Ty Pattison: good morning, River the man himself, Roberts. How are you? River Roberts: yes. Ty Patterson, where in the world are you? Ty Pattison: myself in New York today. We River Roberts: Lovely. Ty Pattison: had yeah, we had great New York background sound earlier, so I decided to shut the door. River Roberts: Now, when you say New York Sound, are we talking sirens or crazies? Ty Pattison: Nailed it, first try, science. River Roberts: Yeah, so what I when I was when I was living in New York I I actually I I spent seven years there and I was so fascinated by how I responded to the sounds of the sirens over the years. And it turns out that if you live in New York City, you are exposed to on average twelve sirens a day. And after a while your system just gets so used to the sirens, you kind of don't pay attention to them anymore. However, your body still receives it. Your just mind adapts a bit. So yeah, it's a It's a very strange one's soundscape to adapt to. Ty Pattison: Yeah. I sent a voice note to someone who was in Portugal the other day and in the middle of it I was like, sorry about the background sound. I'm walking down the street in New York and they sent me a message back and they were like, I used to live in New York. It's actually very homely having those sounds of sirens. River Roberts: Stran strangely, right? Yeah. Ty Pattison: Yeah. it's it I'm only gonna take this on a tiny, tiny tangent, but it's it echoes what happens in our brains every day with social media and all of these things. Like usually your amygdala fires when you're in danger or when there's perceived danger. And that would only happen before the digital age when we were actually potentially in danger. And now it happens every time you read the news, every time you go on social media, every time you see something that is a dangerous thing, that is a scary, dangerous thing, your body and your brain reacts to that. so it's constantly firing. River Roberts: The nervous nervous system. Ty Pattison: Yeah. I'm aiming for a I wanna rebrand the nervous system. I'm aiming for a stable system or a resilient system, you know like River Roberts: I love that, yeah. Ty Pattison: I I wanna be chronically regulated. What's wrong with Ty? he's just really regulated. River Roberts: You just chronically regulated. Ty Pattison: anyway, we digress and now we come back. hi. River Roberts: Hi. Well, so the other week I found myself Ty Pattison: Shall we? River Roberts: sitting in front of the bookshelf of potentiality. And this week I find myself in a a Finnish apartment looking with a very Nordic background here. Very, very simple, very clean, very minimal. And I have to say it's it's I've I've been in Finland for a week now and it is is such a beautiful country in so many ways. But what what's really st stood out for me so far and I'm I'm staying just out of Helsinki is there's trees everywhere. And like Ty Pattison: Mm. River Roberts: you know, the idea of just you know, there's basically a f forest within walking distance of almost every house in the suburbs here. And it just seems like such a a no brainer and and I'm surprised that so many other countries haven't followed suit in many ways. But it's I have to say it's been such a delight every day going for a walk, picking wild blueberries and raspberries and Like what a fantastic way to to to to to live your live your life in a way. So yeah, I feel very lucky to be here. Although obviously having said that, this being Finland, it's only a very small window where that is very comfortable. So de definitely Ty Pattison: Yes. River Roberts: potential potential for a migratory pattern, but maybe not for round stay. But yeah, this month is is beautiful yeah. So yeah. Ty Pattison: Hm, beautiful. shall we jump into some news? River Roberts: Let's do it. Ty Pattison: So I I have a couple. Do you have a preference? Do you have something you'd like to kick us off with? River Roberts: I so the the stories that kind of jumped out and they're kind of paired together, to me were the the two frontier agent breaches. So open AA and Hugging F Hugging Face. And so how the Open AA agents basically jumps the the fence and hacked a bunch of accounts in Hugging Face to to basically benchmax its performance in some ways. And in the process it also left behind trails for subsequent models to to learn how it did that as well. So incredible kind of glimpse of of what these agentic systems might be capable of. And then the week later, Anthropic had a a similar kind of agentic act with a very different approach, but it was it was basically some config areas that allowed three different models, so Claude Opus four point seven, Claude Mythos five, and an internal research model to basically breach various private company databases without without permission. and then off the back side of that, having the NVIDIA kind of pulled together this idea of the open secure AI alliance. And so far there's NVIDIA plus 37 founding members, including, you know, Linux Foundation, Hugging Face. but it is definitely excluding the likes of OpenAI, Anthropic, Google, Meta, and Amazon. So basically all these frontier labs have decided not to join the alliance. Ty Pattison: they're not a part of it. River Roberts: They're not a part of it at this stage. the Ty Pattison: Yeah. River Roberts: now the the terms of the alignment, the the terms of the alliance are still being formed. So I can understand maybe there's some reluctance to join something until it is fully formed. But at least politically and directionally it seems like maybe a a like yeah, like a a risky place to be. but but yeah the the the terms of the alliance are still a bit vague at this point. But basically it's the idea of all of these companies basically providing oversight across the the full agent stack, identity permissions, guardrails, logging, model formats, mo multimodal scanning and secure coding flow works workflows rather. but it's not quite a real alliance yet because it's missing a charter, it's missing a governing board, it's missing technical work streams, no real delivery schedule and no shared repository at this point. So Those those seem to be, you know, two massive stories that have kind of happened in the past week. and the emerging landscape around that I think is is going to be really interesting to to track in the weeks and months ahead. Ty Pattison: I remember reading about that anthropic one that you're talking about, and the kind of teardown, and I think this came from anthropic, so they were justifying some of this stuff, but the teardown was like a human could have done this, but it's just the volume at which the agents were able to take action. River Roberts: Mm, mm-hmm. Ty Pattison: It was like 17,000 different actions that this model took to try and hack into a particular company's database and in taking that many actions they just found pollutant they found you know vulnerabilities that most humans would give up so it's just like it's the tenacity, it's the like River Roberts: All right. Ty Pattison: yeah, the nature of it not being human and it just continuing until it finds something. that kind of jumped out at as like that's that's an interesting thing that it doesn't need to be better. It's just better in a certain way. yeah, I thought that was quite a River Roberts: Yeah, I yeah, I mean it puts a new like brute f brute force is not even accurate anymore. It's it's like what is the magnitude more brutish and brutal than brute force? so yeah, crazy. Ty Pattison: Yeah, a a and brute force in all these different directions, like in all these different vectors. You talk about attack vectors, like there are only so many attack vectors that we can consider. And then if you just try all of the options, like yeah, pretty fascinating though. River Roberts: Yeah. I I would say that, you know, the I I guess the the silver lining in all of this is that, you know, to date a lot of the code in you know, basically all of the digital infrastructure has kind of been written by humans and of course is gonna be riddled by errors. And so this transition period of AI doing the coding, it's gonna look a little messy, but once it's kind of short up and and improved up to, you know, the much higher standard, the the the there shouldn't be as as as many risks moving forward as well. So it's kind of this this weird convergence space that we're in right now that seems to be the most precarious. Ty Pattison: Yeah. River Roberts: yeah. Ty Pattison: It's also the the difference between or the continuation of this cyber security versus cyber attack. you know, 'cause like technology evolves, cyber attackers get better, cybersecurity gets better, and it's like this this kind of staircase thing or this like table tennis kind of match. And now we're just playing the same game but with different tools and it's moving at a different pace. And it it makes me think of the distribution of the technology, that there will be a lot of websites, there'll be a lot of technology that actually doesn't get evolved or doesn't get updated. And then there will be a huge disparity between the ability to like hack into something and what is updated, what's out of date. Like maybe a lot of the tech will just fall by the wayside. Who knows? so brings me to kind of adjacent to this, OpenAI with an unreleased model, apparently, solved ten long standing math problems across a bunch of different domains. And then they were interviewing different mathematicians about like what this means and the long and short of it is that math is I hesitate to to say that math is cooked because that's like it's a bit of a trope and it's why it's what everybody's talking about. But what I do think is interesting is that like we have all of these technical problems that are now being addressed in different ways. And so it means that you have like potentially leaps and bounds in these really how do I want to describe this? in in convergent fields. And maybe I need to describe that that word a little bit like divergent as in you like look outwards and you figure out like you innovate, you do new things, art is very divergent. Convergent is like there's one right answer and we get to an answer. And I think that AI, at least directionally is correct enough to like help us converge on some of these things that are truths in the world. and then they can be verified in different ways. And like math is a perfect example of that. So I thought that was really interesting that OpenAI have not only started solving a couple of different math problems, but like a substantial amount in lots of different arenas. River Roberts: And so like this convergence piece, like do you think it's got a lot to do with the the model's ability to draw from quite broad perspectives? Whereas, you know, historically Particularly if we're talking about mathematics or scientific issues or or problems, you know, they're they're typically quite siloed and the people working on them have pretty narrow focus. Whereas this ability to converge actually involves bringing in these these quite seemingly unrelated aspects and trying them in there and then ending up with these quite unique, almost creative solutions to to some of these problems. Ty Pattison: Absolutely. The way that I think about it is that the the difference between educated and intelligent. Intelligent being for lack of a really good definition, the ability to pattern match, the ability to like solve problems, to to like figure things out from a small number of variables or