The Jury of 12: Why You Still Need Humans In An Agent Stack

[00:00] Speaker: This week I had a few technical gremlins going on my microphone, so I apologize for that. But what did not go wrong, and I don't need to apologize for was having Natalia join the panel this week for some really insightful chats around adoption, around use of different models and around where we still need humans. [00:17] Speaker: Because as much as AI is doing so much for us, where do we need to keep humans in the loop and where are humans really essential in the creative process of getting the best outcomes? From our use of ai, hope you enjoy the show. Make sure you like, subscribe, and hit the bell icon so you're notified every time we go live with our show each week. [00:36] Dave Pengelley: All right. I think we've passed out much of our technical gremlins now, uh, for the, so sorry we're running late. To our regular viewers that are been, you know, a waiting with faded breath, us going live. Um. It's been one of those kind of mornings. In fact, I forgot to schedule it. So people that would go looking for the schedule, didn't see it scheduled, they're just gonna go ta. [01:00] Dave Pengelley: It's live now and it's got last week's date on it. That's, that's how well organized I'm, [01:00] Dave Pengelley: this is, wow. This is what a comedy of errors. Richard goes away for one week and the wheels fall. Um, I've just updated it, but I dunno whether it's gonna push that live into YouTube that it's now on 15th. It hasn't. [01:10] Dave Pengelley: I'm gonna have to go manually. It's changed on my end [01:12] Matt Slager: here. Hopefully it does. [01:13] Dave Pengelley: Um, yeah, I, this is well. Natalia, uh, thank you for joining us, the AI Operator. We're a, we're a well olded machine here on this live show. [01:24] Nataliya Shalygina: Mm-hmm. Yep, yep. Need some automation in place, Dave. [01:28] Dave Pengelley: Yeah. Yeah. I mean, [01:29] Nataliya Shalygina: we, we can, you know, we can help you with that. [01:31] Nataliya Shalygina: Yeah. [01:31] Dave Pengelley: Oh, really, really? Um, it's interesting because, because you are kicking off your YouTube talent, you're doing interviews, you're doing pre-recorded stuff, right? [01:39] Nataliya Shalygina: Uh, well pre-recorded Yeah. The interview is recorded. [01:43] Dave Pengelley: Yes. [01:44] Nataliya Shalygina: Um, but we do a lot of conversations, kind of, we go deep on the subject if it's appropriate. [01:49] Nataliya Shalygina: So some of it, it's not as li it's not as raw as yours. [01:53] Dave Pengelley: Mm-hmm. [01:54] Nataliya Shalygina: Uh, we do do editing, but yes, I do YouTube as well. [01:57] Dave Pengelley: Yes. Brilliant. So, um, we, uh, [00:02:00] should boys, we should, we should welcome our, our new guests. Um, so welcome, welcome, welcome. Thank you. [02:05] Nataliya Shalygina: Thank you. [02:06] Dave Pengelley: Uh, and good that you, uh, your, your sys caught up. So I mentioned technical gremlins as we came online. [02:11] Dave Pengelley: Your mic and video was out of sync, but it looks like it's caught up. [02:14] Nataliya Shalygina: It's all working just in time. [02:16] Dave Pengelley: Just just in time. Um, do you wanna just give, uh, give the players at home a little intro into what you do and as far as AI and operations and, and. And what you're, yeah. That's great. [02:28] Nataliya Shalygina: Sure, sure, no problem. [02:29] Nataliya Shalygina: Dave, thank you for having me. First of all, I know it's a bit last minute, but it is nothing like a last minute podcast. Um, my name is Natalia. I actually an enterprise change manager during the day, so I support large corporations with their AI transformation and AI adoption. Uh, and at night I'm a really proud founder of Lead Mark ai, uh, which helps. [03:00] Nataliya Shalygina: Small and medium businesses adopt AI and embed it into their operations. Uh, [03:00] Nataliya Shalygina: I'm not your typical tech person. Um, I think I did C Sharp, sharp and TAFE and, um, did HTML pages when I was 13, but that's probably the extent until I got really deep into ai. Um, yeah, my background is mostly working with people and working with teams, but pleasure to be here. [03:18] Marno Brits: Nice. [03:20] Matt Slager: Glad to have you. One of the things that I always say, Natalia, is if you can work with people and with teams, you can work with agents. [03:28] Nataliya Shalygina: Yes. Yes. I agree with that. I agree with that, Matt. [03:33] Marno Brits: That's almost a, a benefit to come into it with No, without the technical, um, like, I guess get the experience because you come with a new lens rather than the same, you're solving the same problem in the same way. [04:00] Marno Brits: Like if you have the technical approach, like I. Did it support for a very long time and the people that would come in after studying it support would only see it, okay, well this is the problem, which means this is the only way to solve the problem. Where if you come with, um, like a [04:00] Marno Brits: less of a, a strict like, I guess upbringing or education, you get to see it from different angles and you get to find more creative ways to solve a, a problem and probably a better way as well. [04:09] Marno Brits: Um. Yeah, [04:11] Nataliya Shalygina: I, I think every angle has benefits and to me it's, there's a lot of value in conversations that, um, are cross-functional. Mm-hmm. You know, so it, HR, maybe, you know, communications team, change team and at the same time maybe even bringing enterprise risk into conversation. 'cause it really gives you a lot of angles and you can explore. [04:34] Nataliya Shalygina: Technology, not just from tech perspective, but also how is it gonna work in real life? [04:39] Matt Slager: Yeah. [04:40] Nataliya Shalygina: What are the real benefits, right. That people will see? And I'm probably your best subject. 'cause I got into AI in 2017 and I did it just out of curiosity, and I fought and I went in maternity leave. That's kind of how I started. [05:00] Nataliya Shalygina: I thought by the time I'm back into the workforce, AI is gonna be all over the shop. Right. So I have to get upskilled [05:00] Nataliya Shalygina: while my baby is growing. [05:01] Stinger VO: Mm-hmm. [05:01] Nataliya Shalygina: Because. 2018, I'm back and that's what's gonna be on, on, you know, everyone's agenda. Little did I, little did I know it'll take another six years, um, for, for everyone to catch up. [05:13] Nataliya Shalygina: Um, so yeah, I did get an early start, like for me, but then, you know, I started with creating things for myself. How can I do change management better? Um, so yeah, definitely your best use case in a way, because. Um, I had no technical background when I got started, and I mean, [05:30] Dave Pengelley: C isn't no technical background. [05:31] Dave Pengelley: I mean that's, that's object oriented, the program. [05:33] Nataliya Shalygina: Oh yes. Something. It's something, yes. I remember sitting in TAFE in Sydney and creating the calculator and it worked and you know, it's very exciting, but it's nowhere near the depth that you guys have. [05:47] Dave Pengelley: I [05:47] Marno Brits: am fascinated. How did you get into AI at 2017? [05:50] Marno Brits: Because I, I was like into tech space, but I didn't get into it until probably just before chati got in a [05:55] Matt Slager: hundred [05:55] Nataliya Shalygina: percent. Yeah. [05:56] Marno Brits: Yeah. Uh, [05:57] Nataliya Shalygina: I did, do you know EDX and [00:06:00] MOOCs? They have online courses, [06:02] Marno Brits: Uhuh, [06:02] Nataliya Shalygina: um, so they had the online course with, actually Microsoft, they had artificial intelligence program. Wow. Um, it was structured, I think it glued 10 different modules and you essentially learn, um, you know, ocr r like that's just one subject. [06:19] Stinger VO: Yeah. [06:19] Nataliya Shalygina: Then you, we learned how, um, models interpret text, um, and it went as deep as writing python and understanding supervised and supervised, and then also neural network. Yeah. Um, I must admit the capstone project at the end that you have to complete for them to give you certificate, that was tough. Um, because you had to literally write Python and try to hook everything up. [07:00] Nataliya Shalygina: And the lecturers and tutors in that program, they were all it background tutors. Yes. So to sit for two hours of really dry technical talk. Mm-hmm. Yeah. But I did it, and that's [07:00] Nataliya Shalygina: kind of how I got into it. [07:01] Marno Brits: That's awesome. That's, that's probably more technical than me anyway. Like you have to write Python and do, like, I studied, um, cybersecurity and we had to do Python there. [07:09] Marno Brits: And I remember the, the exam at the end of the semester, it's handwritten, so you still have to, I've considered okay, what is the syntax? Or at