EVERY Market Needs Hyper-Personalization

Brian's whole don't die philosophy and blueprint philosophy is like, zero to, like, 5,000 really quick. So we need to meet people where they're at, like, zero.

I mean, I think the kind of philosophy we're talking about, it has a history of, like, there's been sort of like this institutionalized analytic philosophy.

I hope, like, this is my ideal society where there's this more explicit relationship in which individuals own their models, models of self.

Hi, guys. Welcome to the ABCs for building the Future podcast. I have today here my co founders, Jonathan and Dean. I just got back from a flight and I'm really tired, but this is going to be a great podcast because we're going to debrief a really big milestone in the project live, which is pretty cool. So it's on brand with the whole building in public. We also learned a lot because I just went to the Don't Die summit hosted by Brian Johnson. I went to the dinner and I went to the summit, saw a lot of partners there, talked to a lot of entrepreneurs there, learned a lot of, got a lot of data points for the types of problems we're solving, and we're going to debrief those, too, and actually brainstorm, use cases live on the podcast, which is something we love doing. And we're also going to talk through, if we have time, some of the architecture diagrams that explain epistemic me, because now we're getting solid interest. And so even just recording that bit is going to be great for us because we have people kind of asking about that. So with that said, I could start off by just explaining maybe the story of all this. I was thinking about this, Jonathan, when you and I were talking on the phone last night, debriefing, I was like, man, the story of just us meeting from Guadalupe, who was common community member that knew both of us in sidebar, and don't die and was like, hey, you're the only two people I know inside bar and do die. You should meet each other or whatever. And so we met and we totally hit it off. And then I had. We had just gone to the summit, the very first San Francisco summit. So this summit, hosted by Brian Johnson, ended up being the last one in California for a while, as far as I know. So I was like, all right, I gotta go. I, like, booked this thing. So timeline is we met in October and we decided, all right, let's like, really go a hundred percent really hard at this project, and. And here's some ideas to go to market. And we really Focused on, you know, hey, we just came out of this Don't Die summit. They seem really values aligned. They're thinking about this very philosophically. This seems like a huge win win. And there's been a lot of data points. We're getting learning about what Brian Johnson is saying about the philosophy and it evolving and him releasing the documentary and all these things. So we decided, okay, let's, let's go hard on this in November with this Don't Die use case. And then Robert's gonna fly in January to LA and try to, like, meet and make this, like, pitch happen and lock in this partnership or whatever. And so that's what we kind of planned. And I was reflecting on that and just the story and the arc of getting here where it was kind of a roller coaster. So I'll just go ahead and tell the story. We got these demos done with Brian's development team and then we were feeling really good about that day of I fly down to la, there's going to be this Don't Die dinner. And I had texted him because we had exchanged contact information and I've been trying to help grow the community from grassroots perspective and giving feedback to the app and now building this thing. And so I texted him, just making sure we could lock in time to remind him because he's a really busy guy, he's starting many companies. And so I didn't get a text back and I'm like, a little worried, all right, maybe this is not going to happen. And I was mentally prepared, like, maybe this just won't even happen, right, but at least I'll learn something. And then went to the dinner, saw him and we shook hands and he said he was excited to see me there again for the second dinner. And I asked him explicitly, hey, are we going to make sure to chat about the thing we shared your team? He said, yes, absolutely. And so then dinner went really late because a lot of people were asking really good questions. And so we just went really late, tried to catch him after. I was like, hey, Brian, do you got two minutes? And he's like, no, sorry, gotta sleep, obviously. And then. And then. So I leave the dinner, I debrief with Jonathan and I'm like, all right, man. Like, didn't go down. Oh. And I had asked him, I was like, all right, can we do it tomorrow? Can I catch you at the summit tomorrow and we do it tomorrow? He's like, yes. So I debriefed with Jonathan on my walk home to, like, the hotel Friday night. And I'M just like, all right, man. I don't even know if it's going to happen. But, like, if it does happen, I'll catch him tomorrow. And if it, you know, if it doesn't happen, then at least we're already in with his development team. And there's a lot of value that we bring to the table. And we're really values aligned and philosophically aligned to grow this. So then the summit day comes and it's super fun. Also, like, the summit was just a blast. Um, it was cool. There was like a morning rave and I saw people from the San Francisco summit and I saw people from the dinner the night before. That was super fun. And anyway, fast forward, lots of activities and stuff going on. I. I find out my bio age in physical activities. I get super fucking humbled. My bio age is like 40. I'm like 32. I'm like, Bro, what? Oh my God, I'm so out of shape. So I gotta. I got some work to do, guys. Anyway, so I. I'm like looking for an opening because the guy's just. He's just walking around and everybody's filming him all the time. There's like a camera crew on him right? There's there. It's hard to get it open. And I finally tracked down one of the people working with him and I was like, hey, Brian said, you know, he promised me a few minutes so we could walk through this thing. We were already talking to his dev team. Could you get me in? And she literally, this person was a hero. Like, she literally just heroed it and walked me straight up to him. I was like, brian, you got to talk to this guy. And then he talked to me. I had my laptop open, like on that walk from Ask. On that walk, I turned on my hotspot because there was no WI fi in the convention center. Unlocked my laptop out of my bag. And then I had the pitch deck, all Those things, the GitHub, all the Figma files, like for the help boost AI. All those things, like, just ready. I was like, oh, please, just load up. And then I catch them. Loads up. Perfect. I showed him the first five slides that were super hyper optimized. Isn't everything we learned. And he was like, yeah, we're super aligned. Your path forward is working with Ryan. And I was like, okay, great. And then I. Then I. What was really exciting on the are we technically aligned and product aligned on the direction of value was I mentioned how the don't die dinner was a series of dialectics. So in for the audience, Brian Johnson is this longevity influencer, and he's starting this whole don't die philosophy and culture around we should go to bed on time and we should be healthy. That's really his thinking. There's a lot more to it, and we'll certainly get into it, but that's what he does. And so the dinner that I'm referring to is when he brought a bunch of people together and we talked about things like why we should focus on this kind of foundational philosophy of we shouldn't die. And humanity is at a point of, there might be a lot of existential risk here with giving birth to super intelligence and AI, which is going to be smarter than our kids. So we should talk about that. So that was what the dinner was about. And so when I referenced this during the pitch, where I literally caught him and I had my laptop literally standing, people filming us, it was a whole. Like, I had to just pitch that in that environment. First five slides. And then he was like, we're aligned. And I'm like, brian, this is what we want to do. We want to take what happened last night. And that was a series of dialectics, and we can calculate somebody's don't die belief score coming out of that, and that will be the way that we could leverage this to scale it, right, to really understand how much is this philosophy growing. And he totally latched onto that and was like, yes, like, when I said that. And then when I showed him one of the slides that showed him the convergence of, like, philosophy, consciousness, et cetera in the pitch deck that I framed very purposely for that, I thought it really fit his mental model. He was just. I could just see it in his body language that he understood. And then he told me something really validating, too, which is he really appreciated from the bottom of his heart all of the work I put in to try to grow the community and show up and stuff, and even just showing up there. So that was validating for all the effort. So that was how the pitch went. And so basically, this is like, you know, nothing's like, inked or anything. We didn't even get to that stage. But we have a clear green light of, all right, we're trying to build this AI. Brian HEALTH COACH and so even during the conference, I got together a bunch of private beta users that would be interested. So now we got users. Now we could form a feedback loop, and now we could fucking run. And I'm just like, yes. You know, when we started this, Jonathan, like, when we. We Started talking about like what's the go to market use case. We started thinking, okay, maybe this could be it. Health is something that is really tied to people's beliefs, something that's very emotional. This is something we could help with. And this community seems like there's a huge need of people that want to be healthy, are longevity curious, but they don't know exactly where to start. And Brian's whole don't die philosophy and blueprint philosophy is like 0 to 5,000 really quick. So we need to meet people where they're at 0 to 1. And so it's cool to just see all this stuff coming together. And now we're putting ourselves in a position where we have our ideal icp. Not even just like the icp, it's like our ideal ICP and the health use case and longevity use case. And we have a short feedback loop to actually build this thing and be able to validate. Have we built the right thing to deliver hyper personalized value in the health and longevity use case? So that's a great position to be in. So that's the high level, really quick update of like the story where we're at. But the next milestone is all right, AI health coach Brian for many reasons. And yeah, so I think that also serves as. As kind of an update. I wrote this update out in, in Slack, but I think this is the first time we all three have gotten on the phone and actually have. Have talked about that. So reactions. How. How are you guys feeling?

Yeah, I love how excited you are. First of all, you know, it took about 10 minutes to get the story out of your head and I see like, you know, just a very excited version of Robert and it's just like really energizing.

So I'm pretty hyped, dude. I'm tired. But I'm also like, dude, it's so validating all the thinking and all your guys hard work. And I'm like, damn, everybody works so hard on this. And yeah, it's just like that Kobe mentality, you know, like the outcomes will be a function of how much effort you put in. We put in a shitload of effort. So it's nice to see and we could be in a position to help people. Like that's so crazy. Anyway, sorry, Dean, I get too excited. Sorry, guys.

Yeah, maybe. Maybe flipping the podcast narrator script a bit on you, Robert, which is like kind of unique for this podcast. What do you think actually resonated with Brian? Maybe you could speak to that. Yeah, I know you mentioned that he was Using a lot of terms that we, that we like to use when we're describing epistemic me, like epistemology and things like that. And yeah, just, just curious to know like what actually resonated either from a philosophical perspective and then also maybe from a product perspective. Yeah. And like where he fell along that spectrum.

