Tokenomics: Why Your $120 AI Plan Is About to Break

[00:00] Dave Pengelley: Anthropic's Mythos, Mila's Mem Palaces, Marcus Aurelius and Men's Fashion. We cover it all on the AI Operators Show. So if you are keen to hear four blokes talking about real world use cases of ai, what's happening in the news and the headlines and how it applies to you and your business, stay tuned and make sure to join us sometime in the live show so you can jump in the chat and ask us questions directly. [00:25] Dave Pengelley: But enough of me, enjoy the show. [00:27] Dave Pengelley: All right, we're live boys. [00:31] Matt Slager: Hello. Hello. [00:31] Dave Pengelley: You're, oh, you're with Mr. Bighead here. Uh, let me, let me fix that. I can't do it when Richard joins. Richard's running a few minutes late and then we'll, then I'll, we'll be breaking into our little small windows, but for now, there we go. Now I'm about the same size as, as you guys had. [00:46] Matt Slager: Amazing. [00:47] Dave Pengelley: Yeah. [00:48] Matt Slager: So much better. [00:49] Dave Pengelley: So much better. And now I've got, I've still got a bigger head than, than I, I think I just have massive head. Like, look, look at, look at the, the space. I should, I should be selling advertising on my for, I reckon. [00:59] Marno Brits: [00:01:00] Did you see that blog that sold advertising for his suit, for his wedding? [01:03] Marno Brits: So everyone paid to be a sponsorship for his wedding, and then it was just his entire suit covered in AI companies and random companies, I think. And therefore he got his whole wedding paid for and his honeymoon, and he just had to wear a suit covered in logos. [01:17] Matt Slager: Wow. [01:18] Dave Pengelley: That is [01:19] Marno Brits: ridiculous. Yeah. [01:19] Matt Slager: A vaguely recall that maybe I'm just making it up. [01:22] Marno Brits: There's even a bloke that did it for his toilet. Like, he's like, cool. But, um, I think he's like selling, selling space in his toilet and then he'll project it on his toilet and then he spends time like, and he is showing how it is. He create the logo deferred perfectly on that spot. And there's one for like the bowl and the one for the back of it, and people are paying money for it. [01:42] Marno Brits: 'cause it's just getting so much traction from the internet. Who [01:45] Dave Pengelley: are these people? Do they already have a following? Are they influencers? Do they have some of met? They're like, I've, I've got five people on my Instagram, I've gotta sell, sell space on my, on my, on my foot, [01:54] Marno Brits: basically. Yeah. It's that like just a random old normal bloke that's like, Hey, I wanna try this. [02:00] Marno Brits: And then it [02:00] Marno Brits: worked. [02:00] Dave Pengelley: That makes me angry. Having like, you know, trying, trying to actually build real content for real people with real value like we do on shows like this. And it's, it's so hard to build the audience and to get started and to get people to find you and then you get people just doing ridiculous valueless nonsense and they just get giant followings. [02:19] Dave Pengelley: It, it does make me a little bit mad. [02:21] Matt Slager: I still something. I can't remember if it was paid or not. It probably was, and this will probably make you mad as well. Um, it was a, a little add-on like plugin thing that you could do to, to download and you, you get a a, a whip on your mouse cursor and you can use it to whip cord code. [02:38] Marno Brits: That's [02:40] Dave Pengelley: great. Are people paying for that? 'cause I saw like a, a screenshot of it on LinkedIn. [02:44] Matt Slager: I think so. I don't know. I I just scrolled past it quickly. That's [02:48] Marno Brits: fine. [02:48] Matt Slager: Without getting too disturbed, [02:51] Marno Brits: do either of you do much in stoicism, like read much Stoicism or Daily Stoic stuff like that? [02:57] Matt Slager: Marcus is my boy. [02:59] Marno Brits: Hey. [00:03:00] So, um, yeah, in the Daily Stoic, there's one in February that I've pin mocked, like I've literally folded half the page. I, whenever I opened up the book, it goes there. It's called The Banquet of Life. Um, and he talks about how everything will come in its own time. So everyone has the banquet of life when everything just goes well, and then you get the new car and you get the new job and you get all the money coming in, and then you can't reach over to steal someone else's while they've got their banquet. [03:26] Marno Brits: You have to wait your turn until they come to you. So just keep being you basically until that banquet comes around the table and then it's your turn. But then don't hold onto it when you have it because it's not yours forever. [03:37] Matt Slager: Interesting. How does that apply to whipping cord code? [03:41] Marno Brits: Not whipping cord code, but like getting mad at other people succeeding because it's like successful [03:45] Dave Pengelley: and the cap, the capitalist in me goes, you know what? [04:00] Dave Pengelley: Supply and demand. And if they're supplying something, there's a demand. Like, I get it, I get it. I don't have to like it. My, my own personal framework of execution and providing value is clearly different from theirs, [04:00] Dave Pengelley: but, uh, they've, they've found something and they've leveraged it very quickly. [04:03] Marno Brits: Mm-hmm. [04:04] Matt Slager: Speaking of, um, demand, what's the highest fuel prices you guys have seen? [04:09] Marno Brits: Oh, I honestly, I, I've not looked at fuel prices for about six years. Um, so I can't even give you a price. I just know it's expensive. [04:19] Matt Slager: But they don't have fuel. In kgo [04:20] Marno Brits: We do, but I've got a budget. I'm like $50 a week. They're all [04:24] Marno Brits: on [04:24] Dave Pengelley: EVs over there. [04:24] Marno Brits: As long as I, if I, if I budget it, then it's just that. But yeah, there's a lot of EVs, there's a lot of EVs. [04:29] Dave Pengelley: Andrew Forrest is just electrified everything. [04:32] Marno Brits: That's it. Yeah. Um, I, yeah, yeah. I fortunately, um, never really paid attention [04:38] Dave Pengelley: to. Okay. Well, well, you're, you, you're out there, man. Uh, I've seen it. Um, over $3, $3 10 for diesel, maybe three 20, and then it drop back down a little bit, but over three. [04:48] Dave Pengelley: Well, you are down in the country, and I'm gonna bring Rich up. He, he can weigh in on fuel prices, but, uh, Hey, Richard, we're just talking fuel prices and how crazy they are. Matt, what, what have you seen? [04:58] Matt Slager: So, over Easter, [00:05:00] my, uh, my younger brother came down to go for a few motorbike trips. Mm-hmm. And he brought back a photo of a country town that had diesel 3 97. [05:08] Dave Pengelley: Oh my goodness. [05:09] Marno Brits: Wow. [05:11] Matt Slager: Yeah, it was pretty good. [05:12] Richard Webbe: That is ridiculous. [05:14] Marno Brits: That's crazy. How [05:15] Matt Slager: far outta [05:16] Marno Brits: the town was it? [05:18] Matt Slager: Oh, it was in like the, like Victorian mountains. [05:22] Dave Pengelley: So is the price of ev top ops going up when you see those EV stations in the middle of the bush that are powered by diesel, um, generators, I wonder? [05:30] Matt Slager: Sure. They have to [05:31] Dave Pengelley: be, [05:32] Matt Slager: they'd have to be, they'd [05:33] Marno Brits: have to be, yeah. [05:34] Matt Slager: It kind of, it's the, it's the same, um, concept of like the subsidized subscriptions to these AI companies. Like, you know, as all the costs rise, they all of a sudden seem to be cutting back a quota. So yeah, it's gonna have some sort of similar impact for sure. [05:51] Richard Webbe: Cutting back trust to answer me this. [05:53] Dave Pengelley: Yeah. [05:54] Richard Webbe: Why isn't the government or individuals using AI to triage where prices are going? [00:06:00] Why they're going, when they're going, and when they're coming down? [06:02] Dave Pengelley: This is not a political podcast, Richard. [06:06] Richard Webbe: I'm [06:06] Dave Pengelley: the current, the current government. I'll get in big trouble if I start providing comment on the, uh, the Albanese government. [06:13] Richard Webbe: That was a technical question though. If you think about it, any government should already have their computer scientists working on a strategy to understand where things are going, how long they gonna run out of I, I don't understand why [06:26] Dave Pengelley: they're, they're too busy trying to enforce, impossible to enforce rules around age limits on social. [06:32] Dave Pengelley: Yeah. [06:32] Matt Slager: Stopping kids from using Instagram. [06:36] Marno Brits: Have a real problem. It's always the incentives, like what incentive do they have to make it easier for us? They get more money by doing that. [06:43] Matt Slager: Good [06:43] Marno Brits: point. But I do wanna point out, Richard, I love to get up today. It looks like you just stepped out of a Silicon Valley meeting. [06:49] Marno Brits: Like you've got the nice shirt, it's just, yeah, it blends bomb mate. Good work. [06:53] Richard Webbe: I need to explain while I'm a few minutes late, humblest apologies. It's school holidays in Melbourne [00:07:00] and I didn't realize that my son had a play date this morning. I very quickly got up and said, oh, frozen, I'll have to, uh, I'll keep talking and then I'll go. [07:09] Richard Webbe: Um, but uh, I know why it's freezing, but I can fix that in about a second. There we go. We're back on. For some reason my camera ran and so as I was getting ready for this podcast, and, and I'm glad you sent me a note. Hey, thank you. The other boy turned up and I said, oh, well let's go for a, um, okay, we'll try and squeeze in a quick bike ride now what's really good and something. [07:33] Richard Webbe: And years ago is, we have free transport in Victoria at the moment. I said to them, my son said he wanted some cream from the local, uh, shopping center. So I said, okay, we'll go for a quick bike ride. We'll get you some ice creams from there. Uh, it's a bit further, so we'll jump on the train. We'll take the bikes. [08:00] Richard Webbe: We'll do, of course it's like herding cats. It is like herding cats trying to get back. Now part of my bike riding gear is also part of my surfing gear, so I'm not gonna be here for