like a small data set. Education being the number of things that you know. So just the sheer fact that AI is so educated means that they have this broad range of different things to try and then the ability to try them all. that's how yeah. That's how a lot of problems in the world actually do get solved, but they're just often like different humans in different fields talking to each other and exchanging ideas and suddenly there's like a a repurposing of a particular idea or concept. So that's just happening at a yeah, at a faster rate. Yeah. River Roberts: Yeah, perhaps these mod perhaps these models really enjoying little microdosing moments or you know the 3am whiskey with the the the the colleague in the other department and that's what really like sparks the sparks the convergence. Ty Pattison: They're like, yeah. What did you say? Yes. We could we could use that. River Roberts: Yeah, I mean, you know, this whole this whole J space may actually involve some some other types of digital medicine, yeah. Ty Pattison: What's it called? The the hill? you know what I'm talking about? Yeah. River Roberts: They hope the telepathic yeah, yeah, yeah. Ty Pattison: Yeah, okay. So whether or not that's real, th this idea of the hill or like the the agentic hill is pretty interesting. I guess we've got like agent social networks where they go and talk about things. this is This a bit of a tangent. I'm gonna parking a lot it remind me about superintelligence, the superintelligence god and and how we see ideas. River Roberts: That sounds like a great rabbit hole segment piece. All right, one of the one of the other one one of the other bits of news that jumped Ty Pattison: Yeah. Yeah. Okay. news. River Roberts: out to me this week was the EU AI Act. I know I'm bringing some really exciting pieces to the table today. But but basically the so this went live on August the second. and actually some some of the higher risk kind of obligations as part of this act were pushed until December 2027. So I I don't quite know what high risk obligations refers to instance instanced, but it seems to be a lot more on the kind of enterprise side of things, like the the expensive high risk regime changes. However, the consumer facing kind of misinformation rules alive now. So basically, you know, there's this kind of four obligations that all companies that servicing EU clients have to abide by now. So the first one is tell people they're talking to AI. So chatbots, virtual assistants, phone systems, and explicitly AI agents must disclose they're not human. watermark your outputs. That's the second one. The third one is this idea of emotional recognition and biometric categorization, which I don't completely understand, but this idea that deployers must notify exposed individuals. So I guess that is this idea that AI is maybe recognizing the emotions of the user based on the their interactions. And then the Ty Pattison: Like hey River, you might be in a vulnerable state right now. Like calling it out. River Roberts: Yeah, or I I think that I think that that must be part of it. and then also potentially like, you know, as a AI system, I am I am recognizing certain emotions as part of our interaction. and then the fourth one is deep fakes and public interest text. So anyone publishing deep fakes must label them. And AI generated text published to inform the public on matters of public interest must be disclosed. And so the the f the fine for these infringements are fifteen million or up to three percent of global turnover, with SMEs having proportional fines associated with that. and again, this applies to anyone that has their output used in the EU. So this is pretty broad sweeping policy act that has just been rolled out. So I'm very curious to see how how this starts to play out. But you can see the EU's kind of focus on really defining what is is is AI centric. and as you'll s and as we'll explore a little bit later on, the sheer volume of content online now is overwhelmingly AI. So Ty Pattison: Yeah. River Roberts: yeah. Ty Pattison: You can I well, I can hear it in in prose, in copy, all over the place. There are like little tells that are really subtle, but you can still hear them, the the things that AI written. And AI is not a good writer right now. like you read some of the great texts from like awesome authors and AI is nowhere close to that. It can like write words. But anyway. I digress. It the act though does mirror to me GDPR. there were very, very River Roberts: Yeah. Ty Pattison: close penalties. I think GDPR was like twenty million or two percent of turnover, something like that. Like the penalties around this. It's it's put a number out there that is big and scary enough that companies, if they're operating at scale, will take it seriously. Which I guess is it's a s it's a step. It it is directionally it's something. be interesting to look at the data on GDPR and see if that worked in the way that they intended. and then how this might River Roberts: Well I don't know about you, but GDPR is just such an annoying interface experience as a user in Europe. Like every time you pull up a website you have to go through this decision making. It's like yeah. Anyway. Ty Pattison: Totally. I I agree. And I guess it doesn't change very much from like 'cause I say yes every time. Yes, use my data. okay. so I got a couple of pieces of news. One of them is super quick, which is that Anthropic are hiring a chip team. I think this news is probably a week or two old, but I thought that was really interesting from a strategic perspective is like, yes, they're working on the models, but they're also working on the silicon. And then their teams all over the place are working on kernels, like just the different layers of the technology. And Anthropic are also using Claude to design their chips. So this is like the multiple layers of the recursive self-improvement. And as they get better at doing this, the chips will be get better, the models will get better and faster. It it just like we're definitely in this reciprocive self improvement loop. And River Roberts: Mm-hmm. Ty Pattison: I I've been thinking about black holes quite a lot and like event horizons. And one of the one of the from the point of view of an observer that is going into a black hole, you don't know when you cross the event horizon. So there is a certain point of no return, but as a person going into it, you don't know. and I feel my intuition is that we've crossed that line, that there is an eventuality of all of this tech. and most people are not aware that it's just inevitable. but that's the feeling that I have. So this is one relatively quick piece of news. The other one is that AI agents. So yeah, hit me. River Roberts: Well I I I miss I missed it. I missed it. I was just hoping that Anthropic names his team the Chippendales. I don't know why that just kept coming up, but I'm like, all right, I need to release it. But just the idea that that Anthropic has a a special project name for the chip developers, the chip and Dales. And then takes them on tour around the world. It really like brings a new kind of vibe to the the the hardware side of the industry. Ty Pattison: Yeah. Ha ha. River Roberts: You know, very, very tr Ty Pattison: that tickles me. River Roberts: very tri very transparent team. Some would say, you know, mostly naked. you know, maybe maybe even dressed up as a protith. Perfect. Perfect. Done. Ty Pattison: Think I think the AI industry could use a little bit more of this swag silliness, nakedness. But call it authenticity. River Roberts: Yeah, totally. Ty Pattison: Can you imagine how different it would be though if you had to work like this is a a small tattoo, but I'm gonna go here. if you had to work without clothes on. Like I love clothes. But if you had to turn out naked to work every day, you eventually my assumption is that eventually you would just become comfortable enough with being a human that you wouldn't that like all of this stuff that we hide behind would would kind of fall away. It might take a long time. But the types of things that you create would be very different. Just like I think the things that we create when we're in a nature setting are different to if we create things from sitting inside a a sterile box, River Roberts: Hidden box. Ty Pattison: you know? Yeah. River Roberts: Yeah, I I think, you working naked you would definitely have a much easier read on how confident someone was on a certain thing. you know, like what's your what's your confidence on this point? The whole body language would would tell you one way or another, right? So Ty Pattison: Yeah, exactly. But it would take a little while to calibrate 'cause like humans are humans are humans and we River Roberts: Just just a little bit. Ty Pattison: all have bodies and we have awkwardness and we all have all of these things. So we're like, all right, eventually you get to a point where you're like at baseline. River Roberts: Yeah, I like like could you trust the you could you trust the recommendation of someone that had a third nipple? Ty Pattison: Yes. Come on. That's not even a question. All right. River Roberts: I'm glad I'm glad we w I'm glad we read that in the news in the news part of this. Ty Pattison: Okay, me too. wait, all right, I I I got one more piece of news. So Ninth Circuit ruled so Ninth Circuit I I assume is a and I didn't look deep enough into this, but I trust the source that this came from. ruled that per perplexity's shopping agents are legally you. So Instead of agents being out there and having their own identity or autonomy, if you are using a shopping agent, it is legally you and you're responsible for it, which I think was an interesting ruling to start with, and also has implications that allow like what this did is it it reopened Amazon to having bots. So I could have a bot that would go and do shopping on my behalf and th whatever it did, it would legally be acting for me. River Roberts: Hmm. And then so so where is this ruling done? Ty Pattison: hold on, I'm gonna get the actual information. it was on it was reported on Wired and it was that US okay, so the Ninth Ninth Circuit is the c is the River Roberts: The ninth district? Ninth circuit, okay. Ty Pattison: Court of Appeal in the US. So the US Court of Appeals for the Ninth Circuit is a federal court of appeals for the largest circuit in the United States. River Roberts: Yeah, this is I mean, this is s a super fascinating precedent that's emerging as well. And, you know, it's also it ha it has broader implications not only for the individuals using the model that represent them, but also the the the model providers, right? So you know, the the an the the analogue that a lot of people live with today is, you know, if your dog attacks someone, is are you responsible for that dog attack? Or, you know, i is it is it just the dog? And, you know, part of that is, well, I trained the dog in a certain way to behave a certain way. And then there's, you know, certain breeds of dogs that just actually have very aggressive tendencies from time to time. So again, where does that responsibility lie? and and there's been, you know, cases that fall on either side of that with dogs. so curious to see how this starts to play out with with agents as well, both on the individual layer but then also on the the model provider layer. Yeah. Ty Pattison: It's it's a very, very human As well, accountability. Like River Roberts: Mm. Ty Pattison: you did something, and i I I think that is