Is Australia Falling Behind on AI? Practical Workflows, Not Hype

[00:04] Dave Pengelley: So this is like the least banter we've been as I've hit go live [00:07] Matt: It's [00:07] Dave Pengelley: like in [00:08] Matt: been a [00:08] Dave Pengelley: weeks. [00:08] Matt: rush. [00:09] Dave Pengelley: I blame me. So [00:10] Matt: We [00:10] Dave Pengelley: it's [00:10] Matt: can [00:10] Dave Pengelley: my [00:10] Matt: ban [00:10] Dave Pengelley: fault. [00:11] Matt: it now. Some [00:12] Dave Pengelley: You [00:12] Matt: ways [00:12] Dave Pengelley: guys got [00:12] Matt: everyone [00:12] Dave Pengelley: all the banter [00:12] Matt: angry. [00:13] Dave Pengelley: out before I joined because I was off making coffee. Um because I was running late because you know I'm trying to it's funny, I don't have access to almost any models, and I'm still running late because I'm still talking to what models I do have. It's been a week. I just hit my limit on Chat GPT, actually. [00:28] Richard Webbe: You [00:28] Dave Pengelley: Like [00:28] Richard Webbe: and your [00:29] Dave Pengelley: not [00:29] Richard Webbe: tokens. [00:30] Dave Pengelley: No, no, no, not even. This is through the web. This is through the web app. This is through regular normal [00:34] Matt: Oh, [00:34] Dave Pengelley: consumer chat GBT. Because I've been using six pro, I've hit my six pro limit. So I can still use five [00:40] Matt: researching. [00:40] Dave Pengelley: point six soul is the the next best option. It's still pretty good. Like I'll be fine doing that. But I'd been just like hammering six pro with heaps of live voice chats and brainstorming and and um yeah, I'll I'll share more in my win of the week around how I've been using it, but yeah, like it resets in two days. So I'm like, okay, I can deal with that. Six Astra in mobile and web chat for a few days and my codex resets tomorrow. And I'm like, can I can I hold out? I held out all yesterday. I've held like I've got the show today and some of the runarounds and errands to do. Can I save my banked resets? Can I just like push through now? Like it's been I've I've lasted 48 hours. Can I last another 24 before? And I think I can. I think it's worth it and keep the reset for, you know, when you've got five days left, not not. One day. [01:31] Matt: Yeah, fair enough. [01:33] Dave Pengelley: Although it resets the time period, right? Like that's the thing. Like when you when you use a bank reset, it actually resets your respawn time. [01:40] Matt: Yeah, exactly. Yeah, I've uh I I I went through that that burn that we talked about yesterday because we could see that there was hints of of resets coming. [01:50] Dave Pengelley: Mm-hmm. [01:50] Matt: Um but then they added it as a banked, so like we didn't need to worry so much. So [01:55] Dave Pengelley: Yeah. [01:55] Matt: I only got down to like fifty percent yesterday and um yeah, I [01:58] Dave Pengelley: That's [01:58] Matt: got the bank [01:59] Dave Pengelley: good [01:59] Matt: sitting. [01:59] Dave Pengelley: then. It's good. You got the bank. [02:02] Richard Webbe: Am [02:02] Dave Pengelley: Got [02:02] Richard Webbe: I [02:02] Dave Pengelley: a [02:02] Richard Webbe: uh [02:02] Dave Pengelley: banker. [02:03] Richard Webbe: am I Max Headroom again? Is my video breaking up [02:06] Dave Pengelley: Yeah. It's actually kind of cool. And you got the jacket for it this week. [02:13] Dave Pengelley: The bright jacket, very 80s. Love it. [02:16] Richard Webbe: my favorite jacket? [02:19] Dave Pengelley: People, is this actually Richard, or have you just like got one of your little AI video tools to [02:24] Richard Webbe: I'm [02:24] Dave Pengelley: send [02:24] Richard Webbe: an [02:24] Dave Pengelley: an avatar [02:24] Richard Webbe: AI [02:24] Dave Pengelley: into [02:24] Richard Webbe: avatar. [02:25] Dave Pengelley: the chat [02:25] Richard Webbe: Uh [02:25] Dave Pengelley: now? [02:26] Richard Webbe: this is not Richard at all. This is his uh, you know, eleven year old son pretending to be. [02:30] Dave Pengelley: This is that new real-time Higgs field model or whatever that can generate um content faster than people can watch it. [02:39] Matt: I'm actually impressed if it is it because like those little glitches, they're like they're like a like a vocal stutter, but with your eyes it's like trippy. [02:47] Dave Pengelley: Yeah, it's just it that's it's great. Like this real this frame dropping thing going on. Um but have you seen that? Have you seen um I don't know, is it foul maybe? I think it's the foul model where it can generate new videos. Faster than people can watch it. And so people are like building these infinite slop engines where you just watch continuous generated video without stop. Costs [03:09] Matt: Yeah. [03:09] Dave Pengelley: a fortune to run, but you can literally stream generated video now. [03:15] Matt: Yeah, one of the um the open models from Minimax could do it, the H three model or [03:19] Dave Pengelley: Yeah, [03:19] Matt: something [03:20] Dave Pengelley: mate, [03:20] Matt: like [03:20] Dave Pengelley: that's [03:20] Matt: that. [03:20] Dave Pengelley: what it is. [03:21] Matt: And oh yeah, it's crazy. Like there's this is sort of what we were saying a while ago, like as things increase now, because they're sort of you know quote unquote slowing the frontier, um there's going to just be efficiency gains. So you're gonna see stuff uh get faster and then ideally get cheaper at the same time. [03:41] Richard Webbe: Which brings me to my point that I make most weeks, and that is the human brain is much quicker in adapting to changes and differences and spotting patterns and anything we've created to date. And boredom, and this is rubbish, will click in. The human brain very quickly and they'll go, I'm not watching this. It's not stimulating me, it's not doing anything for me. So this whole race to create content without enough context or need, I think, is going to create video slop as opposed to all those, you know, image slops we see of, you know, some friends who I love them, they're great, they're trying to create some advertising. And I remember when Dave and I were first messing around with the imaging for syllogism and it like to create these very complicated. pages but they don't work there's too much on them people don't read them we try to get it simpler and grab it and the same thing's happening with this all this ridiculous video content i