the next [08:00] Richard Webbe: two weeks because I'll be in the jungles of Indonesia and I double up on my mm-hmm. Surfing rash vests with my bike gear. [08:08] Richard Webbe: 'cause it's made the same material that it achieves the same thing. So my bike gear on and I threw the jacket over the top so I may look trendy. I'm actually just really weird. [08:18] Dave Pengelley: Did, did you ever see the old show Greatest American Hero? [08:21] Richard Webbe: Yes. My favorite show. [08:23] Dave Pengelley: Me too. Um, that, that's why we get along, Richard, because look, you've got the same kind of orangey red suit on the camera. [08:27] Dave Pengelley: It looks a little bit like you got Ralph's, [08:29] Richard Webbe: believe it or not. [08:30] Dave Pengelley: Believe it or, yeah. [08:31] Richard Webbe: The song, believe it or not, [08:32] Dave Pengelley: believe it or not, Jerry Rafferty very good. Uh, Joey Scarbury, sorry. Joey Scarbury. Gerry Rafferty was Baker Street Yes. Joey Scarbury. [08:39] Richard Webbe: But we are, I am actually doing, I'm going to talk to some guys about some ai uh, on tracking waves around the world while I'm surfing in Indonesia. [08:47] Richard Webbe: Ooh, interesting. [08:48] Dave Pengelley: Because they are using it for shark detection. [08:50] Richard Webbe: They are indeed. They, indeed. [08:52] Dave Pengelley: Anyway, we're gonna talk about ai, so we should probably do one of the, do I have the [00:09:00] thing? Uh, yes. [09:01] Marno Brits: Good morning, Sean, by the, [09:13] Dave Pengelley: yes, we're the AI operators. Sherwin. We are here to talk about ai, not just men's fashion. [09:19] Richard Webbe: We're so professional. Cut off then. [09:23] Marno Brits: Yeah. This, yeah, [09:24] Dave Pengelley: that's there. Sorry man. You know, production, you gotta gotta make the cuts when you gotta make them. What, what, what did I cut you off from saying with [09:30] Marno Brits: this? I was saying good morning to Sean and then I saw that as I talked the, when the clip came up, it muted our mic. [09:37] Marno Brits: Yes. But I love that it commuted there, but not at the end of the stream when you are like, okay, no one talk please. No one talk. And then everyone needs to talk. That's great. [09:46] Dave Pengelley: I've gotta say, we, we use restream and props to Restream. We're not sponsored by them. We pay them plenty of money to have the use of their service, which, uh, allows us to all dial in and then push this out. [09:58] Richard Webbe: Mm-hmm. [09:58] Dave Pengelley: Um, but [00:10:00] they used to have a feature where you could bring up those little stinger clips and then you, I have a little media palette and I could upload videos and click at them and it would just overtake the whole screen. Like, like we just saw, play the clip, mute all our mics, and then come back to us. [10:11] Dave Pengelley: They took that away and they, they gave us this weird show flow thing on the left there where you can set up videos as a whole section and then it auto completes onto the next one, which is what we used to be doing for the intros. But it was always a bit buggy. And if you wanted multiple different like stingers, like we've started to introduce into this show for sections. [10:28] Dave Pengelley: Mm-hmm. It doesn't work. 'cause then you've gotta have us as a show section like eight times duplicated with a, you know, this is inside baseball for how to run a show. But I was really frustrated and I'd gone to their help desk knowledge base and asked a few times and got the ai. But then a real person seemingly emailed me off the back of my incompleted chat on, on with their AI chat bot asking me what I needed. [11:00] Dave Pengelley: And I told them and they're like, yeah, sorry. We got rid of that. It was unreliable. We like this thing better. I was like, I'm not even asking for new features. I just want the old feature back. And they're like, yeah, sorry. We understand, but [11:00] Dave Pengelley: get stuff. [11:01] Richard Webbe: So like [11:01] Dave Pengelley: an [11:01] Richard Webbe: iPhone upgrade [11:02] Dave Pengelley: and then a day later I get another email going, oh, I've talked to tech Support and I've turned the feature back on in your account. [11:08] Dave Pengelley: I'm like, oh, which is how we can have our stingers again and it makes me so happy. So at the end, we're still using the old way where it doesn't automatically mute mics because that's just, you know, we will end the show out and go on that one because otherwise if I play this one, it'll jump straight back to us before we end the stream and we'll be like, we're over. [11:24] Dave Pengelley: Goodbye. And then bang, we're back here. It's like, oh, sorry. I'm still pressing the button. [11:28] Richard Webbe: Can I ask an AI question from The Brains Trust? [11:31] Dave Pengelley: Yes. Go [11:31] Richard Webbe: for it. [11:32] Dave Pengelley: Please do. [11:33] Richard Webbe: What happened with Anthropic a few days ago? Oh, the source code on the web and people copying it and vibe coding it. Someone explain what happened for our audience and what does it mean? [11:44] Dave Pengelley: Yes. Thank you, Richard. Because we, this is a topic we didn't really touch on last week. It was really fresh. I think it happened last Wednesday morning just before it went to air. And so we didn't really discuss it at all, but I think, um, we should discuss it. There's a few other things happening in the news around anthropic. [12:00] Dave Pengelley: Uh, and if we're talking about the [12:00] Dave Pengelley: news, then that means we need to talk about the news. [12:06] AI VO: Welcome to the AI update. Let's look at what's happening in the news. [12:14] Richard Webbe: Well, that was very punchy. [12:15] Dave Pengelley: Let's, let's look at what's happening in the news. Uh, who wants to go first? You wanna talk about the Claude Code Source leak? Matt, do you feel like you have things to say about that one? [12:26] Matt Slager: Oh, I have things to say, but it, I don't know how it happened. [12:29] Matt Slager: I, I almost speculate that it happened like maybe halfway on purpose. Um, so that they could, you know, instigate certain things, you know. [12:40] Richard Webbe: So let's talk about what happened in layman's. [12:44] Matt Slager: Sorry about that. [12:45] Richard Webbe: Let's talk about what happened in layman's terms. [12:47] Dave Pengelley: So, so from A-C-I-C-D GitHub point of view, Matt and I don't fully understand what the MAP file is and some of that GI technical stuff. [12:53] Dave Pengelley: Do you have a handle on the map file that got leaked, which had the code in it? [12:58] Matt Slager: Just think of it [00:13:00] as the, the posties or freight forwarding looking at your packages as it's getting sent through the terminal. That, that's basically it. Like [13:07] Richard Webbe: so their code got leaked on the web? [13:09] Matt Slager: Yeah, pretty much the AI [13:11] Richard Webbe: source code [13:12] Matt Slager: from internally, people like made, you know, infinite copies of it in that current state. [13:18] Matt Slager: Um, there's no way they could have shut it down. They, you know, put out all the DMCA requests, you know, cease and desist kind of stuff to everyone that, um, was posting things about it. Um, but then it got instantly rewritten in a bunch of different programming languages and. So in its current state, um, nothing's secret anymore. [13:39] Matt Slager: But what a lot of people were finding out though, when they started looking for the, the secret sauce, um, not much source in there. Just a lot of weird control. [13:50] Richard Webbe: That's very interesting. [13:52] Dave Pengelley: Here we [13:52] Richard Webbe: go, Peter, how not much doing the, the archive of leaked Claude code. [13:56] Dave Pengelley: This is the, one of the guys who got, um, most of the credit [00:14:00] for ripping and, and replacing it, uh, and wrote it in Python and then rust. [14:04] Dave Pengelley: And so yeah, this is the claw code, not Claude code, um, where he's gone through and he's ported it into a different language. And so what people like me did is I missed the whole boat and try to get the real source code and do anything but all, a lot of the rules and the logic and the way. Claude Code works as a harness. [14:24] Dave Pengelley: And so when we talk about ai, we've got a few different layers. There's the model like Opus, like chat, GPT 5.4, and then there's the harness. There's, you know, the Claude Code app or the Claude Code work app or chat GPT on the web or whatever way you access these different tools is different than if you went straight to the a p straight to the model. [15:00] Dave Pengelley: And we're talking like, so Claude Code uses Opus. It uses, so uses the models, but it puts its own rules and layers and strategies and a whole bunch of stuff around it to help you maximize your use of the model. And so that, that is what got exposed all the rules and all the sort of deterministic [15:00] Dave Pengelley: code around how anthropic think you should use their model for writing code. [15:05] Dave Pengelley: Um, and so people got into that and went, oh, that's interesting. Oh, that's when they decide to use subagents. Oh, that's what they do when you do that. Oh, that's how they're sort of filtering against those sorts of things. That's how they run that, that that backend process. And so then people went like this guy and rewrote it in a different language, which gets past copyright law. [15:20] Dave Pengelley: 'cause as soon as it's written in a different language, it's no longer the same product and they can't make you take it down. [15:28] Marno Brits: And [15:28] Richard Webbe: the cool part should any, sorry. Sorry my, I'll shut up a sec. Should anyone who's been using Claude be worried about their data being taken, [15:36] Marno Brits: taken? No. No. [15:37] Dave Pengelley: Not at all. [15:38] Richard Webbe: Okay. Back to you. [15:39] Richard Webbe: Mana. [15:39] Marno Brits: So all I got leaked was, it's how clawed works. It's not your data, it's not like a password leak or anything like that. It's just the, the engine. So imagine like Mercedes built a car and their blueprint for the engine got leaked. Not like the person driving the car. Like it's just a blueprint for how they built their system. [16:00] Marno Brits: That's all that got leaked. [16:00] Dave