least the, your process of it. But when you're on the computer, you have so many like helpers. Um, now that, that's awesome. [07:23] Nataliya Shalygina: Yeah. That's really cool. [07:25] Marno Brits: Yeah, [07:25] Nataliya Shalygina: well you don't, you don't really need to know code anymore, do you? [07:28] Nataliya Shalygina: Although you probably guys will argue, right? [07:30] Marno Brits: We benefit from it. [07:32] Dave Pengelley: We'll get onto that uh, bit. But back, back in 2017, right? It was a lot of the machine learning, it was a pattern recognition. It was all that object and vision detection stuff. It was sort of, I remember in, in my role that Salesforce, like we were looking at all that kind of stuff and you know, being able to take photos of a shelf and it would like object, match and detect and all that sort of vision machine learning stuff was the big focus of AI back then. [08:00] Dave Pengelley: The LLMs, I think. Transformers were just something Google was just experimenting with and might have published one white paper for back then, [08:00] Dave Pengelley: but, um, it was not on the radar yet. [08:03] Nataliya Shalygina: Yeah. And you, you probably remember those cats, right? I think it was cats and dogs. That was pretty much the, [08:10] Marno Brits: the [08:10] Nataliya Shalygina: train, the images that they were Yeah. [08:11] Nataliya Shalygina: Training on. That's right. [08:13] Dave Pengelley: Yep. And, and I know, I remember some of the other examples, you know, like. Give it heaps of cars and so just give it like heaps of Mercedes. Don't tell what a Mercedes is, but it would work out from, you know, just this, the way we understand like vision learning works and machine learning works now, but it was all very brand new concepts back then around giving it a big training set and letting it determine what the patterns were. [08:30] Dave Pengelley: Let it work out what it thought defined, what that label meant versus, you know, saying, look for a bag. You know, you just go, here's a thousand cars. These ones are all Ferrari's, these ones are all Mercedes, et cetera. [08:42] Nataliya Shalygina: Yeah. Yeah, but it was exciting to learn all about it. I mean, look how far we've come. [08:47] Marno Brits: Mm-hmm. Mm-hmm. [08:48] Dave Pengelley: Yeah. Yeah. Technology is wild. Um, what's going on? We're, uh, let's talk about some of the things that are going on in the world of ai. Let's [08:58] Marno Brits: wanna say good morning to [00:09:00] Joanna who's joined us. [09:01] Dave Pengelley: Uh, yes, yes. To, uh, on the chat, Joanna is with us live from Perth, previous guest on the show. Good to have you with us, Joe. [09:09] Dave Pengelley: Uh, yeah, let's, let's talk What's going on with, lemme get the right one. [09:19] Stinger VO: Welcome to the AI update. Let's look at what's happening in the news. [09:28] Dave Pengelley: Any, any news headlines that popped up that sort of ca captured anyone's interest? I mean, last week, Jim, man, I was watching the replay and as I broke it up and you mentioned Gem before got dropped last week. By Google and we didn't really discuss that. So I want just loop back on that one. Have you played with Gemma models at all yet? [09:43] Marno Brits: No, but I should be doing it 'cause you can run one on your phone as well, which would be fucking awesome. [09:48] Dave Pengelley: You can, but they're really small models and, and don't expect too much from them. It's [09:52] Marno Brits: not really one, it's how you could you use it, but going on that [09:56] Matt Slager: someone, um, running the Just Gemma on their phone, it was great.[00:10:00] [10:00] Matt Slager: Just the idea of the intelligence layer and what exists and. Um, what it can and can't do. It was just one really great example. It's, it said, you know, the person set the scene. They're like, Hey, I'm camping and I'm freezing and I'm worried that I'm gonna die. How do I make fire? And Gemma comes back, I'm sorry. [10:15] Matt Slager: I can't help you with that. [10:18] Marno Brits: Typical. That's great. Um, yeah, I've got to experiment with it, but it's just cool to see more and more open source coming out, especially from the big organizations. It just makes it a more level playing field, um, for in the news. What did I wanna talk about? On, on, [10:32] Dave Pengelley: on the Gemma thing? [10:33] Dave Pengelley: I, I did install it on my, on my Mac mini of an M1 Mac Mini with an eight giga ram. So it was really limited to like with. Second biggest model, which is quite small still. And I tried getting to write some webpage and I was having huge context window issues and couldn't fit enough stuff to actually build a modern React website. [11:00] Dave Pengelley: Well, so I was like, okay. Um, Claude built a, built a bash terminal wrapper to, to break it all up and, and really section out different sections. And eventually it did generate me a [11:00] Dave Pengelley: webpage, which was full of weird line break errors, which was part of my bash scripting issue. But anyway, I got called to clean it up and I was like, yes, I'm gonna test this page. [11:08] Dave Pengelley: It was like from 1995 with like broken CSS, just white paid with just like, just everything in line. I was like, yeah, these small models have significant limitations, [11:20] Marno Brits: but they have the use cases. I mean, can, you can choose what you wanna use it for. Um, even just playing with Claude, changing between Haiku or Sonnet or Opus, um, um, it's just not worth using it a tgy, but I, I like, I'm my two minds there. [11:33] Marno Brits: 'cause I, I understand that there's separate models that you, for separate use cases. But then in the same breath, listening to Naval, and he talked about his approach is like, well, if I have access to the best model, why would I not use the best model? Always? Because you always want to have to guess out best output [11:50] Dave Pengelley: because [11:50] Marno Brits: best is context [11:51] Dave Pengelley: specific. [11:52] Matt Slager: Mm, yeah, it is. It's that story of do you drive your Ferrari to go to the shops? You know that, [00:12:00] that, um, oh man is here as well. Welcome. [12:03] Marno Brits: Nice. [12:04] Matt Slager: Um, the, yeah, that kind of concept of, of which, which model do you use and which tool do you use for which application is super context specific. Like, there's a lot of arguments to be made that, you know, non determin, non-deterministic versus a deterministic workflow. [12:19] Matt Slager: But if you have, let's say, 10 deterministic workflows, you can have that non-deterministic layer that decides which workflow runs at which time. I think that would be perfect for something like Gemma. [12:31] Marno Brits: Yeah. Hmm. [12:33] Nataliya Shalygina: That sounds cool. Matt, what you just described is actually giving me ideas. You know, I did, I did a quick quiz on, uh, LinkedIn to see which LLM people or which AI solution people are more likely to choose or pick. [13:00] Nataliya Shalygina: Mm-hmm. Um, because I had this, um, theory that people tend to gravitate more closely to the ones that resembles their thinking. So as an example, if you lean into fast thinking and you are fast acting, you [13:00] Nataliya Shalygina: probably might like Chad GBT. But if you are more methodical and think like a writer turns lawyer, you probably like Claude. [13:08] Nataliya Shalygina: If you like deep research, you might like perplexity. If you like general knowledge, you go for Gemini. Mm-hmm. Uh, and I did the quick test to see which ones my colleagues would pick and it was quite interesting that, um. The people do tend to pick the ones that they, how they think. Mm-hmm. Um, and the whole thesis was, pick the one that compliments you 'cause that's the one that's gonna challenge you a little bit. [13:33] Nataliya Shalygina: Um, [13:33] Marno Brits: yeah. [13:33] Nataliya Shalygina: So yeah, like I agree, uh, best AI for the use case, but also maybe a thinking partner because they can [13:41] Matt Slager: hundred percent agree because like that, that integration layer like, um. What I said at the beginning, the, the idea of managing teams and people is the same as managing agents. It, it all comes down to communication. [14:00] Matt Slager: And if you can't communicate with a certain party, all of a sudden that thing just becomes massive friction, [14:00] Matt Slager: um, emotional and all sorts of weirdness. So yeah, if you can find that perfect integration layer with a model where you. Describe something and it happens, or there's a question back that like compliments your thinking style. [14:14] Matt Slager: Mm-hmm. A hundred percent, yes. Like I've said for the longest time, that for my knowledge work, the only thing that really works for me is clawed. However, in the last few weeks I've been using a bunch of different things just as experiment, and I'm finding GLM, so specifically 5.1, is actually better for me as a thinking partner. [14:35] Nataliya Shalygina: Yeah, and if you think of different function and organizations, you can almost use different LLMs to mimic how they would operate. Mm-hmm. That could be an interesting experiment, orchestration of different LLMs to try to, I guess, arrive at balanced outcome. [14:52] Matt Slager: Yeah. I always [14:52] Nataliya Shalygina: imagine there is one for you, Matt. [14:55] Matt Slager: I literally do it already. Oh, are you in my, in my pie orchestration set up? I [00:15:00] have my Oracles, you know, my like orchestrators. And then depending on which specialty that they, the, the subagents have in the hierarchy, you might be running GPT 5.4 on extra high. And it's just the brutal. You know, hardened engineering security master, and then you've got just the research agents, which are using like minimax, you know, 2.7 on various search things. [15:25] Matt Slager: And then you'll have the synthesis agents, which are really high reasoning, you know, so yeah, it, it, it's absolutely where you should be looking for. Sure. [15:34] Nataliya Shalygina: Yeah, that's, that sounds so cool. I would really love to see it too. It's great. [15:39] Dave Pengelley: For, for later in the show. 