x402's $24M Month, Anthropic's J-Space, Navigating Online/Offline Worlds, Training Rivers With Rocks

Ty Patterson, good to have you back. Good to be back, Ruma Roberts. So tell us a little bit, where have you been? What's been going on? Where have I been? I've been in the offline world. Fantastic place, this. It's amazing how you can spend, yeah, just days without looking at a screen and it completely puts you in a different headspace and I've been living in a different world. So traveling a bit, been traveling to the United States, traveling through New York and going to festivals and retreats and things like that. How about you? Where do I find you, these things? Well, I am sitting, as you can see, in front of the aptly named Bookshop of Potentiality, which you actually named quite well 30 seconds before we hit record. I'm in Portugal, outside of Lisbon, in the Cintra Mountains, actually half-sitting, a couple of cats, a horse and some street cats. And I have to say, it's been really delightful just stepping into someone's house for about 10 days and being here to support, but also trying on different lifestyles and seeing, oh yeah, this bit's nice, this bit's nice, this bit maybe not so necessary. But I will say it's been very nice getting out of the city, but still being within arms reach of it. Yeah, I like that. Not having the necessity to be right in the city, but having the potential. Yeah. This bookshelf of potentiality, it reminds me of that there are two distinctly different activities. One is buying books and the other is reading books. And the delta between the two, buying and reading books, these people have a very small delta apparently between buying and reading books. Whereas I love buying books and then I read like the first chapter or two and go like, oh, okay, I got it. On to the next one. Yeah. Well, I mean, maybe they're just realistic about what they can actually read. But I like that ambition. Let's do the whole wall. We'll do the whole wall bookshelf. But I certainly don't want to talk down on another man's bookcase. But I just imagine the potentiality here. So I think I'm going to have to surreptitiously leave a book behind. Speaking of books, it's helping hold my computer steady here, but I'm going to take it out for a second. This was gifted to me recently. I was traveling through Milan and I was staying with my friend Rodolfo, who is a wonderful human being, a designer and a design thinker. And he's very Italian. And he gifted me this. I know this book. The book of gestures, right? This is fantastic. This has got all of the Italian hand gestures and then it's got a translation for what it means. It's got it written out in Italian, in English, in French and in German. Very good. So with that being said, have you actually learned any new hand gestures? I learned this one the other day. This is like, this is like, ah, come on. This is like, I'm bored of this. Like, kind of tapping the chest. It's like a board seal or something. But I haven't had a proper read through it. They've got this one as well. The Italian one, this one. What does this mean? Can't remember. Oh, hold on. I have the book for it. You can probably find it out. I think talking about hand gestures is probably one of the best things to start with in the podcast, particularly for those that are listening. But just to share one more hand gesture, I got this one from Thailand actually. He's putting his hand in his pocket and pulling out a Korean heart, I call it. Oh, Korean heart. I thought it was a mini heart. I didn't know if it was like Korean or Thai, but in certain parts of the world, it just lights up a room. It's crazy. Particularly coming from a foreigner where they don't expect it. They're like, what more thing? And you go like this. It's great. It's so good. I like the difference in hearts between generations. Maybe someone listening can understand what I'm talking about. So you have like the, you have the, the swim where you put your thumbs down, two-handed heart, thumbs down, fingers touching at the top, which is I think like millennial heart. And then you have the, where you put your fingers, your index fingers down. And is it your thumbs form the top of the heart, which is very like Gen Z. But I'm not quite sure how to do it. No, I think you, I think you, I think you nailed that one. Yeah. And then, yeah. The mini heart. The mini heart. I mean, the mini heart's great. It can kind of come at it anyway. Well, with that being said, how about we get, how about we get started? Let's crank into some, some new weeks. Yeah. The weekly, the weekly check-in. So I know it's been a few weeks since we've spoken and there's a lot to catch up on. And having said that, I think let's just for this week's news, because there's just been so much movement. And I've got a few points I can kick us off with. One in particular that I think I actually haven't played around with just yet, but I've heard some of the examples of it so far is ChatGPT Live. And so they've actually figured out a voice model that can listen as it speaks. So the fluidity has shifted entirely. So its ability to actually understand the conversation and understand when someone is actually thinking and the subtle vocal cues for that. So it reduces the interruptions. And this could be a really pivotal point for a voice as an interface layer between kind of online and offline as well. So, yeah. Again, I haven't played with it, but it was released this week at ChatGPT Live. Interesting. It's fascinating how when we start to pass these things out, things that are very uniquely human, how complicated they are. Like having a conversation, listening to, getting the voice cues, getting the body language cues, all of these different pieces of data that we hold like that. And then when you try and do it in a digital fashion, you realize what barriers you come up against. Yeah. Well, I mean, I'm not sure if you recall the Babelfish from Hitchhiker's Guide to the Galaxy, but it feels like we're just on the edge of that, where not only do you have this voice agent that can listen and talk at the same time, the ability to traverse all these languages as well. So, I think, you know, huge news, but again, I haven't kind of looked into the weeds of that. Do you have anything that's jumped out this week, even though you've only been back kind of in the online world for a couple of days? Yeah, I haven't been back for very long. I have been kind of resisting getting back on a screen because it's been so lovely being offline. But what has jumped out at me is like, when you're witnessing somebody grow up on a day-to-day basis, you don't really see the changes. But like, I've been gone for a month and when I come back now, I can kind of see the broader strokes of what's happening. I see that the models that are all being released, like these latest models from OpenAI, The Soul, Lunar and Terra. People are raving about those. One thing that jumped out at me in the news is Fable is back, back for what seems like, well, what is being told is a limited time. And that seems to be extended and has been extended a couple of times. So I think that was interesting news that there's some kind of tension going on there, trying to like release these models and not have the US government go back on their word. Or it seems like they're trying not to go back on their word. Anyway, the news piece that jumped out at me was that this company Cognition were posting about. And I believe the guy's name is Jun, who was writing about this, Jun Lee. He was saying that they're using Fable to decrease their, basically their overall AI cost. And they're doing that because the trade-off between, they're able to do that because the trade-off between the intelligence that Fable gives them and the kind of token usage versus Opus 4.8 means that their overall cost has gone down, even though they're using a much more expensive model. And it's just, to me, jumping out because it speaks to that we're now reaching a point where actually having a more capable, intelligent model do most of the orchestration and then having some kind of harness where a cheap model does the execution is a really viable strategy. I'm also noticing that for a lot of people in a lot of tasks using the frontier models is kind of plateauing or like you don't need to use the most intelligent model for most tasks. And that's a really interesting place to reach because to me it says that these things have reached a point where it's good enough for a lot of things. And it's kind of a, I don't want to say it's a moment that we kind of leapfrog, but we're going into the unknown from here as far as intelligence goes. Like when you can't tell the difference between what two models are doing or like there's, yeah, when the cheaper models are good enough, then we reach this point. Oh, okay. From here on, and that's a bonus. So I'm interested to see how all of this keeps playing out. Yeah, it's going to be crazy because the intelligence isn't just going to stop. Like it's going to keep going from here on out and way past what we've got capacity for. And then having said that, it seems like a lot of tokens are being optimized for revenue generation. So, you know, I think in a very short period of time, there's going to be a lot of trade-offs between do we use these tokens to produce another film or do we use these tokens to do groundbreaking medical research? And unfortunately, the one that generates more revenue is going to be the one that wins. And it's kind of like we're up against this cultural system that we're a part of that will play into this as it continues to evolve too. So, yeah, just watch those two parts kind of grow out over the next two, three years. It's going to be, yeah, a pretty interesting bifurcation actually. But also a really interesting point in like how much intelligence do we actually need access to for a lot of the things that we're actually working on in our current culture and systems? Not necessarily a lot. In the whole scale or the whole spectrum of what intelligence could potentially be. But I actually wanted to get your thoughts because you're still kind of fresh from offline and maybe you've got a better sense of the signal versus noise right now. But like what have you maybe had time to reflect on? And what was maybe more of the inner news that came through during the offline period rather than the outer news? Good question. What's been coming to me a lot is what is important. And the realization that we're really good at making tools. We're really good at making things that can capture our attention. We're really good at getting stuck into the weeds and like doing a thing. Whether or not that thing is worthwhile is the question that comes back to me quite a lot. And a piece of news actually that jumped out this morning when I was reading is the title was something like AI is the most human technology that we've ever created. And that landed really well because the argument was that we're in a point in time where we can be creative, where we can take a problem and say, hey, I want to try and solve this. We can instead of having