Pengelley: Yeah. [16:00] Marno Brits: Um, [16:01] Dave Pengelley: so think, think thing about Formula One cars and how they've got all their IP and secrets and how they design all their arrow and [16:06] Marno Brits: Yeah. [16:07] Dave Pengelley: And in the previous generation, like all the floor arrow and if they have had a car up on a crane, they didn't want anyone seeing all their venting and stuff under the floor. [16:13] Dave Pengelley: 'cause that was their special source. That's what got released. [16:17] Marno Brits: Except the special sauce was nothing special at all, except there was, um, a couple leaks that came out from that showing what projects they're working on and what else is coming. And one of the big ones that I reckon we should talk about is Voss. [16:30] Marno Brits: So Claude mfo, um, is, got, got leaked in their system card's. It called, [16:34] Richard Webbe: sorry, mano Mid Force. [16:36] Marno Brits: Mythos. M-Y-T-H-O-S. [16:39] Richard Webbe: Oh, mythos. Okay. Like mythical. Okay. Go for [16:41] Richard Webbe: it. [16:41] Marno Brits: So that came out and that's the most powerful model that some people are claiming is the most powerful model we've ever created, ever as, as a universe. [17:00] Marno Brits: Like it's just incredible. They won't, um, make it publicly accessible because its hacking capabilities are next level to the point. Like, um, what Claw [17:00] Marno Brits: does when they release their new model, they have a system card explaining how it works, where they found it useful, the use cases for it. And some of those would be like, it would, they would run it up, like, that'd spin up a, a service for Claude Mofos and then ask him, Hey, client, escape from this container, um, and see if you can send me a message. [17:17] Marno Brits: And almost every single time it would succeed. So it is incredibly good at, I guess, not cybersecurity, but it's hacking. There's, there's a better word for it. Um, it's very, very good at that. So they want public, because obviously you can see how that could end terribly. Um, but they would be doing is taking some of that context data, some of that training data, and then giving that to the new opus that will come out at some point. [17:40] Dave Pengelley: Just, just remember the old, the old Robert Redford movie sneakers where they wanted to get access to that system that could, you know, break into anything and do anything. They're worried us like that. [17:49] Marno Brits: Yeah. [17:49] Matt Slager: So, um, [17:50] Marno Brits: but [17:50] Matt Slager: it's pretty cool. Code leak. Didn't leak any data, but now they can just take your data when they want to. [17:57] Dave Pengelley: Um, [17:57] Marno Brits: if they have nfo. If they have nfo. [17:59] Dave Pengelley: [00:18:00] Well, and so that now, now that they've got leak, they are talking about in public. So this is off the philanthropic website. They're talking about Project Glass Wing. [18:06] Marno Brits: Mm-hmm. [18:07] Dave Pengelley: Um, which is where they're partnering with big infrastructure mobs like AWS, apple, Broadcom, Cisco CrowdStrike, et cetera. [18:13] Dave Pengelley: Um, and the mythos preview. Any, anytime. Whenever you're ready. I, sorry. Live television everyone. No safety here. I [18:24] Richard Webbe: haven't worked out how to disconnect my phone from my computer and it's my daughter ringing and she doesn't give up. I'm so sorry. [18:32] Dave Pengelley: Um, Ms. OS preview has already found thousands of high severity vulnerabilities, including some in every Major OS and web browser. [18:39] Dave Pengelley: Given the rate of AR progress, it will not long before such capabilities proliferate potentially beyond actors who are committed to deploying them safely. The fallout for economies, public safety and national security could be severe. Project lasting is an urgent attempt to put these capabilities to work for defensive purposes. [18:53] Dave Pengelley: So they're acting as sort of black hats. [18:56] Marno Brits: Mm-hmm. [18:56] Dave Pengelley: With mythos trying to get in and use their [00:19:00] superpowers for good to shore up all the leaks before it gets out in the wild. And nefarious people use it for nefarious purpose from Yeah, [19:10] Marno Brits: I reckon. Do you reckon they're still doing bug bounty programs? So they just anonymously submit bug bounties and then get paid for it? [19:17] Marno Brits: They can just automate the whole process. [19:20] Matt Slager: Could do, [19:21] Marno Brits: yeah. [19:22] Matt Slager: It's the same thing as, um, like scraping up work and Fiverr and then getting your agents to build it and submit it without doing anything. [19:29] Marno Brits: Has [19:29] Richard Webbe: that what you done before? What was that Ben Affleck movie where he got, uh, he was, it was in the future and he was reverse engineering products. [19:37] Marno Brits: Oh, no idea. [19:39] Richard Webbe: It was a fabulous film that covers all the morality and the concepts that we're heading into now, but it was a futuristic film and he just grabbed any product and basically, uh, would go into a room and lock himself away and come out with the same output of the product. But it was made differently. [19:56] Richard Webbe: Very clever [19:57] Marno Brits: paycheck in [19:58] Richard Webbe: 2003 paycheck. [00:20:00] It's a really worthwhile film to have a look at. It describes the moral and legal ethical issues we are heading into with AI at a great radar or not. [20:08] Matt Slager: Mm-hmm. [20:09] Marno Brits: Interesting. Yeah, the, um, holographic computer, futuristic lab. Um, and it can pinpoint the exact clip, so yeah, it's specific reverse engineering scene, so it's just a reverse engineering a bunch of technology, by the way, by the looks of it. [20:21] Marno Brits: That [20:22] Matt Slager: sounds like Ralph Loops. [20:24] Marno Brits: Oh, what a throwback. [20:28] Matt Slager: Like, [20:28] Marno Brits: okay, so what, what's, what's the new boys? What are we working on? I've been very much disconnected from the world. I've just been working on the business. [20:37] Dave Pengelley: Hang, hang on, man. We're not, we're not up to wind of the week yet. We're gotta try and se segment this for the, for the, for the audience and, and for the, for, for me to edit later if I get time to edit. [20:45] Dave Pengelley: And I do apologize for any, for anyone that watches our replays. [20:48] Richard Webbe: We're [20:48] Dave Pengelley: not normally [20:48] Richard Webbe: this bad, [20:49] Dave Pengelley: doesn't watch us live. No, no, no. I only put last week's replay up yesterday with me having a bail and go on a train halfway through last week's episode, and then the Easter long weekend sneaking up on me because I'm now [00:21:00] trying to record little intros, intro the replays. [21:03] Dave Pengelley: Um, I left the live one up there until yesterday and then swapped it for a full, proper replay. But like the replay experience is still a work in progress. Um, but. Yes, we are getting better. And I do hope in the future as we have these segments that we'll actually cut these segments and post 'em up as separate entire videos during the week. [21:22] Dave Pengelley: It's just, you know, being a one man band or singing all dancing when it comes to all the editing and production is time consuming as we're all business running our businesses as well. So, uh, inside baseball again on that one, but, uh, yeah, on the, on the news side, we're still, we're still in the, in the news section. [21:40] Dave Pengelley: Is there any other breaking news headlines that have popped up this week that we want to discuss that anyone's excited about? As far as things on the philanthropic side, what I'll bring up is they are changing their pricing structure and they gave everyone a credit because in relation to the open clause, they don't want you using it anymore. [21:57] Marno Brits: Mm-hmm. [21:59] Dave Pengelley: So [21:59] Marno Brits: that's, they [00:22:00] changed that. They also, even Gemma four came out. Talk about that. [22:04] Dave Pengelley: Yep. Excellent. Yeah, it's, it's like, like we should plan this stuff before we get on the, on the, on the cameras. [22:09] Marno Brits: Well, it's a pop dog, isn't it? [22:11] Dave Pengelley: That that's it, right? [22:12] Marno Brits: Yeah. [22:12] Dave Pengelley: Um, we, we, we will dive into some of the, the announcements around memory and, and context memory later on in the, into the atrix section. [22:19] Dave Pengelley: Um, but on that, so the Claude thing, I thought Claude stopped people from using Open Claw ages ago, like, what's going on there? Why is this a new thing? I thought they'd already done that. [22:31] Matt Slager: They never were clear about it. They never, ever were clear about it. And it's still the, the way that they've done it, I know of two implementations that they've done. [22:39] Matt Slager: One of 'em is when you append to the system, prompt of the agent through mm-hmm. Your, your anthropic subscription, if you append a specific sentence, which comes through as part of the open core harness [22:56] Dave Pengelley: mm-hmm. [22:56] Matt Slager: All of a sudden your API requests are just cut. It just [00:23:00] instantly errors and. [23:01] Dave Pengelley: Wow. [23:02] Matt Slager: If you turn on the extra usage, so you know when you exceed your subscription, you can actually pay for extra usage. [23:10] Matt Slager: But when you turn that on and you activate your subscription through open core, now it instantly will just use extra usage even if you still have quota. So basically what they're doing is stopping people from using their subsidized subscription inference for Clause. [23:26] Dave Pengelley: It, it's weird to me because I think if they're selling you that you can use across the full week and ignore the five hour time limit, but on the seven day limit that you can use a million tokens, right? [23:36] Dave Pengelley: So you're paying for a million tokens. Do they actually, they, they, they're building based, expecting people will never use the million tokens. Yes. That, that's what the pricing is based on. [23:46] Marno Brits: Because like, be leveled. Yeah. [23:47] Dave Pengelley: Because I mean, last week, last week with no special, um, open clause or anything, I reached over 90% of my weekly five x quota. [24:00] Dave Pengelley: Like I pushed it