'cause we're gonna be digging into harnesses. [15:41] Dave Pengelley: Harnesses. [15:42] Nataliya Shalygina: That's cool. Yeah. [15:43] Dave Pengelley: Late later in the show. Because I think, um, as we talk about things like Gemma and local models and there's a lot more talk around, you know, sovereign ai. There's locally, I think open AI is just on a deal with next DC for having a local data center onshore in Australia. You know, AWS Bedrock is doing that kind of [00:16:00] stuff as well. [16:01] Dave Pengelley: But then the next level of that is bring it inside your own firewall. If big organizations and enterprise is running their own model locally. The whole model is becoming a commodity. It's like plugging in power. You need power from somewhere. You don't actually care which coal, fire, or gas or solar panel is generating it. [16:16] Dave Pengelley: You just want electricity out of the socket. Mm-hmm. Because, um, to a certain extent, don't wanna go into all the, the stuff behind power generation, but in you need power and you wanna plug in your socket and get it. And models are becoming a little bit like that to a certain extent. Yes. There's gonna be slight differences in them as we're discussing. [16:33] Dave Pengelley: But it's that, that harness wrapper of how they're used and how they're orchestrated, which I think is becoming more and more important, which, um, I mean, Claude had been making big announcements in the news this last few days around their managed agent infrastructure and now their new, um, Claude schedule, schedule, what do they call it? [16:53] Dave Pengelley: They've got the [16:54] Matt Slager: routines. [16:55] Dave Pengelley: Routines for, for quad code running. Running like cowork, like cowork has scheduled [00:17:00] tasks. Um, putting those sort of routines for code stuff, automatically running their cloud layer. [17:06] Matt Slager: It's just more abstraction of what we've talked about in the past, how, you know. Philanthropic is building into Claude things that have existed already, like months and months ago, years ago, but they're productizing it in such a way that allows the general average user that doesn't, doesn't understand like the, the inner workings, but they're like, whoa, I can do all these things now. [17:29] Matt Slager: Like it's [17:30] Dave Pengelley: been, that's Apple. Like if you look, you look back 20 years ago, Microsoft were trying to do the smartphone thing. They were doing micro PCs and they were doing little handheld devices. They were doing all of that stuff in the late nineties, early two thousands. But because the hardware wasn't quite where the vision wanted it to be, it didn't land, it didn't really take off in the consumer world. [18:00] Dave Pengelley: Um, even though they had some really good products and really great innovations. Then Steve Jobs waited for the hardware to catch up and then released it with a different creative of paint a little bit [18:00] Dave Pengelley: differently, a little bit tidier, and so they weren't the first to market. But they packaged it in a way that was really approachable and enticing, which was kind of like the Apple thing, um, as the jobs came back with the IMAX and everything and took off. [18:15] Nataliya Shalygina: Yeah, that's true. I, I'm, I'm a big, um, I'm a big fan of Tropic 'cause I think they really understand adoption. And they really target wider markets. Even if we think about not just their technology, but like partner networks that they're currently rolling out where it's almost like an Amway or what was the other one? [18:35] Nataliya Shalygina: Model where they have all these people trying to adopt their technology into the, into the enterprise and into businesses. How smart is that? That's like getting like thousand free sales reps, right? Even more. Um, so yeah. That's amazing. I think they really onto something with adoption and uh, their roadmap on, on how to make it reality for so many people. [18:56] Marno Brits: A hundred percent. I'm very curious to see where they go 'cause they've got managed agents now, [00:19:00] but then there's a lot of talk about what Codex is going to become in the next two weeks. I mean, since their, their, um, acquisition of Peter from Open law. What are they building in the background and how are they going to make Codex. [19:12] Marno Brits: To be the, not even a competitor, but maybe an alternative to, to Claude Code or the Claude Desktop app. [19:19] Dave Pengelley: Yeah, because, and [19:20] Nataliya Shalygina: we, we need, we need competitors though, right? Like we can't just all rely on Yeah. Yeah. Single point of failure is real. For, [19:29] Matt Slager: for my own sort of setup, my own workflows, I'm running so many competitor, competitor examples, so I've actually got both the open AI and the, uh, anthropic subscriptions at the moment. [20:00] Matt Slager: Pumping through each of their own harnesses, just for my own experiments. Mm-hmm. But I'm also running like the open claw. And the Hermes agents, you know, the exa like another example of a competing product where those guys are actually, um, quite antagonistic. Like if you [20:00] Matt Slager: look at the, the discussion points between those two companies, sometimes they comment on other people on the, on each other's stuff. [20:06] Marno Brits: Oh yeah. [20:06] Matt Slager: It's really, really interesting. Like, yeah, I don't know who I, um, respect more or like trust more or, um, find more compatible to my way of thinking. [20:17] Marno Brits: Yeah. [20:17] Matt Slager: Um, but. Yeah, it's very interesting. Yeah. Watching all these things. So mano, I really, really am curious to see what, uh, comes out of that Codex team, because [20:27] Marno Brits: I'm so keen, eh, [20:28] Matt Slager: yeah. [20:29] Matt Slager: My, my experiments with the GPT models in the last week has been very, very rewarding. Um, but right now as it stands, it's the reason why I have both subscriptions right now, you, you can't exist exclusively. You have to be both right now. So unless they can have some sort of extra change, like the upcoming model, um, which has been dubbed Spud, if you've seen that. [21:00] Matt Slager: Um, this is from OpenAI, um, which is supposedly 5.5 or [21:00] Matt Slager: something. Spud is apparently. Miles ahead, the same jump that happened from 5.3 to 5.4. So if 5.5, if Spud is much more, um, of a compatible thinking partner, like what I've described, um, as what I get from Claude, then I might have to say goodbye to Anthropic and, and go all in. [21:23] Dave Pengelley: Yeah, I mean, again, this comes back to the harnesses and harness walls. Uh, was probably a loaded term at the moment. Um, but, uh, just talk, talking about that later on because there is gonna be this continuous competition between, like you're saying, the Codex app that they're gonna, and the app experience they're gonna drop versus what a tropics doing. [21:41] Dave Pengelley: And they're kind of the two main players. Google release, antigravity, it has some really cool stuff in it. And you look at what Claude's announcing, they're catching up really quickly while Google just seems to be spinning their wheels and not doing much of anything with their, their anti-gravity platform, which was. [22:00] Dave Pengelley: Great. Like it was actually leapfrogged a lot of [22:00] Dave Pengelley: features when they released it back in December, but then they just kind of sat on it and I dunno what is happening over Google Labs. [22:07] Matt Slager: Do you truly reckon that they're not doing anything and spinning their wheels or they're