to put so you used to have to put so much energy into creating a thing that it was much easier to consume. But actually consumption is not a particularly human trait. It's something that comes easily to us and it's a survival thing. But it's not something that we're inherently designed to do, I don't think. Like I have a real drive to make things, to create things, to bring things into the world. And I think that that's a very common experience. And that's what I was getting really clearly in the offline world. And then when I come back and I start working with these tools, it's like, okay, I really want to create things. And now the question isn't, can I? It's what should I spend my time creating? Because whatever I get pulled into, I'm going to put all of my time and energy into it. So that, yeah, what's important and how do I make the decision to work on a particular thing now that there are all of these things available? And now a lot of people can do this. Like we have co-conspirators in creating like amazing new technologies and discovering things. And like, I feel just like more is possible now. And in parallel, we're in this kind of dangerous period where it's also easier to slip into just doing anything. But that was always the case. You could always go and, you know, spend your time counting rice. And sometimes that's a really productive use of time. And sometimes, especially when there are things to be done, it's not a productive use of time. So it's like, yeah, that's what's come through. Well, so actually one of the pieces of news that really jumped out for me this week, which you kind of had a thread of, which is, you know, some of these AI systems are actually starting to become quite human. And Anthropic released a research paper this past week, actually July 6th, on this idea of the J-space. And so basically what's happened is the VED model has developed this. And they're calling it the J-space after the Jehobeian lens, which my understanding has actually got more to do with named after the Jehobeian lens, which is more of like a mapped to a human cognition pattern. And now we're seeing this in AI agents. So basically the J-space is this middle space between input and output. And it's kind of the idea of it's somewhere between intuition and thought. So the J-space, it doesn't, it's not output, like it's not actually writing it anywhere, but it comes up in its workspace, basically. And so what they're finding is, you know, like one of the examples was they gave it a big chain of protein sequences of like raw amino strings. And without it outputting anything in this J-space, it came up with this idea of a protein, this idea of fluorescent, and this idea of green. And without being told anything, it kind of intuited that this was from a jellyfish DNA. And so for a long time, a lot of the agent work has been black boxed, but now there's some visibility around it, which is just for me seems crazy that we can actually have kind of insights into these kind of like agetic thoughts or agetic intuitions. So I think this is a huge space to keep looking into. And the fact that it maps so closely to how our system works. Yeah, it, it, like, it makes sense to me when we, when I think about, I'll speak from a personal place. When I think about my own cognition, when I think about my metacognition, um, we're making patterns, we're connecting dots. Um, there's this idea of, like, intelligence is not very, um, clearly defined, uh, these days because we, we use the word so much, artificial intelligence. But, um, this really, uh, thoughtful, informed AI, um, professional that I know is talking about the difference between intelligence and education. He's talking about the AI models are highly educated, as in they know a very broad spectrum of information and then can, um, connect dots between all sorts of different things. And choosing which dots to connect is arguably intelligence. Being able to connect different dots is education. And because we're widely read and we know a lot of different people who bring a lot of different ideas, I would say that we have, um, a broad spectrum of education. The depth of it, for me, at least in different fields, definitely varies. But then this J space feels like that kind of, um, dot connecting layer where you start to, and then for humans, yes, we have this like J space personally, but we also have interpersonal J space. And we also have like, um, um, every time I interact with a different piece of information or a book or a piece of music or a piece of art or, um, anything really, another human, we have the potential for this extended space. Um, so that makes sense to me. It's interesting to be able to visualize it. And, um, well, they've been experimenting a little bit with it and have, um, you know, shared a few things, but one of the experiences, one of the experiences they did is they closed down, but killed the J space for a few requests. And they basically found that, you know, their best models could still do very simple tasks, multiple choice, grammar, sentiment, that kind of thing. But everything that required reasoning, um, composition, uh, translation, all of that like collapsed to, to below their, their lowest model or their smallest model rather. So this kind of like intent space, this thought space, this workspace for the model is proving to be a really powerful space within how, um, intellect actually functions. Um, yeah. So it reminds me of the marshmallow test as well. Yeah. Like how much patience do these agents have? Curious. Basically how much, like we talk about creating space. We talk about holding space. Like how much space do you have to get things wrong, to play, to, to make mistakes? And then to let something form out the other side. And if you have to like make a, give an answer really, really quickly, it's like, ah, it'll give you something. If you allow the time and space and different inputs and more information to percolate, then you get like some real juicy answers. Totally. Yeah. Um, I've got one more piece of news. Do you have anything else that's jumped up here? I have a couple of small things. Um, why don't I jump in with, with my next one and then you can go. Yeah. Please. So, um, Manus AI have, uh, this auto publish feature that they've released. Um, and this was an interesting piece of news because it basically allows the automatic deployment of successful builds into live URLs. Um, so it, it sounds like it's a coding agent and I haven't gone deep on this, but it made me laugh too, because this is how I do my, um, agentic coding. Um, and I know that most, um, developers would be appalled at this kind of, oh yeah, it works. Let's ship it live straight away mentality. Um, but now Manus have this kind of automated pipeline where they go, they apparently have like a testing function and then it auto deploys and builds the, whatever you're building to live URLs. Um, which is I think a step in how the stuff just kind of gets compressed and, and how we ship things really quickly. Um, and from experience, it will probably lead to things breaking, uh, more often. Definitely things breaking more often. However, I get the sense that there's also going to be quite a lot of self-repairing mechanisms emerging in some of the next, uh, rounds of model releases. That seems to be just around the corner. So we're kind of struggling with a lot of like buggy things and we're trying to patch them with, you know, human in the loop pieces. But, um, you know, already I'm, I'm experiencing that a little bit with Hermes, like some of the self-repair functions are starting to become really, really good. Remind me of this when we get to tools. We will. Yeah. Um, so the other bit of news that really jumped out for me this week, and it's a little under the radar, but, um, I think it's, uh, pretty foundational in many things. And it, um, looks at more of the agent to agent landscape. And so this is kind of a collaboration initially between Cloudflare and Coinbase. Um, and Linux has now kind of recently been involved as well, but it's the, it's called the X402 protocol. And so 402 used to be the error that you'd get if there was a payment gateway that was denied. And so now they've actually turned that into a protocol for agents to be able to pay for APIs, content and services automatically. So I think, you know, it's small, but it's a big thing. It's like, uh, here's, here's a floodgate, but I also love kind of the, the nuance of converting an error, uh, protocol into, uh, actual, uh, open, open, open pathway for these. So, um, yeah, I think we're going to see a lot of paywalls, a lot of micro payments, um, a lot of bot traffic, unfortunately getting, getting rewritten, um, for these, for these protocols. Yeah. For this protocol in particular. That's so funny. I, uh, I, uh, I don't know if you've seen my LinkedIn. Um, I was posting about this as well. And, uh, the volume of this is crazy. So they launched it in, uh, June, 2025 with basically nothing. Um, and now the 30 day volume as of June 26 was $24 million. Um, and 95% of that volume is now above, uh, dollar per transaction, which means that it's gone from mostly testing this protocol and mostly like just seeing if this thing works to actually having, um, agents have their own wallets. And have their own like way to pay for things. Um, and the way that, yeah, I see this as a massive space. I think the AI agent economy, agent agent economy is just going to dwarf the human economy, at least online. Totally. Yeah. Beautiful. Any other bits of news? The only other thing that jumped out at me was, um, this website, which was the, the agentic economy, uh, treaties. And, um, it started talking about the different ways in which, um, agents interact with each other and with the world. Um, but specifically what jumped out at me is that the way that you interacted with this is that, um, you went to their website and you could either read about it, listen to it or watch it. But it was like, uh, um, consume this content in any way that you like and in, in any, um, size that you like. You could, you could consume the whole thing or you could get a summarized version. Um, so it's not specifically the piece of news, but it was the way that you could consume this news that, yeah, that was really interesting. What was the website? Uh, it's called the agentic economies treat, the agentic economy treaties.com. Um, and subtitle the convergence of intelligence and the economy by Jeremy Allier. Beautiful. Um, yeah, I'll share it in the show notes. Great. I haven't. I wonder if it's, has he created it as a skill yet? Cause I'm still starting to see skills and, um, I'm seeing the emergence of also like not only skills, but bundles of knowledge as well. So actually people sharing the bundle of knowledge that they've accumulated in a wiki and like, that's the transfer of knowledge now. So it's not for us to consume. It's for our agents to consume on our behalf. And then it's like, well, you know, do I trust, um, my agent to ingest another agent's output? And how, what's that trust layer between, um, the, the agents as well. So that, that