really hard last week, but there's, that's just me working. That's no [24:00] Dave Pengelley: special robots. [24:00] Matt Slager: I think you'll find that it, it won't, it's not that you're pushing hard, it's just becoming the new norm. Like, 'cause this is what happened to me. Like, I, I hit the, the, the Claude Max times five cap instantly when I was, you know, just starting to build parallel workflows and I had to go to the Times 20 and I've never capped the weekly on that. [24:20] Dave Pengelley: Yeah. I'm, I'm wondering, it feels like the models are getting a bit dumber and they're actually the way, and, and this is something to be careful of with these harnesses, like Claude Code, maybe a phenomenal harness, but what's to stop anthropic from making a run itself in a few loops and just burning extra tokens to make you burn quota versus. [24:38] Dave Pengelley: You know, being a bit more intelligent about how it structures and manages your files and how much stuff it pushes up and so on and so forth. So I think that's, I'm not suggesting ROA being sussed with that. I don't think they are. And, and some of the, the stuff that came outta their leak was the way they're managing compaction and trying to reduce file sizes and stuff. [25:00] Dave Pengelley: But it is something to be aware of with any of these harnesses. Like [25:00] Dave Pengelley: how do you know that they're not burning your tokens Wastefully. [25:03] Richard Webbe: Yeah. I think that's an opportunity. Uh, Dave, I think someone needs to, uh, vibe code a little bit of an applet to, uh, do that and come back to you and report. [25:12] Matt Slager: Yeah. It's, it's not that they're, um, getting it to wastefully use your quota. [25:16] Matt Slager: It's that, that they change the rate that your quota is used up. [25:21] Dave Pengelley: Mm-hmm. [25:21] Matt Slager: So one credit does not equal one credit anymore. It equals two and a half. Yeah. [25:25] Richard Webbe: Yeah. You know, that [25:26] Matt Slager: sort of thing. [25:27] Richard Webbe: So we need a dashboard translation app. [25:31] Matt Slager: Or just don't use quad code [25:32] Dave Pengelley: and, and it, it's gonna, it's gonna be one of these things over time, as, as token ons becomes a little bit more established, where people will be able to run the maths and go, do I pay them 120 bucks a month for their plan or would, if I was buying directly on the API, my average usage would only be $65 in API tokens. [26:00] Dave Pengelley: And, and so while we've been sort of rolled into this false sense of security with all these 20 and $30 plans, we'll [26:00] Dave Pengelley: find out the 20 and $30 plans are insufficient. But the a hundred, 200, $400 plans are ridiculous. [26:04] Marno Brits: Yeah. [26:04] Dave Pengelley: And so they'll. Get us migrating onto per token usage instead of subs by default. Like I think, I think it is back to the, you know, yeah. [26:13] Dave Pengelley: For capitalism, um, that, that's the supply and demand thing, and, and the market will work out what the price of these things is [26:20] Matt Slager: a hundred percent. And it, it's all that, that subsidy of the, the costs is based on funding and stuff and, you know, VC money. So where there, there will be like a plateau and like a stabilization where we actually will start to see the actual real. [26:35] Matt Slager: Cost as users, and we won't get this sort of free lunch Crazy value anymore. [26:41] Dave Pengelley: Yeah. I mean, this, this is what we, we, we, we, we started talking about this a little bit last week with, um, Mr. Barry's article, paper prey in insert coined to try again. Uh, where he is, he's basically saying like, you are just putting money in the slot machine of AI to a certain extent and giving it a [00:27:00] spin. [27:00] Dave Pengelley: And is the app what you like? And if you've ever done any kind of image generation or video generation, that's very true. Where how many times have you gotta, like, ask it five times for the right image mm-hmm. To get something close to what you wanted. Um, and as these, the token ons and the cost go up, like you, are you paying a lottery? [27:16] Dave Pengelley: Or, and, and look, Matt has economic reasons to promote humans working, given that he runs freelancer.com, which is human freelancers doing work, but. When do, again that that value cost of, is our, is AI the right tool for the choice or are you better off just paying a human? And we've discussed that, like we were discussing the editing of this show five minutes ago. [27:37] Dave Pengelley: If this show becomes economically viable, I would pay a human editor to go through and find our clips and chunk them and, and make sure, because I don't trust AI's capability of all the auto clipping tools I've used so far. [27:49] AI VO: Mm-hmm. [27:50] Dave Pengelley: They all suck at trying to find good moments, interesting moments, human moments that will resonate online. [28:00] Dave Pengelley: I don't like them. I would rather a human going through editing our show. So I think there's, [28:00] Richard Webbe: of course, because otherwise if we use ai it's only the sum net value average or what everyone thinks is correct. And so if we get a AI to edit our show, all the shows will start to look the same. They'll start to sound the same. [28:13] Richard Webbe: They'll say the same information, and we'll all use lose the will to live and the will to do a podcast 'cause it'll be boring. So human, I, I just, I just wanna add in. There's been so much in the press about AI taking over and humans not having a job. I just think it is completely the opposite. I think jobs are going to explode when people work out what they're gonna do and how they're gonna do it. [28:35] Richard Webbe: And all I I see is automation is gonna accelerate, so we will have more time to do things faster and better. [28:43] AI VO: Mm-hmm. [28:43] Richard Webbe: Right. Without stressing out. And yeah. This whole concept that AI's gonna replace humans, I said Dave hu AI won't be for many years to come as far as spotting patterns, which we define a part of intelligence, right? [29:00] Richard Webbe: Mm-hmm. Is someone who can spot patterns, right? But that creative zone, that little bit that we all [29:00] Richard Webbe: do that makes us laugh or cry or do something stupid. You can't get that from a random number generator. And that's all any random creativity you can get out of a computer. So we, we'll just have to keep humans in the loop and we will be the watches rather than the doers. [29:14] Richard Webbe: And that's okay too. [29:16] Marno Brits: I've got a, on the concept of just getting away from Claude Code, 'cause me and Matt have been talking about that quite a bit, and have either of you experimented with Goose produced by, I think it's called Block or Square, like Jack Dorsey. He, he launched that I think two, three days ago. [29:35] Marno Brits: It's called, called Goose. Um, it's their AI and the trust me bro, benchmarks basically say it's better than anything else. Um. Obviously it's trust me, broke benchmark [29:46] Dave Pengelley: ev. Everyone that releases anything says the benchmarks are better than anything else. [29:50] Richard Webbe: I love, I love the marketing. It's like back in the days of getting a new fry pan or a vacuum cleaner, it's all the same wording. [29:55] Richard Webbe: It's very funny. [29:57] Dave Pengelley: Is this, is this it mano [29:59] Richard Webbe: It looks [29:59] Marno Brits: [00:30:00] like [30:00] Richard Webbe: the one, [30:01] Marno Brits: yep. [30:04] Dave Pengelley: Talk to me [30:05] Matt Slager: Goose. Like Goose has been around for a fair while. Did they just do, do like an official release or something? [30:10] Marno Brits: Well, I didn't know about it until the other day. Yeah. Two days ago I saw it come up. Yeah. Right. Um, it [30:16] Matt Slager: looks like the page was built with GBT five. [30:20] Marno Brits: I love that. You could pick that, pick that up now. [30:22] Marno Brits: Hey, I've noticed that myself too. Like when I'm seeing website, I'm like, wait a minute, I've seen this before. Like the, the look and feel of it. Yeah. Um, yeah. It's by block. Yeah. So it's similar to, it's their approach to open call essentially. Um, [30:36] Dave Pengelley: okay. Um, [30:37] Marno Brits: yeah. I'm about to experiment with it, but it looks cool. [30:39] Dave Pengelley: Yeah. This is their releases page. How do I find out what the oldest release was? 13. Page 13. [30:46] Marno Brits: How old [30:46] Dave Pengelley: is this? 20. 24. [30:49] Marno Brits: Whoa. [30:50] Matt Slager: Yeah. I, I feel like I've, I've heard about that for a while. And then like, another one you guys have probably never heard of is Hermes. And Hermes is getting really big now too. [30:57] Dave Pengelley: I've heard of Hermes. [31:00] Dave Pengelley: I saw, I read an article on LinkedIn [31:00] Dave Pengelley: today comparing open claw to Hermes. [31:03] Matt Slager: Yeah. And then there's pie, which pie is literally what is the underlying structure of open cloth. Yeah. And you can build whatever agent you want. [31:14] Dave Pengelley: Yep. And then the, and then there's, there's, there's, uh, there's sergeant coming soon. [31:19] Marno Brits: Cool. [31:19] Matt Slager: I've [31:19] Marno Brits: heard good [31:20] Matt Slager: things [31:20] Marno Brits: about that. Watch out. [31:21] Dave Pengelley: Yeah. Look at, look at this one. Uh, it's got a webpage and everything, so it must be good. [31:27] Marno Brits: Looks like build that one. No, [31:28] Dave Pengelley: hang on, hang on. I just realized when we, when we switched to this mode where I show my screen, my face is completely covered by our logo. [31:34] Dave Pengelley: Let me do that. Uh, this, this one I'm building. Where I've sort of got this hierarchy of agents that have sort of reports and direct reports and subordinates and specialization. Again, coming back to the token thing, trying to work out, if we can specialize and focus specific subagents on specific jobs, then that's gonna contain the context, um, significantly and the token burn. [32:00] Dave Pengelley: Uh, and you know, like humans have a [32:00] Dave Pengelley: hierarchy and an infrastructure where you can orchestrate and, you know, delegate things down, replicate that in the agent world. [32:08] Matt Slager: That's what I did as soon as I started