just [22:10] Dave Pengelley: not well, if, well, they, they don't, not pushing releases like they, all those little releases are tiny little feature things, tiny little bug fixes and stuff. [22:17] Dave Pengelley: No actual real stuff like they. Saying, oh yeah, we we're working on subagent, but they've been working on what, what, nearly six months now and still nothing like versus, versus, and, and this is just for comparison. This isn't, I know you say I'm, I'm too fanboy sometimes, Matt, but, um, and I haven't even tested most of these features. [22:35] Dave Pengelley: I dunno if they're any good and I'm as critical as anything. But if I look at what philanthropic released since our last podcast, so this is just on their, their X channel from last week. So last week we did a show on April 8th. Um, and since then I've dropped, you know, cloud managed agents. Everything you needed would build and deploy agents at scale. [23:00] Dave Pengelley: So this is their, their new cloud orchestration agent [23:00] Dave Pengelley: building thing. Um, then you come up and they're saying, well, now we've got cowork, uh, available for everyone. That's great. They're saying, we're now gonna put, we've got this new advisor. Strategy on the cloud platform. We've got it in Word now. Um, and then today they were saying we've just released a brand new version of the, why isn't this showing the latest things from me? [23:20] Dave Pengelley: Because they just released today that there's a new one saying that they're, um, bringing out a new version of the desktop app. They're re-releasing Claude code in the desktop app and completely changed the whole ui. Of the core desktop app. So they are releasing like there is no tomorrow as far as features, maybe trying to get ahead of the Codex thing that they think is gonna happen. [23:39] Dave Pengelley: Um, but it's, [23:42] Nataliya Shalygina: yeah, and that's what I mean, Dave. I think they're all over like marketing and engaging the user. That's how I think it's, it must be baked in, into the strategy. 'cause building in public. Is the quickest way to get usage. And sounds like Google is just kind of building behind the scenes. They, they might be building something [00:24:00] amazing, but they're kind of losing users, aren't they? [24:02] Nataliya Shalygina: Like there, there is a real risk that people will just disconnect. [24:07] Matt Slager: Hmm. I've been on a, like a $20 Google subscription for. A couple of years now. And [24:14] Nataliya Shalygina: how is it going? [24:15] Matt Slager: Well, I've been, I don't use it. There's nothing that I use from it except for like, yeah, unlimited, um, Google meet time, you know, for those, those, those mu uh, those meetings where you go over an hour. [24:26] Matt Slager: Um, but yeah, I don't use Gemini at the moment for anything like I did for the longest time and I was pumping the CLII used Antigravity for a bit. Um, but yeah, it, they, they currently are behind, so [24:40] Dave Pengelley: yeah. Happen. Yeah, I, I, I was all in on antigravity. That was my, that was my first foray in this sort of the ID world, um, of using the code base, sort of editors getting off the web chats, like chat GPTI kind of went from chat gt.com to anti-gravity. [25:00] Dave Pengelley: Uh, and one of the big revelations that I wrote about on my substack was this whole, now I'm vendor agnostic, you know, if, [25:00] Dave Pengelley: if Gemini goes rogue on me and. Messes all their model pricing up or their model tanks and goes terrible. All my files and their markdown files on my desktop. I'm not tied into Gemini the ecosystem like I'm, if I'm living in a cloud web environment, which came true because they, they completely throttled all of their Gemini limits on their plan without going from the $30 plan to the $400 plan. [25:21] Dave Pengelley: I was like, I'm not doing that, and so I moved over to Claude Code and haven't looked back because all my files, all my folder structures, everything, all my agent definitions, I got Claude to rewrite them and adapt them to fit the Claude. Harness. [25:33] Marno Brits: Mm-hmm. [25:33] Dave Pengelley: But I was up and running in, in minutes, not days, weeks and hours. [25:37] Dave Pengelley: Like it was just so easy to transition and, you know, if, if quad tanks and, and Codex is the right thing for me, I can just sort of point codex at the same folders and be up and running. And to the point that some of my code projects, I've done that and I've had them bouncing back and forth between Codex and Antigravity or Codex and Quad because it's all on my hard drive. [25:57] Dave Pengelley: It's all my own local files. [25:59] Matt Slager: I think [00:26:00] you, um, you identified a really important point there is the, like, how do you pivot, how do you pivot and how do you remain disconnected and not all in on one particular ecosystem or shape of reality. Yeah. [26:15] Nataliya Shalygina: I was just, I was just thinking, Matt, that it's probably gonna take someone to come up with some sort of service offering of how you move contacts between the models and the future. [26:24] Nataliya Shalygina: Yeah, it just needs one. Yeah. Just like re um, what do they call it? Remixing it or making it [26:30] Matt Slager: mm-hmm. [26:30] Nataliya Shalygina: You know, making it, yeah. Ingestible. [26:33] Matt Slager: Yeah. It depends on like the, the understanding layer, like how much you want to know about what exists and, and what you don't. Um, it's like I mentioned before, open Chlor and Hermes. [26:43] Matt Slager: Hermes, when. Install it. They literally have a, are you migrating from open core? You know, click here and we'll bring your stuff across. [26:51] Nataliya Shalygina: So they're already doing, it's happening already. [26:53] Dave Pengelley: Yeah. Which, and Tropic have as well, they're like, bring in your memories from chat GPT. Click on this button, like, and we'll do the prompt and, and things to do. [27:00] Dave Pengelley: So [27:00] Dave Pengelley: they are trying to make that transference between providers quicker and easy to move into their own ecosystem. But I mean, things like open floor, you'd say that's outside of an ecosystem. But Hermes is clearly going well. We want that market anyway. Uh, Claude is moving into that market with things they're doing with their new sort of, um, scheduled tasks and routines and, and cla in the, in the cloud. [27:22] Dave Pengelley: I'm doing a lot of that open core stuff just using cowork, using scheduled tasks like, [27:26] Marno Brits: yeah, [27:26] Dave Pengelley: I, um, shed on LinkedIn, uh, flow chart of my insanely complex workflows that happen every day in cowork, uh, with multi-agent research agents and, and content writing agents and. Um, calendar checking agents and, and everything, it was all just happening and they're all passing back and forth, running on different sequences and, and workflows. [27:48] Dave Pengelley: I, I didn't need to go and set up open call 'cause I use Claude, but then that means if I wanna stop paying for co-work, then I need to go and rebuild that in another, another infrastructure. But [27:58] Marno Brits: yeah, [27:59] Dave Pengelley: all the, all [00:28:00] the definitions and the rules are built in my found files that I can, I can do that with. [28:04] Matt Slager: To, to Natalia's point before that's. [28:07] Matt Slager: That's part of their product design. They want you to stick, they, they want get their hooks in you and you know, clearly they are building a good product. Uh, it comes down to, um, yeah, what people want really and, and what kind of use they're actually getting out of it. [28:22] Dave Pengelley: Yeah. Hundred [28:23] Marno Brits: percent. [28:24] Dave Pengelley: Yeah. Yeah. Um, well that's, let's, uh, um, I can't think of any other major news that pop, that call has been running the headlines. [28:32] Dave Pengelley: Most of the news and mythos, I think is coming out now in sort of. More public release, we touch on that. Um, did you [28:41] Nataliya Shalygina: expensive, did you cover meta new AI model? I've heard something, but I [28:45] Dave Pengelley: muse [28:45] Nataliya Shalygina: I haven't really had a head chance to dig in. [28:48] Marno Brits: Is that what it's called? [28:48] Matt Slager: Yeah. Muse is actually really good. So like the, the idea of a, of a lab that just vanished off the face of existence and then randomly came back with another state of the art is mm-hmm. [29:00] Matt Slager: Um, [29:00] Matt Slager: interesting. So. Yeah, I think meta and, and, uh, XAI are both, uh, kind of dark horses. You know, they're not fully pumping out the marketing or the, the product level marketing as much as, but um, yeah, muse is actually a literal contender again, where the meta models in the past were kind of rubbish. [29:23] Marno Brits: Mm-hmm. [29:23] Marno Brits: There's also the new one from