Really, really good question. And so for the podcast, because I know the podcast is full of, you know, entrepreneurs, founders, other people trying to start things and we want this to be, you know, educational as well for you. That's part of us building in the open. We reflect on our lessons learned, share our wins, share the lows. There's going to be lows for sure. There's always going to be highs and lows. You're catching us at a high. So if we look at it from the lens of this was kind of like a B2B. Can we land here? You know, business to business, can we land here? We're operating as open source project, but it's essentially one entity with IP and stuff that they've built and we're trying to grow that and grow that with another entity. It's not like direct to consumer. Like they're a legal entity that would have their own developer that probably has like a system user API key that hooks into their system. Right. That's what I mean by B2B. So looking at it from that lens, right. I guess zooming out, there's multiple layers in terms of how I'm thinking about Brian's perspective on what is value to don't die. I think that's a really big question. And how is epistemic me overlapping in that value? That's a really, you know, like just in business terms, like it's like, why would it even make sense that I should take a call from you or give you three minutes, like whatever. And so what we proved to his development team was for the audience, we did a demo to his development team and we showed them, hey, what we can do is we built this really great processing engine as an umbrella term to make it simplistic for creating models of self quantifying belief systems, quantifying subjectivity in those and then getting structured data and being able to version that and time series. It's so we've done a lot of that heavy lifting and you can get hyper personalization as a result. Right. And that evolves as you get more data from your users and your users interact with you more. There are more touch points. So that's like the basis of everything we've Built. And we built it in a very let's integrate the fundamentals to philosophy kind of way, and let's implement predictive processing, which is a blend of multiple things. So then going back to the question, how does this overlap in value to Brian. So Brian, from all of his content and going to the previous dinner, we had this set of hypotheses that he's taking a philosophical approach to this movement. Yes, there's like the funny tagline of this is a cult of let's go to bed on time. But foundationally, what he's saying is there is a fundamental belief that we all have that we kind of ignore, which is, hey, I don't want to die right now. We are all playing that same game right now, which is, we all don't want to die in this moment. We all don't want our children to die in this moment. No organism on the planet wants to die in this moment. Biologically, we are fighting to exist in this moment. So there are all these games on top that we play, capitalism, et cetera. And that creates, divides. And so his thinking from a philosophical lens is this could be the type of thinking to really elevate, to go back to the fundamentals of we're all just trying to exist, to unify us in shared important values. Going into this next age of AI and humanity. To summarize his philosophy, that's kind of where he's at. So back to Dean's question during the dinner. It was really cool. They were about to set the scene. There was about 30 people or 25 people at, like, a spa, I think it was Remedy Spa in la. And we were all in a room. We had some, you know, Brian Johnson blueprint food served to us right before. So super healthy. We're all feeling really good. We're sitting there. His first question he asks is, if you had an algorithm that you had to follow and just execute whatever it's said to do, like exercise on time, sleep on time, go to bed on time, and you had to follow it to a T and you had to give up all your control, but you follow it. The outcome is you feel the best mentally, physically, spiritually ever in your life. Would you do it? And so half the room was like, yes, half the room was like, no. And that ended up being a philosophical question to kind of prime the brain on. Like, are you wanting certainty or are you okay with uncertainty? Essentially? Are you wanting to know or are you okay with not knowing? And the series of questions next, I won't go into the Details. But I'll up level again because I think it's important to highlight for the audience. Like, this is really like getting at the hypothesis of like, what is Brian's thinking on this move and what's our value? Right? So he used words like philosophy. He literally said, I see myself as a philosopher. And that's when I was like, okay, that is a huge data point that confirms our hypothesis that that's how we thought. And because there was this thinking we had which is like, well, okay, we want this open source project we're building to help the world. And you know, Brian has blueprint. It's kind of a centralized entity to make money. How does he think about don't die? Like, how decentralized? Does he really want that? Like, there were questions we had, right? And so when he said that he really thinks of it as a philosophy, that was really big. Another thing he said was he referenced that he has read a lot of biographies of different people and he named one of them Magellan and talked about how circumnavigating the globe at one point in the world was a belief that was like, crazy. And so his point is that right now, believing that going to bed on time, doing these crazy things that society tells us not to do because society incentivizes us to not go to bed on time and make it really hard to eat, right? Or whatever, that those thoughts seem crazy. But also, at one point in time, people didn't think that germs existed and they were just killing babies in hospitals because doctors weren't washing their hands. So he made all these different points about how the entire human race at certain points in time had to evolve the way we think about the world into, hey, we know a lot to actually, we didn't really know about the germ thing into now where it's like, okay, maybe we don't really know about the AI thing. And fundamentally, to answer your question, Dean, that's where I think the storytelling and positioning of the pitch was really good. Because we had those data points and epistemic knee is positioned really well. Because the question is like, okay, if don't die is a torch, that kind of lifts up this fog of what we know and don't know and it's a unifying philosophy, then what's the thing that scales that if we agree that this is a philosophy in which we should think about the world? Knowledge. Knowing at this point in time is not the most powerful position. It's accepting the fact that we don't know everything is the most powerful position. So philosophically, that's where Brian's at. You really confirmed that in the dinner. And that's philosophically. You know, as viewers know, we talk a lot about how do we know things, how do we know our parents are real, how do I know I'm not in video games? Like, those things. So we think about that a lot. And so the two philosophies that we have, like on our team and Brian's really align. And that's what that dinner told us. So when I pitched it as, hey, you know, epistemic me could be scaling the torch that lifts up this fok, it made it really easy, I think, for him to understand exactly what we're trying to do, especially being able to summarize it in terms of like. And we've had to test this language for a while now. Like me, Jonathan Dean, we talk about epistemic uni to other people. So we have to really refine our approach to how do we talk to people about this. We found effective ways to do it. But yeah, that positioning and then saying, hey, we're quantifying subjectivity, hey, we're modeling the self. We are modeling human belief systems and allowing versioning time series control of that. That evolves with you being able to position in that way through all of our learnings. The past few months, I think is really what solidified the thinking that went into the pitch. So that was a lot, maybe a lot. Way more background and thinking that you originally asked for. But hopefully that is a comprehensive answer of like, what did we learn from the dinner? How does that inform our positioning and more macro view? How does that align with the value of that don't die wants and the value that we bring to that community to ultimately help people go to bed on time and make people healthier? Like, I would like to go to bed on time. Like, shit, I'll use this AI health coach, because my sleeve, I got messed up from these travels, man. So, yeah. Any reactions to that? Guys?

I was wondering if you want to show that slide that you made from the Canva pitch to that. Yeah, the one about the philosophy and like, where we are today. I think it tells a story of like, conversations we've had that kind of like bubble up to, you know, maybe partner facing conversations.

Yeah. Wait, are you saying let's share it on the pod or let's share that this right now? Yeah. Oh, okay. So, yeah, this is the slide is, you know, how is don't die the unifying philosophy? And I just had it Positioned as this metaphysics, epistemology, axiology, the branches of philosophy, cultural consciousness, shifts pre modernism, modernism, postmodernism, ages of humanity, Stone Age, industrial age, Information Age. Obviously there's a lot of in betweens that I don't show here. But really just trying to say, hey, I really get you that this is the unifying philosophy for the next age of humanity and AI. And then that torch example I guess I could show don't die is the unifying philosophy. Brian holds the torch to lift the fog and then positioning that, okay, well that's a solution. Then the next problem becomes, well, how do you scale that solution? And that's really where we come in. And we do that really well with solving for a term that is resonating with people is hyper personalization. I've been using that term. I'm a little careful to think about what should the de facto term be used for talking about this personalization layer that we're essentially creating a category for, right? We zoom out from a macro lens. That's essentially what our hypothesis is. We are creating a new category. We think a category should exist that connects all these different systems together to hyper personalized experiences and value and decrease eliminate time to value for users for any vertical or use case. So just quick flyby, like everybody that I talk to about the problem at the Don't Die Summit and what our hypothesis is, we are creating a new category. We think a category should exist that connects all these different systems together to hyper personalize experiences and value and decrease eliminate time to value for users for any vertical or use case. So just quick flyby, like everybody that I talk to about the problem at the Don't Die Summit and I talked to a few different entrepreneurs, which I'll tell you guys about in a little bit. And we could debrief that and talk about their use cases in real estate and finance, mortgage lending, as well as gaming and education. But personalization is where everybody understands through every vertical. It's like, oh yeah, I want to personalize something. Hyper personalization adds another layer which is like, yeah, okay, personalization. And actually Dean, when I reviewed the video that we had made last time about walking through epistemic knee at a high level, like the technical direction, that Systems Design interview, when you mentioned that part about, hey, you know, if you're typically trying to personalize things today, you probably are putting in a hard coded rules engine with certain weights and stuff like that. This is a whole different way of looking at it and it will save you a lot of time as well as a development team. And you get way more value. That's really the value talking at that level that epistemic me brings. So hyper personalization, when I talk to somebody technical about it, they really grok that like, oh, I don't gotta solve that rules engine BS anymore. I don't have to go into there and figure that out. Great. And it's better. Cool. And so I think hyper personalization could be talked to that kind of person, which is like engineer, who is business savvy, product savvy, that also understands operational efficient inefficiencies in solving the problem. And so then I just positioned it like we take people's thoughts and, you know, their beliefs and we get quantified subjectivity. Something we can do in the future would be like, don't die belief scores and stuff like that to help grow the community. And then, you know, from all the learnings of the the dinner, I actually did change this slide. I didn't get to even present this slide, it was so fast. But I did change this slide to really emphasize this is the philosophical toolkit for building AI models of self. Anyway, I know that was a few more slides, but just to round out what actually was presented. Jonathan Dean, any thoughts on what I learned at the dinner, how it informed the pitch and this slide. You asked me to bring up, Jonathan, on the unifying philosophy thing.

Yeah, I thought it was cool. I think it was a few months ago, Dean and I have been talking about philosophy and then he had found this video that kind of synthesized the history of philosophy in these periods of metaphysics and epistemology and axiology. And just like it was resonant to our conversations and our background. And it was also like a really helpful simplification. And I think that made it through our conversations and onto something that's like a partner facing slide. So that was just cool synthesis. And we actually have a document in the repository which is a history of philosophy that kind of talks about the different problems during pre modernism, modernism, postmodernism, and goes into metamodernism, which is where predictive processing is trying to set and solve the problems that we have today. So I just thought this was a cool synthesis of our collective collaboration and just how it shows up in external facing conversations.

I thought this was a money slide for sure. When I put it together. I spent a lot of time thinking about these slides. I probably spent more time in my life on these slides than I have on any slides. But yeah, I, I do agree that that video, it's Funny how information works. But I think you guys mentioned it in a previous podcast and then I listened to it and I was like, oh, that's perfect. And then as I was synthesizing all our thinking for this deck that made its way into this. That, that is cool. Like the little origin story of that.

I think we should go leave a.

I think we should go leave a.

Comment on the guys. Your YouTube video.

We're totally going to do that. Yeah, yeah, I'm going to add it to the podcast show notes for sure. We'll add, we'll add the link to the YouTube video. That is a 30 minute video for anybody to kind of just high level understand the story of the evolution of philosophy or the story of philosophy. Really good. Like anybody that likes the topics that we're talking about would totally enjoy it. Really.

I was also saying we should go to his YouTube channel and leave a.

Comment and actually comment and.

Yeah, just explain that he inspired. Required like a lot of our. Yeah, totally, I'll do that and I'll.

I'll totally do that. Another thing I want to call out because I know we're sharing this slide right now. These are not the. I understand that these are not. This is incomplete. I think, I think that's what I'm gonna say that the ones selected here are not in terms of, hey, these are the most important philosophers. I'm sure you have your own, you know, very important philosophers or people that you, you respect or admire. I didn't include everybody. I just like look things up really fast, do them on the slide. So just caveat, because I know people can get, you know, maybe they might see this and think, oh, is this. Are these the only philosophers you think are important? No, that's certainly not it. There are many more. I just couldn't fit everything on a slide. And there are many more ages of humanity as well. Okay, I think one other thing to cover from the dinner that I slacked, you guys. So he said the word epistemology and I was like, damn it, I wish I recorded this because we say that word. I never hear it said in normal human conversation. Never really hear a word said. That's weird, right? Wait, like you guys do? No way. So anyway, he said it once and I forget the exact context, but it was similar to when he was talking about biographies and how he sees himself as a philosopher and he mentioned the word epistemology. I think it was in relation to like how we're at this stage of uncertainty and not knowing is actually more powerful than thinking. You know, and I just wish I recorded that. I wish I recorded the part where he said, I see myself as a philosopher. And he said epistemology, that would have gone down in, I don't know, like our GitHub as legendary. Like that was so cool. Yeah. Any other thoughts from, I guess that whole story of the pitch or anything to double click on? Happy to do that.

I really like that question that he provided or that he posed to the group about. Yeah, would you rather take 100%, I think it was clarity in your approach or would you kind of like want to randomize your approach?

Yeah, it was, specifically, it was. Would you want to give up all your autonomy and follow an algorithm guaranteed to be good for you? Like you only thing you know is this is going to give you the best health, spiritual health, physical health, mental health, emotional health of your life if you just do this. But you just have to listen to what it says. So.

Sorry, go ahead. I mean, yeah, I think unpacking that question is actually very interesting because I actually think that is the foundational question of epistemology. Right. Like how do you go about knowing things? Like what, what, what bounds of certainty do you place on things? Where, where do you place your authority? You know, if I were to say like the scientific method is the algorithm that I want to follow for approaching my life, you know, that's a pretty expansive idea, you know, and that can encompass a lot of different things. So I love, yeah, I love that question. I think it's very philosophically aligned with how we've structured our system.

What would you do? Dean and Jonathan, I'm curious if you answer the question.

Yeah, I tend to, I tend to approach the world a bit more non linearly than other folks. Like I always assume that I'm operating from a place of limited information about the world. So I'm always on the lookout for things that, or structures that I only see a partial representation of. And that if I were to try and explain cause effect in that domain, I might operate successfully on the information I had. But ultimately I'm missing the wrong abstraction. Right. And for me I see the way to kind of exit that bounded rationality is to try to find these like beautiful explanations that explain many different things at one time. That's usually my approach and that's kind of like non linear and it's, it requires like kind of peeking around a lot of corners and you know, it doesn't always follow a straight path, but that tends to be my, my, my approach is trying to find those Explanations that explain many things. And I find that once you've reached that point, a lot of clarity follows.