piece in particular, I think is, um, really crucial. Trust between agents. I do think we're jumping the gun a little bit here. Everybody's excited about this agent to agent stuff. I'm also excited about it, but I realized like, I mean, we work in this every day and, you know, the like little intricacies that, that trip us up. And then you see like how that agent agent, it just like errors multiply. Um, yeah. Um, and also maybe it's, you know, this, I don't think this infrastructure is necessarily us for us to create in any, in, in some ways, I think, you know, maybe we, we put the first stone down, but like the rest is, is going to probably, uh, look after itself in some regards. Let's, um, let's shift over to tools then. Yes. Nice. Yes. Um, you know, maybe you've been working with some more offline tools, which is also really good to, to go into. Hi, it's tool time. Uh, offline tools. Yeah. I built a sauna last week. Um, that's great. Tell me about that tool. Which was amazing. I mean, the sauna as a tool, um, there's a big difference between knowledge and wisdom. Um, and like we can learn all about these tools. I, I realized this working with physical tools is like, I've got quite a lot of experience building things. Um, I'm definitely not a master, but I have enough experience and I was working with someone, I was working alongside someone who doesn't have the experience working with tools, but he's very intelligent and he was learning things along the way. Um, and I realized that there are a lot of points that like, I know how to use a nail gun and I know how to use a skill saw and I know how to use these tools that are like these power tools and intellectually knowing how to use them as one thing. And just having the experience of like, you have to make your own mistakes. You have to learn by doing things. And so some of the tasks I would be like, oh, I know how to do this, but actually I'm going to give this job to this person because they will only learn it by doing it and by fucking it up a couple of times. And we get to talk about it and they go, oh yeah, like I realized that if I hold the nail gun in this way, it's really easy. And if I hold them this way and then it's really hard and you're like, yeah, exactly. Like you don't learn these things without doing them. And I think we kind of bypass that when we're, when we're talking about intellectual tools, but a lot of it is just repetition. Particularly a lot of this online stuff, it's disembodied intelligence. Um, and, uh, there is kind of this idea of, uh, really knowing and it's that embodied knowing that really gets, gets you there and gives you a certain level of, uh, quiet confidence as opposed to this intellectual. Confidence, which feels very loud and shouty. And like, oh, I know it. So I'm going to tell the world. Whereas, um, when you work with, uh, actual masters that build things, typically they're pretty quiet and confident and just, you know, go, go about that way. Um, and particularly when you're working with tools and wood, you know, the tool is one part of it, but actually understanding your, the relationship to the material that you're working with, particularly in the physical world changes a lot. So, um, you can take the same tool and work across, you know, 17 different types of wood and it will respond very, very differently. And it takes that embodied knowledge to get there. Again, you can read up about all the different types of woods and hard and soft and green and, and, and seasoned, but still like, uh, it takes actually time to, to embody that. Yeah. Yeah. It's the, it's the marshmallow test that's coming back to me again. It's like, we can do it really quickly. We can like go in and, and do a bunch of stuff, but actually like slow is smooth. That's the phrase, slow is smooth and smooth is fast. Like let's take our time. Let's, uh, choose what we want to do. Let's think about it for a second. We can talk about it, but even talking about it, like the difference between somebody that really knows what they're doing and somebody that is pretending like they know what they're doing is the pretender often. And I find this with myself, if I'm like not confident in something, I'll want to talk it out and explain that I can do it and like really talk it through. If I know what I'm doing completely and I know that I don't need any help with it, I'll just go cool and I'll do it. Like there's no, um, we don't need to have a whole lot of back and forth dialogue. There's a difference between, uh, communication for like bringing everybody into a project and having them on board and having them like doing their part in it and just communicating to like, for nothing to, to waste time or to pretend like you know something that actually you don't need to communicate that. You know, I don't know. I just, um, yeah, it made me appreciate this week working with physical tools when I really did know something versus when I didn't and actually having the humility to go, oh, I haven't done that before. I don't know how to do that. Can you talk me through it? Yeah. Particularly with a lot of these, even if you haven't done it in a while, it takes some, some refreshing sometimes. Yeah. Yeah. I don't know about you, but every time I like, I don't use a chainsaw enough to really be, to really be confident, particularly when I'm picking up someone else's chainsaw. Like I've, I've worked with them quite a bit, but you know, it's sometimes it's, you know, six or 18 months in between and you're like, well, I'm going to ask for, um, kind of a handover from who is ever chainsaw and is first rather than just going in. Cause you know, these tools, um, can sometimes leave pretty big mistakes. Yeah. They can be intense. Um, I, I also noticed the difference between, um, creating space to be creative and just getting in and doing stuff. Like the difference between practical making things and artistic making things. Um, I was, uh, helping to make a sculpture and I like went into this environment with a lot of people. And, and sometimes there was more discussion than like, I could see the vision. I knew what they wanted to make, but they were also trying to create space for emergence, um, for the art to emerge. And there was this interesting thing where there was like a time crunch from a, um, practical standpoint, but then they were trying to create space for the emergence of the art. And I went in there with a very practical mind, like, all right, I know how much needs to happen here. I know how long it takes to like, we're putting essentially cladding like, um, scales, so to speak on the side of something. Sort of shingles, but yeah. Um, and it, it was, uh, um, it was supposed to look quite rustic. It was shingles, but it wasn't supposed to be like formulaic. And there's a difference between being artistic and thinking things through and then just going, getting things done. So like at some point I was like, all right, let's just get it done. And now we go like, and you know, just not think about it, just do the job. Um, and noticing the tension between, uh, how much thought you need to put into something and like how much sometimes something being your artistic baby might get in the way of actually just doing it. Um, yeah. So what about for you? What tools are coming up? Uh, well, so one of the tools I was exploring using this week and I did get it up and it's working. Um, it can definitely be optimized a bit, but it's, uh, something called OmniRoot. And so OmniRoot is, uh, basically aggregating a bunch of, um, different, uh, model providers and having one endpoint that you point your models to. And then it, it, it's basically like a extended version of open router in some ways. And so you can kind of set up a whole bunch of smart routing protocols based on different requests that come through. And, um, it's, it's, it's really quite comprehensive. Um, so open router is a bit of an open standard, but this one kind of, uh, aggregates a bunch of these others as well. So, um, it's proving to be super helpful, particularly when you're working with, um, the open source, uh, you know, uh, um, landscape, um, and trying to get the most out of those. I was finding that, uh, so for instance, working with Hermes and open router, I was finding that you'd reach API limits, uh, quite quickly. And I've kind of set myself at least with one of the Hermes agents, um, the positive constraint of let's just use free open source models and see kind of what we can get, what performance and, and what agents can really kind of, uh, work well under those circumstances. And, um, yeah, and finally OmniRoute being a really powerful tool to, to support that, uh, that effort. So what, what, uh, the, um, API limit, like who has the API limit gate? Is it the model provider? Is it open router? Is it Hermes? Like where is the, yeah. So the limits are typically coming from the model providers themselves. So, um, basically, uh, also, so OmniRoute also compresses a bunch of different inputs and outputs. So the token usage isn't as high. So there's a huge kind of efficiency mechanism there. Um, and again, you're just working with one endpoint as opposed to switching across multiple endpoints. Um, which seems to be something that if, if you're not doing it that way, you have to manually change your endpoint in a prompt or an output. Um, or, you know, the agent kind of gets stuck somewhere along the way. So the idea is just that it gives it, you know, up to, uh, 290 plus, um, model options to, to kind of work through. And it's all plug and play. So you just, uh, plugging this or you're setting this up for your, um, ever present agent that you communicate with via what telegram? Well, so at the moment I've only got it running locally. Um, so every time I fire up my, my M5, I'm, I'm, uh, using open, uh, OmniRouter now. Um, and then also I'm using some, some local models. So a combination of those two seems to be working quite well. Um, yeah. Yeah. And so, you know, this, so OmniRoute also has the ability to do kind of intelligent routing between, uh, local and cloud providers, which is again, another good thing. Um, and so I think if we're looking for like a true headless, uh, agentic system, then this is a really good step in that direction. Um, yeah. Cool. Interesting. Um, makes me think of, so this is the tool that I wanted to remind me of. Um, like how do we make tools or how do we make systems that we already have that are self repairing or that are self, um, um, editing. I built a website for a friend a few months ago. And, um, then they asked to change a couple of things and like, I can make a backend for it so that you log in and you can like update the stuff standard CMS type, um, build. But I was thinking, well, I want to build this and I think that it's not very difficult. And I think that this is kind of the way that a lot of this stuff is going. Like imagine if you had a WordPress website and when you log into the backend, all it is, is an agent. And you say, Hey, I want to make these changes to my website and you talk to it and it goes and edits the code. So it would be connected to, um, maybe, uh, an