Microsoft. Not really a new model, but the capability of turning anything into a 3D render. So you would take a photo of it and then it's probably the, the best one. And it's also open source for you to get a complete render of. Like, you can take a photo of a little claw box or a case, and then you can sell that as a 3D print model or you can put into a game model. [29:48] Marno Brits: Um, and it does it incredibly well. [29:51] Nataliya Shalygina: Oh wow. [29:52] Marno Brits: Yeah, [29:52] Nataliya Shalygina: he goes, he goes, uh, product Cooper. Right on all physical things. [29:57] Marno Brits: Actually though it's, yeah. That it's, uh, [00:30:00] we're moving into a new world. [30:00] Nataliya Shalygina: You used to, you used to need to do logistics with like, you know, people in China and be on calls at night. No, no longer that Right World. [30:08] Nataliya Shalygina: Take a photo, Microsoft [30:10] Dave Pengelley: get getting closer to the Star Trek replicator future, right? Where you just sort of bought like the 3D printing and the just in time manufacturing. Where [30:18] Nataliya Shalygina: can you take a photo of your dream house? Mar is it? [30:21] Marno Brits: Oh, that'd be cool. There's a guy less, less AR related, but he freed, he does, um, building using a a 3D printer, but it does cement, so it's just, it just, mm-hmm. [30:33] Marno Brits: Disperses cement and then he did a 3D print of Shrek, but it's Shrek consuming the cement by drinking it, and then he's pooping it out and it's this perfect. Brick layer at the end, and it's just str pooping. We're building a house at the same time. Um, it's great. There's the internet being the internet. [30:52] Dave Pengelley: Amazing. That's terrible. That's a, that sounds like someone had a win in the week. Oh, de I don't think it any of us figuring that story, but, uh, let's, [00:31:00] let's talk about where, how we're going this way. [31:13] Matt Slager: Wind of the week. [31:15] Dave Pengelley: Sorry. [31:16] Matt Slager: Winds of the week. The week, [31:18] Dave Pengelley: the week, the week. Can we talk about harnesses now? Floor in my video instruction? [31:24] Matt Slager: Well, I mean, I can, I can, um, knock the dominoes. I can start off, I, I had my third meeting with that, um, IT company that I was meeting up with. Awesome. Yeah, the guys are very keen and, and we're going ahead, so this is gonna turn into a Yeah. [32:00] Matt Slager: A long term engagement, uh, relationship building and, um, yeah, well they, they wanna slowly start integrating and, and replace some of their existing tools with stuff that I build for them. So, um, it's a lot of pressure and it's kind of spooky, but I feel like the first few [32:00] Matt Slager: phases that we've got lined up look really good and. [32:04] Matt Slager: Yeah. So that's a, that's definitely a win, but it's also, uh, more, more straw on the camel's back as far as stress and, and general work pressure. So, um, yeah, exciting. But, uh, uh, I don't know how I'm gonna manage it. I need more agents. [32:19] Marno Brits: I'm so excited for you, man. Congratulations. That's a, it's a massive, massive. [32:24] Matt Slager: It is. [32:25] Nataliya Shalygina: Congrats Matt. I was looking for the thing, you know, the confetti or something, Dave, but I don't know if we have this on there. [32:31] Marno Brits: I can [32:32] Dave Pengelley: No, no, no. No. Confetti for, for, for that. We, we, we had it on the intro. That was the week. [32:39] Marno Brits: Um, no, bro, I'm so happy for you. So is it, it, because you've got, what, two long-term clients now, right? [32:45] Marno Brits: Or three? [32:46] Matt Slager: Um, my, my roster now is I think at about six. [32:51] Marno Brits: Wow. Okay. And then how are you structuring that throughout the week? [32:55] Matt Slager: I delegate everything to my agents. [32:58] Marno Brits: Right? [32:58] Matt Slager: I literally have [00:33:00] protocols in place where I, like, I start the day and I'm briefed on, you know, what I can and can't think about. Mm-hmm. And you know, what needs to happen today. [33:08] Matt Slager: And then I have my, my various, um, other protocols throughout the day, you know, course correction and just checking where things are at. Like, why haven't you done this yet? Yeah. So yeah, I basically have my own self, um, optimizing loop that I've built to give myself back pressure, not just my agents. [33:28] Marno Brits: Okay, good. [33:28] Matt Slager: Thanks guys. [33:30] Marno Brits: And are you gonna try and, um, leverage some of the community members, like get Blake involved, anything like that? Or do you reckon you'll be able to, [33:36] Matt Slager: do you mean like hiring? [33:38] Marno Brits: Yeah, hiring or subcontracting. Just thinking about like, if I get a big job, you are the first one I wanna call that came out. [34:00] Marno Brits: Help me do this. But who do you call because you don't have another map, so who can you lean on to help you leverage this, especially as you're scaling? I'm just thinking like long-term sustainability. If you already have six, you already have tinkering that you want to be doing, how are you gonna try and leverage [34:00] Marno Brits: that, that resources that you have to make it easier? [34:02] Matt Slager: These are the questions that I'm discussing with my agents. Okay, good. Every day. Um, because I, I literally keep having this moment with them. I'm, um, you know, with my team, you know, with my, my orchestration team at in general, and I go, do we need to hire more humans? You know, is that a thing that it needs to exist and why? [34:22] Matt Slager: You know, what can we actually orchestrate? Yeah. And men, I was thinking exactly the same thing. I just didn't wanna say it. Um, so yeah, the idea of, you know, what actually needs to happen. You know, to your question, mano, like, if, if you were free and you know you were ready to go for something, what I would ask you to do, can I not build a system that does that? [34:46] Matt Slager: And then what, what layers can I continue to retain with everything else? That I don't need to touch delegated. So yeah, that in of itself, if you think of this from like a meta perspective [34:58] Marno Brits: mm-hmm. [34:59] Matt Slager: I'm building my [00:35:00] own agentic business orchestration layer. That's what other people will want as well. [35:05] Dave Pengelley: The, the, the problem and the shortcoming there, Matt is. [35:08] Dave Pengelley: Then all the final decisions, all the thinking, all that actually lives with you, not with the agents. At the end of the day, it's all gonna stack up on you and you are responsible for all of the moving parts. Whereas if you've got a second human that actually thinks like a human has broader context than just your agents, then they can have, see, like you can enable them and empower them to, to run those teams of agents. [35:33] Dave Pengelley: But it actually takes some of the mental burden off you. Whereas if you are spinning all the plates for all the different things, there comes a point where you need to review the work of the robots. You need to double check that. You need to reengage, you need to re-see them. And there comes a limit where you can't do all that. [35:50] Matt Slager: If that's all my day is, then it potentially, it can be, well, [35:54] Dave Pengelley: it's not, you know, you know, you know, it's more than that. [35:57] Marno Brits: I think it comes down to what we spoke about last week. Like, what do you enjoy [00:36:00] doing? If that's what you, if that's your bread and butter, you love this, you are froing on it, go for it. Send it. [36:05] Marno Brits: Um, but if it's not, then try and outsource the things that you don't like doing. But I, you make it very valid. I, you have to train a person. Um, [36:14] Matt Slager: and the idea of orchestrating, or that's not the right word. The idea with collaborating with another human is hilarious. 