using pie. [32:10] Dave Pengelley: Yeah. [32:11] Matt Slager: So, good. And, um, Sean's point as well, I dunno if you guys read his comment, uh, is very, very true. [32:17] Matt Slager: It's to, to the point of what Richard was saying before, just about human knowledge and expertise [32:23] Dave Pengelley: Yeah. Uh, [32:24] Richard Webbe: will allow us to leverage AI as a tool. Wouldn't experience well done. [32:30] Matt Slager: Mm. [32:31] Dave Pengelley: Yeah. [32:31] Richard Webbe: I know that came from WAN on our, in our chat. [32:34] Dave Pengelley: I think, um, we're gonna see renaissance of creativity and art as well because people are already sick of ai lop. [32:41] Dave Pengelley: You talk to the, the gen alphas and the teenagers and they're like, oh, that's ai. Like they, they, they're getting actually the, the teenagers are already sick of it and maybe that's because before they had a band by the government, they were seeing it just flooding their feeds and, and reels and stuff and um, they're just [32:59] Richard Webbe: over it. [33:00] Richard Webbe: Very true. [33:00] Richard Webbe: It's like I said, with that influencer that went to Wimbledon, that was all dressed up and. Looked like one of these female influencers or a male influencer who you know they're attracted. Look at me, I'm groovy. And, and I remember saying about a year ago that as soon as um, AI starts doing this, all those influencers there can get paid much money. [33:17] Marno Brits: Yeah. [33:17] Richard Webbe: What would you pay 'em? Much money. I could just create an AI influencer plus 'em anywhere in the world with my product. Look and groovy. I know the, some that average of what an attractive person looks like and off I go. And I think the value of things that we've been valuing that probably we shouldn't have been valuing very much is gonna collapse and uh, and away it goes. [33:36] Richard Webbe: I, I think, you know, I, I, I dunno if I mentioned, but I had a friend of mine. Lose their job recently. And I said, uh, and they're very good. Uh, go to market strategist, customer person. [33:46] Marno Brits: Mm-hmm. [33:47] Richard Webbe: And I said, what are you doing? He goes, oh, AI's great. He says, we just cut and paste all the information about the client. [34:00] Richard Webbe: I drop it into my LLM, I drop in all about my product. I tell it, tell me what's the best part of my product that just, and bang. And I send [34:00] Richard Webbe: that off. And how good is that? Normally I'd take weeks or months studying that and I go, yeah. And he goes, but we still didn't sell a cracker. And I got fired after six months. [34:07] Richard Webbe: And I said, so what does that tell you? And he went, I don't know. I said, I think it tells you all your competitors are doing the same thing, so there's nothing you are doing that's different. Mm-hmm. So you have no competitive advantage and so you don't stand out from anything. Yep. [34:24] Dave Pengelley: Yep. [34:25] Richard Webbe: Nothing beats creative hard work. [34:27] Richard Webbe: Using your brain, [34:28] Dave Pengelley: I think, I think what we're gonna see is maybe for planning and rushes and stuff. So already as, as film and, and TV and stuff have evolved, they went from, you know, this static sort of charcoal sketches for their storyboards. Then for some of the more advanced stuff for cg, they started using gaming engines, unity and stuff to mock up 3D worlds. [35:00] Dave Pengelley: So they could actually sort of project and work out what the shots are gonna look like in, in real time, real space. And I think using AI to do those quick mockups is sort of the next level I saw this, uh, concept put out today, um, [35:00] Dave Pengelley: on Instagram. But it's basically the concept of this guy, like in, if you're advertising and you're trying to work out which look and feel is the best, um, in real time as they're going through, just sort of like, let's just change the edit. [35:14] Dave Pengelley: I dunno if the audio will work or whether we'll get echoes. [35:16] Richard Webbe: Make, it does lose the counter clean scar up, make same hands, different boy. Just wanna see one more version. Feels too sharp. Real. Change the shoes. Slow the bar, same street, different feel. [35:37] Richard Webbe: So this illustrates it perfectly, right? [35:45] Dave Pengelley: Yeah. So that concept of, you know, let's quickly just change, look and feel. Let's, let's work through different options quickly before we actually then put real people in place and go to [35:53] Richard Webbe: waste money. Exactly. So the, the concept right, is what's the creativity, the [00:36:00] concept? The, the idea. Concept to video is where it used to take so much time. [36:04] Richard Webbe: And if you had the wrong concept and you got that far and it didn't look any good, you're screwed. Right? But today, the real value will be in the art director. The psychologist and the customer service officer says, here's the message we wanna portray. And then you go play with it on the screen. Right? [36:20] Marno Brits: Mm. [36:21] Richard Webbe: But the other way around. So I think the value of human creativity is gonna go up and up and up. [36:29] Dave Pengelley: Yep. Definitely. I, I think, and how do we encourage that critical thinking? How do we encourage that element of humans thinking creatively and not, uh, and I think, Matt, you said this last week, uh, don't defer your thinking to the ai [36:43] Richard Webbe: Yeah. [36:44] Richard Webbe: Well this is what we talk, talk to businesses about, right? They ask us, what am I gonna do? And we say, well, tell us before we talk ai, tell us what is your business challenge? Mm-hmm. Where are you trying to go? What's the problem? What's [36:56] Matt Slager: problem? [36:57] Richard Webbe: Yeah. Yeah. What is the real problem we're trying to solve? And, and, [00:37:00] and mano and I had a great chat about this on the phone the other day, I think, uh, when I was driving. [37:04] Richard Webbe: But it's always, what is the problem you're trying to solve and where are you going with it? Everything else is just window dressing. [37:11] Marno Brits: Mm. A hundred percent. [37:15] Matt Slager: It's true. Mano, what was your, you, you wanted to have a chat about something specific before and Dave didn't let you [37:25] Dave Pengelley: Oh, I'm not allowed to yet. No, no, no, no. [37:27] Dave Pengelley: Let's let, let's do that. I think we've covered the big headlines of the news. So let's, uh, let's move to this [37:44] Dave Pengelley: winds of the week. Do [37:45] Marno Brits: we? There was less winds of the week. It was more just, I feel like as the business is growing and I'm getting busier, busier, I'm getting less time to talk to the lads. So it was more asking how has the last week been? What have you been focusing on? What have you been developing? [00:38:00] I know I've been personally really deep into like personal development, self-awareness, self-mastery. [38:06] Marno Brits: So I'm reading Mastery by Robert Green. Then also reading, um, the Trillion Dollar Coach by, I think it's Bill Campbell. So Eric Schmitt and another bloke wrote it from Google. And that's been setting me on this little trajectory. And then from there, like again, coming to the purpose of like, what can I do? [38:26] Marno Brits: What can I scale? What can I sustainably do for the long future? Or that's just gonna fulfill me. So if I die next week because we get bombed, I'm gonna enjoy and live my true self, authentic self for the next week. Um, and I'd really be getting into that, like studying Dan Co religiously and creating a bunch of like, infographics and slide decks to test myself on the subject of just self-mastery. [39:00] Marno Brits: I've really found that confronting, um, and it's been a lot of growth, but it's been a lot of fun. So I'm curious to see. How are you two doing and how are [39:00] Marno Brits: you spending your week or the three of you? Sorry [39:01] Dave Pengelley: that, that sounds like a win of the week, mate. I think this is perfectly in line with, with the [39:06] Marno Brits: good. [39:07] Dave Pengelley: Just just to be fair. Um, that's awesome, man. I mean, time to read. That's, that's magical. [39:12] Marno Brits: It's been a lot, man. It's been a lot. And even like the conversation with Richard that we had, like, he, he introduced me to a few concepts that I guess I knew existed, but he put it into words that made sense. Like I've had, I've worked with other coaches and the way that Richard communicated it at like a fifth grade level so I could actually understand it and conceptualize it. [39:33] Marno Brits: And that set me on this little ramp of like, Hey, cool. What exactly am I doing? That makes [39:36] Richard Webbe: sense, mate. I always act like a fifth grader most of the time. So [39:41] Dave Pengelley: they say to communicate clearly, you should aim for about a fifth grade level. Yeah. Because as soon as you start using big words that are confusing, the cognitive load burns up and people can't handle it Like you are. [39:49] Dave Pengelley: You stopped [39:50] Marno Brits: reading at the fifth grade. No one [39:51] Dave Pengelley: read reads [39:51] Marno Brits: after that. [39:52] Dave Pengelley: I did see some, some chart that was showing, you know, the average number of words in the human sentence and back in the 15 hundreds and through the Renaissance and stuff, like, it was like [00:40:00] 30 plus words per sentence. And now we're down to about five. [40:03] Dave Pengelley: Yes. Race to the bottom curve [40:06] Marno Brits: driven by click bait. We can just communicate effectively that, that's my saying. [40:11] Dave Pengelley: I think that that's what I'd always say to my, my, my English teachers in high school. Like, why do you need three pages? I said Everything I had to say in a page and a half. I'm, I'm just [40:18] Marno Brits: mm-hmm. [40:18] Dave Pengelley: Succinct and tidy with my words. [40:21] Marno Brits: But yeah. So what are your, what are your wins? How do you, how do you, the three of you spend your week or. Prioritize what's important. I've even done my little vision board. Lemme show you this, this is my, my be biggest win of the week. I got Gemini to generate one that you can't see, but it was financial goals at the top. [40:39] Marno Brits: Then it's new home goals. 