agent API. And then it's connected to the GitHub repo as well, or however it's orchestrated. Um, and so you can, instead of going to the backend of a website and editing the website manually, you can go and log into your website's agent and say, Hey, I want to change the color scheme from blue to green. And I want you to increase these fonts and I want you to decrease these fonts. And I want you to change this copy from this to this and it will all go and automatically do it in the backend. So just like this, um, not necessarily self editing, but user editing version of a website where you can change anything that you like about the UI, about the display. Um, yeah, I've been starting to see a few of these tools emerge. So one of them is called ploy, P L O Y. Um, and the founder of, I think web flow put it together and his whole thing is like these ever evolving, uh, marketing websites. So it's very much optimized for kind of, uh, um, um, conversion rates, um, and it kind of changes itself. And then, yeah, the updates are very, uh, um, agent based. Um, I played around with it a little bit, but, um, yeah, I think this is definitely the direction that a lot of these, uh, digital interfaces will start to go. Um, I mean, I hope I never have to touch another line of CSS code or, uh, you know, pixel nudging and things like that. Again, having said that, um, you know, there does still seem to be a bit of a gap in terms of small design refinements, um, so that things don't look just super sloppy, um, uh, AI output. Yeah. Yeah. And culturally, depending on where you are, like sometimes, uh, I'll take, um, the UK and the US as examples. Culturally in the UK, if you are in uniform, if you fit in, you belong. Culturally in the US, if you stand out, you belong. So you can do the same thing with websites. You can say, oh, if this website is just like all of the other ones or similar enough to all of the other ones, it ticks the box in my brain to say, yes, this is a professional website. I trust this. Yeah. Uh-huh. Uh, on the other end of the spectrum, if there is something that is different enough and not homogeneous, then in some cultures like the US, people go, wow. Oh, interesting. This is a different experience. So I think we do need, yeah, that ability to, I think that there will be a space always for the new interesting thing. And yeah, like you look at Apple's website, Apple used to have, or you look at, uh, notion and Apple as examples. Notion have, like, um, uh, illustrations. They have, like, these things that, that used to be difficult to do because they were artistic and they needed to be consistent. And, like, the modern website kind of changes as the difficulty of doing different things change. Um, so you've got, like, the, the notion era stuff where if you have, um, if you need artistic talent that is expensive to create this kind of narrative that is really sought after. And then you look at Apple. Apple doesn't have any illustrations or anything. Apple's mostly photos. Like, they're, it's more expensive now to go and maybe expense is a weird way to look at it, but like, it's more difficult to go and take really, really beautifully crisp photos and to, like, really bring that realness onto a screen. Because it's much easier now to make cartoons to have that. Um, so like, as these things get easier, these companies that are trying to differentiate themselves by doing the difficult, expensive, artistic, grunt, uh, like leading edge things will continue to find things that are difficult to do because that's what we like potentially. Any other tools? Tools, tools, tools. Tools, tools, tools. No, I've been downsizing, uh, on my phone, taking away as many applications as I can. I think I've got it down to less than 50 on my phone. Are you, are you, are you like Marie Kondo in that process as well? Are you like, do you spark joy? Yes or no. And then saying thank you, or are you just being like ruthless? You're like, bah. I'm being ruthless. I'm glad, I'm gladiatoring this. You're gladiatoring. I, I would hate to see, um, it was a Russell Crowe and, uh, Marie Kondo in a podcast together. Or I would actually really like to see it, but. Yeah, we can, we can, we can, we can tee that up. Yeah, no, I have though, um, done this. So this is, um, supposed to be my frog. Mm-hmm. Because if you hold, uh, reptiles for too long, they overheat. They're cold, cold-blooded creatures. So this is to remind me to not hold my phone for too long. That's such a beautiful visual cue. Put the frog down. Yeah. I love, I love my phone. I love my frog and I don't want to, you know, I don't want to hum. Uh, how about for you? Any, any other tools? I feel like it's, it's kind of getting into building stuff anyway. So, um, we can, we can kind of switch into building and, uh, see where, see what goes from there. Um, so yeah, I've, I've just been focused on building out the, the, uh, MacBook while you've been, while you've been offline. I've been, I've, I've been practicing kind of offline as well, but like local offline. So how can I build my interface to be running locally as much as possible? And, uh, yeah, I've, I've been working with LM studio. I've been finding that really effective to manage my local models. I was started off with Ollama and for whatever reason, the end points and things were just super buggy. Um, and then, uh, running a local instance of Hermes. And what I'm finding, one of the tools within that that is kind of unlocked, uh, uh, a bit of a shift is this idea of the Kanban board. So Hermes now has Kanban board where you can put a task or a goal in there and the agents will automatically, there's like an orchestrator agent that will decompose that task into lots of different subtasks and then allocate it to a bunch of agent models that you've created profiles for. And so they go out and kind of go off and do that thing. And so I've created a researcher and a designer and a librarian and, uh, you know, a critic, uh, and like, uh, you know, a, a tie, like a tie, like, um, super like, um, minimal, like how do we just keep this to the bare basics? And, um, it's been beautiful to see how it, um, how it operates actually. Uh, I haven't quite been able to compile a lot of the output artifacts in a way that the outputs are still a little kind of scattered. Like they ended up being outputted into a bunch of random local folders. And so there isn't quite as a nice interface where you just kind of see it all there. Um, but I have been following a few YouTubers around the agentic, they've built these kinds of agentic operating system, but kind of these web UI interfaces where they have their clawed, they have their Hermes, they have their other agents and they just run them all from one web, uh, user interface. And it seems to be potentially a really good path to, to explore, um, and kind of bring all of these things together under one roof. And then also have it as a self-improving thing so it can kind of cut things out quite quickly as, as things go. Um, yeah, that's, I'd say those are the, the, the, the big things I've been building this week. Um, I've, I've got, uh, more of an offline project that I've been putting some energy into as well, but it's just a little bit too soon to talk about. So we, uh, should we move along to rabbit holes? Rivers rabbit holes. Do you have a jingle for this one? Ah, that's right. Rivers rabbit holes. Um, boing, boing, boing. So, so, I mean, this is a bit of a tangent, so which is fitting first for this particular segment, but this is the perfect place for it. I did, I did a, a voice activation workshop this week and it was very much about, uh, letting your creatures out. And, and actually I find that that is such a unique tool to, to practice. And it's just like letting these different noises out in a way that they, they want to be expressed. And, you know, how even with that in our own language and cultural context, we're so used to just, you know, keeping the sounds that we make in these quite constrained boxes and just playing around with that quite a lot. And, uh, I don't know. I saw some rabbits yesterday. So I like, I started making sound effects to them into landscape. And one of my favorite, favorite questions to ask somebody when I first meet them to get like a, a sense of how weird are you feeling or how weird could you be is, um, can you make a noise that you have never made before? And exactly. And sometimes people just come out and try. And one of my best friends, uh, I asked her and she just looked at me and she was like, no, I can't don't, don't have it. Yeah. I mean, I, so I'm actually, I'm wanting to extend the practice a little bit now and ask people, like, um, make a sound with how you feel. Like, I don't want to, I don't want you to tell me how you feel. I want you to make a sound. What's what sound would reflect how you feel? I haven't quite got the right question for it, but I'm working on it. Yeah. Yeah. Uh, vocalize how you feel, but don't use a word that I understand. Yeah. Um, rabbit holes. So, I mean, we can go anywhere with rabbit holes really, can't we? Um, I've been conscious of rabbit holes, like trying to shine a flashlight down rabbit holes. Um, and maybe this is, uh, typical because I feel like I always say this when you start talking about rabbit holes, but noticing like even opening my computer, this week has felt like, okay, I know that that's a rabbit hole and I know that it might feel, uh, I know that it might suck me in maybe. Hmm. Um, I don't want to change it from rabbit holes, but the, um, term that's coming to me is wormholes. It's like portals into pulling you into something. Um, so the rabbit hole that I found myself in yesterday was, uh, going back into terminal, uh, going and, um, realizing that even when I'm working with an agent and I'm going and getting involved in, in these projects that I'm going to need to open a bunch of other tabs and tasks and tools, because my agent will say, Hey, we're working with this tool and we're working. There's all of these services that, um, sometimes, uh, down the rabbit hole and around the corner and trying to understand what those are before going into the rabbit hole has been a really, um, useful practice for me. Yeah. I like that, um, idea of converting a rabbit hole into a wormhole as well. It's like, this is a bit rabbity, but actually at the other end, we've crossed so much time in space and we've done so much, um, through going, going through this hole in an effective way. Uh, it's, it's worth, it's worth the rabbity start of it. Um, I did actually find myself this week, uh, exploring, designing, uh, spaces for animals and kids to interact more as kind of like a, a play exploration learning space. And part of that was actually designing these, uh, rabbit, uh, tunnels, but also tunnels kind of alongside where kids could, could, um, kind of, uh, witness the rabbits without going in with the same ones and getting that kind of rabbit perspective. And I don't know, I feel like there's something quite, quite, quite nice to that. Um, yeah. So I don't know, actually I spent some time thinking about rabbit holes more