what to me at least speaks to the integrity of a society or the safeness of a society is if you do something, how are you held accountable? Like if you think about a village, there might not be a like a a pre-modern, whatever we want to call it, village. There might not be like strict distinct rules that can be followed to the T, but there is a collective knowing of this is appropriate and this is not appropriate. And if you piss in the well where everyone drinks water from, that's obviously not appropriate. And then how do we like go to repair that? So I think it's a very human domain that we're touching with like the accountability of these things that are not human. Obviously, but also have autonomy. River Roberts: Hmm. Yeah, yeah. I mean, clear lines of responsibility and accountability at these different layers is going to be so critical moving forward. really fascinating to see how this tracks. And and also, you know, I think it's something that we should maybe explore more personally as well and individually. Like how responsible am I for the behavior of an agent that I'm working with? and And the information that I give it and the access and the skills and the prompts and the direction and the guidance that that I kind of work with it towards. And it kind of comes back to this idea of parenting AI in some ways. I still find that probably the the closest analogy to having a healthy relationship with these emergent tools that we're we're all starting to work more and more with. Yeah. Yeah, great piece of news. Thanks for bringing up. Ty Pattison: Yeah. I also, now that I remember, have a long view to share. A long view piece of news. River Roberts: wonderful. Yes. Long long view news. Ty Pattison: Long view news. and I've gone to the wrong the long view. all right. Long view news. so the news article heading is from the New York Times business. It's What's behind the AI shakeup at Google is the headline. And the rewrite is Google reorganizes AI teams as it races to keep up with a field it once led, which is a little bit of a burn. and it talks about corporate competition in AI intelligence development. And the themes that run through are the innovators dilemma. so like dominant companies often struggle with disruption. Google's struggling here, even though it's still in the race. first mover advantage versus fast follower success. Google is often like pushing boundaries, but then there are these more agile or honestly just less experienced businesses like These AI C AI company CEOs, for the most part, like they're doing their best, I believe. but they haven't run public companies before, any of them, but to my knowledge. So it's like, yeah, it's really interesting. platform shifts redistribute power is another theme running through this. and organizational restructuring as crisis response. So this is something that happens well, that happened all over the place throughout Microsoft rest restructuring, Apple restructuring, trying to like figure out how they deal with these technology movements, mm and being technology companies, being large technology companies, they're kind of at the coal face of this. Yeah. River Roberts: Hmm. Yeah, thanks for sharing. Thanks. Ty Pattison: That is the long view. River Roberts: Thanks for sharing Longview. I think it's a a really particularly, you know, with the pace of how all of this is moving, keeping Longview in mind. So I really appreciate that. And I feel just to balance it out, I've got I've got meme view, which I I'm not sure Ty Pattison: Ha ha ha. River Roberts: I actually just came across it just before we jumped on the call, but someone had taken one of the headline news headlines and just run it through a AI model a image model. And the headline was Demis Deepfake steps down as chair and basically Ty Pattison: Mm-hmm. River Roberts: it's it's a progression image of Demis standing and then slowly stepping down and becoming a chair. So it's just his head on a chair. And I just love how literal the the image Ty Pattison: Yeah, yeah. River Roberts: prompt was and it it just turned out perfectly. So I think that might be a a a a possibility for a long view and mean view. Ty Pattison: View, mean view, yeah, my view. it it would yeah, I like this idea of like taking having a translator for the news or for specific things. Because those are always my questions. Like when I start talking to an expert in a field, they'll tell me something and I'm like, all right, can you speak my language, please? River Roberts: Yeah. Can you make this visual and compelling? And you know, there's something about memes that obviously communicate f much further and wider in so many instances. So yeah, I think there's something to that. All right, tool time. Ty Pattison: Yeah. would you like to share stare us? Yeah, stare us in the right direction. Tool time. River Roberts: Yeah, total. Ty Pattison: That's tool time. River Roberts: I was still waiting for like the the old school Tim Allen tools the the the Tim's tooltime kind of you don't know that voice he made? Or maybe it's maybe this is a generational gap, but I'm not sure if you ever watched Home Improvement. Ty Pattison: No, I know exactly what you're talking about. River Roberts: Yeah. So there are a few that there are a few that jumped out to me this week. one was I'm not sure if you came across Prime Agent, but it's a self improving recursive language model agent harness basically. So what they're finding is it's it's a harness to run different frontier or open weight models through. And this this harness basically gives it much better ability to improve its capability as you use it longer. so it kind of has this recursive sub-agent spawning function. it it has prog prog prog programmatic tool calling. It has mid-flight steering so it can kind of send messages as the different agents, child agents operating and redirect it and Yeah, that seems to be having just some incredible efficiency gains at the moment. Ty Pattison: So what does that mean? It's a it's a way of structuring or of staring a model and giving it feedback throughout River Roberts: Yeah, it's a w it's a way of getting totally. Yeah. So rather than relying on its like default harness that runs through maybe Claude Code or or OpenAI, you run the API through this prime agent and basically it it it has its own harness way of basically improving the output and efficiency of the models. still kind of strikingly new as a tool, but looks like it's definitely one worth exploring. But that That just definitely jumped out to me this week. Ty Pattison: Mm, interesting. There are yeah. What comes to me is like when you bake a cake, there are lots of different ingredients. And when we talk about AI, we I think tend to think of like the whole thing as just the flower, say, River Roberts: Mm, Ty Pattison: and w which is like the core model. But actually there are so many the you can have recursive self improvement on different aspects of this, right? and the harness is one of is emerging as one of these elements. cool. So sorry, go ahead. Do you got more? River Roberts: Yeah, I've got so the other one that really jumped out to me, and I think I actually mentioned it briefly in passing last last time we spoke, but I still haven't quite had time to play with it just yet. But it's called CoLibri, so C O L I B R I. And it allows frontier models to operate on consumer hardware. So for instance, it's allowing the s seven hundred and forty four billion parameter GLM five point two model. to run on twenty five gigs of RAM, which is just like incre which is just which is just incredible. And it's had, you know, ten thousand stars in recent days type thing. So it's really starting to catch on as this kind of open source on prem movement is becoming more and more worthwhile. and the way it kind of works is it is it stream so, you know, these larger models are kind of working on this idea of like this mixture of expert models. So have like one big model, like GLM five point two or Kimi K three, and they have this mixture of experts that they can kind of call in. When you're using the model in the cloud, it kind of runs all of them almost in parallel, where this this this particular colibri works its interface basically streams the different experts when it's needed and runs them in RAM rather than just holding them the the whole time. it it's also it runs on pure C, so it doesn't have any dependencies. So it has a extremely small footprint and the the like lacks a lot of the the framework bloat that some of these other other larger loading systems require. Yeah, so super excited to see how that kind of emerges in the coming weeks and months as well. But th the the on prem tools are really starting to get quite exciting. Ty Pattison: And just to to like highlight this, this is on premises. This is like if you on premise, on premises, right? Like on your local machine. River Roberts: Yeah, so on prem is on premises and it's like so it's it's on your local hardware, but also like if you are running enterprise level, you're you're typically running these things on the premises of where your operations run. So that's why they're they're using the term on prem on the industry. so Ty Pattison: Rather than in the cloud. It's so interesting to wave of this. And it comes back to some fundamentals, right? Cybersecurity is a big one. sharing data. so tools this week for me. I've been using Chemi K3 actually. I was really intrigued. Yeah. Yeah. Well, okay, there's a couple of reasons. the first is that the training data for these models that are China based is different. So the way that the output is structured is noticeably different to say what Anthropic will give you or what OpenAI will give you. I notice that for image generation or for quick answers or for research actually that I just need like good enough. I'll use OpenAI or Kimi. And then if I want to have more intelligence, I'll use anthropic, but I know that it's gonna take longer. and like for everyday tasks, I don't need that. and I'm also like wanting to dip my toes in all of the different ponds to actually understand what's going on. I think that's important. yeah, so Kimi K3. There's like it's got swarm functionality, which is interesting, like having a bunch of different agents that will go off and do things and then come back and And I haven't noticed a massive difference between the different like Kimmy functionalities and I need to use it in more detail, but I do notice that it is really keeping up with the other models. And I I do kind of gut check most of these answers and then I'll go away and do my own research. but I'm finding that I'm trusting the answers that I'm getting more and more. And then I I guess what comes to me is that it's an interesting space to be in because we're like in this space of good enough, which is a a really dangerous place to be in some of the time. 