So.

Yeah, so the point of. The point is to try and gain clarity.

Yeah, so like for the AI algorithm question, like, the fundamental question is like, would you be okay giving up control then to like, something where this does the sense making discovery for you and optimizes for you?

I probably wouldn't, because I think I would. I would be skeptical of one just like following any arbitrary authority unless I fully understand its mechanisms for actually, like understanding the world. Right. If I don't understand how it's understanding the world, like it may be operating in incorrect on epistemologies I don't agree with. So I think. Yeah, I think, I think. No, I think, I think that's where I'm coming from. From, from that question.

Yeah, that makes sense. Okay, so if you got to see the underlying mechanisms, then would you say.

Yeah, okay, and if I could fully understand them?

Yeah, yeah, perhaps. Yeah, yeah, yeah. Okay, cool.

And, and, and if I agree to its underlying philosophy, of its underlying epistemology.

Yeah, yeah, yeah, that makes sense. That makes sense. I mean, basically, if, if everything was built out of epistemic me, like. Okay, cool, perhaps. Yeah, yeah, with, with many caveats.

But I'll let Jonathan go. I'm just for the viewers. I'm going to step away to go to the restroom, but Jonathan should go ahead.

Yeah, I think those are some good terms, like having explainable, you know, principles that kind of, you know, even if I don't understand the decision at the moment, that I can go back and understand, like the principles that were used to derive it, that's important and that's, I think one of the key goals of epistemic me is to sort of make explainable the human condition. And how are we looking at a model of self? Exactly. Why is this the best thing? And I think, yeah, it's. Bounded rationality is a good term here. This is Herb Simon's kind of approach to economics, is that we, as conscious beings, we only, we, you know, we treat people as rational actors, but they're only like you're, you're limited to like a set of decisions that your cognition, you know, your prefrontal cortex synthesis can sort of like manage in a given situation. Nobody's rational. And I don't, I also don't know that free will is really kind of like a helpful construct. We're all kind of like, you know, bounded by, by the, our, our, our societal constructs. The fact that, you know, there's so many things that we are bounded by in a given time that are implicit. And I think that's. That's, like, already true. So now you get to something that really can understand me better than me and tell me, like, this is. This is the action that you should take right now. And do I think that AI will get there? I absolutely do. And I think before this conference, I was writing the paper about, like, what I thought AGI should be, you know, as in what's the most general version of intelligence? That's actually sort of helpful because it's normally defined as. As like, human experts, that AI can do what human experts do. And I thought it was more important that AI evolved to understand not just what human experts can do, but sort of like all conscious experts can do or all conscious beings can do in the sense that it could, like, look inside of us and understand how the homeostasis of our tissue systems and the homeostasis of our cells. And if you imagine an AI that can understand you and all the layers of you and, like, you know, in this detail, in real time, that's much deeper than you can understand yourself, and then it can. Then it can come up with, like, you know, a principled, explainable recommendation based on all of that. Yeah, I think that's. I think that's the future that we're looking at. And if you're looking for a core principle as to, like. Like, what should that thing do? I think keeping yourself alive has always, you know, been kind of just like a teleological principle that, you know, it worked for all our ancestors. So let's keep doing that and then figure out from there.

Wait, so what would your answer be, Jonathan, on giving up control of the algorithm or not?

Well, I think. I think I have to reframe the question a little bit to answer it, but, you know, it's ultimately. Yes.

Brian would say you can't. Sorry, you don't get. I think that's what he did during the dinner. It was pretty funny.

But, yeah, I think. I think it's. Yes, but the explainability component is really important to me. And I'd say if we shouldn't develop algorithms that we don't know explainability, like, it should just be a boundary on how we regulate our technology evolution. So I would consider that, like, kind of a weapon. And, like, yeah, I would. I would think, like, if that kind of thing existed, I would have to sort of, like, behave according to it, but I would not sort of appreciate the direction that humanity was going at the same time.

Yeah, yeah, yeah, yeah, that makes sense. I mean, we're kind of amidst this whole TikTok ban too. That's kind of interesting to intersect with. Right. Like that. That really intersects with this conversation pretty deeply. What do you. What do you think, Dean? Sounds like you've heard of it. Maybe you have some thoughts.

Yeah, it's a. It's a. It's a tough problem because I think when I view it from the perspective of, you know, freedom of speech, the ability to access information. Yeah. We are essentially shutting off, like, a huge amount of the population's ability to speak and express themselves in a certain medium and in a way limiting freedom of speech in a pretty profound way, you know, and. Yeah, but at the same time, we also have these large national security concerns. So. Yeah, not. Not to relate everything back to epistemology, but I also think that's a philosophical question around. Right around, like, how do you. How do you explore. How do you. How do you approach this philosophy of I want to remain open to every possibility, I want to allow new information to expose and uncover the truth, but at the same time, how do I add restrictions to that because I'm operating from some teleological principle, as Jonathan put it, around the. The nation state and the need to preserve its national security. So that's a. That's an interesting philosophical problem. And frankly, I would probably leave it up to people who have experience in the policy domain. But, yeah, I can see the lever kind of being something that you pull on that spectrum.

Yeah, it is interesting. It is interesting. One thought I have, and I'll leave it to the experts for sure, I don't have, like, it's complicated. I can't answer this of, like, what's right or not. Like, that's. It's complicated on this. But yeah, I agree with everything Dean said. Another thought that popped up as Jonathan was talking is just how algorithms exist today, where we don't know why they are doing the. Their outcome. Right. Like just the Facebook feed or Instagram feed. Like, those are close source today. Right. We know their. Their objective, which is get. You just keep your eyes on the screen to get to the next ad or the next post and just keep getting you addicted. So, yeah, that's actually, you know, foundationally, I think us three coming from tech and kind of understanding some of those challenges for a user to be able to want to know what's under the hood, like, do you really know me? And with our own philosophy of, like, we should really be empowering the user to Know what is truth or what is knowledge or do sense making on their own. We should allow everybody to be their own scientist, not just like force feed them, spoon feed them things. We should be able to let you pop up what's under the hood and kind of see what's going on. So that actually maps directly into, yes, that's our purpose, that's what we want to do. And we think that's a values aligned way for us to go about solving that problem leveling back up. So that was, we spent 40 minutes on that debrief just to transition it and to keep it moving. That was the story on the pitch. That was really awesome. We are now, you know, now we actually can get users who would be interested to have a Brian AI health coach, which is really great. Real quick update. We had originally said in a previous podcast that we're going with this healthboost AI as a go to market use case. As startup life goes, that's no longer the case. We're killing that idea. But that idea evolved from the hypothesis that, that, hey, people are missing personalized health and longevity advice. People are missing personalized roadmaps of where to start to meet where a user's at in their health journey to get better and longer life. So the plan has kind of evolved to, well, we'll see where this goes with Brian. We're going to focus on AI. Brian, I think something else to talk about with his development team is do we want to build this as a platform? Like literally just give them the healthboost AI use case. It's like, this is a. We thought this through. I think it has some leverage. Do we want to go to market with this? And there's some pros and cons to either direction. So we got to figure that out. But I think we're primed to be in a position where we have shortened feedback loops with real users and real developers of those users, which is a great place to be to grow the value of our project. And once we knock down some more milestones and then I think we can get into a place of sustainability. But right now it's like kind of like startup mode. Fighting for our life for the next milestone is how I see it. But we got a big win. So another thing that happened at the Don't Die conference was I got to meet a lot of Don't Die people and I realized. So Brian asked this question in the audience, super funny side quest, but so he asked this, this question in the audience and he was like, how many people Here are founders. It was like 40% of the audience. It was during his two hour talk. And I was like, oh, this probably does draw the hyper optimization freaks of the world like me. The people that are like, I'm going to get every inch out of every second of every day. And so I was like, oh, that's so interesting. Of course I'm part of some ICP for this. That makes sense. And then another thing happened that I thought was really interesting to debrief. And we're going to talk about use cases. There's other stuff that I learned. One woman asked at the very end of his big two hour talk was and this was like the biggest Brian talk of the day. She was the last question. I think it was very memorable. She said, hey, number one question, when are you making a don't die dating app? Everybody in the audience died laughing. And she said she just went on a rant story about how she's such a go getter and she needs to find a man that can keep up, but she doesn't. She feels like we're in this age of the Internet where people don't approach each other in public anymore and she's getting trying to get after it with her own entrepreneurship dreams and she wants somebody that keeps up with her. But it's hard to find somebody that also lives this like, don't die, go to bed on time, all this like it's tough, right? Like that, that's that tension with social life. And so everybody, as she was saying this like rant story was in on it. They were like, oh my God, yes. Everybody was clapping, people were cheering. And so at the end, Brian cheekily was like, all right, I guess we got to build this thing. The reason why this is funny is because when we were talking about what would be the most valuable use cases in society right now, we have this apps channel in our slack where we just dump ideas as we think of them. We're like, well a dating app would be perfect. It's people's beliefs. You could figure out their values, you could figure out compatibility like versus like commoditized dating where people are our products just about, you know, surface level attraction. So that's another data point guys, that this would be valuable. Like people want this. Like a lot of people cheered for it. I think people really want that. Another data point was I talked to a real estate, real estate lending banking CEO. I think it's called E Mortgage Capital. I think they're the biggest mortgage lending based services in la. I believe I'LL have to double check that, but I'll throw a link in the notes. Met the CEO. Super good dude. And we got to talking and we talked about our use case and one of the biggest learnings I had is it's kind of easy to talk about our use case when you ground it in personalization. That's another big learning is every vertical has personalization problems. So we got to talking and I talked about what we're building and he's like, oh, I really want to get my business into the AI game. We're really services based, we're super not futuristic. And I'm looking for ways to get an edge. And I was like, okay, cross referencing that with what I know from Moors crossing the chasm technology, life cycle adoption. It's like, that's kind of like an icp, right Brian? And don't die. They're trying to do something brand new. They're looking for edges wherever they can to scale what they're doing. This guy's, I think they're at about 100 million or so in revenue and they're trying to grow that's, you know, kind of expanding their footprint in the banking sector. They're trying to grow. So how do they get an edge? So that that's giving us data points on like who is the icp. And personalization, I'm realizing is kind of a juicy problem to these. What's forming for me is, is companies that are trying to get an edge on competition, trying to do something new or expand markets or create new markets. That's the running hypothesis. And another one that resonated was I met this gentleman who was early at Riot Gaming, so. Oh, Riot Games I think is what it's called, so helped build League of Legends when it was early and has a portfolio of companies and is interested in improving the bar of education in Brazil in particular. That's where we were values aligned. And so he saw hyper personalization as a problem in that space. So with both of those teams kind of scheduled, locked in interviews or sorry, locked in demos to their teams of like what epistemic me could do. And I showed them our GitHub of like how we fractalize beliefs and I showed them what we did with the content that I had that you put through the system Dean and then got my beliefs model. I showed them that and basically kind of explained how our pipeline handles all of that and has these benefits. And so that had a light bulb of like, oh, it's not just words. Hyper personalization. Okay, you have some thoughts behind this. Oh, okay. It's grounded in a use case. Oh, cool. Like, I could see how that improves my operational efficiency and I could see how that could be a better way of doing. Doing things. So a bunch of use cases there that I got a lot of data on. Dating. There's a need for that in the market, as funny as it is. Number two, personal finance and real estate. Commercial or personal real estate lending. So the use case we kind of talked about is like, that's all about risk, right? How do you view financial health? What are your beliefs around financial health? And then education. How do I get personalized learning? Meet me where we're at. So happy to unpack any of that. Maybe we can. Let's maybe put dating last and then let's start with the personal finance, real estate one. What do you guys think about that use case? I'll just explore generically.