specifically this week, which is pretty much a rabbit hole piece. Um, but yeah, I'm excited at the possibility that exists there and, you know, talking about embodied knowledge and learning. And I, I, I feel like we as humans have so much to learn and so much to gain from our relationships with other species of animals and, uh, particularly outside of those obvious ones like cats and dogs where we've kind of, um, we've kind of adopted them into our lifestyle and they're just kind of in it as opposed to taking people into a more natural setting. And then how do you create the space so that the animals might actually want to interact with us in a way, rather than us just going in and picking up the rabbit and being, oh, it's a cute rabbit. Um, so yeah, I find it with cats, cats, uh, great teachers in there. Um, you can't go in and pose yourself on a cat. If you go and try and pick a cat up or you try and pat it without it being ready, it's like, nope. But if you wait and you let the cat come to you and your patient, then I think animals in that way can be really interesting teachers. Yeah. There's been, I've been watching the, the birds circle, um, here in the States and especially on like really hot days where the, um, columns of air are rising and there are these birds of prey. Oh, actually this is a little bit of a rabbit hole, but I'm going to go there. Um, I was in a park in the lower East side, uh, a few days ago and I was sitting under a tree, um, writing in my journal and my friend is like standing there and pointing up into the tree and this is in the middle of the city. There's this huge hawk sitting in the tree, like right there. Um, we found out it's a, it was a red tailed hawk, I think. Wow. Um, yeah. There's actually quite a lot of condors in, in New York as well, particularly around Washington square park. Yeah. There's, it's, there's, there's actually a whole, um, series of, uh, episodes on, on what, what wildlife is kind of reclaimed urban context. And I know they did an episode in New York that's, but it's fascinating because, you know, there's a lot of rodents, particularly in New York or the rats and mice. So there's actually a lot for these birds of prey to feed on. Oh, that makes so much sense. And it's relatively easy for them to like spot, you know, these, these guys. So, you know, the trash feeds the rats and the rats feed the, the hawks. Hawks amaze the humans. Not that the hawks need to do anything for the humans, but yeah. Yeah. Um, I was reading about. I saw, I saw in Sweden that they've trained the crows to pick up cigarette butts and place it in a receptacle and then they get food. That is awesome. Is it though? I mean, it's manipulative, but. But it's. Like, but it's. I think it's brilliant in so many ways. And then it was like, well, man, like really? Yeah. We're training other species to pick up our trash, which probably has a huge amount of like toxins for these birds to, you know, be exposed to repeatedly in exchange for probably not necessarily high quality food. But, you know, this is still Scandinavia. So chances are it's better than, you know, most, but still it feels. It was, um, yeah, a bit. Okay. I'd like to restate my exclamation. That is so clever. Very clever. It's so clever. And like, it's, it is, it is a source of food. And you got to think that the, if the birds could, like, if the crows had a better source of food than doing this, then maybe they would go after it. I mean, I think as far as a public program goes, it's incredible. Yeah. I, I remember seeing a guy, um, somewhere trying to, maybe it wasn't Scandinavia as well. I think he was, uh, using euros, but he was trying to train crows to go and collect money for him. Um, and he would like have these notes or yeah, he had euro notes and he would, um, every time the crows touched the notes, he would feed them. And then he was like going through this process of trying to train an army of crows to go and collect the money to, to feed them. Yeah. I mean, food seems to be a really, um, easy way to, to, uh, train animals and creatures. I'm not sure if you ever saw that YouTube video of, I think he's actually Swedish as well. There seems to be a bit of a theme here, but the Swedish, Swedish guy teaching an octopus how to play the piano. It's, it's incredible. Yeah. Um, but again, like food was, food was the gateway for it. And then the octopus, you know, started to play without necessarily, um, needing the food, but, uh, yeah. Fascinating kind of, uh, way in which we help to, I don't know, shape our relationships with, with various animals and creatures. Yeah. Yeah. Well, like to, to bring this back to future tools, um, and I think actually I've mentioned this before, uh, it's Max Tegmark. Is that how you pronounce his name? Yeah. Um, and super intelligence talks about the best way that we can conceptualize our relationship with a super intelligence would be to think about how our relationship, to think about the relationship between humans and animals or humans and pets. But it's also probably like the intelligence gap between, and again, intelligence is a very undefined word to use here, but the, the gap between, uh, humans and animals and, uh, humans and super intelligence is going to be stupid. Uh, it's going to be very different if things continue to go in the direction that they're going and also realizing that we don't understand the exponentials very well. Human brains are not very adept at understanding large numbers, um, exponents. Yeah. Compounding. Exponentials. Yeah. Yeah. It takes, it takes a real, it takes a lot of practice. And even though I feel like we're on the edge of that, I still find it's very easy to kind of fall into linear thinking. And, um, actually one of the rabbit holes I found myself in this week was starting a new project with a bunch of non tech savvy people. I was like, Oh, how do I set this up? Like what, what is the interface we're all working in? How, like, and there's a bit of a rabbit hole to kind of figure out what that was. Um, and I ended up falling back on notion for now, but still I'm like, ah, so close, but not quite. Yeah. The tools we have versus the, maybe they are the tools we deserve, but the tools that we, that we want and need. Open questions. Open questions. Do you have any open questions this week? I feel like you've come, you've come from the offline with some beautiful open questions. My open questions, uh, what's important. What's important to put time into, to spend time doing. Um, yeah. What's important and what's the, not the most direct path there, because I really believe in the winding path is the best way that the following the river is how we get to, you know, experience all the twists and turns and the, um, that the path really does lead us to the things that we need to find in order to like, feel like the journey of getting somewhere is, um, enjoyable and fulfilling and all of these things. And knowing when to pull yourself out of a rabbit hole, like knowing what the appropriate amount of time to spend. Yeah. I don't know if it, if it belongs in rabbit holes or in open questions, but I've also got this open question around time. Um, speaking it publicly here for the first time, I'm going to write a book, perception of time, how we experience time. Um, what the, um, scientific basis of human perception of time is, and then how we can, uh, like a practical guide for how to, for myself, for how to stretch and compress time and, and how to really feel it or experience it. Um, because it's such a wonky concept at the moment, uh, like, and it's just continuing to become more and more. Um, I, I think that time is kind of an irrelevant concept at some stage or it's becoming that way, at least for me. Um, and then it's the only thing, like it's really becoming this, this like central point of, um, so I've got a lot of open questions. It's around time, how different people experience time, what the, um, what the cultural perception of time is in different places, how time finds its way into language and in different languages. Like in French, there, there isn't a word for this. There's not this and that. Like in English, we have this like inbuilt time perception piece where we have, um, where we can communicate time subtly. In French, you have sa, that. Like there's, um, it loses the time element. And that is the case in a lot of other languages as well, that you have this difference in how people perceive time. Sorry, this is a lot of open. So one of the, um, one of the lessons I was exposed to this week is, uh, part of reading, reading this book, how to read water. And I was, I was, um, was really curious to learn that part of how you train a river is not by reinforcing the sides, which is what most people and lots of governments actually focus their attention on, but actually putting rocks in the middle of the river. So then what happens is the water flows around and comes in behind it and it kind of train you. So you can train the direction of the river by putting rocks in the way. So even as a metaphor for life, it's, it's quite a, quite a strong, powerful one. So, um, I do feel like that's definitely one of my open questions, um, this week as well. It's like, what are the obstacles that are in front of me right now and how are they training the path that is in front? And also how to appreciate the obstacles. And I, I, I, I don't know you, but I, I, I do feel quite frustrated at progress sometimes and having this like, all right, this is the future vision. That's very clear. And then the kind of slow, arduous way to get there sometimes like, ah, man, let's just get there already. Um, and yet, yeah, finding the, um, finding the practice to, to recognize it, you know, everything along the way is, is a gift. So, um, one of the, one of the practicing questions is how is this a gift? Um, so that's one of my open questions. It will stay an open question as a practice. Um, and also the other one that's quite present, which I feel is very much connected to what you were talking about earlier is what is enough? Um, and enough in a whole, lots of different layers of it. Right. So, you know, again, I'm, I'm living someone else's lifestyle for a week. And I was like, well, what aspects of this would be enough for me and which aspects would kind of like too much. You know, for instance, I'm driving around, I'm spending the week driving around in a convertible BMW. And I was like, this is more than enough for me. I don't, I don't need this. This is a nice experience to have, uh, for a week or two, but like, this is not something that I need day to day. Um, and there's kind of an appreciation of it now, um, and not an attachment to it. So that feels actually quite healthy in many ways. And then the other side of it is what is enough time and energy and spirit to put into exploring a various direction to identify is their flow here? And are these obstacles that are in the way, like actually training to, to, to lead to that flow? Or is it just like, um, not your best use of energy right now? So the, the open question of what is enough, um, keeps coming up. Yeah. Those are my