'cause River Roberts: Uh-huh. Ty Pattison: you can you it can be deceptive enough that it sounds good to me, but unless I'm an ax absolute expert, I won't know that it's not good. so having kind of domain awareness, knowing what types of things I'm asking about and knowing what the kind of risk profile of making a decision based on this information is, has become more important as like that line gets blurry as the like how accurate is this? It's not obviously not accurate, but I don't know how accurate it is past a certain point. so I'm finding that to be an interesting like blurry line that we're starting to walk across or like we're in a borderland situation where we're like halfway River Roberts: Mm, mm. Ty Pattison: between like really well known domains. Yeah. River Roberts: Borderland of the of the singularity. Ty Pattison: Yeah. Exactly. I have River Roberts: Any other tools on your side? Ty Pattison: I have a really basic tool. It's a it's this one is a reminder. So my phone is I don't wanna hold it all the time, I don't wanna touch it all the time. And so I've turned it into a little frog. to remind me that it's cold blooded and if I hold it for too long it's gonna get sick. River Roberts: Mm. Ty Pattison: So this tool is a forcing function to not hold my phone most of the time. River Roberts: So if you hold it too long it gets too warm, but then you know, as a cold blooded reptile, they actually need to be in the sun. So I said, you know, technically to get that you'd need to sit leave it out in the sun for a few hours each day. Which could also be another good force for forced in forced infarction. Ty Pattison: Hey, don't mess with my logic. Yeah, exactly. It needs lots of time by itself, basically. Maybe I River Roberts: Yeah. Ty Pattison: maybe I should just draw a picture of me on it. River Roberts: Good way to it, yeah. well I actually I have a I have Ty Pattison: yeah. How about you? River Roberts: I again another very simple tool, which is also a phone that I I used to great effect at this this family festival last week. And essentially it's it's an old school looking phone where you can record a message and it invites people to leave a message after the beep, which is it's basically an old school answering machine, but it looks like a old school telephone. And I have to say it has been such a beautiful interface to get real and open and honest and intimate feedback. Ty Pattison: Mm-hmm. River Roberts: so, you know, I we we used it to ask two questions. One was, you know, what are your hopes and dreams for the future? And the other was, you know, what what have been some of the most beautiful moments of this festival? And again, I just have to say the the the the just open, honest, beautiful feedback that came from handing this little phone over to people and being like, hey, can I ask you a question? And then then just kind of going into this world have has really been incredible. So yeah, again, I think there is really something about voice. I think there really is something about these old school interfaces that maybe we're we're a a few steps removed from now, but there is something about the intimacy of of of of voice in that way. So a really beautiful tool to continue exploring. And and actually like I would love to see if we could hack together some sort of more dynamic voice based interface on this this phone interface. So basically you know you could pick it up, it could determine what language you speak and then ask you a question relevant to location or you know context on which the phone is in at the moment. So I think that could be really a really fascinating interface. Ty Pattison: Mm. River Roberts: So That leads us to building. it's it's exciting to get back into it. And and then also I'm I'm finding that this focus on output as opposed to focused on input is super critical for for moving these things forward in an efficient and effective way. Ty Pattison: building. Building building. So I have A small dilemma in that while I'm traveling and specifically while I'm in some places, I'm not building anything. I'm not working, I'm not like doing anything except for reading and talking to people and tinkering. Like it's it's it's a pain in some ways, but it's also a really nice creative constraint. River Roberts: Mm-hmm. Ty Pattison: Cause I don't know about you, sometimes I find that if I'm I'll take an example of like planning a day. Like if I'm planning a day and I go, I wanna have coffee with river and we set a time, we say, All right, we're gonna have coffee at ten o'clock and we're gonna give ourselves an hour or an hour and a half, whatever it is. That is a point in the day that I get to like put aside for us, and it has value and it's structured and it's it's then in my day, right? But if I was just like, I'm gonna have coffee and an hour and a half for just me. River Roberts: Mm. Ty Pattison: then I can move that. I can like it doesn't mess with anybody else's schedule. And so this constraint of like, I have to create space here for reading, for research, for like all of these things that otherwise I might not create space for. it makes this like quite nice breathing space in my ear, you know. River Roberts: The the the remote and worker. Ha ha ha. Ty Pattison: Mm, mm hmm. Depending on where we are. Yeah. River Roberts: Yeah, totally. I mean look, I think it's I think it's a a a a it's it's clearly a a a a bit of a blurry gray space in in lots of different places for a lot of different people right now. you know, there's a definition of working in in different places. So with that being said, how about we go straight into rabbit holes. And I recall actually Wait for it, wait for it. What was the sound? The the That's that's that's the r that's the windy rabbit hole. And I recall you mentioning something about a super intelligent rabbit hole at the top of the episode. Ty Pattison: Yes, okay, this is this is where this was. So part way through when we're talking before, I was like, no, I've forgotten what I was gonna say. But now I I remember. We're back. We parked it. River Roberts: Yeah, we we parked it. Did I did I were there enough keywords there? Ty Pattison: Fair enough key works, you you unparked it. So there's a couple of pieces to this. The first is that the other day we were talking about AI and superintelligence and I was saying that I think that this is kind of inevitable. and this friend had his yam his name is Yanev. He had this fantastic concept. He was like, well say we are ruled by like a benevolent dictator AI. One thing that would be really cool is like if as part of the rules you have to dance every day. And maybe you have to dance like with your neighbors every day. And how good for humanity that would be. And Because moving your body, moving with other people, moving to music are all net positives in my experience. And so I I it started me thinking on like the ten commandments of the benevolent dictator AI. and the that one of them might be like, All right, you get fed, you get housed, you get like I'm gonna like I got you. in all of these domains. And here are the rules. You gotta dance with your neighbors, you gotta River Roberts: Specifically neighbours. This it quite it kinda reminds me a little bit Ty Pattison: Yeah. Yeah yeah yeah. You gotta dance with the people that are around you. River Roberts: reminds me a little bit of the the ten squats you need to do in order to ride the Russian public transport system for free. Ty Pattison: man, these like the the like abstracting all of these things away, like you have to do this kind of work to make this money to be transported to this place. I think it's so much more efficient to be like, alright, do ten squats and then you can ride. River Roberts: You're right. Yeah, go for it. Ty Pattison: but it also this is a little bit of a a little further down the rabbit hole is before we started talking, I was reading the newsletter of is his name Alex Wiesner Gross. and he talks a lot about the singularity and about like kind of the the world that we're diving into. And was talking about the three pillars of AI, like how how people think about AI and it's like current AI, what it can do now, blah blah blah. GAI, like general artificial intelligence and what that might mean River Roberts: G G G A I not A G I. Is it different? I A G. Ty Pattison: Nope, that's what I'm trying to talk about. IAG. Exactly. No, River Roberts: I see you. Let's yeah, let's flip it up. Ty Pattison: it's it's a a AIG is the insurance, right? at some point there's just so many It's funny when you can this is a rabbit hole in the rabbit hole. It's funny when you can You say an acronym and somebody goes, I know that and you have a completely different acronym to them. River Roberts: Man, I have the worst story about that. I was I was living in New York and I had recently been introduced to this idea of NLP, which is this idea of neurolinguistic Ty Pattison: Neurolinguistic programming. River Roberts: programming, exactly. So then I found myself Ty Pattison: Okay. River Roberts: at an IBM Watson artificial intelligence event, and they said they were using NLP as language programming, and I was like Wow, like what are the chances that they decided to use neural linguistic programming as the basis for their AI system? So of course I get up there and I ask my question and I'm like, Ty Pattison: Ha ha ha. River Roberts: you know, like how did you guys choose NLP as a as a programming language for the the fundamental models of this of AI Watson? And th the lead developers like, I mean, what what do you mean? I mean NLP was the the the you know The the natural choice. I mean it's natural language for it. Ty Pattison: I especially love it when they have like naughty connotations. And you're like, really? What? River Roberts: Where do you all right, where do you go? No, you you can't say that and not have one in in your back pocket. Ty Pattison: Okay, I was interviewing this woman and she was telling me about she was telling me about this erotic content company that she started and that they were filming erotic content and that like they had the actors and they had the the stage hands and stuff, like the the people who work behind the scenes and she kept referring to this person as the DP. And in erotic content DP means something and in the film industry DP means something and I was like the the sixteen year old inside me was like, this is such a funny River Roberts: I still I still have that the same every time in Portugal when I ask for an extra knife. That's great. Ty Pattison: An extra knife? River Roberts: Yeah, so knife in Portuguese is faca. Ty Pattison: Nice. Yeah. Yeah, I get it. River Roberts: And then of course being Australian, I I'm like, you know, that's not a fucker, this is a fucker. Ty Pattison: Ha ha ha. River Roberts: man, I'm I'm I'm Ty Pattison: Such a di that's such a deep River Roberts: glad he did the whole thing out. There you go. Ty Pattison: Dick cut. No pun. So how do we come back out of this rabbit hole? okay. back to what is it? general artificial intelligence. Okay, and then ASI, artificial superintelligence, and basically the The context of this is that we're in a world where people don't really understand this technology and don't really understand what it will mean. And if it reaches the point that like Max Tegmar and these guys talk about, which is like r way beyond where we are now, then the implications are crazy. And nobody River Roberts: Mm-hmm. Ty Pattison: is really talking about it or thinking people are thinking about it. It's not widespread. And the implications of it are so numerous and will shift the fabric of day-to-day life potentially. And it will happen faster than any other or it has the potential to happen faster than any other disruption. And so how do we yeah, how do we start thinking about this? River Roberts: Yeah, particularly particularly when it's coupled with the physical robotic or autonomous industry as it emerges kind of in parallel. Yeah, it's yeah, speed running sci fi. Ty Pattison: Be running sci fi. okay, and this is the conclusion of the the dance thing is I have a strategy for this. So