By the way, Dean, I'm getting a lot of feedback or background noise from you.

Oh, sorry about that.

Yeah, let me move. Yeah, for the viewers, we have like four other people in this house right now because we had a little party last night. So.

Struggling to find some.

Some places where I can talk about myself.

That's hilarious. So you just have people kind of waking up groggly.

Yeah, exactly. Yeah. Eating donuts. Maybe that's what you're getting. All right, no, but sorry, go ahead and let me know. I can just mute while others are talking.

Oh, that good, man. Jonathan, did you have something to add? So those, those use cases I thought were interesting to learn about.

So personalization of financial risk, huh?

Yeah, right, like. Like that makes sense. Right? Like you go to get a loan or something like that, or apply to a mortgage loan, and you have X amount of money in your bank account, 401k. Like maybe some debt. Like lots of. Lots of things to figure out. And then you have. I don't know how you grew up and how your parents viewed money and how much scarcity you're coming from or abundance you're coming from. I think that matters too. So I'm curious how you guys think about that at the intersection of epistemic.

Yeah, I mean, I think on the lender side, you know, they have these actuarial. Actuarial tables. And then they have like a model of you, you know, that is like very much fill in the blank. You know, like a banker's job a lot of times is just to like, fill in the form and then put it through the actual model, actuary model, and then say, yes or no? When I first became an entrepreneur I was like 23 years old. I wanted like I thought you could just go to a bank and get a loan, you know. And what I learned was that like they, they don't really want to hear your story. They don't want to hear like why you're doing what you're doing or how you're capable or how you're going to conquer the world. They're just like what's your cash flow this month? You know, what are your net receivables? What are your assets in the bank? You know, and they kind of like characterize you as a straw man that helps them streamline their lending process and there's not like any room in, in there for understanding your small business and like you know the nuances of the actual risk profile in the world. They rather it's just like kind of a strawman risk profile of the world. And it's one thing I met early on in my business. Just like, yeah, if you start something that's sort of like assets specific and you build up assets, then you can always borrow against it. Like real estate for instance. I think that's why real estate is such a popular business, is because it's a business lenders can understand and if it's a risk profile and I think subjectivity and lending profiles could probably open up more opportunities for more people and help characterize risk better for lenders.

Yeah, for sure. That's great that you drew in that. You know, first time 23 year old Jonathan entrepreneurship experience coming full circle. Maybe we'd solve this in the future. Yeah. Dean, did you have any thoughts to add to the personal finance use case? Lending use case?

Yeah, definitely. Lots of thoughts around the personal finance use case. I think the actual actuarial use case I find a little bit depressing in a way, you know, in terms of like, you know what the ways that that could be abused. But actually I, I actually do think there's a path forward there as well. I do actually do think that hyper personalization as a, as something that businesses demand has already been proved out in the market. You know, with Facebook and TikTok, like we have really good hyper personalization algorithms that are really good at figuring out what you care about. But the problem is it's centralized by a few entities which are the ones which like manage users data. So if you as a business want to make use of this personalization, you have to operate on the top of the rails of these existing systems.

Right.

And you have to hope that they are going to do distribution appropriately for you. I think that what's going to happen is that eventually that model is going to flip, right? And eventually people are going to realize that they're going to want to own that model of personalization for themselves or their models of self. And those models of self are going to want to give particular sets of information to, let's say this actuarial or this personal finance use case. So you kind of have this more explicit relationship that I hope, like this is my ideal society where there's this more explicit relationship in which individuals own their models of self and when they go and they sign up for a person personal finance software, they have the ability to share the subset of beliefs that they care about and restrict the set of information that then passes between the system. So there's a win win here in which finance products are able to explicitly define the set of beliefs or information that they might care about and then there's a transaction that takes place in terms of how the information is shared between the systems. And this goes back to this idea of not just how do you build a model of self, but how do you build a protocol in which models of self exchange information with each other. So I think there's a way to build a win win scenario in which these kind of companies aren't as reliant on the metas and the tiktoks of the world and users actually feel more secure in sharing their data. And I think that's the really interesting story to build here. Whether that's through an open source protocol or some company that comes and fills that gap.

Yeah, that's a. I think you're bang on with just iterating that unified vision for the future where hey, users have more power to understand and control and maintain what's going on under that personalization algorithm hood of complexity. I see a world where hopefully this is aspirational, but hopefully my 60, what is it? 76 to 8 year old mom could understand what's going on under the hood of these things. Like remember when the iPhone came out and it was only just cool kids and rich kids that had them and people were like oh, what's this thing? It's kind of cool though. And old people weren't using it yet and now it's like grandmas are using iPhones. Right? It's just the technology adoption life cycle and hopefully in the future, future we've gotten to the point, the use cases we hear now that I'm Talking about like B2B hyper personalization, you know, trying to get better, upsell better, matching whatever personalized value. Hopefully it evolves into someday it's like democratized. Cool. Like I'm, you know, Robert's mom and I totally understand what my model of self is. A lot of education, gaps in between. We, we'll get there famous.

I thought about the current way that say Facebook or TikTok understand us right now. I think the distinction of how they use data so well to grab our attention is maybe not a representation of how well they understand us because it's just like it's a race to the bottom of the brainstem, you know. And so like, yeah, they understand what material they can put in front of us to grab our attention so that our higher level thinking doesn't have to engage in the process. And it's like kind of an optimization algorithm that I don't know is going to help understand people more deeply, that type of personalization. And you know, I think if, if you're like really wanting to like use technology to nurture somebody's well being, you're not probably trying to grab their most valuable research resource of their attention so strongly. You're, you're trying to say like, how can I understand this person to a depth where I really understand, you know, like their, their cognitive like what they want to learn how they can better engage with other people on the planet and, and like you know, direct their attention towards those things. So you know, and I do think it's just like a whole different paradigm of, you know, like from the TikTok attention grabbing algorithm to like a holistic well being model of self that is, is being nurtured and like living, living a better life.

Yeah, you're totally right that the current algorithms are incentivized to optimize to steal our attention. I liked that quote you said. That's so quotable. It's a race at the bottom of the brainstem. Like dude, is that already a quote like that.

That's Tristan Harris, I think is the.

Oh, okay. Shit. I was like, bro, you just dropped a bomb. That was sick. We gotta, we gotta put that in the show notes. That's a great question quote. I love that. That's a good perspective, Jonathan. I, I'll frame this for the viewers because I know we're talking about like algorithms and, and stuff like that, but I think of it like a therapist. Is your therapist actually trying to get to know you and all your deepest darkest secrets and core fears and dreams to help you be a better person or Is your therapist trying to get to know you? All your deepest, darkest secrets and core fears and dreams to try to get you to pay them more money and keep coming back? Right. Like that's, that's really like we just personified or amorphized, you know, the closed source kind of algorithms versus, you know, what we're trying to do. Building the open. It's, you know, a user could see their own belief system philosophically. That's what we're designing the system from that lens. That's how I see it. What do you guys think of that analogy? It's super simplistic, but that's how I view the algorithms in it. In an analogy. Yeah.

I think he's a clean way to look at it. You know, that relationship with you and your therapist.

Sure, yeah. And I think like in, in the context of epistemic me, let's say we were trying to model a therapist. You could argue that those are two therapists operating on completely different epistemologies. One in which the therapist is trying to self preserve, the other one in which is trying to maximize information learning about an other entity. And so if we could, you know, build a system in which you could actually publish those or make them available or understand what epistemology is your therapist operating from. Now you can go out into the marketplace and select the ones that you think are best for you. Right. It's this, it's selection of process that I think is really cool. And if we can build a system that makes it available or evangelizes this idea, that you really want to pick the person whose process most aligns with your philosophy. Yeah, I think that's a really powerful concept. And it's kind of like, how do we make this idea of epistemology more consistent, consumable, more understandable to others, and more accessible so folks can actually pick the epistemologies that will improve their lives the most.

Such a big thing you said. And it goes back to something, you know, I've heard you say before, Dean, which is knowledge as a process or truth as a process is way more interesting than as like, this is just a thing. Like there is this thing as a process that's way more interesting. And our whole thing is like, we want to help everybody be their own scientist. Right. I think that's like a kind of sweet tagline of we help you sense, make or find knowledge or truth on your own. And that kind of goes back to, you know, how we started off the podcast, we talked about the Brian Johnson, don't die Dinner, his whole framing for those dinners in two hour, heavily facilitated session, he's done hundreds of these, right? So at this point, he knows the playbook. He knows what people are going to ask, he knows what people are going to say. And that was a series of dialectics. And so when I framed it to him, he was like, oh, yeah, we're totally on the same page. Yeah, that was a series of dialectics. So first of all, that was really cool. But going to this. What's interesting is his whole facilitation was to give people the tools to question what they know and what they really don't know or what. And that mental experiment that we went through at the dinner was just to expose you to even thinking about, hey, do I actually know anything? Like, what is knowing? And so the next step is like, all right, here's epistemologies. Here are steps in which you could figure out what, like, I suppose, implement knowledge as process. Yeah, I think, I think this all layers and aligns very nicely. So we talked about the personal finance use case. You guys want to touch on like, high level brainstorming for, you know, the gaming use case or the dating use case or the education use case. Since we're at an hour, we'll just pick one of those, maybe riff on that a little bit and then get into, I think, the architecture diagram. I think that would be good. What do you guys think?

Yeah, I think that sounds good.

Very cool.

I think it's like a. There's like a roadmap of, you know, like, what can we impact the most? Which is, you know, where. Where does subjectivity kind of come into play the most? And that's. Yeah, that's what I think about, like, I don't know. Of the cases that you ran into yesterday, did you get a sense of like, just what, what should be top priority or, or where can we impact the most given the problem at hand?

Yeah, I think for that, that's a great question. And as, as a product person, I always think about, like, where do we prioritize our time? Like, and there's impact in terms of value to society. There's impact to terms of like, can we get sponsorship dollars from this? Right. Like, impact, I think has. Has different heuristics under it from the most, like humanity, social good impact. I love the education use case and I also like the personal finance use case in particular of like, can we help? You know, because my parents were immigrants and they had trouble, you know, with finances, banks, like credit, credit scores and stuff. And so I do think that that could help a lot of people. Education is close to my heart. Dating I guess is technically close to my heart because I've been on dating apps. It's been a bad experience. But if I had to pick in terms of social good, I would rank it like education, then personal finance. This is after our number one is already health and longevity just be ultra clear. That's what we're super focused on. But if I think about okay well epistemic me could land there and then expand to other verticals for hyper personalization. What are other fast follow laggards or use cases? I'd probably put education first. More primed to be disrupted. In fact it's being disrupted as we speak. Personal finance, less primed, slower to be disrupted, although that is also slightly being disrupted as we speak. As we see more money flowing into Web3 and Bitcoin, even from institutions. But there's just more compliance challenges on that end versus there are clear market trends of billions and billions of dollars leaving private institutions like Stanford, Harvard, et cetera, traditional school education, higher education institutions flowing into personalized learning marketplaces, platforms like school, platforms like Coursera. So that's a clear trend similar to how there's a clear trend of money going into health, longevity and especially at the intersection of AI and technology to scale those industries out and solve really big problems and progress them. That would be my top. The other ones would just be like fun side quests that noodle on and whatever. It's something we always love. Like speculative speculating is just kind of fun thinking about the future. So with that said, that's kind of my breakdown of what would be most socially good. And then also where is the money going? And oh, I guess another market I should mention is enterprise productivity is another big market. When I analyze all the different markets where personalization could be a big thing, I know several people trying to build employee sentiment AI apps collect all unstructured data from your slacks, your teams, your outlooks, your whatever, create this model of a user and their sentiment score. How likely are they to turn over what's their risk score, stuff like that. I certainly think that we could provide value there. That makes a lot of sense right from modeling epistemic me as any piece of information you make as an enterprise user for a company doing knowledge work, for example, we could tie it to a model of self. We could probably get you better quantified subjectivity. We could probably get you intersecting that with some level of like health or well being data. We could probably get a good360 profile of an employee for you that. That. That management type person in HR would care about for, like, a business. So those are the industries I would. I'd think about. Jonathan. Cool.