mean [04:39] Dave Pengelley: Here we here you go. [04:42] Richard Webbe: infinite slop awesome [04:47] Dave Pengelley: And people so people are just putting requests in and it's just just generating the stuff in the queue. [04:52] Richard Webbe: see we're gonna get bored pretty quickly [04:58] Matt: I don't know. [04:59] Dave Pengelley: I don't know. I can [05:00] Matt: really [05:00] Dave Pengelley: I [05:00] Matt: there's [05:00] Dave Pengelley: can watch [05:01] Matt: a [05:01] Dave Pengelley: that [05:01] Matt: whole [05:01] Dave Pengelley: screw up for Alice. [05:03] Matt: there's a whole like series of content that exists that's just pure brain rot like I I think the the idea of it not being stimulating enough is possibly not uh a concern I think it's gonna be that it's too stimulating and that it literally does just captivate and become you know it's it's it's the black tar heroine of of digital content you know [05:28] Richard Webbe: Brian Roch and Black. I think they're the two new phrases I've got from you today. They're great. [05:35] Matt: Yeah, [05:35] Dave Pengelley: It's [05:36] Matt: so [05:36] Dave Pengelley: I mean in the in this world of infinite generation stuff, it is it is weird and and where are the kids at it? And I've I'm seeing different posts from people. Actually, we're getting we're getting deep into the show, so come on, let's let's follow the rules and do this. [05:51] Dave Pengelley: Because we have [06:03] Dave Pengelley: Standards, people. We have standards. Uh we're producing producing television here. I mean, YouTube's got a whole thing where they're they're pushing shows and they want regular relatable content as shows. So um I don't know whether that's why we're getting a bit of a leg up on our live streams after the fact now. Uh like last week's did really, really well comparatively. Um we got a good lift a few weeks ago because I used Steve Jobs in the thumbnail, and I thought that was a bit of a hack because I mentioned him. Use one of his lectures. I was like, I'm gonna put him on the thumb there. I'm gonna put him in the title. Lessons from Steve Jobs. Uh, and that worked. But last week it was like Doomerism and stuff, and we still did pretty well on the live show performance, just to talk a bit of inside baseball there. Um, but yes, we're we're making a show. Uh yeah, so we're we're we're kids on AI. Let's let's jump into that one. [06:52] Matt: How do you [06:53] Dave Pengelley: And [06:53] Matt: feel now, like after last week, just out of curiosity, you guys, like because we we did go kind of a little bit do Jimmy, like i i are you still in that mode, or are you feeling more calm about it? [07:06] Richard Webbe: I think AI is copping a lot of criticism at the moment and it's picking up speed. And there's, you know, I think there was something that said 73% of people said AI is going to be dangerous. And again, I flash back to when my parents were advised that black and white television was going to rot their children's brains and they should never do it. Well, after a thousand repeats of lost in space and Maxwell Smart and F Troop, I'm doing okay, I think. So I think it's all unwarranted fear and a lot of complaining for nothing, but hey, we're gonna pass through that needle. [07:42] Dave Pengelley: It's it's interesting. Like you talk about those all those TV shows, and I think depending on your mindset and attitude towards it, nothing can be not like nothing has to be wasted. Uh I mean, arguably people would say video games are a waste of time, but some people might say watching movies are a waste of time. Some people might say Reading comics books is a waste of time. All these things, like if it's not producing, like what is what is productivity? But at the same point in time, if it's enhancing creativity and helping you form new ideas and make connections, that then you can then build into something that adds additional value somewhere else. Like I've been building a video game myself and watching lectures on YouTube from video game design conferences and stuff. And like one of the guys is saying, one of your jobs is, I don't think that is Sid Meyer, the guy who made Civilization. He's like, one of your jobs as a game designer is to play lots of video games. Like that's actually part of your thing. You need to know the reference. You need to know if you're making a kind of game, what do you like in those kind of games? What do other people like? What is the culture around those sorts of games? What do you need? And I'm finding like I'm pulling from like eight different game genres into this thing. And it's a real creative process of brainstorming and back and forth. And I'm soundboarding off GPT, which is helping, you know, coalesce and make sure ca there's consistency in my ramblings. But it's been One of those things. And if I didn't have so many different influences from experimenting with different things and ideas, I wouldn't be able to form the connections today. So where do you like people get confused between invention and innovation? Invention is like coming up with something brand new that no one's ever done before. And that one is really, really hard, especially in the modern era. Innovation is taking something and adapting and improving and innovating upon it. Um, and people would get those two confused. But I think if you don't have a base catalog of things, You've learned from or experienced, then it's hard to innovate. [09:34] Matt: There's a really cool concept that came up in [09:36] Richard Webbe: Good [09:36] Matt: one of [09:37] Richard Webbe: explanation, [09:37] Matt: Blake's [09:37] Richard Webbe: too. Sorry. Come on. [09:39] Matt: um this is just piggybacking on that same that same thing. That that idea of innovation, um, you know, how do you how do you actually innovate in the first place? You know, like how do you come up with something new? So most of the time it's a combination of things, you know, it's the fusion of different ideas or or perspectives. And uh there's a there's a concept that one of um Blake's students brought up in his school community live calls recently. Very, very smart guy. He's always amazing to listen to. It's called TRIZ, T-R-I-Z. And that would be definitely one to add to the the bookmark list or the uh the the the the GPT or Claude research um just to sort of get an idea about what what TRIZ actually means. And it's kind of like the the combination