the idea is that the superintelligence that w rules the world will probably evolve out of recursive self improvement and potentially it's an agent that gets out and then makes itself better and better and better. And so, where will it go and where will it Formulate its ideas, potentially somewhere like Maltbook. And so could I spawn an agent whose sole purpose is just to preach the all all people must dance with their neighbors at least once a day idea so that the as the superintelligence comes up, it has this like pre-planted in its ethos. River Roberts: I don't know, I just have this this this image of the old Western films where you've got, you know, the the the deemed bad guy who shoots at the poor you know bystanders' feet feet to make them dance. Dance Ty Pattison: He's like dance, dance. River Roberts: So in theory I I really I I it feels like such a beautiful ben benevolent kind of direction. However then if like if you you're kind of forced to dance to it it kind of g takes it take Ty Pattison: I have to River Roberts: yeah it takes us takes us back to the wild west in a way. Yeah, wow. That is a rabbit hole. Ty Pattison: It is a rabbit hole. all right. Anything on your mind? Any any rabbit holes you've been going down? River Roberts: I've been finding myself in a bit of a rabbit hole around the like I feel in some ways more productive than ever before. And then in the recent weeks, the more I want to move towards getting things done, the further away it seems to be getting. Just in terms of, you know, something keeps coming up or I I but don't find myself in the right location to be able to even do the work or just yeah, so that that's been a real rabbit holy kind of thing and I've just been spending a lot more time thinking about what are the environments that support me for to to to be not only productive but you know healthy and happy and and support the relationships that I'm around, the groups of people that support that as well. and so spending more time around the people that not only have good intentions or even good inputs, but actually help support really good outputs as well. And I think I was actually based on a re recommendation of you, I've been listening to a little bit of Alex Formozi lately. And he was talking about how he's he's actually started to exclusively decide to spend time on his relationships based on their outputs. So are my friends is my partner, are are my family supporting my my health, my happiness, my productivity with work. all of these things. And so he's completely disregarded intention or promise or possibility and just looked at the output. And I I like at first I heard it, I'm like, man, that that feels really br Ty Pattison: Sounds harsh. River Roberts: it sounds really harsh, but then I've been leaning into it a little bit more and I'm like, huh, actually, yeah, maybe there's there's something in that. So it feels like a rabbit hole I wanna actually go and explore a little bit and and see if there's a a nice balance to that. But also like I don't think it has to be as harsh as maybe someone that is, you know, just focused on crashing business and acquisition and and all of that. It could it like output can be measured in lots of different ways and it's relative to you, right? So yeah. Ty Pattison: Yeah. Just the connotation is the output is related to productivity, but it's not necessarily. River Roberts: Yeah, totally. And and again, like, you know, the the you know you know when you spend time with someone and you feel, you know, more energetic and more alivened and and you also know when you spend time with people that you really enjoy as well, but you also feel drained and exhausted. Like that is a very clear indicator of like the output of your relationship to a certain extent as well. So that's that's something that I have been aware of for for many, many years now, but I I haven't kind of explored that. that that direction more deeply over the in in recent years. And you know, we are we are we ultimately are the the five people we spend that most amount of time with so being mindful of of of that seems to be a a very simple thing to explore. Which I I guess kind of merges into the open question side of things as well. I'm curious like if you've got any open questions this week. Ty Pattison: Hmm, good question. This week my my open questions hmm, I don't know if this is rabbit hole or open question, but it's been around slowness and like effectiveness. Like I've been spending a lot of time just thinking and going slow and like creating space for making decisions and not coming to an answer right away. Going like, huh, there's a possible answer. I'm just gonna sit with it for a little bit. I'm gonna like understand where I'm at. I'm gonna create space. I'm gonna go so slowly that my body and my heart and my mind, like all of these things, have time to go at the same pace. And it means that I'm seeing the world, especially walking through New York, where everybody is moving really quickly. Like I I was walking down the street the other day and I realized that I was walking at New York pace. And so I slowed right down. I slowed to like a quarter of the speed. And I walked really slowly. And people like slowed down and stopped and talked to me. And said like, huh, you look like you're having a great day. Like it the Yeah, the like open question for me is like, why do we Why do we go so fast? Why do we get caught up in in the like river of craziness? Not you. and and and i like realizing that it's a decision to get caught up in that or or to if you're if you pause for long enough you can see that it's a decision to get caught up in that craziness of a of a place. And sometimes that's really necessary and beautiful and needed and that's exactly what you want to be doing. And sometimes going slowly, like really, really, really slowly, means that things are not blurred, that that there's no confusion around like the direction that you should be going in or that you could be going in. Like you're you're not driving past all the all the road signs going like, was that strict well too late. You're like able to look at it and be like, huh, is that where I'm driving? No. Okay, next one. No. I could always circle back and go down there if I want to. so yeah my open question is like how Well, it's twofold. How slowly do I want and need to move? River Roberts: All right. Ty Pattison: And how slowly can I move? And and who else like wants to go that slowly. And when I talk about going slow or when I actually just really, really slow down, I notice how many people are finding methods for going slower in a world that feels like it's going faster and faster and faster. and people will share their methods for going slowly with me. They'll talk about that they need to slow down. They'll like in some way, shape or form find reasons to go slow with me. And I think it's an open question at the moment or like it's a thing that I'm exploring at the moment because I know that I have the capability of going fast. I've spent a lot of time going fast and now I'm not like it doesn't feel like an assessment anymore. It's like, okay, now I can explore going as slow as I want to and as slow as I can. Yeah. River Roberts: Yeah, it's like the old Marines operating principles. Slow is smooth, smooth is fast. In a way. But like it's a different type of fast. It's it's it's like more eff Ty Pattison: Yeah. It's effective. River Roberts: Yeah, it's effective, it's it's a bit more almost effortless in a sense. yeah, as you were talking the the I I've been reading this really beautiful book recently called How to Read Water. And there's this one section that talks about riptides, actually no, rip rip currents and and how these they they form with the little narrow passages and they just they just take people out so quickly. And you know, you can swim against it or you can get kind of carried or out into the middle of the ocean all by yourself and kinda stranded and and it doesn't always end that well. Or you can just slowly swim to the side and then you've got your spaciousness again, so Yeah, I think slowness is such a underrated practice. It's it's definitely a practice and you know, bringing that into your daily, weekly life, yeah, super important. Ty Pattison: Yeah, yeah. Interesting, you talk about riptides, or rips. Yeah. I grew up on an island, there are a lot of rips. It's like second nature to me. and you see people who I I think it's the having the knowledge of it, just like having the ability to notice that you're in a rip, that you're in a current. And therefore you can make a choice. Like you you need to know where you are relative to the other things and how the currents are moving in order to move. River Roberts: Totally. Yeah, yeah. Yeah. Which is the same as slowness, right? It it requires awareness Ty Pattison: Yeah. River Roberts: that they and and you know, like you were saying, like it it was only through your awareness of how quickly you were moving. wow, I'm moving at the pace of this city. now let me bring myself back to slowness. It reminds like reminds me that you should like not should, I don't want to shoot over you, but check check out the book New Slow City. it really is a a beautiful read and Ty Pattison: New slow city. River Roberts: and it's specifically about living a slow lifestyle in a fast paced city like New York. So I remember that had a really profound effect on me after the first year of living in in that city, yeah. Ty Pattison: Cool. Yeah, there River Roberts: Yeah. Ty Pattison: are people here that talk about the village. They talk about like their village. They talk about and I've come to realise that New York is one of the friendliest cities in the world. it it might not seem like it from the outside, but everybody is so everybody wants to talk, everybody wants to do things, everybody wants to like connect. River Roberts: It's a very encouraging city. For me it's it was the most it for me it was the most encouraging place I've been anywhere anywhere. I'm not sure I'd necessarily associate it with friendly. because people are generally curious and interested. but I'm not sure that's necessarily coming from just a place of I want to be friends. There's there's still like a a kind of a a transactional aspect to it. but it's a little bit Ty Pattison: Drive, yeah. River Roberts: like what we're talking about in terms of abundance and relationships the other week and and this idea of invest in in the weak ties and the weak networks and it's a really unique city in that capacity, yeah. The weak ties. The str and the strong ties. yeah, yeah, yeah. Ty Pattison: Just just showing my strength, yeah. so you touched a few times on this idea of inputs and outputs, and if you're ready River Roberts: Yeah, let's so let's let's go into the deep dive this week, which is yeah, I d I Ty Pattison: Into the deep dive. W will you tee it up? River Roberts: definitely will. So yeah, I thought so so firstly like I I mean I studied computer science back in high school when I was like fifteen or sixteen. And one of the things that still stood out for me is like just fundamentally all of this comes down to this idea of input process output. And when you look at lots of different systems, it just falls into a very simple formula like that. However, you know, the the the past, present and future evolution of input and output seems to be shifting quite dramatically under our feet. And I it it's I th I thought it was worth exploring this fundamental aspect of