Yeah, very interesting.

I.

The overarching conversation makes me think that kind of need to solidify a theory of change. You know, like, I think we've talked about maybe like, the three horizons approach. And then I think that would give us ways to say, like, this is. These are shared beliefs around the things that impact the most and what. In what order and, like, articulate a clean theory of change.

Yeah, yeah. Do you have any formative thoughts on that? I got a. I gotta think that through a little bit more, and we could riff on it on the fly right now, but if you have any formative thoughts to lay a seed, that'd be great.

I mean, I think the kind of philosophy we're talking about, it has a history of, like, there's been sort of like, this institutionalized analytic philosophy that, you know, I think was especially, you know, popular in the United Kingdom. And then there was continental philosophy. Continental philosophy was kind of a pejorative term to the French and German philosophers who are responding to, like, the holistic holes in the philosophical method. And I think when you look at something like metamodernism, it gives you, like, a lens on everything. It's like a theory of everything, and it kind of like deconstructs a lot of the current philosophical tools. I think if you look at a philosopher like Brian Johnson, that's what he's doing. He's saying, like, look, we have to look at everything through the lens of itself keeping itself alive and not dying. And like, that kind of like a primary. And I think that, you know, if you apply that to a culture of people, you know, like, I think Brian's kind of creating a religious vehicle for that where there's. There's theories and ideas of how everything lives and dies and then what the rituals should be on a daily basis to sort of, like, do the protocol, do the algorithm. This. It becomes a religion. And I think that. And I mean that in actually a really healthy way. I think like a religion of health as a theory of change is a really interesting way to look at where we are today and whether that's what sort of like, the moment needs for readiness and AI and AI alignment. Because thinking about AI alignment to explicit, like, quantify the level, you know, like the religions of health also gives you distinct, differentiated belief systems where matchmaking becomes kind of natural. Right? And it Sort of solves a set of problems and then like, it also bifurcates how, how research might be done to like, meet the, the needs of each, each religious system or each belief system. And then now you have like a, you have a reason to. And a way to evolve education. And, and there's like, you know, I think Gramsci is a good philosopher that talks about like, you know, the progression of how economic change gets made through, through cultural change, through education systems at, through the cultural elite taking on certain things and the populace sort of, you know, and like, I don't know, I'm sorry to, to dive, you know, to.

Go off on really, like, it sounds like there's layers of theories of change, right?

Yeah.

There's this, hey, like, what if, what if we do end up. Everybody unifies on the don't die philosophy. That's a running hypothesis right now, and we are actually players in that hypothesis, trying to make it more true. Yeah. But then at the same time, we operate in our own self of us three, and we have our own individual selves of like, oh, I gotta get married and have kids and I got financial stuff, like, and then, so the intersection of that, those goals and like, how does this project change? Like, what's that horizon look like and what does it map back to? What's the next three months look like? Six months, nine months, 12 months? If I had a riff on it. And actually before I do, I just want to be conscious because I'm. I'm talking a lot this podcast and I'm a little conscious about it. Dean, do you have any thoughts about what Jonathan brought up, theories of change, before I give my take?

Yeah, I think I'd need to get a bit more clarity on maybe what we mean by theories of change there, because I think that could mean a lot of different things. But going back to your earlier question about the, like, the use cases, not to jump into any particular use case, but I think one interesting way of viewing the multiple use cases problem is that each use case cares about a particular self and that any given self, for example, me, you, Jonathan, have a set of beliefs, and if you take some subset of those beliefs, you can almost treat it as a sub self that is pertinent to dating or a sub self that is pertinent to personal finance. So, like, each of these use cases can all be powered by the same model of self. Self that itself splits into smaller models of self particular to that use case. And it's like, how do you figure out those lines either if they're shared explicitly by the user through some interface or if they're figured out dynamically on the fly. I think that is interesting. And then as that relates to change, you have all these smaller subsystems changing as the whole change. Maybe to tie back to the theories of change idea. But to me, that's really interesting that this idea of, I think that relates to this idea of like impermanence of self as well. That there is no true self or that the self is like really everything and that depending how you draw the lines, you find a self that is pertinent to the given content.

Yeah, that, that makes a lot of sense. I, I'd love to actually double or triple click on this because I was thinking about this too on my travels home. I was thinking about how even for myself, when I look at, I have a little document that has all my identities. Like, I, I, I see myself as a rock climber, I see myself as a dog dad, I see myself as a leader, I see myself as a entrepreneur, I see myself as a brother, a son, a reader, a tea drinker, things like that. So these are all like identities or models of self, right? If you think about to ground the audience, I think of it like you have a video game avatar and there are maybe different profiles of it, or how much you invest in a particular trait, like charisma or intellect, intelligence or strength. Just like, how much do you invest in being the best son? How much do you invest in being good at tea drinking? What are heuristics like? There are different heuristics for all of them too. So I think you're right in that just looking at some of the data points of like, then the question becomes hyper personalization is that all of these systems, as you say that form the system of your overall self. They all interact with each other. It's all quite fluid. So then the question becomes, how do we solve that? Like, how do we actually combine these together? So I had a thought because I was looking at, for like day job stuff, I was looking at Salesforce's technical architecture for their app exchange. And one of their core pillars of their architecture is they have a metadata driven framework. And I'm a little familiar with metadata driven frameworks for my time at work day. So I had this thinking, I was like, I wonder if metadata might be a way to solve that problem of like connecting models of self together. The identities. When I say identities, I map them to models of self, like rock climber, son, partner, whatever. You know, your identity as a person negotiating a Car deal or a personal loan or whatever, your identity, exercise it, that kind of stuff. Those are all models of self. And so I started thinking like, oh, do we need to be thinking this early on about this? Like, because if we don't, do we miss an opportunity where we'll need some giant refactor at some point?

Point.

So this is, this is probably good to talk about. What do you guys think about? I like that use case. Just picking at it a little bit about how do we cut across all this to give you, like a holistic view across all your models itself, and how do those interact together? Jonathan or Dean, I'd love to hear from you on that.

I don't think it would require a huge refactor and maybe Jonathan can talk about that. I actually think the observation context is one of our first forays into that. I don't know if you see that as related, Jonathan, but I certainly do. This ability to kind of like, say, I want to pay attention, I see as like, observation context as almost like equivalent to attention. Like, I want to play, I want to place attention on this specific subset of information, and then I want to update the belief system and contextualize the belief system in that attention block that I'm placing on something. So, yeah, I'm curious if you agree with that framing, Jonathan.

Yeah, yeah, exactly. I think each. Each observation, the idea of the observation context, which is within our predictive processing epistemology, is meant to, like, bring your attention in a certain context, bring a certain set of beliefs that are relevant to that context into view, and that itself could be like a subset of yourself or a distinct version of yourself, depending on where you want to draw the lines. Because, yeah, the whole idea is that, like, there's a unique set of beliefs that are needed for this situation and that are relevant in the situation. One way I think about it is like, if you think about just like, you know, waking up every morning, going through your day and like, what. What version of yourself do you need to bring into, like, your morning routines and, you know, like, maybe making breakfast or, or doing exercise, having your first call the day. Right now you're switching out of the observation context of like, how am I going. Going to get nutrients to nourish myself to like, hey, I need to connect with this person and I need to bring up our common history and you have working beliefs sort of like, you know, ready, ready to go and to answer the questions that are, right, relevant. And even that switch, like, is a different version of yourself. So observation context is like you know, a way to. It's a, like a unique attention model. It's meant to kind of like embody a certain version of self that expressing itself in the moment of its environment.

This is a highly. This is like a very Buddhist idea. Like, like, like exactly what Jonathan just talked about. Like, if you go to like any like, leading meditation teacher, like, like, like they're going to say something akin to the same idea, right? That like, attention is this mechanism that creates objects in the world and instantiates selves particular to given context. And that oftentimes it's. It's seen through that illusion and learning to place attention on the larger self or seeing what the true self is. So from, from. From just like a nerding out perspective. I loved hearing Jonathan frame it that way.

I love that. Like, that's. Yeah, I love. I totally love that. And I always love how you bring your perspective and ties into Eastern, like meditation into it, because I'm a huge fan too. Also, quick side tangent is I started this daily meditation challenge, and I think I'm on the hook forever now. I started this daily meditation challenge on Don't Die, and I'm at like day 39 of meditating, 30 minutes of day. And so I'm like, okay, I'm going for it until I'm 60. Get the 10,000 hours, right? Like, I'm. And so I'm like, all right, got 30 days, 365 next. And yeah, so that's just the quick plug that. When you said the meditation thing, that reminded me. I didn't get to post my meditation yesterday, so I missed the day. But I got the streak still because the Don't Die app only allowed me to post like five posts in 24 hours. And I was trying to like, post stuff from the summit and I went to post my meditation thing and I was like, no, no, like, people will think this thing's broken. No, it'll be okay. It was just a catastrophe for like literally a half a second. But I thought it was funny. My share because now I'm like, people are holding me accountable to it. I'm putting it out there. Right. Okay. I think this is a solid point as any to transition to because we're talking about, you know, the model now. It might be good point to transition to our architecture diagram. Jonathan. Or maybe. Yeah, Jonathan, do you think you could bring up the Excalidraw we have from our midlife documentation? And I think what we could do is maybe even just start at that observation context and talk about that multiple models of self item and how that might flow in a given user's day. They get up, they have the morning routine. Let's say in the world we fast forward and we have sense making. We have in data devices that give us data through all that and we're not worried about the interfaces and somehow it makes it epistemic. Me for your morning routine, that might be a good way to show people how does the data move in our architecture.

Yeah, I think it's a good segue. Cool. I'll happy to share screen here. I have to leave and come back.

Yeah, actually I'll just cut the previous part and I think it will still flow away just. Just thinking about it. Okay, cool. So let's take a look at the architecture diagrams and walk through that. Multiple models of self slash identities, morning ritual use case. That's a fun one.

Sounds good. And you want me. Yeah, walk through the diagram and kind of explain it in the day in the life of a person managing their health. I'll start here with our documentation. And this kind of. This model of predictive processing, which is coming from the free energy principle by Carl Frischen. So this is actually from the textbook. And so this is yourself over here, this head, it contains beliefs, is what we're working with here. And there's a cycle of perception that's happening on an ongoing basis where we make predictions about the world, we act in the world to change. It collects evidence about what, what the state of the world is after taking that action and then compare the discrepancy with our prediction. And depending on what that discrepancy is, it updates our belief system. So the way I've modeled this in the predictive processing context of epistemic me is in this diagram which is also in our documents. But it's sort of like not. I don't know, there's a lot going on here, a lot to cover. So just let me know when you have any questions. I can't. I can't actually see you while I'm screen sharing.

Yeah, just walk through that morning ritual for your health use case. Just like. Yeah, I think that's a good one to ground the audience on.