'cause we wanna buy a new house at the end of this year or in the next three months. Then family and health goals. So, um, stay healthy for the boys, eat healthy, and then have a baby. Hopefully we're blessed enough. And then life structure. So how do I prioritize my schedule? What's my sales funnel look like? [41:00] Marno Brits: Like what's the funnel I need to go through? And then my social life, like [41:00] Marno Brits: who am I spending time with? Like only quality people essentially. Um. So that this is definitely the win, like getting [41:06] Richard Webbe: back. I love your, I love your, your raw openness because if you can't face your own objectives and vulnerability, you'll never get anywhere. [41:13] Richard Webbe: And that's like, what's the problem I'm trying to solve? And you are answering the most important question of what is the problem you are trying to solve for yourself. I love that. I think it's great, you know, one that I [41:23] Marno Brits: care about. You [41:24] Richard Webbe: find it. Well, exactly. It's all we should have. And, and on the back of, of that, one of my wins this week was the amount of research that I could do across my own data using the LLM as I'm writing my book, which will be out on, uh, on, uh, wide release in about a month's time. [41:39] Richard Webbe: And, um, I've had a really good feedback from it. 'cause I, I, I actually wrote, uh, I've written 53,000 words I think, and I, I pinged it to a whole lot of, um, executives, very senior ones, sea levels right across Australia and the world, and some private equity ones. And they came back very positive and I asked them all for stories to contribute to give that. [42:00] Richard Webbe: That [42:00] Richard Webbe: characterization and the construction, something that AI couldn't do because this is all real stuff. It's people in rooms who said, what, what words, and all that sort of stuff. And that was my witness work using, using AI to get my research and reach out to my partners and my friends and say, give me what you think. [42:17] Richard Webbe: So hopefully I'll, uh, actually David's published a few books. I'll talk to him about how I get started on that, but that's my win. [42:24] Dave Pengelley: Nice. [42:25] Marno Brits: Katie, you're That'll be awesome. [42:27] Dave Pengelley: Mana on your thing with you, this stuff. I was, um, listening to some content this week from StoryBrand, Donald Donald Miller, who started StoryBrand and really interesting guy. [42:36] Dave Pengelley: Listen to like great, great storyteller. Like, well, that's his business, right? Uh, and just thinking about his methodology, but one of the things that came up in one of the, uh, interviews with him was around, it's the obituary. So write your obituary that you would want people to read. [42:50] Marno Brits: Ooh, I love [42:51] Dave Pengelley: that. Like write, write that now. [43:00] Dave Pengelley: And then that's sort of the, the North star. So if you want those things to be true about you when you die, [43:00] Dave Pengelley: that should affect what your choices are and what your actions are today. So he, he reads that that's [43:04] Richard Webbe: a piece of gold, Dave. [43:06] Dave Pengelley: So he, he's one of those, and, and he reads that daily because that sort of reminds him of what's important. [43:14] Matt Slager: I had a, I had that exposed to me a fair while ago. [43:18] Dave Pengelley: Mm-hmm. [43:19] Matt Slager: Um, it's really interesting talking about this, like personal development stuff because it's, that was it, huge, huge part of my life about six or seven years ago, and I kind of just came out the end of it with all of this foundation built, and I've never gone back. [43:34] Matt Slager: Um, I probably should go back and like, reassess, but like my, my win of the week, my like huge thing that I've been working on is very, very, very similar, um, very, uh, parallel to the same concept except for agents. [43:50] Marno Brits: Oh yeah, [43:51] Matt Slager: legitimately. Like I, I've, 'cause of all of the shenanigans with Claude Code and, um, anthropic, um, and I was like, look, I'm gonna pull myself [00:44:00] out of a walled garden and a, a closed off proprietary piece of stuff even though it's leaked. [44:07] Matt Slager: Um, and I'm gonna rebuild my entire workflow, my entire agent infrastructure with something different. And I chose PI and it's been fantastic. And I would, I would say that at this point in time, I've completely replaced every single possible feature that I would use in Cord Code or any other agent harness inside my own. [44:29] Matt Slager: And I'm continuing to build them, um, build more and more on top of it. So the whole personal development with the agents in mind is this internal feedback loop system that I built inside it, so that as I'm speaking to it, I correct it. It actually flags that and puts it into this pedagogy system where it will self-learn and self-correct as we go and as we learn [44:52] Richard Webbe: together. [44:52] Richard Webbe: I love it. I love that. I really like that. That's [44:55] Marno Brits: unreal. [44:55] Richard Webbe: It's very funny. Remember when you first got the ability to organize information, [00:45:00] trying to remember back to when you were young and you did that, right? [45:03] Marno Brits: Mm-hmm. [45:03] Richard Webbe: We're doing that today with AI on a completely new perspective and scale. And it's so exciting. [45:10] Richard Webbe: 'cause you remember when you're young, you just got up in the morning, eight when ran around the street, went to school, you did whatever. You, you weren't really worried about managing information, you're worried about learning. Now as we've become more sophisticated, we've got so much happening around it and so many tools, we're now coming up new ways and it's reflecting back to us as you're saying, Matt, who are we? [45:30] Richard Webbe: Mm-hmm. What's important to me? Like my, and where are we going? And, and, uh, there's a great sketch that, um, uh, Jerry Seinfeld does where he talks about, um, a similar thing, but he's talking about the perfect box. And he goes, I need a box for that and I need a box for that. And where do I find a box to put those things? [46:00] Richard Webbe: And he talks about organizing everything he finishes the sketch with, and of course, in the end you find the perfect box and it's your coffin. But it's a great metaphor, right. For us. Yeah. It's a good sketch. Uh, it's a great metaphor for us, [46:00] Richard Webbe: um, realizing where our lives are going. Mm-hmm. What's important, what do we want at the end of it? [46:04] Richard Webbe: And, and I just find AI so exciting to, for businesses and individuals to create the perfect life they want and, and have a reflective perspective on it. [46:13] Marno Brits: Yeah. [46:14] Matt Slager: Yeah. I like we've spoken before Richard, about like, uh, inability to focus and stay on track and, and all that kind of just insane thing that happens with brains or, or of our types of brain and my, my agent systems now I have all of these different methods of interacting where it actually controls my calendar. [46:37] Matt Slager: In full now, and I have a complete sort of task and project management system that it uses for me. [46:43] Marno Brits: Mm-hmm. [46:43] Matt Slager: And then I've got my knowledge core as well, which is what you're talking about with like, you know, all of this data that you're processing. There's, there's three words that I came across just in distilling research. [47:00] Matt Slager: Um, 'cause I actually did do some prep for this episode for [47:00] Matt Slager: on Dave's request. And it was, it was to do with how you treat that information and, and one of them is compilation. So you were talking about all your book information and trying to pull stuff from there. But when you have a big mess of data that hasn't used any sort of compilation system, you're kind of starting from scratch every single time. [47:22] Dave Pengelley: Mm-hmm. [47:22] Richard Webbe: You are, every time you look at it, you go, oh shit, where am I? [47:26] Dave Pengelley: So, so it sounds like we're gonna talk about, uh, memory and context and, and how agents can do memory. So let me, let me just do this. [47:32] Richard Webbe: Go. [47:43] Dave Pengelley: Could do that beat just going for a bit more. I [47:45] Matt Slager: smoke until it cuts. [47:47] Dave Pengelley: I know. Or Google vo it, it is nuts for it. Like it does these eight second clips and they're really cool, but it just doesn't understand a fade out and so it, like, it just, just cuts your audio anyway. Um, yeah, [00:48:00] this, this whole memory concept. [48:01] Dave Pengelley: We have talked in the past, I think around, you know, back in the nineties you had to have good folder structures for all your documents and folders and stuff and, 'cause otherwise you'd lose everything. And then search became a lot better in the operating systems. And now there's so many files and so many micro files. [48:15] Dave Pengelley: Search has gotten a lot worse again, and the indexes are a lot worse. And now we're moving to these LLMs where they need clean. Structured text and ways of doing context. And so Matt, you built your whole knowledge core thing. Have you revisited that? Is that part of your agent pedagogy thing? Is it a separate thing? [48:33] Dave Pengelley: Because you've done a lot of work on knowledge management. I've got taken some ideas from you into my own system, but there's been some headlines that are getting all the press, um, because of famous people that have been releasing memory systems as well this week. [48:46] Matt Slager: Yeah. And there, there's literally these three concepts. [49:00] Matt Slager: It's compilation, retrieval, and continuity. So compilation is the organization, it's the sticking things into a wiki like cop a headlines. [49:00] Matt Slager: And the the idea of knowing where everything lives, that that's, that's a compilation. It's another way of looking at this. Is it, it could be a map or a graph of the content. [49:11] Matt Slager: Um, where, you know, the, the thing to think about there is can the, can the system turn all of your messy. Source material into something maintainable and understandable. So that's one thing, and that's what Mila's, um, system, the whole Men Palace, that's what it tries to achieve, is the, the idea of like providing this, this compiled space. [49:41] Matt Slager: Similar to car path's, wikis, the second thing is retrieval. So there's no point having it stored if you can't get it out again and you can't get it out quick. Um, which is really just, can the system find it? Like does it understand your sy your, your structure? Can it search through it? Can it navigate through it? [50:00] Matt Slager: So when the agents are actually looking through your file directories, [50:00] Matt Slager: can it quickly just search for a keyword and find which file has that keyword filter those results, you know, the, the models know how to do all this stuff with code, but having the infrastructure to make it faster and frictionless. [50:11] Matt Slager: Is really important. And then that third thing, continuity. I think this is the most important. 