open questions this week. What comes to mind is I remember hearing Barack Obama talk about his job being decision-making and that he's obviously was in a high intensity job, um, had to make decisions in a timely manner, but he would collect enough information. And for him, enough information was more than 51% and then make a decision. And that's enough. And then move. Think about that. How do you know what 51 is though? Yeah, exactly. I don't know. Maybe he had criteria or maybe it was just an intuitive sense. Um, shall we deep dive? Let's deep dive. Can you, let's, let's tear, let's tear us up. Um, so we were talking about this deep dive as, um, what is the, what is the difference between the online and the offline world? And what is that space in between them? What is this intersection between the online and the offline world? How is it changing and how is it coming into view? I think this is something that doesn't get discussed very much. And these days it's, um, an important topic because there are such strong pulls, especially toward the online. And then we have this like real, at least I have this real sense of being a human and what it means to be human and, um, to, to not be binary and to not be condensed to bits. And, uh, to retain this context and to retain the nuance of what it is to feel in all of these different dimensions. And then there's this kind of space in between the online and the offline that is, um, being designed whether or not we choose to actively do it or not. So, um, that's kind of where I'm thinking a lot at the moment. But we have all these, you know, these are all these different angles to, to take to it. So I guess to give it some positive constraints, um, we can look at, uh, yeah, behavior and devices and getting things done and kind of how each of these, um, create this, this bridge between the online and offline. Is there anything that jumps out at you initially or shall I dive? Yeah. I mean, there's, there's, there's lots of things. However, I, I, I kind of would love to, to hear where your flow is initially with this, because I really value the offline time. Right. And, you know, even having spent probably the most offline time in my life earlier this year, your time in a dark room where you're, you're not only you're completely offline as much as possible. There's a clarity that comes from that. So I, I, in this instance, I'd love for you to kick us off and thanks for setting us up. And I'd love to, to hear what comes up initially. Yeah. Interesting that you say about this dark room experience. Um, I think about it as when you're offline, you just have like in a very simple way offline, you just have less inputs. You can have the extreme of being in a dark room or having your eyes closed, having earplugs in. Like you, you can have no sensory inputs and still there's something there. Still, there is you still, there is all of the memories that come up. There's all of this stuff. Your brain is a meaning making machine. It will create something out of nothing. Right. So it's not like there's, it's not, we're never in a place where there's nothing, but you can have less inputs. And in that way you can have a simpler existence. It doesn't mean that it's easier, but you can have a simpler existence. Then you have this online space, which is, um, when it started, you know, 90, 93, 94 was a very, very simple thing. It was, oh, I have this ability to get information that is in another location and I can have it now. Uh, I can read data from a data center somewhere else and I can see it on a screen and I can see it in all of these different places. That was still a very simple concept. Um, I, I could get data that way. I could get data if someone sent me a letter, I could get data if I went for a walk and I saw new things. I could, I could go and collect outside stimuli very easily. Uh, but it was simple. It was, you know, there were small bits of data. Now we have the ability to get, um, not only the ability, but now in the online world, we're so plugged in. And there are so many ways to experience and sense the online world that, uh, our brains can be very easily overwhelmed. Like this, this idea of noise. Um, and not only do we have the ability to find it, but there are all of these, uh, actors who's, it's in their best interest to take our attention and our decisions and to impose things on us. So now we're in a space where, um, we've trained ourself to the point that we have these inputs coming in from notifications, from, um, like every time I go the online world, I'm bombarded with different, with ads, with all of these things. It's the, I, if I'm lucky, get to choose when to engage with that. But most of the time it's subconscious. It's a learned behavior that I'm going into the online world. And so I'm at the same time choosing to go and do something, but it's a dangerous environment because I could be distracted by something else. And especially apps like Instagram are very good at hijacking your attention and stealing your time in that way. You're getting all of this extra noise by giving you inputs that you didn't choose. So that's kind of the spectrum that I am thinking of it in. As you're talking, I feel like it actually goes back way further and it's almost like offline is what you perceive yourself. And then online is what is being presented to you. So I'd almost go back to the printing press as maybe the first analog version of online where you had this kind of, uh, enough, uh, um, distribution for a thought to kind of circle around a bit. Prior to that, it was like all offline. And one of the other predictions that I would actually make is that we've reached peak online at the moment in terms of the amount of time that we're glued to a screen. Um, the average adult is I think still over eight hours a day. So more than half their waking hours. And, um, I would say from here on out, we're going to start to, uh, reduce at least the screen time. And we're going to rely more and more on a few interfaces that are connected with the vastness of what online has actually become. And we're in this like really awkward middle space where, um, you know, like this device here, the phone is the bridge between online and offline. And the fact that we can fit it in our pocket and take us with it everywhere, um, is, uh, both a crutch and a complete, uh, distortion field of, of on and offline. And so many of us have grown up without the understanding of, of how it's actually hijacked our hormones, um, and our reward patterns. And, you know, you see it in a lot of teenagers right now. It's a, it's, it's really, um, affected their ability to interact. I mean, teenagers are a bit hard to interact with anyway, regardless of era or culture, but, you know, particularly when there's certain, um, offline abilities that haven't been cultivated in any meaningful way. Um, it's, it's really, really apparent. So, um, that's how I would kind of, kind of put the two and then that space in the middle, I feel like what we're talking about again is a little bit around this idea of interfaces. And so how do you choose to interface with online and offline? And, um, I feel, I don't know about you, but I would love to hope that I'm at the tail end of my, uh, lifetime that involves, uh, big screens and that, and that serious amounts of time are committed to screen time like this. Um, and that there's a whole bunch of things that can be automatically done in the background to support offline interactions. And then quite ironically, these agentic systems that are coming online could actually help us to become better humans again and better as a broad term, but maybe like more human, um, in relation to the planet. However, I'd say the, the, the force at the moment or the, the, the attention is still very much going towards optimizing for revenue generation. So there's this kind of distortion between everything that is done online has to generate revenue in some way. This is the, the word that jumps out at me that you used as choice and the distinction between choosing to do a thing and, um, being coerced into doing it or being, uh, manipulated. These are, these words are actually too strong in connotation. But what I mean to say is that, um, doing something because somebody else chose it versus because you chose it. And a lot of the time revenue generation is the main driver. And so the, um, um, energy, the force, the cognitive, um, work that goes into creating things that will take our attention, take our choice away and give us some kind of reward in, uh, in order to, to generate revenue is kind of the crux of the matter. But to me, that space, that likes choice space, we just need better alternatives. We just need to like be able to actively make choices, be able to pause and say, Hey, actually I, I don't need to go and watch this Netflix movie. Or it's not necessarily that I don't need to watch the movie. I need to be able to myself make the choice. Um, and so instead of like, Oh, it's the Netflix movie or it's the, the thing that takes my attention. We need better alternatives that are not necessarily online or that are not on screens. Like there's not a problem with being online, right? With being plugged into more information or more resource or more, more. It's just being able to choose when to do it and when to not. Yeah. And the healthy balance of that. I mean, to me, the, the, the more, the most compelling alternative right now towards offline is going to build at a festival. Like consuming the festival is fun, but it doesn't quite feel as wholesome. And I don't know, I feel like maybe we've been trained into needing to be productive in a sense for things to really feel, uh, purposeful or, or to get that, that spirit of fulfillment and to be working alongside someone, even if it is, you know, putting up shingles on a sauna. But somehow that feels much more alive and present than like an offline that is like, let's just sit on the beach for a week. Yeah. You know what it comes back to for me is loops opening closed loops. The reason that a lot of these online things are so addictive is because they keep loops open. Infinite scroll, um, agents that keep asking you questions. Yeah. These are all open loops. What if we just built online things with closed loops? Congratulations, River. You finished the, you, you finished the internet. Good work. Go outside. Imagine if every YouTube video ended with congratulations. You've accomplished your YouTube goal for today. I feel like that should just be a website in and of itself. You've finished the internet. This is the finish line of the internet. You're done. There's no more you need. At least, at least for today. Shall I just make that right now? Yeah, yeah, exactly. Dictate it. Um, okay. That is actually going to be a five minute task. I think we could probably sell it. We could probably, uh, pitch it to Finland as well. I feel like there would be a great, great country to earn that. Do you know what their, um, their subdomain, it's not a subdomain. What do you call the suffix of a domain dot com dot fi? All right. That is right. Yeah. That's Finland. Yeah. It