Um, cool. Okay, so I'll talk in terms of like, you know, how I usually wake up in the morning and I think like. But one of the first things I do when I wake up is I measure my HRV with a heart rate strap that has a pretty accurate measure of hrv. And that's through the Morpheus training app. When I went to bed the previous day. I'm already thinking about what my HRV measurement is going to be that ends up coming into this predicted observation here. The idea is that the system is, is within my model of self is like calculated in ontology. That takes my beliefs in that context and sort of like calculates a probability. And so this, you might imagine that I've had a history of like day after day I have like, like gone to bed and I wake up and I, I say, well based on the sleep that I got and the stress that I'm under and the exercise I did yesterday and you know, what do I think my HRV number is going to be? And all that information is stored in these belief context table here where there's a history of like in this context of measurement in the morning, what were my predictions, what were the accurate actual. And there's a set of discrepancies that are stored here. And what those discrepancies give us over time is like a set of conditional probabilities that we can use to generate into this observation context. Context. So yeah, we take our beliefs in the context and that's in the observation context and we generate an ontology. So we have a prompt engineering based system that kind of takes all our history here and then creates an ontology. The shape of the ontology is like a tree of different belief context. And it says these are when you wake up in the morning, these are the situations you might be in given any of those situations here the probability distribution of what you might expect to see and so what this might look like when I'm waking up and it's the context of like, my observation is that I'm going to look at my phone after my HRV is measured. What should I expect to see there? The probability distribution is going to be like a set of HRV numbers. You know, like maybe my average is like 75. And you know, there's a probability distribution that predicts like based on the history of what you did yesterday, as in how much did you sleep, what kind of stress are you under, et cetera. Here's what the probability distribution would be and that would go into a predict observation. So I'm predicting the observation of what I'm phenomenologically going to see on my phone after I collect data by measuring and that's the action I'm taking. Action I'm taking is to like go get my phone, to put my heart strap on my chest to lay down and like for two and a half minutes and Just kind of like, I kind of borderline meditate during that period. Just like really try to, you know, listen to my breath and make sure that I'm not moving and be consciously aware of like everything, you know, where my body is. And this, this creates a, we're labeling, like a source and a resource. And the idea is that the data coming back from my phone as I take this action kind of gives me a way to observe the information. And then I interpret that data through the same observation lens. Like, maybe there's all kinds of data on my phone, but I'm actually just going to be looking at the HRV score and my resting heart rate as it comes back. And so then that's an observation. I compare my predicted observation with the actual observation and that gives me a discrepancy. That's one example that happens for me every morning is I just compare my prediction with my phone as I take this action, gives me a way with my observation. And that's how the model handles this, this information, observing the information. And it. Then I interpret that data through the same observation lens. Like maybe there's all kinds of data on my phone, but I'm actually just going to be looking at like the HRV score and my resting heart rate, you know, as it comes in the different kind of, you know, aspects of the, you know, what we call the, what we call these things in our data models that lets us store the information in this way. Does that make sense?

Can you talk a little bit about the probability score? I think that would be interesting. I feel like that might be somewhere where somebody would have a question. Yeah, the probability score and how you calculate that. Just anything around that would be good.

Yeah, Well, I think when you have a predicted observation, what it's doing is there might be hundreds of observation contexts that it could pull from. So what we're trying to do is say like this perception cycle is in the specific observation context where we're going to look at hrv. Or maybe it's like I'm going to look at HRV and resting heart rate, but those would be their own unique observation context. And then the probability distribution is what are the expected values that you could see. And so this is probably a familiar thing that most people who do any kind of machine learning, you just like you have a probability distribution and it tell you like the most likely and least likely value back. And so then that's an observation. I compare my predicted observation with the actual observation and that gives me a discrepancy. And so that's one example, you know like that happens for me every morning is I just like, you know, compare my prediction with my observation and that's how the model handles this information that you're going to observe. And so then we mentioned in the different kind of aspects of what we call these things in our data models that lets us store the information in this way. When you get back an observation this is actually captured as a state distribution which is also a probability distribution. And so these two numbers go into the discrepancy and we use a calculation called the KL divergence. So you have two inputs which are probability distributions and the KL divergence is kind of like an entropy measurement against each of the probability distributions. Does that make sense?

Can you talk a little bit about the probability score? I think that would be interesting. I feel like that might be somewhere might where somebody would have a question.

Yeah, the, the probability score and like how you kind of produces point wise KL terms.

Calculate that just anything around that would be good.

And those get, those get stored in our belief context as like discrepancies. So if, if I'm 100% accurate, you know it the, there's not going to be much discrepancy if I, you know, am, I'm off by like, you know, maybe within a standard deviation or something like that. Less discrepancy if I'm like way off. And that would for me be like if, you know. Yeah. Well I think when you have, when you have a predicted observation what it's doing is there might be hundreds of observations. Maybe, maybe I took yesterday easy and I didn't think I had any straddition context that I could pull from. So what we're trying to do is say like this perception cycle is in this specific observation context where we're going to look at like hrv. Or maybe it's like us is in my life. And then I measured my HRV and it was like really low. There would be, that'd be a large discrepancy and that relates to some of the epistemic emotions. Um, like surprise. I would be surprised you know if, if my HRV was, was much lower than I had predicted it to be. Um, and we can calculate and look at HRV and resting heart rate but those would be their own unique observation context. And then these epistemic emotions from the discrepancy of the scale divergent probability distribution is like what are the expected values that you could see? And so this and the epistemic emotions are things like this is Probably a familiar thing to most people who do any kind of machine learning. You just belief I predicted this and it was true. So I confirmed it. I predicted this and you have a probability distribution and it would tell you the most likely and least likely valued it was different than what I thought it was. And that would be a surprise. There could be a confusion which is like I predicted there'd be going to be a set of values that I was going to see here. And like, you know, I didn't even get back any, anything in my data set that I predicted as a possible value. Right. And that, that kind of KL convergence is copies that you're going to observe. And so then when you look at it as a confusion and so these, this kind of, when you get back an observation, this is actually captured as a state distribution which, which is also a probability distribution. And so these, these two numbers like show where the learning system's at in any given point in time. You know, is just how accurately does my ontology predict the world? And you know, and then based on the discrepancy, how can I sort of characterize these epistemic emotions? To understand what is the next question I can ask next of this in this particular observation context that will help me like learn.

So, so a surprise.

Say your HRV is go into the discrepancy and we use a calculation called the KL divergence. So you have two inputs which are probability distributions like 5 or 10 out.

Of what you thought it would be run through.

And the KL divergence is kind of like an entropy measurement against each of the probability distributions and it produces point wise KL terms.

Um, this what you just said about what you're surprised would be like and how the data would flow the calculations.

Yeah, I mean I think you have different, you have different variables that can sort of like be put in the observation context. And so you know, like a, a simple like variable would be like how much sleep do I get? And what is my hrv? And then, then I might introduce variables like how much sleep did I get, what kind of exercise? And those get, those get stored in our belief context as like discrepancies. So I do, and what is my hrv? How much sleep did I get? How stressed I am, I about work? And what is my hrv? If, if I'm a hundred percent accurate, you know, it the, there's not gonna be much discrepancy if I, you know, am, I'm off by like, you know, maybe within the standard deviation or something like that. Less discrepancy if I'm like way off and on and on. So you could probably come up with dozens of factors and understand what are the correlatory variables, the confounding variables, and what really matters and that we can kind of cast all those into their own unique observation context. And if I'm getting, I sort of want to be surprised because, you know, like, that means that there's something in the system. And that would for me be like, if, you know, that I can control the learning around, like, that my prediction is accurate, but I can take more, I can collect more data in that same observation context and sort of like reduce the amount of surprise. Maybe. Maybe. I took yesterday easy and I didn't think I had any stresses in my life. And then I measured my HRV and it was like really low. That would be a large discrepancy. And that relates to some of the epistemic emotions. So surprise is like a healthy epistemic emotion as opposed to confusion. And so surprise might invoke curiosity and saying, like, hey, there's a gap in our knowledge here. Do we need to start collecting more variables about this observation context or do we just need to collect more data? And either one of those things are tools that you can use to kind of like reduce the amount of surprise.

Could you tell me about the confusion aspect? Like, you touched on something I think maybe people would maybe want to hear a second time. Like, why would it be, quote, good for the user to have surprise? Oh, okay, I get to have reinforced learning loops and.

Like surprise. I would be surprised, you know, if my HRV was, was much lower than I had predicted it to be. And we can calculate these epistemic emotions from discrepancy of the chaotic divergence.

How does that compare with confusion in the system?

And the epistemic emotions are things like a confirmed. Yeah, like, so surprise versus confusion. And then how do I update my beliefs? And you probably just have a felt sense of like when you are confused, it is belief. Like, I predicted this and it was true. So I confirmed it. I predicted this and it was different than what I thought it was. And that that would be a surprise. There could be a confusion, which is like I predicted there'd be going to be a set of values that I was going to see here. And like, you know, it. I didn't even get back any, anything in my data set that I stressful. Right. And that's because, like, we have this kind of belief system update down here. And when you're confused that means you might have to scrap a set of existing beliefs that you have in observation context and completely rehypothesize what is going on right now. Whereas when you're surprised, you just might need to add some more variables to the observation context to kind of make it more accurate. And those are two inherently. Being confused requires a lot more resources to kind of like recalibrate yourself to your sense making environment versus being surprised, which, which is like, you know, maybe an enjoyable feeling of learning. And I think it predicted as a possible value. Right. And that that kind of KL convergence is calculated as a, as a confusion. And so these, this kind of comes down to like how heavy is the lift here? Like now that ended up.

And that's what the epistemic emotion like.

Show where the learning system's at in any given point in time, you know, is just how accurately does my ontology predict the world? And you know, and then gives us.

Is that surprise confusion contextualization for the update belief system processing.

Yeah, exactly. So it's meant to update the way to learn in terms of like how the belief system might be updated, how, how we kind of like test new.

Hypotheses and then the observation context concept. That's the thing. I'm thinking about it like a lens. Wonder if you can confirm that's a good way to look at it. Is, hey, the lens in which I'm asking myself about my HRV is through the lens of health. And then you move about your day and you go to the next ritual.

Based on the discrepancy. How can I sort of characterize these epistemic emotions to understand what is the next question I can ask next of, of this in this particular observation context that will help me like learn to train a surprise.

Let's say your HRV is like 5 or 10 out of what you thought it would be run through. After you do your two and a half minutes of mindfulness of HRV tracking, then what do you do next, Jonathan? And then how does that change with the observation context in our model?

Yeah, sorry, can you repeat the question?

Yeah, so just try to demonstrate the power of the observation context as being a lens on a particular belief. And so you're shifting observation context now of like, all right, got my measurement done. What's the next thing in your morning ritual and how does that affect the observation context?

Yeah, totally. Yeah, you're totally grasping the concept here because observation context is highly phenomenological.

This what you just said about what your surprise would be like and how the Data would flow the calculations.

Yeah, I mean, I think you have different, you have different variables that can sort of like, be put in the observation context. And so, you know, like a, like it comes back to your senses and your attention and like, what objects in the world are you going to pay attention to? And like, you know, sort of generate or, or, or understand the causal relationships of, in a given time. So, so it relates directly to attention. What am I looking at? What are my senses? And so then like a simple like variable would be like, how much sleep do I get? And what is my hrv? And then, then I might introduce variables like how much sleep did I get? What kind of exercise did I do? And what is my hrv? How much sleep did I get, how stressed I am, I about work and what is my hrv? And, and on and on. So you could probably come up with like dozens of factors and understand, like, what are the, the in my morning, you know, like after I measure my hrv, I, I'll get up, you know, I usually use the restroom at that point. Um, I don't need to go in that observation context, but then I'll come out and have my like, morning shake and my pills, my supplements, you know, and at that point in time I, I'm paying attention to. Cause like, you know, I maybe take like, like 60 or 70 supplements a day. And so I'm, I'm really thinking about like, what am I putting in my body right now? And like, you know, I'm making sure that like, I'm sort of setting up all my, all any of the other supplements I need for the day. I have a morning shake, which is meant to like, specifically give me energy. Correlatory variable variables, the compounding variables and what really like matters and that, you know, we can kind of cast all those into their own unique observation context. And if, if I'm getting, I, I sort of like, I sort of want to be surprised because you know, like, that means that there's something in the system but not, not trigger digestion. Digestion, digestion. And so there's, you know, like, there's a number of different observation contexts I'm carrying into like my morning routine as I like start to put things in my body as in breakfast and you know, and again, like, what am I going to measure there? Like, I might be measuring, like, how am I qualitatively feeling? Am I ready for the day? Am I feeling groggy? Am I waking up, you know, and am I sort of like ready to meet my first, like Email or Slack message or, you know, open my phone into apps and in all those things like, you know, I'm carrying like that I can control the learning around like that my prediction is accurate, but I can take more, I can collect more data in that same observation context and sort of like reduce the amount of surprise. So surprise is like a health, a healthy epistemic emotion as opposed to, as opposed to confusion. And so surprise might invoke curiosity, a specific belief system that's helping me cope and like understand exactly, you know, how I'm, how I'm moving my hands and arms and like, what I'm putting into my body and that. And obviously there's a lot of like, habits built in there. Right. Things that I might not need direct attention for, but that I, I'm. There's a trigger cue and routine that is that, that's kind of like modifying.