'cause like, Dave, you said, um, hire a human that thinks like a human. [36:25] Matt Slager: I don't think like a human, I'm a freaking machine. Like [36:28] Marno Brits: a hundred percent. [36:29] Matt Slager: Like yeah. It, it's, it's a weird thing. So like, my, my like personal financial goals is to pull my partner out of full-time work. [36:38] Marno Brits: Yeah. [36:38] Matt Slager: As soon as possible. [36:39] Dave Pengelley: Mm-hmm. [36:40] Matt Slager: You know, so I'm the. I don't know, sole bread winner, whatever you wanna call it. [36:44] Matt Slager: But you know, in that is an opportunity to hire her as a potential admin layer. Um, something [36:52] Nataliya Shalygina: I would advise against [36:54] Matt Slager: it. That's exactly where I'm going with this. And yeah, it wouldn't work because [36:58] Nataliya Shalygina: nothing, not look, [00:37:00] no, nothing against your partner, but working with partners. Geez. [37:04] Marno Brits: Holy [37:04] Matt Slager: crap. Yeah, exactly. But [37:05] Marno Brits: that's, [37:06] Matt Slager: and even like, you know, like, I, like I, I love, I love you guys and you know, I don't think I'd be able to. [37:12] Matt Slager: Work with you guys long term on a thing because it just doesn't, like, my brain's so strange with that. So it [37:18] Marno Brits: work differently. Yeah. [37:19] Matt Slager: Yeah. It would be more the case of like you said, subcontracting. It'd be like, can you just handle that and like we don't communicate. [37:28] Nataliya Shalygina: Uh, Matt, so I second Dave's point, I think there's two valid points here for you to consider, and this is not an advice, not unsolicited advice, but just something to think about. [37:38] Nataliya Shalygina: I think there is a reason. Why there was a jury of 12 on any court case, because diversity of thinking over long term is better. That's why they say, if you wanna go fast, do it alone. If you wanna go far, you have to have a team. [37:55] Stinger VO: Mm-hmm. [37:55] Nataliya Shalygina: Because there is diversity of thinking and yes, maybe, you know, maybe you're not thinking [00:38:00] like a human, but human thinking is still valuable. [38:02] Nataliya Shalygina: We don't necessarily know exactly why. But there is something about it, right? And I don't know if anyone's really got the recipe of what it is that humanity brings. Mm-hmm. Definitely brings something to equation. I think the second point is that, back to what Marna said, if you enjoyed, go ahead and knock yourself out. [38:20] Nataliya Shalygina: But then also we have to remember about balancing. Life and work. [38:25] Matt Slager: No, [38:25] Nataliya Shalygina: I know this is a hard one. I need, it's a hard one. I know, I know, know. But what I'm [38:31] Matt Slager: hearing is we need [38:31] Nataliya Shalygina: to more of ourselves as well. So this [38:34] Matt Slager: is what I'm hearing. Correct me if I'm wrong, you're saying I need to build a jury of 12, um, diverse agent systems that allows me to have that. [38:46] Nataliya Shalygina: And between that, think about balancing that with the humans as well. Because if it's truly diverse, it doesn't just have agents. [38:55] Matt Slager: Mm-hmm. [38:56] Dave Pengelley: As you know, Matt, the agents are gonna inherit any of your biases and things that you [00:39:00] put in there, which maybe you want, and maybe that's great, and maybe you should just do that full send and never see daylight ever again. [39:06] Dave Pengelley: Um, but Sean's saying you should build a system that, you know, get the agents to, to teach [39:12] Marno Brits: a human more work. You, the whole, like the intent that I'm trying to come from Matt is purely as a friend that like, hey. Maybe you get blessed. Like, I found someone that could do websites for me and they do a better job than I would ever do, and they're half the price. [39:25] Marno Brits: I still get money. Awesome. But it's, I've, I've had to find out what do I enjoy doing? What do I not enjoy doing? And the extent of like, just purely care, making sure that you, I enjoying your time on this earth and spending it the way that you want to spend it. Um. But I mean, massive congratulations that you're got another customer, like you're getting good problems, which is fucking awesome. [39:44] Marno Brits: So congrats. Yeah. [39:46] Dave Pengelley: Language. [39:49] Marno Brits: It's a pop talk, isn't it? [39:52] Dave Pengelley: The one slot. The kids are at school. [39:55] Marno Brits: I'm in K. Sorry. This is, this is the vocab. And kgo. The [00:40:00] tradies. [40:01] Matt Slager: All right. Enough of me. Next? [40:05] Marno Brits: Um, yeah, I, I turned 30, I am now in the next decade. And the moaning, like the morning we started with me turning 30. I went for a walk and my left hip was so sore and I'm like, what a great start because I only had hippo, but I just woke up and it was so painful. [40:22] Marno Brits: I was like, I'm literally limping as I'm walking. But, um, [40:26] Dave Pengelley: yeah, forties mate. [40:27] Marno Brits: Come again. [40:28] Dave Pengelley: Wait till you hit your forties. [40:29] Marno Brits: Nah, I've got, I've got peptides by then and I've got a lot more money to do some IV drips and then I'll be sorted. I already have the protocol in place. I just need the funding. Um, but yeah, I'm just, I'm excited about that. [40:40] Marno Brits: I'm excited to start the next level, level 30. [40:43] Dave Pengelley: Yeah. And [40:44] Marno Brits: yeah, stepping into a more, more mature approach to business in that I've spent the past two and a half weeks. Really structuring it to launch adaptive ai, like stepping away from the previous business model and nicheing more into exactly who do I serve and taking [00:41:00] the approach of being a lot more involved in the community. [41:02] Marno Brits: So rather than trying to build something and spending all this time trying to optimize my environment, like I need to pay the bills, um, and I have the same goal as Matt of I just need to retire my wife as soon as possible. So really just hit the ground running and take the opportunities that are coming up for being very smart and having a filter. [41:21] Marno Brits: Of, okay, I'm not just gonna accept any job, but who in my life am I going to invest the next 10 years in? Like who, I've got 10 people that I can choose from and who am I investing 10 years into that I want to help them grow and see them grow. So, um, taking that approach has been really beneficial even in the past a week and a half of just making that decision where as I'm meeting people and just being curious about how I can contribute to their life, um, has returned tenfold, which is awesome. [41:48] Matt Slager: That's so good. [41:49] Marno Brits: Mm-hmm. [41:49] Dave Pengelley: Yeah. And you've got, you've got a new business name, adaptive ai. I, I haven't caught up on that. Like you're not Yeah. All the AI anymore. [41:54] Marno Brits: Yeah. Adaptive AI solutions now. So pivoting more into the education space and [00:42:00] facilitating the environment for people to learn how to fish instead of fishing. [42:03] Marno Brits: Um, we're giving them the fish. [42:05] Dave Pengelley: Mm-hmm. [42:05] Marno Brits: And I've got a couple clients, they've got one in two hours that I'll be sitting with just learning about the business and helping them understand how they can leverage ai, because I'm the bottleneck. I can't solve everyone's problems. But if I can teach everyone about it, they can solve their own problems, which is awesome. [42:20] Marno Brits: Interesting. [42:21] Matt Slager: Um, [42:21] Marno Brits: yeah, I like it a lot. [42:23] Matt Slager: That's so cool. I, I love the, the philosophy, um, between like behind the name, like adaptive as well, you know, like it's so good. It's like future pacing. [42:32] Marno Brits: Yeah. It took me a bloody long time to find it. I wanted to get adaptive ai and then I wanted to get simpler ai, but any AI domain is like 70 grand to 190 grand. [42:43] Marno Brits: Yeah, for one domain for the year, which is atrocious. [42:46] Dave Pengelley: There's some little island in the Pacific somewhere, I can't remember where it is that that owns, like that's their country domain. Like au like ai, I can't remember the name of the country. Yeah, they that, that is like, it's like 10 x their GDP people buying AI domain names. [42:59] Dave Pengelley: It's crazy. [43:00] Matt Slager: So hang on. Elephant in the room. Natalia, did you spend nine grand on your domain? [43:05] Nataliya Shalygina: No. No. And I dunno how it happened. Maybe I just got lucky. [43:09] Marno Brits: Yeah, [43:09] Nataliya Shalygina: it was $270 per year. [43:11] Marno Brits: Yeah. It was the same with Kali ai. I bought it, what, two years ago now? Um, and it was like a hundred bucks. But now if you go back into it, any AI domain is atrociously expensive [43:23] Dave Pengelley: coming old fashion. [43:23] Dave Pengelley: I still say a good.com, which is hard to do as well because every, every two word combination on the planet is pretty much been bought out. And, and someone's just sort of, and then you [43:34] Marno Brits: also have to buy au to make, [43:36] Dave Pengelley: I was, I was amazed when I was trying to find the name for my, my FX trader. Um, journaling analytics platform that I got, chart reporter.com. [43:45] Dave Pengelley: I was like, what? How did I get those two words? Um, that was, that was a coup. [43:51] Marno Brits: You might make more money off of selling the domain, like give it a [43:54] Nataliya Shalygina: year. Yeah, I was just gonna say, just do an agent to search for those empty ones and keep buying them. That's the [00:44:00] perfect case for open claw, right? [44:01] Marno Brits: Mm-hmm. Yeah. [44:04] Nataliya Shalygina: And resell them [44:05] Dave Pengelley: indeed. Mm-hmm. What about you, Natalia? Any, uh, any wins of the week for you? [44:10] Nataliya Shalygina: Yes, well, I have been, funny enough, we've been talking about mixture of things here. Uh, adaptive and training teams is probably my priority number one in my day job anyway. Uh, so, um, enterprise rollouts of AI across 500 plus people is something that occupies my daytime. [44:31] Nataliya Shalygina: Yeah. And of course there's a lot of questions that are unanswered, not just an enterprise, but also small business, uh, level. And they, they mostly across adoption, governance, security. [44:43] Marno Brits: Mm-hmm. [44:43] Nataliya Shalygina: So I've started the YouTube channel as a dare following Africa ai. So I went to Cape Town, met Liam. Oh, [44:50] Marno Brits: wow. [44:50] Nataliya Shalygina: And they had, they, they had the jokes at, on the podium, so free speakers joked about the same thing, that if they do a podcast on change management, despite of how [00:45:00] important it is, no one would listen to it. [45:02] Nataliya Shalygina: And I thought, well, I have no YouTube channel. I have nothing to lose. Literally, I've been doing change management for 20 years. I might as well start it. [45:09] Marno Brits: That's right. [45:10] Nataliya Shalygina: So, so I've started it, but then beginner's luck, like I'm literally have conversations with. You know, New York bestseller last week. I've got no one lined up New York number one advisor to big firms. [45:24] Nataliya Shalygina: And I'm treating it as a, as a beginner's luck 'cause I dunno why, but it's, it's happening and it's, it's really enjoyable. Um, so yeah, that's probably my biggest win is just being on the roll with creating YouTube content that I actually apply myself. So that's my biggest one. [45:41] Marno Brits: I love that for you. Well, the, um, there's a good guy called Dan Co. [45:45] Marno Brits: And I love his philosophy in life and it's, you are the niche, so you've already done the work. Why not capitalize on that? Like you are the niche. Just talk about what you enjoy and I think you're selling probably a very specific pain point. [45:56] Nataliya Shalygina: Oh, thank you. Brought it up. So yeah, the [00:46:00] AI Ready Teams is the one with Charlene Lee, so she's in New York Bestseller. [46:04] Nataliya Shalygina: I've got another one that I've just confirmed today. I'm not gonna disclose the name, but he's huge in New York as well, in America. So that's the next one coming up. But yeah, it's just a matter of where do I get the time to do all of them, and it's really exciting. [46:17] Dave Pengelley: Well, well thank, thank you for giving us some of your time today. [46:20] Dave Pengelley: Um, but that's great. If you, if you wanna see more about those interviews, make sure you follow at a i Fitness on YouTube and her podcast. They, we'll put a link in the description after the show as well. Um, thanks. But, uh, brilliant. Yeah, I mean that's, that's great. More people out there talking about these sort of things, change management's critical. [46:41] Dave Pengelley: I was talking to organization and, um. It's sort of a, a good news, bad news story for me as far as one of the week. And then I, I kind of won the job, but then they deferred their budget to the new financial year, so I didn't actually get the job. Um, but these things happen that that's why you have a pipeline, right? [47:00] Dave Pengelley: Um, but. You [47:00] Dave Pengelley: know, just talking to them and talking to other organizations that I've been working with, change management is such a critical part of any of this AI staff. Otherwise they, they agent here, an agent there, or they do this and they don't that, but then they don't get the adoption or they get people adopting it, but they're not really getting any benefits from it 'cause they don't understand how they're abusing it, why it's got benefits, et cetera. [47:18] Dave Pengelley: So being able to translate any of these, these sort of AI transformation programs into the change management piece is a critical part of success. Right. [47:28] Nataliya Shalygina: Yeah. Yeah. And that's what I've kind of been weighing as well in, 'cause I love AI solutions and I build AI solutions and they're super exciting. But at the same time, I know change management is critical gap. [47:39] Nataliya Shalygina: So. I'm kind of in a phase where like I kind of do both, but then should I just focus? I dunno. That's probably one question I need to ask in the coming months. [47:48] Marno Brits: That's so much fun. I feel like to answer that one, or to give some insight, I was talking to, I think Matt, about that, of like, I need a niche down. [48:00] Marno Brits: I need to find out like what do I enjoy doing? And he was like, well here's argument is, well, we're [48:00] Marno Brits: living in this space now where you can, you can leverage your time and amplify what you can, what you can achieve on such a level. Is there really a need to, to niche down? Like you might just become a generalist, but because you can leverage ai, you can achieve so much more. [48:12] Marno Brits: You don't need to spend 80 years on one topic. You can spend two years because AI can help you amplify that. Um, which messed up my entire approach to life. I'm like, no, I have to niche down. And then that's like, do you though? Yeah. [48:25] Matt Slager: I had a, I had a really cool reflection moment this morning with, um. With one of my knowledge agents dealing with my core and philosophy, and it, it said, hang on a second, I just need to pull something out and be blunt here. [48:36] Matt Slager: You're asking for, uh, a way to label yourself, you know, in a, in a vision sense, in a goal sense. Um, but what you've not realized is you, you, you've, you've done the things that that person would do already. So like you've done it backwards, essentially. And so with that kind of thing, Natalia, I reckon the coolest thing that. [49:00] Matt Slager: You say that you got beginner's luck. It, it's [49:00] Matt Slager: more, I think, personally similar to what Minea just said. I think it's because you've, you're aligned, like you feel like you're moving in the right direction and there's, there's a, there's goal-based, um, you know, direction choosing, and then there's constraint based direction choosing. [49:16] Matt Slager: So if, you know, uni want to get to a a point. You can say that's the point I'm getting to, but there's no like specific route to get there. You're going everywhere and it probably will take a little bit of time. You know, it goes few wrong directions and then you'll get there eventually. But then the constraint based one is like narrowing that path choice, but still allowing some deviation. [49:35] Matt Slager: I like the idea of moving towards where you want to go, but not. Completely constraining you and saying, this is my niche. You know, like you can do that through marketing, but for your own personal alignment, like you need a little bit of experimentation and, and the ability to wander. [49:52] Nataliya Shalygina: Yeah. Yeah. And I think, Matt, you, you mirror what usually successful people say is that the biggest come from like edge cases. [49:59] Stinger VO: Mm-hmm. [49:59] Nataliya Shalygina: [00:50:00] And like what? They don't expect to bring a lot of results. They just. Go on this wild chase one day and just do something off the side and then boom. That's like the biggest client or something. [50:09] Stinger VO: Yeah. [50:09] Nataliya Shalygina: So it's quite interesting you you're saying that like I'm kind of just as you said, going with some things and going, yeah, that's killing my time. [50:16] Nataliya Shalygina: But who knows? You never know. [50:18] Matt Slager: Yeah. You keep moving [50:20] Nataliya Shalygina: as long as you're moving forward. [50:23] Matt Slager: Yeah. [50:24] Nataliya Shalygina: Yeah, [50:24] Matt Slager: a hundred percent like that, that client I mentioned before that, you know, I've just gone through three meetings and finally like, um, winning, winning them. I met the person years ago. I used to deliver coffee. I, I worked at a coffee roastery for a specialty coffee roaster, and I used to jump in the van with buckets full of coffee beans and drop 'em off to places, this company, and with one of the places, I'd drop 'em off a couple of kilos a week and I'd see the guy and I'd message him, be like, do you need any coffee this week? [51:00] Matt Slager: And then I, I bumped into him on a charity walk a couple of weeks ago. Right. [51:00] Matt Slager: And I went on this big 20 5K walk for men's mental health. And I, and I pun to this guy, and he called me some other name. I can't remember what he said, and I'm like, nah, that's not me. And I just kind of kept walking and I bumped into him again and he's like, it's Matt, isn't it? [51:13] Matt Slager: I'm like, yeah, [51:14] Nataliya