, but at least I know having an idea. I think the only thing I was kind of confused was probably just like, when the API keys activated, like, when it reads, when the model reads, is it reading like, does it read the whole, like, if I had like 10,000 notes, does it go over and read all the notes, or is it able to sort of pick the relevant notes? So that’s probably, I didn’t get an understanding of how it’s actually capturing those notes.
🌴 Brian
I can answer that real quick. Basically, the local processing that I was mentioning just a minute ago, that is what is going to find the relevant notes. So this way, all 10,000 of your notes are not being sent to the server, and instead it actually reduces that to maybe ten notes that it considers as most relevant.
Matthew Oatts
That’s a common question that a lot of people have, is like, oh my gosh, openapis. Is my bill just going to start going skyrocket? And it’s been designed to not have that big of impact on it. I mean, there is an impact. My understanding, Brian, is one of the drivers as to why you sort of wanted to create the green path was so that people could not necessarily use Openapi or your calls to interact with your notes. Because essentially, if you have a subscription, you get as many calls as you want for free until you get throttled. So that’s one of the big appeals why you start seeing people sometimes go this direction. Like, for instance, I’ve paid for this path and I have a ChatGPT pro subscription. I have not used Smart Connections Obsidian plugin nearly as much as I used to, because I interact more with it through ChatGPT, where I don’t have to worry about throttle. And again, I was, I used it extensively, and it wasn’t a high bill at all for, like, the calls, because it’s designed the way Brian described.
🌴 Brian
But anyway, yeah, that’s correct.
Tap mab
Okay. Yeah, because I think I put the two of them because I got the green path and have the blue one, but I think I’ve just been using interchangeably. But I think I’ll have a read and just go through the green one like what you’re saying.
Matthew Oatts
Cool. All right, sounds good. I’m going to put the recording details. The 6th person on this is my AI bot that I use to capture recordings. So you’re going to see in that little green box down below, you’re going to see a link to the recording and then links to the different types of parts of the conversation. So you’ll be able to quickly go back and look at different parts of this conversation. Here. Like, if you click on, you know, one of these kind of things, it’ll go to a little thing here, and it’ll talk about that with the transcript. So if you want to go back and review anything that we talked about, you’ll have a nice, easy way to see that on this mirror board, and we’ll post that information as well.
🌴 Brian
Thank you, Matt, for facilitating. I really appreciate it.
Matthew Oatts
Appreciate you guys. Take care.
Tap mab
Bye bye bye.
Offsel Offalseller
Thank you. Bye. 😊🌴