The one, one way to ground the audience. If so, observation context helps us with our model for epistemic me. You know, bridging into the quantifying subjectivity part, which, that's the part that's really complex.

But the observation and saying like, hey, there's a gap in our knowledge here. Do we need to start collecting more variables about this observation context or do we just need to collect more data? And either one of those things are tools that you can use to kind of like reduce the amount of surprise.

Could you tell me about the confusion aspect? Like you touched on something I think maybe people would maybe want to hear a second time. Context gives us that another lens on which to look at this dialectic or new information coming in. Update belief systems. And so shifting from the observation context of okay, cool, I'm measuring my heart rate variability. That helps me understand my recovery rate, understands how much more strain that I need to put on my body today or things like that. And like, why would it be, quote, good for the user to have surprise? Oh, okay, I get to have reinforced learning loops. And how does that compare with confusion in the system or how I slept. And now you're moving into, okay, I'm, I'm taking supplements.

Yeah. Like, so surprise versus confusion. And then how do I update my.

Belief systems around these? And even one thing I think about is you've already put in some of the mental load to determine these systems. So your observation context for taking these supplements is going to be different than when you first bootstrap these systems of taking these 70 or 60 pills. So even if we look, thought experiment, if we did have epistemic me power, Jonathan Model of self in the past before you did the 60 pill, 70 pill thing and then now we have.

That version noose and you probably just have a felt sense of like when you are confused it is now what.

Would we be able to do with those two? That that would be my next question. Like and just open question and maybe I invite Dean to stressful.

Right. And that's because like we have this kind of belief system update down here and like when, when you're confused that means you might have to scrap a set of existing beliefs that you have an observation context and like completely rehypothesize.

Explore here if you wanted to. I think that might be interesting to get your perspective on it.

Didn't fully understand the. The question but Jonathan, I think yeah, so.

Oh yeah Jonathan, if, if you want to go on that too you could ask it again.

Sorry. So what is going on right now? Whereas when you're surprised you just might need to add some more variables to the observation context to kind of like make it more accurate.

And yeah we have model of self Jonathan right now of like okay cool, I'm going to be taking my pills.

Oh the learning versus like.

Yeah, yeah.

Habits.

Right.

Okay thanks. Yeah yeah. And I think that's a great example. Right. Like say you're testing a new. Those are two inherently like you know, like being confused requires a lot more resources to kind of like recalibrate yourself to your sense making environment versus being surprised which a set of supplements. Right. Like you're paying a lot of attention to what dosages you're paying attention to. Like am I getting the outcome that I'm expecting from this? And then you know, once you've habitualize something your your attention changes. Like you don't need as much to collect as much data on on that and you can kind of just habitualize it. And so I think that's where like you know I kind of underline under which is like you know maybe an enjoyable feeling of learning. And I think it comes down to like how heavy is a lift here like now that I'm.

And that's what the epistemic emotion understanding these.

These layers of habits that we have. You know. And I really like the way James.

Clear gives us is that surprise confusion contextualization for the update belief frames, habits.

And how each like little micro habit is just like part of your identity. Right. So you're kind of like learning to build identity and then once the identity.

Is kind of system processing.

Yeah exactly like there you don't need to actively Put attention. You don't need an epistemology, you know, because. But. But then your identity becomes a context for the new things that you're going to learn next. And then you have a triggering bill.

That'S pretty cool to wrap it up into, you know, close the loop on. All right. In the typical health journey I have to take on new habits that's going to be different than my model of self today.

And so it's meant to update the way to learn in terms of like how the belief system might be updated, how we kind of like test new.

Hypotheses and then the observation context concept. That's the thing. I'm thinking about it like a lens, epistemic me. What we're really trying to do help you in that journey. Like we built the tools to be able to express an AI health coach with a certain philosophy, have it be interchangeable, have it evolve. So it's nice to close the loop on kind of that you can confirm that's a good way to look at it is. Hey, the lens in which I'm asking myself about my HR storyline and the data model. I wonder, I wonder if I could.

Jump in for a second there, please.

Go ahead, Dean. Yeah, please.

Yeah. I think that idea that Jonathan brings up is also very analogous to the concept of explore, exploit, if folks have heard about that. Basically it's the idea of how do you operate in an environment where you.

Have B is through the lens of health and then you move about your day and you go to the next ritual after you do your two and a half minutes of mindfulness of hrv.

You know, low certainty. So this actually applies to like our every tracking.

Then, then what do you do next, Jonathan? And then how does that change with the observation context in real life as well? Right.

So imagine I'm in a new city and I'm trying to figure out our model. Where's the best place to eat? My best strategy in this given moment is to explore a lot of places. But after a while, once I've explored enough places, I can start to exploit the information that I've learned and just agree on a few places and regularly.

Eat really good food.

So this idea can like similarly apply to our beliefs or our way.

Yeah, sorry, can you repeat the question?

Yes. So just trying to demonstrate the power of the observation context or our habits.

Right. We form new beliefs, more exploratory beliefs to try and gain more information, information about environment. And once we've established the right amount of information, we can just expect as.

Being a Lens on a particular belief. And so you're shifting observation context now of like, all right, got my measurement done. What's the next thing in your morning ritual?

What them. And reduce the amount of learning because it's not particularly useful to the task at hand. Yeah, and that's like a. I think it's a really powerful concept and it shows up in many different ways.

And how does that affect the observation context? Yeah.

And I think it's a. It's a great analogy to like how reinforcement learning thinks about the same problem.

Really great perspective in addition to the conversation. Dean. I think that's a great way to frame it. Jonathan, did you have anything to add as commentary to what Dean said?

Yeah, it's really additive language. Yeah.

Cool.

I think it's a good way to look at it. Maybe, maybe in a learning system, you know. Totally. Yeah, you're totally grasping the concept here because observation context is. Is highly phenomenological. Like it comes back to your senses and your attention and like what objects in the world are you going to pay attention to and like, you know, sort of generate or, or, or can. If you can think about measuring that threshold between Explore and Exploit, that's. That's where those epistemic emotions kind of come into play.

Yeah. In. Sorry, go ahead.

And like, I just like we. And we talk about asking the next best question. Right. And so you know where you are on that threshold, then like, you know, should I ask or Exploit question? That might be one way to conceptually understand the causal relationships of in a given time. So. So it relates how that works.

Yeah, I think as we start to socialize these ideas with like the larger like research community, like that. That parameter, for example, that's used in the Explore Exploit framework is often called the epsilon parameter. It's really cool. I think we can.

Could start to directly to attention. What am I looking at? What are my senses? And so then like map some of.

That language and help to like bridge the gap and get more folks involved in the language.

In my morning, you know, like after I measure my hrv, I. I'll get up. You know, I usually use the restroom. At that point. I don't need to go in that observation context, but then I'll come out.

That'd be really cool. I actually was thinking once we get this to a point where we. We benchmarked it against that study because actually that's what I learned from this entrepreneur I met. He's kind of curious about different and.

Have my like morning shake and my pills, my Supplements, you know, if there.

Are different companies solving for hyper personalization and how they might benchmark, makes sense, right? Because at the end of the day it gets to that level. And so I think that'll be a next step class follow after we figure that out and then we could start bringing this to the research community and science community, maybe get Michael Levitt.

And at that point in time I, I'm paying attention because like, you know, I maybe take like, like 60 or 70 supplements a day. And so I'm, I'm really thinking about like, what am I putting to my body right now? And like, you know, I'm making sure.

That like on the podcast that'd be like a dream goal. That'd be so cool. So, yeah, certainly in the future. Okay, so this was like a really quick double click, triple click into the use case of multiple models of self for the first couple parts of like a morning ritual using epistemic me. So if you're an application developer who's thinking about implementing an AI health coach.

Or I'm sort of setting up all, any of the other supplements I need.

For the day, trying to understand like how epistemic me you could leverage it for hyper personalization for your app, regardless of its health, regardless of whatever. We talked about education, we talked about.

I have a morning shake, which is meant to like, specifically give me energy, but not, not trigger digestion. Digestion, digestion. And so there's, you know, like, there's a number of different observation contexts I'm carrying into like my morning routine as I like dating.

We talked about personal finance. Hyper personalization cuts through all of these. So one thing I'll say is like, you know, look at our docs, look at our GitHub. If you want to be a contributor, be a contributor. If you have feature requests and stuff like that. We have documentation on how to contribute and how to reach out to us. So one other thing that I'll just say is based on where we're at now, my synthesis of what should epistemic me's immediate roadmap be from a product perspective.

Start to put things in my body, as in breakfast and you know, and again, like, what am I going to measure there? Like I might be measuring, like, how am I qualitatively feeling? Am I ready for the day? Am I feeling groggy? Am I waking up, you know, and I am I sort of like ready to meet my first like effective is AI Brian.

Cool details underneath. We got to get into sub cycles with this dev team. There's Got to be some discovery of like what are the features. You really got to get some clear analytics ones, some low hanging fruit ones like okay, well we just need a way for just constantly updating Brian's stuff from his content that he creates. There's clearly a little loop there we got to establish but we got to get like a PRD on that. So that's the clear next step. And from a strategy or organizational strategy perspective I think it's about time we like incorpor incorporate actually officially because it's like getting pretty official and I want to run this if we vision this as a global open source project.

So there are many email or Slack message or open my phone into apps and in all those things I'm carrying a specific belief system that's helping me cope and understand exactly how I'm moving my hands and arms and what I'm putting into my body options.

And I've been thinking through how we might form so we could talk about that as well offline and then on the pod we could talk about you know, kind of our decision making process later. But kind of examples I'm thinking of are like Hashicorp, Linux, React and there's even ways where you could just offload the administrative and funds and stuff like that to. There's a little app that I found that does exactly that. Some open source projects just don't even worry about governance and structure and VIP.

So And obviously there's a lot of like habits built in there right. Things that I might not need direct attention for but that I I'm. There's a trigger cue and routine that is that that's kind of like modifying the observation context.