is. Okay. Um, you call it the, yeah, I think it's just the suffix. There's a prefix and the suffix. Prefix is at the start. Suffix is at the end. Yeah. Uh, amazing. Hmm. I mean, so, so where do you think the future of on and offline is going? Um, like in, in thinking about this overnight, I kept going towards, uh, robotics, like robotics as an interface for on and offline. Like if you have a robot standing, standing next to you that has access to everything you need and you just ask that, like what point of your interaction is online versus offline? And because it's very physical, it's, it's, it's, it feels like it's very connected to the offline experience. Or if we start going down the pathways of ingestibles where, you know, you have these, these online aspects or self-healing properties that any number of injectables might actually be able to take on in the future. Um, that feels like a really fascinating bridge between on and offline. And then of course there's the neural, uh, interfaces where you're not even interacting with, uh, anything anymore. It's all through, uh, mind space. And almost like a cyber telepathic interface. What I would like to see is, um, like a really human embodied, like we're talking about user interface, right? I would like to have an interface that is not just intellectual. I would like when I, when I want to learn an idea, I will read about it. I'll learn about it. I'll ask questions about it. I'll do research. Then I'll go and try and talk to someone who knows a lot about it so that I can like really understand it from different angles. Then I'll try and teach somebody. And ideally I'm going to move my body. I'm going to do something embodied. I think that that is how we're going to like interact and feel like we have some, um, some closure and a lot of like interactions. Like, can you imagine if, if, uh, the end of writing an email and sending it was like high fiving, uh, it doesn't need to be a robot, but like, you know, high fiving some kind of thing, like a really physical activity where you're like, yeah, I did the thing, you know, like that as closure. I think that we're going to see, I would like to see each of these experiences being actual experiences and having them felt on a body level and whether or not that's like an ingestible and it gives you like a specific feeling. Um, I, I don't care so much about, actually that's not true. I do care about how we get there. Cause I think that this is far more nuanced than, um, just like when we try to pass out, um, communication, like we were talking about at the, at the start of this podcast, when, when we're trying to pass out what it's like to have a conversation and have an AI model pause and let you contemplate things and all of that, I think there's so much nuance that is specific to the human animal. Um, so I'd like to see us create interfaces that are specifically designed for, um, kind of a human in the loop, as in we do human things in this loop. We, we give a high five, we have some kind of cathartic experience to, to round things out. And just like things evolve, like, uh, I was watching a video on how interfaces have evolved and how we've got the chat chat interface, which basically doesn't show any information and how we have like the, the graphical user interface, um, that like has, uh, on all of these apps that have all of these different options and menus and buttons to push is the space in between, which is very much, um, experienced done. Uh, so like when you're deleting an item on a computer, you don't, um, type delete and have it deleted, you can't do that with a terminal, but you pick it up and you drag it and you put it in a little make believe trash bin. Like this is a thing that we do and we just do it on there. Like, I think that the user interface will be a much more worldly embodied experience. Um, and then there will just be like a digital twin version of it that happens in the offline. But I actually see that the space, the online, the offline becomes, um, more. And I think we're quite a long way of this. I think this is a space that will evolve for a long time, but I think there's a lot of nuance and the difference between what we see in a two dimensional, two dimensional is a funny way to put it. Cause we can have 3d in screens, but like in a flat way on a screen versus what we have in the quote unquote real world. Yeah. Anything comes from you. Regardless of where we're going, it really feels like the convergence on on off, there's going to be convergence of on and offline and, you know, the closer you kind of get to it, it's like, can you even separate from it at any point? It's almost like there's, there's so few places on the planet that are left where you, uh, can't hear any sounds that are made by man, like this, this acoustic ecologist. He has this project where he's like, I think he's just trying to get like a square foot of space or some ridiculously small size of land. And he's traveled around the world with his microphones to try and identify these places where there's, you can't hear any man-made sound, no planes, no chainsaws, no, like any of this stuff and, um, this, this convergence of the, the technological and the human biosphere. Um, and then particularly with the, the satellite infrastructure that is, is increasingly starting to surround us. It's, um, crazy. Yeah. I'd say the, I mean, I actually was just looking into see if there'd been any, um, leaks on Johnny Ives device, but I think we're going to, the next iteration will be very much kind of voice orientated. It's probably going to be leveraging a lot of earbuds that a lot of people are using to both interface online and offline. I don't know about you, but I'm finding myself using that quite a lot in terms of podcasts and conversations and feeling like I'm offline. Like I'm, you know what I mean? I'm in my world, but I can still bring someone into that and converse in some meaningful way. Um, like I would love to do this podcast with just, um, earbuds while we're out doing our, um, while we're out offline and see how that kind of, um, tracks as opposed to just being like glued into the screen and see how that shifts the, the output, but also the thought process as well. Um, I often find when I'm even talking on the phone, having my eyes closed, I'm not having visuals changes the depth in which I can actually connect with someone a little bit more like intuitively or telepathically a bit as well. So, um, it's a little bit like all of the lights of the cities have crowded out our ability to see the stars. And so all of this time online has crowded out our ability to perceive the world offline. And so when you're working with different parts of nature and a lot of indigenous cultures perceive that knowledge and receive their knowledge from sitting in nature, observing nature, like listening to it on a more fundamental level. And there's kind of like a sense of connecting with, uh, a bigger sense of online in a way we kind of pass through that, that, that, that veil, so to speak without getting too esoteric. Um, but I can certainly comment on that even just in the darkroom experience. Yeah. There's no such thing as completely offline. There's always some input. Um, but our ability to access these, these quite big spaces is there. It's just waiting for us to, to listen. Yeah. Yeah. Listen to the quiet voice. The listen. Yeah. The ego shouts and intuition whispers. Hmm. Listen to the whisper. Any, any kind of, um, yeah, any kind of final burning points on the, on this deep dive? I feel like there's, there's, there's still so many different ways to go with it. And, um, I feel like we could probably spend all week, um, exploring this. And I think it would also be a fantastic book. So potentially after your time book, we do the on and offline book. I think this is what a lot of the work is. Over this next year, two years, five years, figuring out how choose to interface with both. Um, I'm really excited to see what devices come out and how the, how the work actually shifts. Um, oh, I did a fun thing, uh, last week that I think you might appreciate. I went and I had a little cough and, um, instead of running right next to each other, we ran to the park and then we got on a call. And I ran one way around the park and he ran the other way around the park. And we had a conversation while we were on a call and we were both running, but we are running in opposite directions. And at some point we ran past each other and we're like, oh, so that was a nice little online offline experience. Like, you know, parallel play doing the same thing, but in slightly different places. I don't know why I just had this image of two hamsters on this reel and then one of them goes the other way and they're like, hi. It's like that. Yeah. It was very cute. No, I mean, I love those invitations though. It's just so simple. Like, hey, what if we just do this the opposite way? AMA? Let's go AMA. Oh, I haven't prepared anything, but I can quickly. Well, while you quickly prepare, I will slowly browse, peruse my longtime listener AMA inputs. I did actually ask my local Hermes agent and it didn't come up with particularly great things, even though it is actually trained on a lot of our knowledge base by now. Now, having said that, I actually haven't given it our podcast transcripts. So that will be next week. I'll set it up probably. All right. I like this one. Long time listener, Kimmy. If you could only use one AI model for the next year, which one would you use and why? Go for it. Sorry, Kimmy. It's not you. Nah, it's just got to be the most intelligent one that's out right now, Fable 5. I've... No, but like if you're with it for the whole year. Sorry, can you ask the question again? If you could only use one AI model for the next year, which one and why? So let's maybe put some constraints around this. So maybe not one model provider, like not one lab, but one model. Yeah. Can I say Fable? Of course. And why? Because it's just the most capable. Because it can do things. Because it's giving me the best feedback. It's still got some weird patterns, but it's very clear in using all of the models from the first ChatGPT that came out through to using Fable 5 that this tool has changed so much. And it's the most... It's not even that it's the most human because it's not the most human necessarily, but it feels the most authentic. It has less pitfalls than other tools that I see, other models that I see. It doesn't fall into the same. I get frustrated with... Maybe this is a nice way to talk about intelligence as well. Intelligence is your ability to improve and to not make the same mistakes. Models tend to fall into these ruts of making the same mistakes. So I know that they will make mistakes, but I know how they will be mistaken and not flawed, but it's flawed in a particular way. This at least doesn't have consistent flaws. Not as many as the others. I've been hearing that Fable 5 has been... Like the guardrails are quite high, so when you're asking it, any questions about biology, I can't answer that. Apparently it's getting quite slow and obviously it's quite