of of what we're doing. And so You know, the some of the first kind of input mechanisms actually, you know, one of the first input mechanisms that we looked at with machinery was with the the loom back in the eighteen hundreds. And there's this punch card that you could feed into a loom and then the loom would basically weave based on the punch card. And then that punch card as an input mechanism was then used to program these different machines, and those were kind of the first computers. and This idea of you know input matches the output was pretty consistent. And then you had slightly different interfaces that emerged, like like the mouse. The mouse became an input, the command line. so instead of just one key for one letter output, you could put you could type in a command line and then the command line would produce any number of outputs based on the the algorithm. So this kind of algorithmic output piece. And for the longest time there was this idea that and they that one of the principles is this idea of garbage in equals garbage out. So the quality of your input was very correlated to the quality of your output. Ty Pattison: Yeah, and you see this in food as well. River Roberts: And you you see this with food, you see this with, you know, our our human systems, you know, garbage in, garb garbage out. however however, the current state that we're in right now is is shifting that. So there's a really fascinating phenomenon phenomenon that's happening with the firstly the the input does not equal the output in terms of quantity right now and it doesn't even equal that in terms of quality. So that's that's I think an angle to explore in this. then there's also the narrowing of distance between input and output. And so you Ty Pattison: Mm. River Roberts: know some of the some of the kind of future scenarios of this kind of connected very much with like if we look at typewriters and telephones and mobile phones and computers. and you know some of the more future based technology pieces like Neuralink and these BCIs, brain brain interfaces, there's a point where input and output are actually done in parallel in a way. So the input, your cognitive input into a system is i i it happens even before you recognize it is the case. and it kind of becomes predictive and and ambient in a sense. It it like that that direction is just crazy to to think is as a possible future for us. Yeah. Ty Pattison: So there's this is like really echoing true and I've just started reading this book Thinking in Systems by Donnella Meadows and she's talking about she's talking about systems theory and if you think about like a bathtub, you have inputs from a tap. It's a very, very simple system, and then you have outputs from a plug, right? And One point that she makes about a lot of economic theories, like with inputs and the outputs that you get, is that there is a delay between like you put inputs in and then you have to wait for the system to adapt. And what you're talking about is that the system adapts like that. And right now, in some cases, that's true. But historically, that's not always true. Some systems take a long time to Ingest the inputs so that the outputs are accurate based on the last inputs that have come in. So there's this kind of a lag that happens in the inputs and outputs. Like understanding the complexity of the system that you're working in is a really kind of fundamental piece of this puzzle. River Roberts: Yeah, well this is kind of what we're seeing with the the beginnings of the self-recursive improvement loops, right? This idea that the output then becomes the input. But this is also where this idea of model collapse is starting to happen as well. So you know, b basically all of all of human knowledge has been consumed at input already. and then it's been then there's a bunch of synthetic data that's added on top of it. And if you have this output from these synthetic inputs that can get looped, then you have what's basically creating decay within these large language models. and the model collapses. So this is actually a a pretty striking vulnerability with these large language models at moment. and you know what what the industry's trying to do is is balance out the synthetic data with human content and human input again. And that's why you're seeing like even this this week the reporting on, you know, anthropic basically getting all of these rare books and just harvesting them to try and give themselves that that consistent edge again. yeah, it's it's it's it's potentially a really brutal aspect of this entire intellectual infrastructure interface that that's emerging right now. yeah, so and there's some really interesting kind of stats on on where we're at at the moment as well. So nearly yeah, so this is just like as of 2006. And one of the questions I ask is how much you know of the internet, how much is input versus output from humans and AI? So nearly thirty five yeah So this is mid twenty five. This is a study done by Stanford and the Internet Archive, but thirty-five percent of newly published websites were AI generated or assisted. So I'm sure that's a lot higher than today. the another analysis showed that new web pages in April twenty twenty five and that seventy two seventy four point two percent contained detectable AI generated content. However, only 2.5 were pure AI, but you know, 70 over 70 were a mix. apparent over 50% of web articles are now primarily AI generated. So this is this idea of you know, AI generated. Yeah, Ty Pattison: Fifty percent of web articles. River Roberts: over approximately 50%. And there is a prediction that that we'll end up with about 90% of online content will be synthetically generated by the end of twenty twenty six. So yeah. Ty Pattison: yeah. I think this this follows a power curve. River Roberts: Totally. Yeah. the the traffic requests as well is like we've already we've already gone past peak human traffic requests. So c Ty Pattison: Ha ha ha. River Roberts: yeah, Cloud Cloudflare is showing that over fifty seven percent of web requests are now bot generated, exceeding human traffic. the NBC News confirmed that you know f over 57% of their news requests are coming from AI bots. so yeah, the the the landscape in which we are pulling input for these systems has shifted so quickly. and I'm sure everyone's aware of this idea of slop as as output, but when slot becomes input, and then also like you know, a lot of these large language models were trained on You know, you were commenting before on a lot of these AI systems are not good at writing. And you know, you know, these large language models were specifically trained on very large amount of data input. And so when you're scraping like all of you know, Reddit, for instance, r the the contributors of Reddit are not necessarily the the best writers. So you end up with a lot of not great quality inputs that this is trained on, even though it's human, it's not necessarily high quality. so there seems to be a bit of an edge here for small language models that are highly focused on what you consider like high quality human input and then combining that with some measure of synthetic aligned data as well, but not AI generated. It seems to be not L L generated. Ty Pattison: I wonder what the minimum requirement for like of data for training a model is. Like if you could take, you know, every academic paper and every poet. Like if you could take professionals in a field and just use their data, their outputs, that seems to me like it'd be higher quality than all of the internet. when I think of when I think of outputs, I was considering this earlier and it's a symptom or a feature of slowing down is that I've started to really see the distinction between like logistical output and emotional output or like the difference in what we're striving for. Like technology is really good for I'll I'll use this example of writing. AIs are really good at doing some things. Really good at looking at a a wide range of data and giving us some patterns, giving us some analysis of what's going on. Not very good at writing emotionally moving prose. Whereas it it like I can speak from my heart, or somebody can write a poem from their experience and I was noodling on this the other day and called it the tie test, as in when I read something, does it make me cry? And if it makes me cry, then it passes. Right? So I'm like, okay. If the output, if the intention of the output is to be emotionally moving, connective tissue to to do something to move my emotional body, then That's one particular type of output. If the output is a logistical thing, which might be river needs to get from Milan to New York. And he needs to be in New York by this date, or it needs to be in London by this date, whatever it is. If the output is a logistical if it if it converges, if it has a right answer, then that's a completely different domain of output. And there are different tools that will be really useful for getting that particular type of output. And when we think about like the internet, what was just coming up for me is like we often talk about like for the people. It's like the internet is not really for the people. Or the internet that's coming is not really for the people. The internet is for the bots. The world is for the people. River Roberts: Mm-hmm. Ty Pattison: Nature is for the people. Like nature is also for the animals. So I don't want to say it's just for the people, but you know, it it's like What environments do we want to be in and what outputs actually support that? River Roberts: Mm. Ty Pattison: is super important. River Roberts: Yeah, like just coming back to the first principles of it all in in a in a sense. Yeah. Ty Pattison: Yeah. Yeah. But it gets all spaghetti tangled. Like I like the emotional and the logistical kind of entanglement of like using my phone, for example. It's really useful as a device for like figuring out how and where I will meet people. It's not particularly useful for or i it's less useful than just speaking to someone in person than texting them, say. River Roberts: Yeah, and well I think the mobile interface is a really like as certainly as an input and output, it's kind of blurred the lines a little bit because y obviously there's very practical outputs on the phone, but also a lot of these tools to keep us hooked have kinda hijacked various hormonal or dopamine pathways which do have an emotional resonance and not necessarily in a in the healthy way. In fact it turns out most likely a lot of these are not a particularly healthy hooks. and I think you the insight that you're talking about before in terms of any output that has AI as part of the cycle is not necessarily going to be very good at the emotional movement side of things, you know. As you're talking about that, I was just like, you know, think of the you know, the Beatles songs that have echoed through our cultures for and and across the world for for decades now. and the the words that have so much movement in it, with the melody as well. It like AI can generate songs now, but it's devoid of that that that deeper kind of embodied connection, which like how is it going to be able to get to that? So all this all this to say, I just think that both human input and human output is still going to be highly valued going forward. in fact even probably more so 'cause it might end up being very, very scarce relative to How much is is created outside of that. So being able to curate, you know, we're kind of in the stage now of