of different Disciplines to gain orthogonal views of things. That you know, if you're an expert in one particular field and you know your stuff and like you're very good at your stuff, but you're you're you're like tunnel visioned on one particular thing and you can't break it, and then somebody from a different discipline comes in and looks at the problem that you're facing and they go, Why don't you just do like this? And it works. And they never would have found it because of that lack of knowledge transfer. [10:58] Richard Webbe: The social construct. Of sharing, as as Dave was pointing out, to innovation and uh and of course improvement and invention. I think you're exactly right. And I think when some of us, like us technical nerds, although not as technical as you two anymore, go down the rabbit hole and you become totally consumed by something, uh, that's where a lot of innovation comes from. You just your brain just goes at a million miles an hour applying it where other people do it in a cursory manner. [11:28] Matt: That's that's basically. Like intuition. And um, I think it was John Wooden in his book, Like Wooden on Coaching. He he talks about basketball players and the idea of them achieving a level of intuition where they can actually react faster than their brain can think about it. So that that's pretty much what creates that excellence in that field where you go deep and you master something, you're no longer thinking about it, you're just feeling and reacting to [11:56] Dave Pengelley: Yeah, [11:56] Matt: it. [11:56] Dave Pengelley: it's that muscle memory that you develop by doing the reps. Um, which is why in so many of these things, it's like you practice free throws for an hour in you know, martial arts, you throw a thousand punches, you throw ten thousand punches over like years and years of training. And so that way, if you do it right and you've been coached to make sure you're throwing the right punches early on and and being constantly corrected, then you build those habits. But um well, I mean, even on the design thing, I think we know AI doesn't have the best taste, and we need that human intuition, that human sense and sense. Sensibility around what looks good, why that thing over that thing. And then every every human has different taste, right? Some people are really into Picasso, some people prefer a Van Gogh, some some people prefer uh [12:43] Dave Pengelley: another artist that I can't think of, which another Ninja Turtle. Um some people like Donatello. Um [12:52] Dave Pengelley: but I mean that and that's okay. But I mean, getting inspiration, like where do you get inspiration from? We were working on um my daughter out of school project. Project and she had to make a little presentation about something, and she had a big empty box, and she didn't want to just stick bits of paper on her cardboard box and look dumb. So I took a photo of all of her things and the box and said to GPT, like, give us some inspiration, give us some ideas of how you could actually make a box art. And it sort of generated things that were based on the assets she had. So, yes, we could have gone to Google and looked at a million amazing little primary school projects, but they would have been harder to like connect the dots because it would have been too. Abstract or too obscure, too beyond. But by going, here's what we're working with, can you actually mash that together in some interesting ways based on the fact that you've seen lots of other things? That gave us a starting point. And what she finally delivered was different from anything we saw. But you know, we were able to take a little bit from that one, a little bit from that one, a little bit from that one, and just guide her into delivering something that fit the aesthetic she she was aiming for. And she she wants to do design. She loves spending time in Canva and compositing little images together. She just made her birthday invites in Canva. So she has An interest in design, but she's also not seeing AI as the enemy. And I think that's [14:05] Richard Webbe: No. [14:05] Dave Pengelley: the thing, like being able to use these tools to steer um is is really important and not just rejecting them whole was bolus. [14:14] Richard Webbe: Totally [14:14] Dave Pengelley: Um is [14:15] Richard Webbe: true, totally true. I mean, Steve Jobs famously said that the difference between a leader and a follower is innovation. And so a leader is always challenging themselves to find a better way to do things. What I love about AI and the Way I interact with it, not just the LLMs. I'm starting to, like you guys, with coaching from you guys, dig down into data, deeper data. It automates our research and brings all of the critical data to the forefront. And I love how the industry has coined the words prompt. Um, I don't say prompt, I just say question. Know the question, and then you'll get to where you want to be. And to get to where you want to be, you need context, so you need to know the questions around that. I honestly sit there. With multiple sources, as I know you guys do, and probably other people do, but I get hyper-focused and I just start asking lots of questions around the barriers of what the objective is and throw other ideas in to see. And it I I do it anyway without AI, it just happens quicker, and I get access to the data quicker. And I love research. So, you know, I go down that rabbit hole on where people have done this before, and how would that affect this? And what if this happened? I just think it's the best fun. And it's going to be the best time to be a human because you'll get the answers you never thought you wanted, and it'll change the way you live your life. And as much as people complain about television or radio or whatever, to be honest, I think it's enriched our lives. Now, the labor issues and who's going to get paid and automation of basic functions, I think they are things that we let government sort out for it. But the idea that we will live more fulfilling lives because we're getting more closer to The coal face or the height of the issue I think is exciting. [16:04] Dave Pengelley: But what what did Reagan famously say? The the scariest words in the English language are I'm from the government and I'm here to help. [16:11] Richard Webbe: Or no, the other one was military intelligence. You're right. So they're they're both they're both the same. [16:20] Richard