expensive. But I guess you actually haven't played around with too many others. I stand by it and I'm not trying to make bioweapons, so I'm fine with that. But yeah, trying to safeguard things. Sometimes you can't go too far, although I don't think that there's too far in this case. I think it's... Yeah, in this early instance, I agree. Yeah. I don't think it's wrong to blanket... I get frustrated with a blanket ban on using something, but I agree with guardrails. And also, it's a difficult thing because all of these guardrails can be taken down within 24 or 48 hours of almost every single model release, somebody figures out how to jailbreak it. It's not that hard to find that information. So, yeah. Yeah. I'm a little torn, actually. I definitely feel like going towards more of an open source model, at least an open weight model. But then, I was also even considering just... I only had a... If I... For a whole year, I could only use one model. I might even go for a local model where I could, just for that year, really refine it for me, for my particular use case, and less about let's go out into the world and all of the corpus of human knowledge and just input all of the more personal inputs, medical data and personal questions, and more as a close-thinking partner as opposed to one that can go out and just access everything. So, I'm a little bit torn, which I know is kind of against the question, but... Somehow, with our online-offline conversation as well, it feels like actually maybe focusing on more of a smaller model with stronger offline capacity and abilities might actually end up being a positive constraint that serves my use cases much better, rather than getting carried away with all the infinite possibilities of what's there. I think having a strong image generation piece would be really important for me, particularly in being able to communicate different ideas and possibilities. Like the image generation piece seems great. So, I still think ChatGPT is a bit ahead on that side of things in terms of image. But I also know it's not something that Claude is focused particularly heavily on so far. So, a bit of a non-answer. It's a bit of a non-committal answer. That's okay. What AMA do you have? So, mine is... I'm really... I've got to say, I'm really unimpressed with the AMA questions that my models come up with, but I do get ideas and derivatives from this. So, the AMA question that is a mixture of me and Opus 4.8 is, what domain or domain expertise do you have that you regularly bring into other domains where they're rare? Maybe this requires a little bit of explanation. Is it clear what I'm asking? No, no, go again. So, if you think about domains that you have a lot of experience or knowledge in, say, you are a profile, somebody who really likes biology and the natural world. So, you know a lot about that, but then you also work in tech spaces. So, what I'm asking you about is domain knowledge transfer. Are there certain domains that you spend a lot of time in that you tend to draw from or that you kind of touch back on all the time? That's just more broadly, what domains do you normally draw from? More broadly, yeah. Where do you find your first principles or your kind of sparks of idea inspiration is usually powerful. Yeah, I do. I do. Over the years, I keep finding myself drawing kind of first principles from nature and observations of nature and stories of nature and poetry about nature and understanding of how the systems work. And, you know, this idea of learning from the design of our planet that's had 4.3 billion years of evolution. And there's something to that that I draw from a lot. I also feel like I draw a lot from, it's not quite feng shui, but this idea of space feeling like harmonious in a way. So, I find myself moved by that a lot. Like, I'm not sure if it's necessarily domain knowledge, but there's like a, almost a sensibility that I like to bring into, you know, pretty much everything that I do. So, I'd say those are the two broad ones. Yeah. Like, almost every situation, those two lenses are very prominent for me. What about for you? That's a good way to phrase that. What sets of glasses do you put on? What lenses do you bring to a lot of things? Um, I constantly find myself drawn to psychology, but more like behavioral economics. I'm, I'm often thinking about how I or other people will experience something, but, um, trying to understand from like, what, what the, what the outcome might be based on like the studies that I've read or the kind of unintuitive, um, behavioral patterns that people have. So, um, I'm often thinking about, uh, yeah, psychology and psychology studies and experiments. Um, that's something that I usually find myself kind of drawn to that lens of trying to understand the world. Um, and the other one that I guess they're a little bit linked, but the other one that comes through is, um, that I often find myself in conversations that are either technical in nature or, um, um, don't have a lot of connection to the spiritual world. Um, and trying to, I find myself using the, not that it's, uh, spiritual, but it is, but I often find myself, um, using the phrase, the human animal, like trying to bring the experience of what it is to be like a fleshy human in all of these kind of futuristic conversations and spaces and experiences, um, and trying to take the perspective of an animal and not think that we're God-like and like realize that our brain is, is a piece of meat with electricity running through it. Like, um, yeah, we're just spinning around on this ridiculous rock, like piece of dust suspended in a sunbeam doing the best that we can. We're fire apes. Exactly. Yeah. Um, so yeah, that's, that's what comes in for me. Beautiful. Um, I have, I have one more for us. All right, let's do it. What's the last thing you used AI for that actually made your day better, not just faster. It's, it's really funny. I've been sitting on this question for, um, a few minutes now and I'm like, just in the back of my mind, obviously we're listening to you as well. And there hasn't been anything that kind of jumps out immediately. I was like, oh, I'm actually quite shocked at that. Um, it's helped me get information quicker and I'd say maybe it's helped me make some slightly better, you know, short term decisions about maybe where to go or what to do. Um, yeah, but I'm just trying to think just like more practically, like what's that? What? Like, yeah. I've got a couple of answers. I, I do a lot of travel as we've spoken about and taking, um, my broader vision of what I want my years travel to look like and connecting Claude to a notion database. There's my travel calendar and using Claude as my travel agent, giving it like old stream of consciousness. I want to go here. I want to do this. These are the dates for this thing. It's all. So it's helped me be more organized and be more organized. It's not in that I be more organized in that I spend less time doing the admin of this stuff and I can use my cognitive ability to, um, to feel into what being in a place for two weeks might be like. And sometimes it's sense checks me as well. I've like given it some kind of guardrails or I know that in this situation, I feel like this. And so when those things come up and I'm, I'll like have an event, I'll give it to Claude. I'll say like, all right, in, in a week's time, I'm going to go away camping for the weekend. And these are the details for it. And I can just like copy and paste or screenshot a bunch of stuff, give it to Claude. It will put it in my travel itinerary and then it'll go, ah, you probably need these things, um, for the camping trip. You need to be at this place by this time. Um, and so I've outsourced a lot of that. Um, it still feels like a lot of that is just like doing things faster, not necessarily like how did, like, how is it making your, your daily life better? Um, like, I, and I think it's a really nice thing to keep coming back to. Like what have you actually used? What have you used AI for that actually made your day better? Not just faster. I've built countdown timers for things that I'm excited for, but yeah, I hear your pushback. So this countdown timer idea is like when I'm really excited about something, I'll make myself a landing page that has a URL that I can click on and go to and get the unboxing of this joyful thing that I know is coming up, um, kind of represented to me. I think that's quite a nice, um, extra added thing. And I'll push back on the logistics thing by saying that my ability to it's creating more space. It's not that it like, it's not that it just makes it faster. It's that I can off board some of that load. Yeah. I get it. But yeah, I get it. But I do think it's like a really nice question to keep coming back to in terms of like, what it, what, like, what is the, what is the actual output of this in term? And it obviously depends on like how you define better as well. You know, one of the use cases I would probably point to personally is I think it's improved my relationship to my own health. But I have noticed that some of the newer labs have reduced my ability to, um, input certain, uh, um, health based questions and get, you know, decent answers. So that's where they are actually. Um, cause I want to know like, um, what, what are the things that I could be doing every day to give me more energy? And then what sequence should I be taking them in? Cause a lot of these things can be quite nuanced, right? Um, and then providing my feedback based on my experience and then that being refined, refined, refined. Um, and what that would cost in terms of going to a doctor and going through, you know, all of these kinds of tests is a very long process. So I'm finding that probably the best use case for improving my life daily. And I'm not going to say like it's a longevity stack or anything, but it's certainly one that I think has enlivened me more in many ways and given my ability to, to, to stay a little bit more switched on and cognitive and, and, uh, focused. Uh, that's, that's certainly helped. Um, I would say the other thing that, um, shouldn't be discounted, but, um, AI has really enlivened my curiosity in, in a way that I wasn't expecting. I'm like so curious about so many things. Finally, I have a tool that I can like really quickly. Uh, and I know it's a fast thing, but I, it can, it can kind of, um, uh, entertain my curiosity in any direction. And that has been at least for my spirit, hugely, uh, beneficial. Um, and there's like a spring in the step every day now is just so much possibility. Um, so in terms of curiosity, I feel like that's a really beautiful piece. And actually I did listen to a podcast with, um, the founder of perplexity and his whole thing was, yeah, like just keep encouraging people to be curious because curious as a kind of primal function of being human is so, so powerful. And the moment you stop being curious, um, is a pretty kind of sad day in many ways. Yeah. So health and curiosity, those would be my two, uh, for that, for that particular answer. Me, that is the perfect bow on the end of this is, uh, stay curious.