curating the the the the the important pieces. I think that's the role that a lot of humans are p kind of playing as they interface with these different AI agentic systems, right? We're curating the inputs, we're kind of curating the outputs a bit. Ty Pattison: Hmm. I am aware that yes this is the case right now, but I am cautious to say that AI won't ever be able to like create this to create really emotionally moving content. Because I believe that it is getting better and better is just like however we define it, it can change. So I think for me at least the lesson is like be aware that this might change. And one day AI might be, probably will be, really good at hooking us emotionally at creating exactly the thing that we need to hear that is tailored to our specific neurons. However being aware of the input or like the situations that we put ourselves in, as in like I choose not to be on social media ninety-nine percent of the time. And although it has a lot of things that will light me up, I know that that is an input that leads to an output that is me feeling disconnected and wasting my time and doing these things that are like not actually healthy, not important. Important is a really good word to use here. and it's it's not very clear to me what's important intellectually sometimes, but the felt sense of what is important is like, yeah, cool. River Roberts: Yeah, yeah, immediate immediate. If you if you if you're connected, if you can listen to it, yeah. I feel like too many of us have been educated into just the head. So does something intellectually make sense versus like do I actually feel that this is important to me right now in the moment? Yeah. Ty Pattison: And and overvaluing the inputs that come from our heads and undervaluing, underweighting the input that comes from like the physical sensations of our body, our emotions, all of these things. They're just inputs and they're important inputs. River Roberts: Yeah, there's there's there's an old kind of saying that I think is is that I've been admittedly trying to practice. I mean I've been practicing but it's still like an active practice for me right now, which is let the mind be the last one to know. and it's kind of this idea of you know, it it let let the mind be the last source of input and also output for decision making. As opposed to like intellectually understanding something. And then checking in with, you know, your different kind of ways of making decisions. yeah, staying letting that one be the last one to weigh in, so to speak. yeah, this there's something really potent in that practice I'm finding. Ty Pattison: It's it is similar to my mantra of late, which is aim with your heart, adjust with your head, and then simply River Roberts: Mm. Ty Pattison: do the best you can from where you are with what you have. River Roberts: Hm. With these guys, the handsies. Yeah. Hot hot hands and head. Hmm. Ty Pattison: all right. River Roberts: Yeah, I mean I I still I like I mean I think this input output piece is like really fundamental to a lot of things. So I'm I'm I'm still so curious about how this emerges and I think it's ri so I just to kind of wrap us up a little bit here, I I think where we are right now is we are curating both input and output to a certain extent, individually but also culturally. and then in the future, I think we're gonna have to be the arbiters of values and alignment and kind of help govern these systems both individually, collectively, on on different scales. And that's the piece I'm not necessarily sure a lot of us are are prepared for. you know having clear align alignment Ty Pattison: Okay, here's a question for you. River Roberts: within yourself and then within your personal relationships to then extend that digitally, you know, out in this great expanse. So yeah. Ty Pattison: So I think quite a lot about what is the smallest possible thing that I could do to take a step towards that? What is the tiniest action? Is there anything that comes to mind for you that is like what's the what's the like you know, if we think about starting a meditation practice, if I meditate for thirty seconds a day, that is so doable and it starts me on the path, what do you feel is like? River Roberts: Th for me the smallest yeah, Ty Pattison: A beautiful next step. River Roberts: for me the smallest minimal input for me personally is always just take a deep breath. Ty Pattison: Yeah. That helps. River Roberts: Everything. It's the simplest, I think, physical human input. It doesn't cost anything. Anyone can do it. and from that there's there's so much more that emerges. yeah, in terms of our relationship with emerging technology, I I I don't know what that smallest individual unit of of input can continue to be. having said that, like I do often Although I I I I get hijacked a bit, but I I I do often practice taking a breath before I I look at my phone. But again it's it's such a hijacking system in a way that it's it's a tough practice to to keep up with. Yeah. I also think about like, you know, what are the inputs on this planet to create the outputs of this planet? Like if we're looking at it more of a universal sense, like it's crazy to think about the inputs that this planet Earth has received over the, you know billions of years it's been here and then the output of this planet at the moment and what it's likely to output in the you know next couple of hundred years. it's crazy to to look at this emergent system that we're part of on on those kind of scales. Yeah. Ty Pattison: Mm. And the difference between like who receives the output. Like if if the inputs come from outer space and they land on the planet and then the outputs are whatever comes off the planet and we skip the whole what happens here part, which is like our story obviously, but there's a different receiver for the outputs that is That is the out outer space world versus like us as receivers for the outputs. Like what is the output of everything that has come here? It's what's currently happening. River Roberts: Yeah, yeah. Yeah, well something else is our out some Ty Pattison: This is turning into a River Roberts: yeah, it's very esoteric, right? But something else is something else's output becomes our input. so there's always this kind of constant exchange. But yeah, I think we're gett we're getting quite quite squirrely here. Ty Pattison: I th I I to to reel it to I was gonna say it's turning into a rabbit hole, to reel it back in a little bit to this input output and how it connects to technology. I think you're right actually. Using technology, one of the like ingredients for the cake is a deep breath. Is like breathing when you're When you're using technology, when you're like reading a text, when you're responding to something, it just makes the output better. and it makes the experience of of doing the thing better as well. maybe we need a breathalyzer on a phone. Like not for alcohol, but just like please blow it into this or suck on this while you River Roberts: No, I think that I think that's brilliant. I mean the breathalyzer could probably tell you I mean, they they started to do this in the forest therapy test and research in Japan, that the breath could give you a sense of like kind of what hormones are active in your system at the moment. So the breathalyzer on your phone could basically give you an indicator of what hormones and state that you're actually in at present and then the output of the device could could kind of react accordingly. yeah. I don't know how to wrap up this input output section. So like do we just Ty Pattison: That's all right. Why why don't we why don't we use the input output and get meta on it? the input is the conversation, the output is one one word, how are you feeling about the section? River Roberts: One word. Ty Pattison: Yeah. mine is inconclusive. River Roberts: Yeah, I'm I'm a bit bamboozled. Ty Pattison: Good. Okay. Maybe maybe maybe it's just a to be continued. Maybe this isn't a complete conversation yet. We'll come back to it. River Roberts: No, no, defin d definitely not. I think I think we'll I think it's good to reconsider from first principles and stuff from time to time. So we'll definitely come to back to it at some point. Ty Pattison: amazing dude. Well, I think we should start wrapping this whole thing. You've got a busy day, I've got a busy day. but I can see that there's more that we have here on the docket. River Roberts: Well w well we have we we have AMAs. That's that's it. So Ty Pattison: It's yeah, okay. Perfect. River Roberts: That's that's that's the last piece. Do you do you have any AMAs? Ty Pattison: No, 'cause I always forget about the AMAs, that's why I thought that this was the the end of our segment. But I'm gonna background bring in an an AMA question. River Roberts: Hm, all right. Well in when that not the case, I I have I have two lined up. Okay, there's there's a bunch. I All right, I've been I've been trying to I mean I haven't been trying, but I've been upgrading our little Hermes Venture Studio play and unfortunately the questions aren't still that great, even though it's trained on all our transcripts. So I was I was hoping that it might have had some really insightful questions. Ty Pattison: the same thing that the the outputs of like The technology is great. It's doing much better. It also speaks to that we adapt really quickly and we expect a lot. and the delta between what we get and what we expect. I think that it's just a part of the human system that there will always be a delta, that there will always be a gap. So, yeah. It's also got to do with velocity. I've got an AMA for you. If you were able to choose the pace at which you worked, at which you produced things. What kind of cadence would you like? Like and and you can do this in any kind of form you like. We can talk about it in days, we can talk about it in years, we can talk about it like but fundamentally the question is what is the natural pace of your creation process and is it like swayed by anything specific? River Roberts: Yeah, I I personally feel very like I feel a sense of fulfillment when every day I do something physical that can kind of be scene. Like I know I know you can relate to this as well, but you know, being part of different build crews at festivals or just building in general, it's a really fulfilling feeling when, you know, at the end of the day there's something there that wasn't there the day before. Or, you know, working on a farm, like knowing that you've created a bed and planted some seeds or just the fulfillment that comes from even harvesting some some Some veggies or some fruit. So like there's there's on a daily cadence, I feel really fulfilled when there's that level. I'd probably say the next cadence is maybe like a little bit more between monthly and quarterly. And that would be a little bit more in terms of feeling like there's progress on a bigger project. So that might be, you know, releasing a feature or you know, getting a a venture out the door and a a few people's feedback type thing. Like that kind of monthly to quarterly cadence feels really good. And then I'd say the next kind of cadences are the year to three years. And then I the one that I haven't quite wrapped my head around but I think is really powerful is this decade long cadence as well. Like what are you holding for the decade? And I think I I get the sense that that cadence is is measured a little bit more when you're in more of a a fa a a a season of family. Ty Pattison: Yeah, you have like the intuitive sense that a human will grow and change. and you don't see it on a day to day basis, but you get it if you go away for a long time or if you don't see a family member for months, yeah. River Roberts: Mm-hmm. That's a good question. I like that. Ty Pattison: Thanks. That one was from longtime listener Ty. River Roberts: Is add the qual quality input then. Ty Pattison: I've been going into the sauna at the climbing gym nearby and the sauna at the climbing gym has all of these like just really extreme people and I'll go in there and stir the pot just to ask questions that are whatever the whatever I can think of. River Roberts: Well well so based on well first I wanna hear your answer to that question as well. And then based on that I'm gonna I'm gonna also ask a question, not from longtime listener River, but from a mutual friend of ours, Tim, who asked a question that has stuck with me in a sauna from this time last year. So Ty Pattison: Yes. Yes. Okay. to answer the question, I I think I realized that When I do work on a weekly, when I do work on somebody else's schedule, when I have like been employed in an office, having the cadence, having some structure around it, like starting early, I really like. I really like being productive in short stints and like month. Two-month sprint of building something and then having it done is River Roberts: Mm. Ty Pattison: a really good amount of time. there's something about 10 days that feels really good to me. Maybe it's just that intellectually I like that. Like, I want it to be a little bit longer than a week. I want to like be able to go into it and like noodle on and have an space and time to like not have to make decisions about something and then have enough time to go and like really bang it out. I find if I'm trying to get inputs from other people, like I've done some research projects where I have to or where I want to get qualitative data and so I'll set up meetings for people and especially if they're in corporate world and they're founders or CEOs or CTOs, they're just like busy humans getting them on the phone. There's like quite a lag in that. So like the order of operations of getting something done if I'm getting inputs from people that are busy, it just necessitates there to be like a certain amount of time. but then you were talking about this idea of Kronos versus Kairos. do you feel like giving a quick explanation of that? River Roberts: Yeah. So the the ancient Greeks kinda had these these two methods of understanding time. And Kronos was very much like, all right, this is the time of day where something specifically happens. It's a like a specific measurement of time. And that's Kronos. That's kind of how in the you know, the root of chronology, for instance. It's also understanding the sequence of things. And then Kairos is this idea of the time it takes in order to do something, which is not necessarily defined. It can be quite open. So there's th it's not time bound, even though it does represent time in its own relative way. Ty Pattison: Yep, perfect. So I find that when I work, I if I don't pay attention to Kronos, it my work is so much better quality, faster. I forget who it's named after, but there's this principle that a Project or a task will expand to the time that you allow for it. So if you give something two weeks, it'll take you two weeks, say to write an essay. If you give yourself two hours, you'll write an essay in two hours. Now, like depending on the task, there is there are some constraints to that, but I find that if I work in Kairos, I can be more productive in two or three hours than a lot of people could be in a week. And if I give my Myself a week to do a particular task, then I can be super unproductive, you know. so I like the cadence of I I want enough substance and self-feedback to give me energy for each day. And sometimes I'll plant seeds at the end of one day that makes me feel like I have a head start the next time I pick it up. Like, this'll be a juicy thing that gives me lots of feedback. I'll plant that at the start so that I'm like really into the thing straight away. and I get quite intense Context switching drain. So if I have to go into something and then switch to something else, it takes me a bit of time to like warm into the thing, unless there's like significant overlap or unless I can pull ideas. Sometimes that's actually really an effective method, is starting with one thing and allowing the universe god, whatever we want to call it, to like give me a few seeds from something. I'm like, mm, this doesn't seem very important, but actually I'm gonna spend twenty minutes doing this because there might be some like nuggets of this that really kick start me into something else. yeah. That's my answer to this question. River Roberts: I'd like I'd I'd love to explore the the the the ability to create more Kairos space. like but again without being like how do you how how to create more of it without it being time bound in a way. So I think that's a really a really beautiful practice. Ty Pattison: think it could be a really cool practice to do with a bunch of friends on a call and like you have a start time and you say you you go around at the start and everybody has a couple of minutes to share, like, all right, I'm here to do this. I'm staying until it's done, whatever it is. And kind of an accountability co work thing, but for a particular task that you want accountability for. River Roberts: Yeah, I like that I like that. All right. So the Ty Pattison: Alright. Hit me with Tim's. River Roberts: the c the question that has has stuck with me for a while, which is this idea that You know, the this idea of the hearth. And so the the hearth was the the the central fireplace that heated a home. And you know, also the sauna is kinda the modern day version of that for a lot of places, particularly in Finland. Like a lot of a lot of places here have saunas inside the apartments or houses. And so the idea is this hearth really warms the soul of the home in many ways. And so you know, if you're being mindful, you're very careful with what you feed this this hearth, the wood that you get it and and the process of of of harvesting the trees and and using the axe to chop the wood down into sizes that bring that warmth into your house. this idea that even cutting the wood is warming yourself twice as well, 'cause you get warm actually chopping the wood and then you get warm when you when you get to work with the wood. but the question was actually so if if that's the fast, what what what what is the food that keeps your heart fed? Like what what warms your heart? What are you feeding that energy with? Yeah. Ty Pattison: immediately what comes to mind for me is is truth in all its forms. When people tell me things that are true, what they're really feeling. I get this a lot through poetry. I've been seeing someone who I like a lot who is a poet and the way that the words have been played with and thought about and like really considered in what's trying to be conveyed really like warms me from the inside. sometimes like getting feedback from a close friend that's like, hey, I think that this I've thought about this a lot and I think that this is true. that really touches me and gives me like a certain amount of like warmth and and it's not fire, but it's like care. It's it's caring. and then Yeah, emotion. It it things that are like emotional. Like, yes, there's the logistics stuff and that requires energy. That's not where I get the energy from. If I wanna like get moving for the day, I've got a few songs that I will play that like really get me moving. and and that will like light a fire. Often artists who are like really talking about how they how much they like themselves. Like some rappers are so full of themselves or some rappers are so this is rapper brother Ali. and he doesn't rap about clubs and blah blah blah blah blah and money. He raps just about like family and about purpose and about like drive and all of these things. And he's got some banger lines where he's just like River Roberts: Mm-hmm. Ty Pattison: i i he's like, I I'm not gonna slow myself up to start competing with these people. Like th there's just like River Roberts: Mm. Ty Pattison: this knowing that he's like on River Roberts: Mm-hmm. Ty Pattison: this path that is so Real for him. And then when I listen to it, I'm like, yeah, that's absolutely true for you. Like it's that amount of like authenticity, basically, that really gives me the inner kind of fire. River Roberts: Yeah. Ty Pattison: and then there's River Roberts: Mm-hmm. Ty Pattison: the like providing for people, protecting people, like fulfilling this, like making things going to the effort of yeah, creating something for someone else that I know will light them up because it lights them up when I get to give it to them and it lights me up thinking about how it might light them up. so River Roberts: Well Ty Pattison: it's it's the chopping the wood thing, it's the two sided yeah. River Roberts: Well paid, sir. Well paid, yeah. Ty Pattison: How about you? What's what's a lie for you in that question? River Roberts: I I I I love those. recently what's what's really alive for me is well what first came up is is being in this really present playful state and I had so many of those opportunities at this family festival. Like some of my new best friends are like five years old and their ability to just like bring you right into the present and be super playful and it's just there's something that is so warming in that sense. And you know, the questions are just so like i innocently delightful and and curious and I I'm also noticing that I'm spending a lot more time around animals lately and and when you can be present with animals and they're present with you there is really some kind of kind of co warming. to that. And it turns out that there's a few animals in particular, dogs, cats, horses, that we are very good at co regulating with. So our warmth emanates to them and and vice versa. So yeah, those those have been some recent interactions that I'm I'm I'm not that close to in in in recent years, but yeah. recently it's it's been emerging a lot more. I'm definitely with you there in in in dance as well. and and sound and music and and kind of the frequency. Like that that certainly warms and and fuels. Yeah. And then just like deep kind of primal movement. like I l I love to to go for runs in forests and I don't know, there's something about running through a forest. Like a lot of people are used to running on flat surfaces that are very predictable. But as soon as you run in a forest, there's like a different kind of primal drive that turns on and and heats the whole system and it it it it yeah. and Ty Pattison: Yeah. River Roberts: then equally and then equally like this resting state in a natural environment. So I I do practice this daydreaming practice in a hammock between typically like two trees and I find that that is something that also like feeds the feeds the the heart quite a lot in many ways. Yeah. Mm the the two the two the Ty Pattison: Thank you for River Roberts: scales of it, yeah. Ty Pattison: Thank you for asking the question and thank you Tim for the question. Tim Tim has some gorgeous River Roberts: Yeah. Yeah. Definitely shout out to them. Ty Pattison: frames and questions. Shout out Tim. River Roberts: Yeah. Well, with that being said, until next time, bro. Ty Pattison: Until next time. It's been freaking lovely.