Webbe: Look, [16:20] Dave Pengelley: Not that this [16:21] Richard Webbe: I [16:21] Dave Pengelley: is a [16:21] Richard Webbe: think [16:21] Dave Pengelley: political podcast. [16:22] Richard Webbe: no, it's not a but I think, yeah, I know what you're saying, and governments will screw up, and they always do, and I listen to the Some of the senators talking about trying to stop data centers or some of the local governments so quickly. It's like you know, a hundred years ago, people saying, No, don't produce any more power or water. It'll upset the balance of life. Well, we need a lot of it now, don't we? So I just think we need obviously need the culture to catch up, and we need some good poet laureates or some good leaders that will speak up about the value of what we're doing and why we're going there. I mean, I always thought Hollywood was quite. Amazing because we would see in movies things that came after the movies, and I don't believe it was a self-fulfilling prophecy. I think those writers and narrators, uh, even people like Carl Sagan, wrote about things that they could envision were coming. That found its way into the arts, was produced as oh my god, that's a great story because it's a new line, and then it helped the audience cope with what might be coming and how to get their head around it. [17:30] Matt: That like the ide the idea of fiction and science fiction, you know, the only reason why it's fiction is because it's not done yet. [17:38] Richard Webbe: That's it. That's a perfect phrase, Matt well said. And that's that book I spoke about, I think you guys know about, where they wrote that most of the innovations that came in our societies come from science fiction writing. And I go back to that thing that uh when a child is learning to walk, psychologists and psychiatrists will tell you, no human can do anything. unless they can imagine it in their mind's eye. [18:02] Dave Pengelley: Yeah. [18:02] Richard Webbe: So we need our artists, we need our creators, we need our crazy thinkers to position in our mind's eye what could be, and then we will build [18:10] Dave Pengelley: And [18:11] Richard Webbe: it. [18:11] Dave Pengelley: you see you see these YouTube creators that are inspired by Marvel and other things, they're fully just these guys independently building little Iron Man suits with the full masks and everything that load up. They're building web shooters. I've seen a guy who's actually been working on how to build web shooters, build working on web liquids and how to get the like the direct shoot or versus. Of the spray web and all this kind of stuff. Like, I saw another guy who was not shooting web, but he was like shooting grapples, and he wanted to like shoot show that he could grapple up into a ceiling on a warehouse and swing across like a foam pit. And he had to be able to shoot multiple grapples and release them and catch the next one. And so all these things that you know seem fanciful and impossible. People go, [18:52] Dave Pengelley: hold my beer. Like, I [18:54] Richard Webbe: No [18:54] Dave Pengelley: reckon I'm gonna give a go at doing that. But they were inspired because they saw it, they didn't just imagine. One day wake up and think, I'm gonna try and make a glove that can shoot spiderwebs. [19:04] Richard Webbe: Well, they're testing that new grappling technology, or I think they're using it in the US now for speeding cars. So police officer gets within a certain distance behind a car that's speeding away. They hit this very sticky, strong rope-type grappling thing that just hangs onto the back of the car and they just put the brakes on. [19:23] Matt: Yeah, [19:23] Richard Webbe: It [19:23] Matt: it's [19:24] Richard Webbe: stops [19:24] Matt: crazy. [19:24] Richard Webbe: the car. [19:25] Matt: I've [19:25] Richard Webbe: I [19:25] Matt: seen [19:25] Richard Webbe: I [19:25] Matt: the [19:25] Richard Webbe: saw [19:25] Matt: one [19:25] Richard Webbe: a [19:26] Matt: that [19:26] Richard Webbe: video. Yeah. [19:29] Matt: there's definitely one. they like they crash into them and that like jams it in it must be like a barb or something [19:34] Dave Pengelley: Yeah. [19:34] Richard Webbe: That [19:34] Matt: um [19:35] Richard Webbe: might be yeah, yeah. [19:37] Matt: the [19:37] Dave Pengelley: Well, [19:37] Matt: uh the the concept of like visualizing something before you can do it i've got a bit of a a brain busting struggle i showed you guys briefly before the show that anyone who uh thinks that they're they're fairly good at typing on a keyboard you know like yeah I can touch type like I felt like I could touch type for for many years now and I went and got [19:58] Dave Pengelley: I a good idea. that's think [19:58] Matt: myself a split keyboard so like This is one half of it. And I I [20:03] Dave Pengelley: Did [20:03] Matt: I'm typing like a child again. [20:06] Dave Pengelley: do you do you find that you actually were like using your index and point for going from left hand and pressing right hand buttons and you didn't realize that you were actually crossing over? [20:14] Matt: 100%. So left hand, I basically never used my pinky. And [20:18] Dave Pengelley: Oh. [20:19] Matt: now I'm discovering like, okay, well, I have to [20:21] Dave Pengelley: How [20:21] Matt: figure out [20:21] Dave Pengelley: did [20:21] Matt: how to [20:22] Dave Pengelley: you [20:22] Matt: do [20:22] Dave Pengelley: how [20:22] Matt: this. [20:22] Dave Pengelley: do you type A? [20:24] Matt: I would use my ring finger for the entire left-hand [20:27] Dave Pengelley: Oh [20:27] Matt: side of [20:28] Dave Pengelley: my [20:28] Matt: the [20:28] Dave Pengelley: goodness. [20:28] Matt: of the keyboard. So [20:30] Dave Pengelley: Wow. [20:30] Matt: because I'm like uh I I was raised PC gaming and stuff like that, like my my hand naturally sits on WASD. That [20:37] Dave Pengelley: Yeah, [20:37] Matt: like [20:38] Dave Pengelley: yeah. [20:38] Matt: my my ring finger always [20:39] Dave Pengelley: Uh [20:39] Matt: sits on A. Anyway, [20:42] Dave Pengelley: so [20:42] Matt: I'll [20:42] Dave Pengelley: you [20:42] Matt: have [20:42] Dave Pengelley: don't [20:42] Matt: to now [20:42] Dave Pengelley: you don't [20:43] Matt: like [20:43] Dave Pengelley: home you don't home on F, like you meant to you home on D. [20:46] Matt: correct. [20:46] Dave Pengelley: So you're a [20:47] Matt: So [20:47] Dave Pengelley: a [20:47] Matt: now [20:47] Dave Pengelley: skew. [20:48] Matt: now I've shifted over one because of this thing, and yeah, man, it's uh it's a lot. [20:52] Richard Webbe: When [20:52] Dave Pengelley: Yeah, [20:53] Richard Webbe: we when [20:53] Dave Pengelley: yeah. [20:53] Richard Webbe: I joined IBM many years ago, it was totally encouraged uh to do a tup touch typing course. uh because we're going into the computer industry. But I have a personal opinion. Obviously smart thinkers like you guys or quick thinkers like you guys need to get it through your fingers. I think typing is going to go the way of writing and it'll all be speaking and verbal [21:15] Dave Pengelley: I think [21:15] Richard Webbe: in the next [21:15] Dave Pengelley: I think it's [21:16] Richard Webbe: few [21:16] Dave Pengelley: horses [21:16] Richard Webbe: years. [21:16] Dave Pengelley: for courses. And I think we've discussed this a little bit, but I've been doing a lot of both lately. Um, and then it's on the but but before I jump into that, on the touch typing thing, I was I grew up in the 80s on DOS prompts and stuff. And so as a kid, I was I knew. The QWERTY layout, but I was a speed typer, but I'd look at it. So I was like a a two two to four finger, just looking where my fingers were going pretty fast, like comparatively, in the especially in the 80s and 90s. Um, but then when I got to and I did typing classes in high school back when they still had typewriter and typewriting classes, and you you'd that was a whole hour of your day learning typing, and then you get your little tip X to like do your uh do your like punch-the erasing and stuff, like old school Richard, you'd appreciate that. Matt's [21:58] Richard Webbe: Did [21:58] Dave Pengelley: like, [21:58] Richard Webbe: you have [21:58] Dave Pengelley: what is [21:59] Richard Webbe: a golf? [21:59] Dave Pengelley: TipX? [22:00] Richard Webbe: A golf ball, uh golf ball type runner. [22:03] Dave Pengelley: No, no. My my um and then then I'd go to my grandparents and had the do ding pfft. We had the electronic ones and stuff. Um, but anyway, I didn't didn't learn a touch type. I still just I just hacked my way through those classes, like typing how I'd always typed at home on my computer. But then when I got to uni and was actually having to write a lot of code and doing long assignments and writing essays and assessments and stuff, it just clicked and my fingers just I just started touch typing without actually thinking about. It. I just my [22:30] Richard Webbe: Yeah. [22:30] Dave Pengelley: brain just worked it out. And all of a sudden I realized, and if I still to this day, you know, 20, 30 years later, if I think about typing, my hands freeze up. I just need to just let my fingers like you talked about the muscle memory for basketballers. [22:43] Matt: I need [22:43] Dave Pengelley: I just [22:44] Matt: to [22:44] Dave Pengelley: need to let my hands do what they do. Um, because if I overthink it. And you know, it's similar now if you think about your phones, your thumb kind of knows the direction and how far to reach. You think you have your thumbs are almost [22:53] Matt: go. [22:53] Dave Pengelley: touch typing in some ways, um, with a bit of [22:56] Richard Webbe: Well, [22:56] Dave Pengelley: autocorrection [22:56] Richard Webbe: those chips [22:57] Dave Pengelley: assistance. [22:57] Richard Webbe: that Elon's working on for, you know, for spinal injury people and and people who are incapacitated. You know, I've seen some demonstrations where just you know the movement of their eyes and thoughts in their brain can actually come out on the screen. Well [23:11] Dave Pengelley: Yeah. [23:11] Richard Webbe: it'd be interesting one day if those people can communicate quicker than we can. [23:15] Dave Pengelley: Maybe and may and maybe like I said, there's a time where you want that there's times where you want to form things internally before you uh represent them externally as well. And so with with chat and again winner of the week, I'll talk a lot more about how I've been using that and having a lot of voice conversations. But also, you know, if If it's late at night and I'm just sitting up in bed and my wife's already crashed out and I still want to keep working and brainstorming ideas, I'm not doing this anymore. I'm back to doing this and formulating things on the go. But you know, I go out for my walk for an hour, and while I'm out doing my 5K, I've got the AirPods in, and I'm definitely talking. I'm not typing while I'm doing that. Um, and so and sometimes, you know, just framing your thoughts up, it is different. And the psychologically, we know neurologically, you compress and analyze. And structure your thoughts differently when you type them or write them versus when you speak them. But then, you know, I wanted to comment on a LinkedIn post, which is relevant and on your stuff around Richard using AI for efficiency gains and so on like that. And I saw this LinkedIn post and went, I've got a hot take. I'm going to comment on that. And I was, you know, making a cup of tea or something. And I was like, I'll just I'll just use the Siri, just the transcribe on my iPhone. So no AI, just the whatever the standard little microphone transcribe buttons on iPhone. Gave it three takes and went. This is bollocks. Just gonna type it. It'll be faster just to type it because I just would not like I'd say hot take and it would write heart take. And I'd be like, oh, I can't do this anymore. [24:40] Richard Webbe: Anymore, [24:40] Dave Pengelley: I'm just [24:40] Richard Webbe: I'm [24:40] Dave Pengelley: typing [24:40] Richard Webbe: just [24:41] Dave Pengelley: because this transcription avenue on this one is not sufficient to help me, and I'm burning time where I could have just written it out. Which I did, [24:49] Richard Webbe: go. gonna [24:50] Dave Pengelley: which I will show you in a sec. But before I do that, any Matt, you look like you want to say something. [24:54] Matt: Oh no no, I I've always got too much. Um I was gonna mention [24:58] Richard Webbe: Go. [24:58] Matt: one thing to do With uh Richard, you said like with Neuralink, which is that chip in the brain thing, um, can we communicate faster through like a Neuralink connection? I reckon 100%. So if you if you picture two situations, like somebody that is chatting to somebody else on the phone, and all you've got is vocal intonation, pacing, you know, word choice, that's a