I mean one one way to ground the audience if so observation context helps us have to talk through that but high level what I just wanted to get your guys's thoughts on and you know, flavor of just being building in the open. My thinking is we hyper focus on the health use case. Like I don't want us super distracted. It's fun to think about other ones because it also informs like all right, are we building the architecture in the right way to incorporate universal concepts so once we land in a vertical could we expand very easily? I think that's always smart but I don't want us super distracted because we have a clear tangible way to go to market and help with our model for epistemic me bridging into the quantifying subjectivity part which that's the part that's really complex but the observation context gives us another Lens on which to look at this dialectic or new information coming in. Update belief systems. And so shifting from the bat community, do science at scale, which is essentially Brian's goal. And that furthers epistemic me's purpose and value alignment too. My thinking and I want to propose this to you guys is we form an advisory council of sponsors who fit a mold of businesses or organizations that want to like pay for a seat to you know, maybe every six weeks, 12 weeks they just get to look at in depth what are we doing in our space. And hey, like what are your use cases? Map this back to your stuff and we learn and that way the observation context of okay cool, I'm measuring my heart rate variability. That helps me understand my recovery rate, understands how much more strain that I need to put on my body today or things like that or how I slept. And now you're moving into okay, I'm taking supplements. I have systems around these and even you know one thing I think about, they have a little bit of, of skin in the game. The reason why I bring this up is because when I met with CEO of E Mortgage Capital, yeah I think it was emortion capital. He was just talking about, like I said earlier in the podcast, he really wants to just get his team thinking about AI. He wants his brand to be more associated with like disruptive technology. And so there's you know, same thing when I talked to the guy that had education in his portfolio. And so I realized okay, this definitely cuts across. I think a good structure for us might be advisory council of like you don't is you've already put in some of the mental load to like determine these systems to. So your observation context for taking these supplements is going to be different than when you first bootstrap these systems of taking these 70 or 60 pills. Right. So even if we like look thought experiment, if we really get to say about anything in our roadmap, but you get to participate, you get to co brand and you have to pay for a seat. We could take some of those funds to help sponsor development as we laser focus on the health use case and then we build the health use case. We keep getting these cycles of getting advice and whatever we're building evolves well overall with diverse perspectives. And we just keep trying to augment that advisory council to grow into the research and science community. We did have epistemic knee powered Jonathan model of self in the past before you did the 60 pill, 70 pill thing and then now we have that version. So then we get like pretty diverse perspectives from high level players, either on the cutting edge of different multidisciplinary fields that we're interested in, like neuroscience, et cetera. Now what would we be able to do with those two? That, that would be my next question. Like and just open question and maybe I invite Dean to. Or we get like the private sector very interested in the hyper personalization problem. How could this be a valuable use case? How could this go to market in these verticals in the future? That's my thinking. So what are you, what are you guys thoughts on that concept?

That's a lot. Do you want thoughts on the concept of like how to attract sponsors in that way?

This is Robert, our request recording for this podcast cut out at this time. We're pretty bummed about this. We will discuss more on these topics in future podcasts. Stay tuned. Explore here if you wanted to. I think that might be interesting. Get your perspective on it.

Didn't fully understand the question, but Jonathan.

I think yeah, so. Oh yeah, Jonathan, if you want to go on that too, ask it again.

Sorry.

So yeah, we have model of self. Jonathan. Nothing right now of like, okay, cool, I'm going to be taking my pills.

Oh, the learning versus like. Yeah, yeah, habits, right. Okay, thanks. Yeah, yeah. And I think that's a great example, right? Like say you're testing a new set of supplements, right? Like you're paying a lot of attention to what dosage is you're paying attention to. Like am I getting the outcome that I'm expecting from this? And then you know, once you've habitualize something you're, your attention changes. Like you don't need as much to collect as much data on that and you can kind of just habitualize it. And so I think that's where like, you know, I kind of underline under understanding these, these layers of habits that we have, you know. And I really like the way James Clear frames habits and how each like little microhabit is just like a part of your identity, right? So you're kind of like learning to build identity and then once the identity is kind of like there, you don't need to actively put attention, you don't need epistemology, you know, because. But, but then your identity becomes a context for the new things that you're going to learn next and then you have it.

That's pretty cool to wrap it up into, you know, close the loop on. All right. In the typical health journey, I have to take on new habits that's going to be different than my Model of self today and epistemic me. What we're really trying to do help you in that journey. Like we built the tools to be able to express an AI health coach with a certain philosophy, have it be interchangeable, have it evolve. So it's nice to close the loop on kind of that storyline in the data model.

I wonder if I could jump in for a second there, please.

Go ahead, Dean. Yeah, please.

Yeah. I think that idea that Jonathan brings up is also very analogous to the concept of a explore exploit. If folks have heard about that. Basically it's the idea of how do you operate in an environment where you have low certainty. So this actually applies to like our everyday life as well.

Right.

So imagine I'm in a new city and I'm trying to figure out where's the best place to eat. My best strategy in this given moment is to explore a lot of places. But after a while, once I've explored enough places, I can start to exploit the.

Are these the only philosophers you think are important? No, that's certainly not it. There are many more.

I just information that I've learned and just agree on a few places and regularly eat really good food.

Couldn't put everything on a slide. And there are many more ages of humanity as well. Okay, I think one other thing to. To cover from the dinner that I slacked you guys. So he said the word epistemology and I was like, damn it. I wish I recorded this because we say that word. I never hear it said in normal human conversation. Never really hear that word.

So this idea can like similarly apply to our beliefs or our ways or our habits. Right. Our. We form new beliefs, more exploratory beliefs. Trying.

That's weird. Right? Wait like you guys do. No way.

And gain more information about environment and once we've established the right amount of information we can just.

So anyway, he said it it once and I. I forget the exact context, but it was similar to when he was talking about biographies and how he sees himself as a philosopher and he mentioned the word epistemology. I think it was in relation to like how we're at this stage of uncertainty and not knowing is actually more powerful than thinking. You know, And I just wish I recorded that. I wish I recorded the part where them.

And reduce the amount of learning because it's not particular. Really useful to the task at hand. Yeah. And that's like a. I think it's a really powerful concept and it shows up in many different ways and I think it's a. It's a Great analogy to like how reinforcement learning thinks about the same problem.

Really great. He said I see myself as a philosopher. And he said epistemology that would have gone great perspective in addition to the conversation. Dean. I think that's a great way to frame it. Jonathan, did you have anything to add as commentary to what Dean said?

Yeah, it's really additive language. Yeah. I think it's a good way to look at it. Maybe, maybe in a learning system, you know, can. If you can think about measuring that threshold between explore and exploit that's. That's where those epistemic emotions kind of come into play.

Yeah. In. Sorry, go ahead John.

And like I just like we. And we talk about asking the next best question. Right. And so you know where you are on that threshold then like you know should I ask explore or exploit question? That might be one way to in.

I don't know like our GitHub as legendary like that was so cool. Yeah. Any other thoughts from. I guess that whole story of the pitch or anything to double click on. Happy to do that.

I really like that question that he provided or that he posed to the group about.

Yeah.

Would you rather take 100% that works. Yeah. I think as we start to socialize these ideas with like the larger like research community, like that, that parameter for example that's used in the. I think it was clarity in your approach or would you kind of like want to randomize your approach?

Yeah, it was specifically it was. Would you want to give up all your autonomy and follow an algorithm guaranteed to be good for you like you.

Only thing you explore exploit framework is often called the epsilon parameter. It's really cool. I think we could start to map some of that language and help bridge the gap and get more folks involved in the language.

That'd be really cool. I actually was thinking once we get this to a point where we benchmarked it against that study because actually that's what I know is this is going to give you the best health, spiritual health, physical health, mental health, emotional. I learned from this entrepreneur I met. He's kind of curious about different health of your life if you just do this but you just have to listen to what it says. So sorry, go ahead.

I mean yeah, I think unpacking my. That question is actually very interesting because I actually think that is the foundational question of epistemology. Right. Like how do you go about knowing things like what, what, what bounds of.

Certainty if there are different companies solving for hyper personalization and, and how they might benchmark. Makes sense. Right? Because at the end of the day it gets to that level and so I think that'll be a next step.

Do you place on things where. Where do you place your authority? You know, if I were to say.

Like the scientific class follow after Wheat figure that out and then we could start bringing this to epic Method is.

The algorithm that I want to follow for approaching my life. You know that's a pretty the research.

Community and science community maybe get Michael.

Levin expansive idea you know and that can encompass a lot of different things. So I love yeah I love that question. I think it's very philosophically aligned with.

How we've on the podcast. That'd be like a dream goal. That'd be so cool. So yeah certainly in the future. Okay, so this was like a really quick double click triple click into the use case of multiple models of self for the first couple parts of like a morning ritual using epistemic me. So if you're an application developer or system, what would you do? Dean and Jonathan, I'm curious if you answer the question.

Yeah I tend to approach the way world a bit more non linearly than other folks. Like I always assume that I'm operating from a place of limited information about the world so I'm always on the lookout for things that or structures that.

I only who's thinking about implementing an AI health coach or trying to understand like how epistemic me you could leverage it for hyper personalization for your app regardless of its health regardless of whatever we talked about education we talked about.

Is the impartial representation representation of and that if I were to try and explain cause effect in that domain dating.

We talked about personal finance hyper personalization cuts through all of these.

So one thing I might operate successfully on the information I had but ultimately I'm missing the wrong abstraction. Right. And for me I see the way to kind of exit that bounded rationality is to try to find these like beautiful explanations that explain many different things at one time.

That's I'll say is like you know, look at our docs, look at our GitHub if you want to be a contributor, be a contributor. If you have feature requests and stuff like that we have documentation on how to contribute and how to reach out to us. So one other thing that I'll just say is based on where we're at.

Now my usually my approach and that's kind of like non linear and it requires like kind of peeking around a lot of corners and you know it doesn't always follow a straight path path but that tends to be my, my, my approach is trying to find those explanations that explain many things and I find that once you've reached that point a lot of clarity follows. So yeah, so the point is, the point is to try and gain clarity.

This is of what should epistemic me's immediate roadmap be from a product perspective is AI. Brian, cool details underneath. We got to get into sub cycles with this dev team. There's got to be some discovery of like are the features. You really gotta get some clear analytics ones, some low hanging fruit ones like okay, well we just need a way. Yeah so like for the AI algorithm question, like the fundamental question is like would you be okay giving up control then to like something where this does the sense making discovery for you and optimizes for you?

I probably wouldn't because I think I would. I would be skeptical of one just like following any arbitrary.

For just constantly updating Brian's stuff from his content that he creates. There's clearly a little loop there we got to establish but we got to get like a PRD on, on on that. So that's the clear next step. And from a strategy or organizational strategy perspective, I think it's about time we like incorporate actually officially because it's like getting pretty official and I want to.

Run this authority unless I fully understand its mechanisms for actually like understanding the world. Right. If, if I don't understand how it's understanding the world like it may be operating in incorrect on, on epistemologies I don't agree with. So I think, yeah, I think I, I think. No, I think, I think that's where I'm coming from on that question.

Yeah, that makes sense. Okay, so if you got to, you know, if we vision this as a global open source project so there are many options and I've been thinking through how we might form so, so we could talk about that as well offline and then on the POD we see the underlying mechanisms then would you say. Yeah, okay.

And if I could fully understand you.

Could talk about you know, kind of our decision making process later. But, but kind of examples I'm thinking of are like yeah, yeah, yeah, yeah, yeah, okay, cool.

And if I agree to its underlying philosophy of its underlying epistemology.

Yeah, yeah, yeah, that makes sense, that makes sense. I mean basically if everything was built at epistemic me like okay, cool, perhaps, yeah, like Hashicorp, Linux, React and there's even ways where you could just offload the administrative and funds and stuff like that to. There's a little app that I found that does exactly that. Some open source projects. Yeah with. With many caveats but I'll let Jonathan go.

I'm just for, for the viewers I'm going to step away to go to the restroom but John mentioned go ahead.

Projects just don't even worry about governance and structure and the ip.

So I think those are some good like terms like having explainable, you know, principles that kind of, you know even if I don't understand the decision in the moment that I can go back and understand like not to talk through.

That but high level what I just wanted to get your guys's thoughts on and you know, flavor of just being building in the open. My thinking is we hyper focus on the health use case. Like I don't want us super distracted. It's fun to think about other ones because it also informs like all right, are we building the architecture in the right way to incorporate universal concepts principles.

That were used to derive it? That's important and that's I think one of the key goals of epistemic me is to sort of make explainable the human condition. And how are we looking at a model of self? Exactly. Why is this the best thing? And I think yeah, it's bounded rationality is a good term here. This is Herb Simon's kind of approach to economics is that we as it.

So once we land in a vertical, can we expand very easily? I think that's, that's always smart. But I don't want us super distracted because we have a clear tangible way to go to market and help that community do science at scale, which is essentially Brian's goal and that furthers epistemology.