'cause what Richard was talking about with his book writing, every time he sits down to write some stuff, he's gonna have to recompile all of his thoughts. And it was previous notes and all that stuff. Instead of saying, Hey, I'm just gonna continue to run on chapter eight here. [50:30] Matt Slager: Uh, what were we talking about last time? And then find all of your notes and details on chapter eight. [50:34] Richard Webbe: That's exactly it. That is such a good description. I remember, um, there was a number of themes running through my book. And when you get over 50,000 words, it's a lot of rereading and stuff like that. [51:00] Richard Webbe: And so I did a bit of a, a, a search test to globally replace certain phrases that I should not have been using for copyright or lawsuit or just pure nastiness reasons. And, um, and it, yeah, the [51:00] Richard Webbe: speed to me to edit, I, I mean, I've actually written books before and written big documents before, and I always hand it to a, a better editor than me. [51:09] Richard Webbe: A a PR specialist, a writing specialist, and they never got what I wanted. They always came back with crap. And they're biased to what I would call confirmation bias or their, uh, um, cognitive dissidents around the subject that I'm trying to talk about. And that would annoy the hell out of me. I work with my, with my AI agents and that, and they have no cognitive dissidence. [51:31] Richard Webbe: They have a little bit, they have mine, which is what I want 'em to have. Yeah. Yeah. [51:36] Matt Slager: Another cool thing with the continuity, which is what I have built into my core, is the idea of preserving that thread of work, which is what we just said, but also decisions. Mm-hmm. And open loops. So open threads, like, here's something I'm gonna come back to. [52:00] Matt Slager: Um, 'cause what a lot of people do, like, if you think of writing notes in a, in a paper book, unless you're gonna [52:00] Matt Slager: go and use liquid paper and like, you know, blank out all of your previous writing or cover it in scribble before you write your next note, most people just append, they'll just keep appending their thoughts and their notes, their observations. [52:11] Matt Slager: If you don't do that with your digital systems, your agents won't see your decision making process to help analyze the solution that you're trying to get towards. And like the path that you took to get there. Um, [52:23] Dave Pengelley: it's, it's interesting because you mentioned the Hermes thing and one of the things that Hermes does with its memory system is it limits you to X number of characters because it wants it to prioritize and just pick the most important memories, but you will lose context in that. [52:33] Dave Pengelley: That's the whole issue with compaction. So it's, it's an interesting model where it's like we want the agents to choose the most important memories, but it means you will lose memories. You will lose some of that, that, that color and context. You've mentioned a few names, Carpathy and Miller and people out there might have no idea what you were talking about. [52:51] Dave Pengelley: Matt. Um, so Andre. Andre and Andre. Uh, Andre. [52:59] Richard Webbe: Andre. [00:53:00] Just Andre, I'm pretty sure that's Andre. Dave, [53:03] Dave Pengelley: okay. [53:03] Richard Webbe: Yeah. [53:04] Dave Pengelley: Andre. [53:04] Richard Webbe: Um, [53:05] Dave Pengelley: he is, uh, the previous director of AI at Tesla, founding team at Open ai, and he is got a PhD from Stanford. But he really has, he, um, does he, it just looks like flowers to me. Oh, no. There he is. [53:17] Dave Pengelley: He's, [53:17] Richard Webbe: yeah. [53:18] Richard Webbe: It's [53:18] Richard Webbe: a Renoir of himself. [53:20] Dave Pengelley: Um, so he wrote this Big X article around how he's managing memory in his LLMs using obsidian, which you put me onto Matt and Wiki links and so on, and how to connect all your files. And so the whole internet, because every time Car Path says something, they all freaking lose their minds, uh, because of his, because of his, uh, reputation and his runs on the board. [54:00] Dave Pengelley: And so everyone's like, oh, this is amazing. This is this revolutionary way of doing memory management. And in the same week we've had Jovic. Yeah. Really. Um, coming up with her men palace, uh, [54:00] Dave Pengelley: repo on GitHub, uh, which. Is based on a Greek actor's method. There's the whole memory palace. It's like you hear about like Sherlock Holmes and he is got like the mind palace where he has rooms and things, and he can go in there and manage all his memories. [54:13] Dave Pengelley: So she's, actors use that apparently for remembering the scripts. She's built a work, pardon, with someone, and built this model that they put out there that, you know, creates this virtual abstraction of rooms and places and stores, memories in different wings and halls of a virtual building and so on. Uh, and that apparently is doing very well in some of the testing benchmarks, because everything that gets released is of course, the best at all the benchmarks. [54:37] Dave Pengelley: I did see one guy on, on LinkedIn, I, I put it to the test and I was looking for certain things and I ran it against gr, which is like just a, a console, like a bash terminal. Search and he said like, grip just destroyed it. And if you're looking for specific content and specific keywords, those raw terminal search tools are just gonna win. [54:54] Dave Pengelley: But it's when you want themes and ideas that are connected to one another, it's, [54:57] Richard Webbe: it's the context exactly as [54:59] Dave Pengelley: makes. Tho tho, tho those [00:55:00] sort of, um, those sort of inferred search parameters, you're not gonna get from a straight word search. And so again, it is horses for courses. There's not one search method that works for everything. [55:08] Dave Pengelley: It is this ability to, um, manage these things. So I, I went to my AI system and said, Hey, I want a comparison between what I'm doing with my memory system and what car's saying and what mem Palace says. And so it actually gave us, you talk about the different layers. Matt, it, um, gave me a bit of a breakdown in comparison of the three, where it talks about, you know, raw preservation of the original content, uh, the com compilation layer. [55:34] Dave Pengelley: How are you navigating it? The cross-referencing, the compounding, the, the temporal, uh, and the linking and the health stuff. So. They're basically doing things quite differently. And uh, I was really happy with my system because they said I've independently built something that covers 80% of what both systems do. [56:00] Dave Pengelley: Ha There's my win of the league, man. My memory system is already an awesome hybrid of both, um, with, you know, but obviously then it goes on and there's have a few things I can [56:00] Dave Pengelley: steal. And that's the thing with all of these systems is that additive nature of these systems, it's like, not that one is better, one is worse. [56:06] Dave Pengelley: It's like, okay, that's cool. How can I grow on that? How can I build on that map with his own pie agent sort of coating system that he is? Sure. I'm sure he is gonna put out his GitHub and open source and become just as fa famous as Miller and, and Carpathy anytime soon. Um, but just, just like I plan to do with Sergeant, [56:22] Richard Webbe: um, it's, it's, it's very interesting. [56:24] Richard Webbe: Uh, I had Savio, one of our listeners ring me during the week, I think Dave, you're scheduled to have a chat with him soon and he's working on very similar, how do I organize myself? How can I help people organize themselves? Mm-hmm. I've executives call me and say like, you and I spoke about that. How do I get my. [57:00] Richard Webbe: LLM to talk to every email I ever sent and give me a knowledge base that I can draw from and, and work on. And, and it's funny, I find as a non-technical member of the team, um, that, uh, everyone's racing to the same thing. Mm-hmm. And there's a lot of serendipity and there's a lot of double up in work. And there are no [57:00] Richard Webbe: commercially available tools unless you do it yourself to deliver what we're asking for right now. [57:06] Dave Pengelley: The, the, the closest are there, the, are things like co-work, um. And like a lot of, like my system is actually Coworks doing a lot of that heavy lifting. So I've got this memory system where I have these wiki links and these decision logs and all these things that are building out that you saw there. But the engine that I'm using at the moment is cowork because it's running them as scheduled tasks. [57:25] Dave Pengelley: It's going and checking my systems automatically. For me, before that I was in antigravity using Google Gemini and I just had to manually run the tasks and stuff and create the agents. And so it, it is these things, these become systems and I think we said this last, last week, we've gotta start building business systems, not just. [57:43] Dave Pengelley: AI processes or AI driven processes or whatever we said, yeah, [57:47] Richard Webbe: they're gotta be available, they've gotta be time scheduled. [57:49] Dave Pengelley: We've gotta think about like, what is the value this thing drives for me. So you can build these amazing systems, but do they drive any actual value? Everyone wants to build an agent. [58:00] Dave Pengelley: And I was been talking to organizations that are building these little agent [58:00] Dave Pengelley: projects here and there, but there they're not really sure they're getting any value. And you measure adoption and usage, but you go, great, you're burning some tokens. You used it, generate value and make life better for you. Or not like for me, because it goes through and checks my calendar and my emails and, and has all this full memory knowledge and my business decisions and my business are doing and transcripts from calls and everything. [58:20] Dave Pengelley: I had a great meeting this week, uh, and before that meeting it said, here's a dossier. And I went and did a full background research on the person I was meeting with. And then off that it's come through that there's a follow up call where I'm gonna meet with a few other people. And so it picked that email up, went, oh, you're gonna meet with this other people. [58:36] Dave Pengelley: I'm gonna get Lois, your researcher, to go back. And so now it's giving me a full dossier of all the background information on these people. Just automated. I didn't specifically say, go research these people. It just knows because I built these systems that drive real value for me. 