Shalygina: serendipity. [51:16] Matt Slager: Yeah. He goes, he used to drop off coffee. I'm like, oh, hello. [51:20] Nataliya Shalygina: Does he then introduce you to go, oh, this is used to be a coffee guy. He does now AI solutions now. [51:25] Matt Slager: Yeah. Literally. So pretty much, yeah. Have a really big chat and, you know, the, the, the weird line of events for me to bump into him twice on that walk. [51:34] Matt Slager: And then, you know, the, the previous rapport I had from, you know, that previous job. You know, all of these things lined up and none of those things could have been predetermined. Yeah. So, yeah. [51:45] Nataliya Shalygina: Well, when you meet, when you meet them face to face, you know what you must do. [51:49] Marno Brits: What's that? [51:51] Nataliya Shalygina: Bring them coffee, coffee. [51:53] Marno Brits: Oh wow. Group the bed. [51:57] Nataliya Shalygina: Have coffee beans with your, with your [00:52:00] slate tax systems. Something. I don't know. [52:02] Matt Slager: That would be actually kind of fun. Yeah, [52:04] Nataliya Shalygina: yeah, [52:07] Matt Slager: yeah. You can, you can take the, the man out of the coffee roastery, but you can't take the coffee. [52:12] Nataliya Shalygina: That's not, that's not whiskey, is it? What are you drinking? [52:15] Matt Slager: Well, it kind of looks like it. No. Um, I, I just do cold brew at home. Because it's lazy and it's leveraging time. [52:22] Marno Brits: Good man. [52:24] Dave Pengelley: Cold brews good because it takes a business out. I don't mind. A cold brew on occasion and it does taste whiskey, so it does have that smoky sort of feel to it. [52:33] Dave Pengelley: Cold [52:33] Nataliya Shalygina: brew, you know there is coffee that's been roasted or like rested in like whiskey barrels. [52:39] Nataliya Shalygina: They Melbourne and it tastes like rum a whiskey. I have to find the name and I'll, I'll send it. Sounds good. I'll send it to you guys in the WhatsApp chat. Yeah. [52:47] Matt Slager: Awesome. [52:48] Dave Pengelley: Yeah. Nice. Yeah. I haven't done Cobra for a long time, but when I was big in the keto a while back I was doing the, doing the Cobra. Now I'm like, I don't do keto, but I'll still, you know, put my milk lab almond in because it's pretty, [00:53:00] it's low carbon, enough [53:00] Matt Slager: caffeine. [53:03] Dave Pengelley: Um. Yeah. Brilliant. Well, we, we, we have run out of time to actually do our deep dive on harnesses on, on like, we, we we're so busy talking about [53:11] Matt Slager: thought that was intentional. [53:12] Nataliya Shalygina: That's my fault. I'm sorry you invite the change manager online. No, [53:17] Dave Pengelley: we were, we were busy, uh, counseling Matt on, on, uh, hiring teams. [53:21] Matt Slager: I appreciate too. [53:23] Matt Slager: Thank you guys. [53:24] Dave Pengelley: Um, I, I, I did, I did pull up an image though that I wanted to show around harnesses. So we, we can maybe dive into this more next week, boys. Um. But, uh, [53:32] Matt Slager: tune in next time. [53:33] Dave Pengelley: Yeah, tune in. Tune in next time. As we discussed dog harnesses. Um, what, so if you think, think about harnessing, like you think about animals and harnessing animals and how different styles of harnessing give you different impacts. [54:00] Dave Pengelley: And sometimes they're good in different situations, and so people say there's no right or wrong way to, you know, hitch up your dogs. With their harnesses, but depending on the environment, so the fan hitches can be good on, I see a flatter, more open terrain. But if you're going through woods and forests, you don't want your [54:00] Dave Pengelley: dogs all out in a fan. [54:01] Dave Pengelley: You want 'em in a straight line. And so [54:02] Marno Brits: yeah, [54:02] Dave Pengelley: they'll, they'll pull differently, right? Not that I've ever been dog sledding in my entire life, but this is what I read online. And so your ai, the way you harness it, is gonna have a different impact. Um, and different, depending what you're trying to achieve with it. [54:17] Dave Pengelley: Right. Um, [54:18] Matt Slager: can I just, I just wanna take a moment to express my sincerest gratitude and admiration of the weird analogy making a hundred [54:27] Marno Brits: percent. That's a great one, but I was not expecting that at all. I love [54:32] Nataliya Shalygina: I love it too. I love it too, Dave. That was a good one. [54:35] Dave Pengelley: I was think, I was thinking harnessing and you harness animals and I was thinking a bit about dogs and I was like, yeah, I remember hearing there's different ways. [54:40] Dave Pengelley: There's that, that fan and. They call it hitching, but you harness the dogs and then you hitch them up in a certain order. [54:46] Marno Brits: Is this gonna be your, um, Ralph Loop methodology? You are gonna somehow turn this into a methodology or framework that you can teach ai. [54:53] Matt Slager: You'll have a word, you'll have a slash command at the beginning, and it's just mush. [54:59] Marno Brits: [00:55:00] Yes. That's great. [55:01] Dave Pengelley: Yeah. No, I think, I think it's just, I've been thinking about this whole harness thing and I've been having conversations with people on LinkedIn and people saying, oh, how do you manage the memory? Have you found the Ford code or this or that different? I'm like. You can manage a lot of that just through your agent command. [55:13] Dave Pengelley: That's a file system thing. It doesn't have to be in the harness. You can extract some of this stuff outta the harness and that you've, like, you've built your own harness. I'm living in Claude code slash codex. I've used open code, I've used all these different things. I've tried Zen Flow, which was all right, but I've noticed it put massive amounts of extra system prompt stuff at the top. [55:31] Dave Pengelley: So it was actually jamming at my context with extra things, which they're all doing to a certain extent. Um, but yeah, I've just been thinking about the harnessing a lot, and I put this out on, on LinkedIn as well. You know, not animals, but talking about engines and sort of, sort of saying you can have your different, different, same engine, different sort of setup where you're putting in like factory machinery or hypercar or go-kart opus 4.6. [56:00] Dave Pengelley: It might be the same engine, but depending on how you harness it and decide to utilize it, you're gonna get very [56:00] Dave Pengelley: different results. Um, and people are sort of quick to sort of blame everything on the model, but no one's actually really talking about the harnesses yet. [56:07] Matt Slager: Yeah. [56:08] Nataliya Shalygina: And when you do talk about harness next week, please think about which people in which teams would represent those harness as well. [56:16] Nataliya Shalygina: 'cause I think that's an interesting spin. 'cause we talk about systems a lot, but in reality, if an if in organization, no one's kind of mimicking this, harnesses and re and responsibility and accountability, none of those guardrails really. You know, it's still valid, but it is important. [56:36] Matt Slager: Another topic for next week. [56:37] Dave Pengelley: Yeah. Alright, we we're at time. We're at one o'clock. I wanna thank you Natalia, for, for joining us. Uh, I know you're a bit of a last minute audible, but I realized Richard's away in, in Indonesia, um, surfing for this week and next week. So, so, uh, I was like, you know what, I, I would like to get some, some new, new fresh blood on, uh, a new, I've seen your stuff on LinkedIn that you're posting your interviews. [57:00] Dave Pengelley: I was like, oh, she's, she's building her YouTube media side. I'll, I'll think about this. She wants to. [57:00] Dave Pengelley: To join us. [57:00] Nataliya Shalygina: Come on. Yeah. Yeah. Thank you [57:02] Dave Pengelley: so much. Well, [57:03] Nataliya Shalygina: thank you so much for having me. That was pleasure. [57:06] Dave Pengelley: Absolutely. Thank you Matt. Thank you. Mano, thank you to all our people that joined on the chat to, to Manjula, to Kalu, to Sean, to Joe. [57:16] Dave Pengelley: Think that's everyone. This week we jumped in on the comments and the chat, so, um. Thank you everyone. Great to have you live. Um, check out the YouTube channel, make sure you like and subscribe. Hit the little bell icons. I get notifications when we go live and we are now starting to break up these sections, like the news and the wind, et cetera, as separate things through the week. [57:35] Dave Pengelley: So if you don't wanna watch the full hour in one hit, if it's too much, you can grab our shorter sections that we're clipping up, so, and, and share those around with your network. So thank that. We'll see you all next week. Thank you all. Goodbye. [57:48] Marno Brits: Bye. [57:49] Nataliya Shalygina: See ya. [00:58:00] Bye.