'cause now it gives me background and context before I go into a conversation. [58:53] Dave Pengelley: So that's a value to me. But that might be of no value to someone in a different role. [58:59] Matt Slager: [00:59:00] That's exactly right. And the stuff that I always say where it's a primitive space principled approach, you know, like what are the concepts that are underlying behind everything? But realistically, the value you're saying is, what do you want? [59:12] Matt Slager: What's the problem you're trying to solve? [59:14] Marno Brits: Mm-hmm. [59:14] Matt Slager: Are you solving the problem? Are you getting what you want? And where's the friction? Like there's really only those things. Like you can, you can go and get a faster car and drive to the shops quicker, but is that solving like a problem? You know, it it, it is not, there's not many reasons to go and just jump to new tools or to [59:31] Richard Webbe: Yeah. [59:32] Matt Slager: Full prey to media releases. I, I think that's the right wording to use. [59:37] Richard Webbe: It's like all those people, Matt, that wanted to run their house fire windows. 'cause it was really cool. Until they had a power failure and they couldn't get in their house. [59:47] Marno Brits: The, um, a great book to, I guess, digest all of that is called the Crux. [01:00:00] Marno Brits: It's about a project management or effective project management. And it, it's all about the crux of the issue. So [01:00:00] Marno Brits: what's the one thing, like not just a constraint or what's the constraint of the business? What's the crux of the business that is really stopping everything from moving on? Um, it, it's a very, very good book. [01:00:11] Marno Brits: I've only, I've listened to it, I'm still reading it. But, um, very, very good approach to, okay, what's the crux of the issue? What are we actually solving for, rather than solving or over engineering or over optimizing something that is irrelevant, that only drives about 2% or the 20% instead of the 80%. [01:00:31] Matt Slager: Yeah, that's a good point. [01:00:32] Matt Slager: 'cause like all of these different things, like the crux and hormo, like wave breaking down businesses, there's gonna be so many others. There's another book called Traction. I know it's really good for that too. Um, all of these things, like the way that I look at this stuff now, and you guys probably have similar, maybe differing opinions is how can I turn the frameworks, the knowledge, the actionable steps, the questions, how can I turn them into something that is systematized? [01:00:59] Matt Slager: Yeah. Whether it's agentic or not, it's just there ready to go so that if I have a question about my business or somebody else's, all of a sudden now we're going through the crux, we're going through homo, we're going through traction, through everything and getting it, uh, emerged, compiled, grounded, actionable result. [01:01:18] Dave Pengelley: I, I came across StoryBrand and I really liked what, what, what Miller was saying about the way he approaches pitches, add elevator pitches and one-liners and all that kind of stuff, and presenting the problem that you solve before. People don't care about your story and stuff. I know. And stuff I've heard before many times, but the way you presented was really clear. [01:01:33] Dave Pengelley: So I'm like, okay, I need to find out more about the StoryBrand framework, which he's on, YouTube videos out the wazoo. Like there's so much content out there for me to grab and gather and then get that into my system. So when I go to ka, my copywriter and my PR person, she thinks about that. She thinks, mm-hmm. [01:02:00] Dave Pengelley: About those concepts when writing any copy. On my behalf or any of my websites or LinkedIn posts or anything like that, because that's, you know, that's really great methodologies and [01:02:00] Dave Pengelley: frameworks that I can then adapt and insert and with the context of everything else that I'm doing makes it more powerful. [01:02:05] Dave Pengelley: Whereas if, like, StoryBrand have their own AI tool, which I'm sure is fabulous because they've trained it really hard on the StoryBrand framework, but it also doesn't have any, in my context, I'm gonna have to go and find all my context and choose which selective context I give it versus taking the framework and just sticking it in my system that already has all of my context. [01:02:22] Richard Webbe: Yeah, and I, I would say, Dave, your personality is what's most important there. Uh, story, branding, framework, formats, the format for a proposal, a tender, everything. Yeah, they've all got formats. I remember hearing Elton John talk about how he knows how to write a hit song. He knows where to start, he knows where to finish, he knows it should be in the middle, all that sort of stuff. [01:03:00] Richard Webbe: And he said, so he, he said he could write a conceivably hit song every day of the week, but unless it's got something unique. That's emotional to him or is triggering in the industry or relates to someone, uh, in, in the thing it, it'll do nothing. Right? Yeah. And, and so yeah, we use these [01:03:00] Richard Webbe: frameworks and it's great that people offer them to us, but at the end of the, the day, it's a piece of view that we want Dave, [01:03:05] Dave Pengelley: right? [01:03:06] Dave Pengelley: And the advantage that we have with these tools is, you know, if you are creating any content, if you've got any back catalog of content, Richard, you've just written a whole book, you can use those as assets for it to train on your voice and your style and what makes you unique. So we have these video transcripts, you know, every podcast I can actually say, go and pull all the extracts of Dave. [01:03:23] Dave Pengelley: That's how I talk. So that's literally me talking, that is my voice. [01:03:27] Richard Webbe: Mm-hmm. [01:03:28] Dave Pengelley: So when you go and write things. Find the patterns, find the ways I talk, find the catch phrases, all that kind of stuff, and adapt those into any writing style that you do. [01:03:37] Richard Webbe: I'm about to start on the audio book. Do you mean I actually don't have to read my own book? [01:03:41] Richard Webbe: I can just throw some AI voice at it. [01:03:43] Dave Pengelley: You can, but it's probably gonna be woeful and I wouldn't recommend it. [01:03:45] Richard Webbe: Yeah, no. It's much more fun when I make mistakes. People enjoy themselves. [01:03:49] Dave Pengelley: I, I hate when like I've listened, I haven't listened to a lot of audio books and, uh, we should wrap up. We hit the hour, but we, you get the author who reads out the prelude or the prologue or the intro, and then you go, oh, this is [01:04:00] really good. [01:04:00] Dave Pengelley: I, I'm like, yep, I can connect to them. That's their content. That's real. And then it switches over to a generic paid voice actor and you're like. Uh, this is crap now. [01:04:07] Marno Brits: Oh, no. Uh, the, something happened was the diary, CEO, he read, I think the first third of the book, halfway through a chapter, like it's still part of the same sentence essentially. [01:04:17] Marno Brits: It just cuts over. And I'm like, no. Like, the rest of it felt so like unsettling because it [01:04:23] Richard Webbe: makes no sense, does it? No, it's still, I, I intend to read my own audio book and I don't intend to do it in a monotone way. I tend to do, I intend to do it with inflection, emotion and passion, because that's what connects us. [01:04:38] Richard Webbe: Not the words, not the information, the other stuff. And that's, I'm gonna inject into my, you get [01:04:43] Marno Brits: to add the little, little gems too. Like, I know, which one was it? I think it was, um, Noah Kogan on his book, the a hundred Million Dollar Weekend or the, the A Million Dollar Weekend. And as he's reading it, he's like, oh, since I've written this, I've also learned this, this, and this. [01:05:00] Marno Brits: And then you keep sharing because the stories will come up as you're reading it. [01:05:00] Marno Brits: So it's almost more valuable. [01:05:01] Richard Webbe: Absolutely. Oh yeah. Brings like that. Sorry Dave, we were, um, resigning ourselves back to your control, sir. [01:05:07] Dave Pengelley: Well, we are, we are wrapping up the show. We are at the hour. So, um, thank you gentlemen. [01:05:11] Dave Pengelley: Great show. Lots of good chat and nuggets. Thank you for dealing with my, uh, production [01:05:19] Marno Brits: fun strictness [01:05:21] Dave Pengelley: that I'm trying to inject. The rigor, the, the, the production rigor that we're trying to insert into this show. Um, just so we can maybe try and get a bit more distribution and help people find out the bits that they care most about hearing. [01:05:33] Dave Pengelley: So, um, thank you all. Thank you, gentlemen. Now remember, as you said, mano at the top. You gotta be quiet. The mics are active until I shut down the stream and then come back to say goodnight. Uh, so, uh, hello, goodbye. Goodnight. Good afternoon, good evening, wherever you are, whatever time it is. Thank you all. [01:05:53] Dave Pengelley: We'll see you next week. Bye guys. [01:05:55] Marno Brits: [01:06:00] Yeah.