that's like if you quantify that as like a bitrate, you know, like a certain amount of data transfer, now take those people and have them standing facing each other, where They now have to think about their body and they're witnessing the other person's body with their eyes. Now you've got eyes and ears that are, you know, it's like double the bitrate, right? Um, what happens if you can feel the person's like emotions and the other thoughts, the subconscious, the things happening in the background? All of a [25:48] Richard Webbe: Wow. [25:49] Matt: sudden that's now exponential. So [25:51] Richard Webbe: Oh, [25:51] Matt: yeah, [25:51] Richard Webbe: wow. [25:52] Matt: yeah. [25:52] Dave Pengelley: I thought I saw that in Pacific Rim, and that that that's not always a good thing. [25:56] Matt: Yeah. Yeah, yeah. You gotta be compatible. [26:00] Richard Webbe: Some of the early Star Trek, uh, some of the early Star Trek uh shows were like that, and they explored exactly what you're talking about, Matt, that uh you know, voice, and then now all of a sudden emotion between some of the alien races. And I saw a very interesting one. Has anyone watched Stuart Destroys the Universe from the uh the Big Bang Theory, [26:19] Dave Pengelley: No, [26:19] Richard Webbe: the [26:20] Dave Pengelley: I've [26:20] Richard Webbe: spin-off [26:20] Dave Pengelley: not watched. [26:20] Richard Webbe: series? [26:21] Dave Pengelley: I I gave [26:21] Richard Webbe: I [26:21] Dave Pengelley: up [26:21] Richard Webbe: was [26:21] Dave Pengelley: on [26:21] Richard Webbe: very [26:21] Dave Pengelley: Big [26:22] Richard Webbe: funny [26:22] Dave Pengelley: Bang many [26:22] Richard Webbe: because [26:22] Dave Pengelley: years [26:22] Richard Webbe: that character [26:22] Dave Pengelley: ago. [26:23] Richard Webbe: on there, Kripke, who's always rude and uh abusing everyone, and they're standing there looking at each other because they got transported. To a universe where everyone communicates telepathically, and they all go, Oh, this is cool. We can talk without without using our mouths. And Kripke goes, Oh, that's not good. And then he starts abusing everyone. [26:40] Matt: I'm gonna find that, like I I would enjoy that [26:42] Richard Webbe: Yeah, [26:42] Matt: if I ever get time to watch [26:43] Richard Webbe: Stuart [26:44] Matt: something. [26:44] Richard Webbe: can't save the universe or something like that. Go find it. It's it's on one of the streaming channels. Very, very funny. [26:52] Dave Pengelley: Very good. Uh, let's the pivot off, you know, neural links and language, and I want to talk about the state of AI. In Australia. Um, well, how are we going for time today? Yeah, we'll we'll we'll talk about this, then we'll talk about news because there's heaps of news, and then we'll wrap up with some wins of the week. Just so you guys have a clue that how much I want to cram in the next half hour. Um, there's been no news, Matt. What are you talking about? News will be nice and quick. Uh, and then we can do wins of the week. But just a final discussion topic is this one. Natalia, the lovely Natalia, who's been on the show with us before, she gets around the world a little bit. Bit and she wrote, in Australia, one line comes up, we are not sure we need AI in our business. Then I get on a plane. The question I get in Europe is, where is the biggest return for us and how quickly can this be done? [27:39] Richard Webbe: Exactly. [27:40] Dave Pengelley: Same week, a completely different question. How far do you think Australia is behind? [27:49] Richard Webbe: Yeah, I oh yeah. [27:50] Dave Pengelley: I [27:51] Richard Webbe: I [27:51] Dave Pengelley: went. [27:51] Richard Webbe: I I I think we're a long way behind because we lack leadership, uh and as you said, we're not a political uh podcast, but when Uh certain leaderships around the world were saying slow AI down, the US government said, stuff that, let's speed it up, because it will create competitive advantage. And I think sometimes the narrative of people saying slow things down is only a message they want to give their competitors while they're trying to speed things up, both politically and commercially, so they get ahead. So I think we're a long way behind. I think the Australian societal view of AI is wrong and it needs to be. Improved, and that won't happen until we start getting proper education into the commerce and into the schools, and the next generation comes through. [28:37] Dave Pengelley: That's what we're doing here. This is education. We're trying to get out there and make people more aware. Um, Matt, what's what's what's your view on why Australia's behind the curve? [28:48] Matt: Why? [28:49] Dave Pengelley: Yeah. [28:49] Matt: Goodness. I I don't think I can answer why. [28:51] Dave Pengelley: Pick [28:52] Matt: I think I think it's a cultural thing. I think um, you know, Australia has no culture, sorry, I can't say that word. [28:57] Dave Pengelley: stop [28:58] Matt: Um [28:58] Dave Pengelley: it. [29:00] Matt: Uh it's it's gonna be that, you know, we're all we're all like I would say most Australian people, like culturally, has that, eh, it'll be all right, you know, someone will think about it, you know, someone will do that. Not my job. I don't have to worry about that, you know. Like even um sorry, Richard, even before you said at the beginning, like the the government can work that out. You know, like [29:20] Richard Webbe: I was [29:21] Matt: it's [29:21] Richard Webbe: very sarcastic. [29:21] Matt: just something it's something that we do. Well, even in sarcasm then. So yeah, I I think it's important it comes down to the individual. You know, this is not a political podcast, but you know, when when people say, you know, that voting makes no difference, you know, it [29:36] Richard Webbe: Yeah. [29:36] Matt: only makes no difference if you think that. And [29:38] Richard Webbe: Exactly. [29:38] Matt: the it comes down to literally the individual, like what what can be done? You know, like you said um that it it needs to be educated. Like what does [29:48] Dave Pengelley: Yeah. [29:48] Matt: education even look like? How do people get educated? And what do they need to be educated about? Because currently they're getting everything from, you know, uh the news channels. Is that what they should be listening to? And [30:02] Richard Webbe: I [30:02] Matt: I [30:02] Richard Webbe: know, [30:02] Matt: know that's a bit [30:03] Richard Webbe: and [30:03] Matt: of a [30:03] Richard Webbe: they're [30:03] Matt: controversial [30:03] Richard Webbe: all selling, [30:04] Matt: statement, [30:04] Richard Webbe: yeah. And [30:04] Matt: but [30:04] Richard Webbe: what do news channels do? They're all selling fear. There's a [30:07] Dave Pengelley: Or [30:07] Richard Webbe: stat [30:08] Dave Pengelley: or no, [30:08] Richard Webbe: out [30:08] Dave Pengelley: they're [30:08] Richard Webbe: that [30:08] Dave Pengelley: selling [30:08] Richard Webbe: I've got [30:08] Dave Pengelley: Jev [30:09] Richard Webbe: in. [30:09] Dave Pengelley: this week, Richard. They're selling Jev. That's what [30:11] Matt: actually. [30:11] Dave Pengelley: they're selling. Um, [30:13] Richard Webbe: 79% of Australians think that the government and the AI companies are not doing enough to protect us. That tells me they've been sold fear, not opportunity. [30:23] Dave Pengelley: look, I think I think Australian the uh culturally Australia has changed, and I watched um a great doco from the group Build Australia, I think, where they talk about fosters and and sort of Aussie culture. They're a they're a new sort of group that's standing up to try and encourage us actually creating uh stuff in Australia. And that we did have strong culture back in the Paul Hogan days and things. Like people knew what Australia was. It was aspirational around the world, the Australian lifestyle and so on. And and over the last 40 years, that has slowly declined. And even the Shield Be Right mindset isn't what it used to be. She'll be right didn't mean I don't have to worry about it. It just means we'll get on with it and what happens happens and we'll work it out. Not She'll be right at someone else's problem. Um, and I [31:02] Richard Webbe: Exactly. [31:02] Matt: Yeah, [31:02] Dave Pengelley: think [31:02] Matt: yeah. [31:02] Dave Pengelley: I think it that it's changed over the decades. And now [31:05] Richard Webbe: I [31:05] Dave Pengelley: to [31:05] Richard Webbe: think [31:05] Dave Pengelley: your [31:05] Richard Webbe: you're [31:05] Dave Pengelley: point, [31:06] Richard Webbe: right. [31:06] Dave Pengelley: Matt, you'll you think that the whole yeah, she'll be right thing means I can offload it to someone else. She'll be right. You used to mean you'd still get in and give it a red hot go. But you know, if it didn't work, you'd you'd work it out. She'll be right. We're not gonna get too upset about it. We're gonna have a laugh, we're gonna crack on. Um, my my hot take that I put on this post, this is the one I had to write out with my thumbs because voice wouldn't get me there. Was uh Australians are addicted to. Work, not to success, not to outcomes, but to being busy working. Busy is a badge of honor. Succeeding is less so for fear of tall poppy attacks. So if I automate tasks and see more success, but I'm less busy, what will people think of me? I don't think people know this is how they feel, but I think it's a pervading issue that will hinder uptake of new tech like AI enhanced automation. Thoughts. [31:50] Richard Webbe: That is [31:51] Matt: I [31:51] Richard Webbe: gold. [31:52] Matt: I really, really agree with that because um before we read Doug's comment, I f at the I'm With this, I have this weird image thing right now where I sit at home and do nothing. I just clack away on my weird split keyboard and and and I just talk to robots, like you know, but but that that is creating this immense value to the company and the people, people's companies and people that I work with, you know. So yeah, with to outsiders, like looking at this, you know, I probably just look like a Doll Roger just sitting at home, but you know, because there is no sort of perception. Like, oh, he's actually working hard, you know, doing stuff. So yeah, I agree with that and I I resonate with that. [32:36] Richard Webbe: Yeah, being busy versus being productive or solving problems is a perspective and a context. And it does relate to that should be right, mate. I always put custom should be right, mate, that is in two ways. Um, one that I'm focused on this right now. And so whatever that is, I say should be right, mate. But at the same token, every time I've ever said that or my businesses have done that. It's because we have a backup plan if it's not right. I think people are saying today, she'll be right, mate, with no backup plan, no strategy, and that's deadly [33:11] Dave Pengelley: Yeah, interesting. [33:12] Matt: I was like just Doug's [33:14] Richard Webbe: planning. [33:14] Matt: quote. [33:14] Dave Pengelley: Yes, let's bring let's bring Doug's. Um good to have you back, Doug. Companies and countries that are behind are in a simple situation of going to wrong with going wrong way down the road and waiting until they see that mile marker that they now need to turn around. Yes, it's now miles down the road. And that not just the next exit. Yeah. Um, and you know that there's a huge cost, not just the opportunity cost, but the actual physical time and energy cost. This happened to me and my wife. We were going up the coast and we had Waze on the GPS, and and my wife got a work call that she had to take. And she was on this call for a long time. And for whatever reason, while she was talking on her phone, Waze froze up. And we missed the exit. And we missed the next exit, and we missed the next exit. And all of a sudden I'm like. Nah, this doesn't fit. Like I and because we just be I can't be relying on the tech. I didn't go, yeah, I know it's this exit off the off the MA, whatever they call the one going up north now. Um I was like, Yeah, I I know which F3 exit I should take. No, I d I was just relying on the little machine and just going, oh, it'll tell me, it'll tell me, because it's probably calculated some traffic blockage or something, and so it knows. And then I realized, nah, this is wrong. So she's still on the phone and I I get off and veer off and take a UE and start heading back. But you know, that cost Us like an extra hour and a half of where we were going, it costs us all the extra fuel. It cost us this in that instance. But if you think about it, you know, at an entire economy scale, if you continue going in the wrong direction, the burn of the lost productivity and the opportunity cost of these things while everyone else gets a competitive advantage and so on, it's quite massive. [34:49] Matt: It's kind of exactly what the the rest of Doug's thing said. It kinda it cut off when you brought it up. He said um he said the the sheer speed of moving at the speed of now, you know, i if they wait Known good or best practices, they've already lost massive market share. So it's the same thing. Like, you know, you're missing your turn-off because the system stuttered, you know, if you if you didn't, if you weren't able to act at the speed of now, like knowing what's going on and being able to predict or forecast, you know, the direction you should be going, like you said, huge economic uh fallout from that. [35:20] Richard Webbe: I love Doug's comment about jaywalking. Uh it's an interesting term being a cyclist. I'm across a lot of the old road history, Doug, and you're right. A jay. People used to think jaywalking was when someone walked the wrong way off the linkage center about what we're talking about versus uh doing things manly versus AI. And a J was a bogan in those days. And Chrysler and Ford brought in this thing, uh, a legal rule to cover a cultural issue and made jaywalking illegal. That is, walking like an idiot was what it was about. That is, you didn't cross at the right point. And so, you know, as AI. Comes in and takes over and automates things, as as Dave was saying, right, we'll start to all behave like Jays or bogans or idiots because we're not using our brain in the right way. My worst example is I decided to ride to a friend's work the other day on my pushbike, which was a good 40Ks away, but I wanted to just get a good ride in. And I had it all programmed into my GPS, but I didn't have the last positioning. I was going to the suburb, not the address. So I said to the friend, just send me the actual address. Well, they decided not to. do that and I'm riding a bike, they decided to send me the name of their work and you know pick it on this map here. And of course it went into my phone. So while I'm riding so I don't die, I just hit the button. And for some reason I turned left and ended up 30 kilometers away from where I was supposed to be. [36:49] Matt: Oh [36:49] Richard Webbe: Because [36:49] Matt: no, [36:49] Richard Webbe: it [36:49] Dave Pengelley: On [36:49] Richard Webbe: sent [36:50] Dave Pengelley: a [36:50] Richard Webbe: me [36:50] Dave Pengelley: bike, [36:50] Richard Webbe: to the wrong [36:50] Dave Pengelley: that's [36:50] Matt: I don't [36:51] Dave Pengelley: even [36:51] Matt: know. [36:51] Richard Webbe: office. [36:51] Dave Pengelley: worse. [36:51] Richard Webbe: Sent me to the wrong office and it was a really hilly part of Melbourne up the back of Blackburn. And man, it was good exercise, but boy I was annoyed. [37:00] Dave Pengelley: Fair enough. Fair enough. Yeah. So for those of you that are in this space like us that are pitching this sort of thing, and I would say we've just got to keep normalizing the conversations and talking about the real use cases of using this sort of stuff, not the hype cycle stuff, but where you're getting those small productivity gains. Um I won't use it in the wins of the week, but this week I recorded my first uh operator stories interview with a lovely uh Tracy. [37:27] Richard Webbe: I [37:27] Dave Pengelley: And [37:27] Richard Webbe: saw [37:27] Dave Pengelley: we [37:27] Richard Webbe: that. [37:28] Dave Pengelley: we had a great chat that's gonna go up in the chat. Channel probably tomorrow. We've just got to edit an intro for it. But she was amazing and she's, you know, um pivoted her career from doing insurance sales and managing sales teams to all in on the AI thing and had a great chat around some of the things she's seeing in those sort of small businesses that she's helping. But I asked her just her general usage of AI, and she said, like as a mother of four, running her own business, one of the things that she does is optimizes her meal prep on a Sunday afternoon. So she tells it, here's all the meals, here's all the people, here's all the food we've got. And it said, she said it even says, okay, get that boiling because that's going to take 30 minutes. While that's taking 30 minutes, chop and prepare these things because they're going to be needed in 10 minutes and then get these next things. So it actually was like time ordering and optimizing the order of execution across five different meals that she was cooking to pack and prep for the week, down to, you know, do this first because it takes longer and doing all that kind of stuff. Like, though those are the sorts of things people don't realize that you can do with it. [38:28] Richard Webbe: So [38:28] Dave Pengelley: It's not [38:29] Richard Webbe: I [38:29] Dave Pengelley: just [38:29] Richard Webbe: only Efficiency and optimization side from both a professional macro level or a personal level. I mean, sometimes I go to my fridge and just for fun, I enter in the last few ingredients that are sitting in my fridge. I have uh, you know, a neurodiverse daughter and a very choosy son, and I've got to work out something they're gonna eat with what I've got left in my fridge. And [38:49] Dave Pengelley: Yeah. [38:50] Richard Webbe: it does it for me and says, try this, try that. Okay, took into account the personalities of my two children and saves me a lot of time. [38:58] Dave Pengelley: Just to close back. On where we started with the inspiration, where do you get ideas from? People get ideas about how to use these things from people like us talking about how we use these things, right? And just sort of giving them that sort of permission to experiment and try because you're they're not the first one, it's not scary, and you've done some of this sort of thing before. Um I've got one of those moments where it just slips out of my head as I'm about to talk about it. [39:23] Matt: I have a [39:24] Dave Pengelley: Um [39:24] Matt: really good I got a uh something I can contribute to that. Maybe [39:26] Richard Webbe: Get [39:26] Dave Pengelley: please, [39:27] Matt: I'll [39:27] Richard Webbe: rid [39:27] Matt: put [39:27] Richard Webbe: of [39:27] Matt: it in [39:27] Richard Webbe: the [39:27] Matt: the [39:27] Richard Webbe: filter [39:27] Dave Pengelley: Matt. [39:27] Matt: wheel. [39:27] Richard Webbe: that I [39:28] Dave Pengelley: Please, [39:28] Richard Webbe: just got for [39:28] Dave Pengelley: please, [39:28] Richard Webbe: it. [39:28] Dave Pengelley: Matt, [39:28] Richard Webbe: Sorry, [39:29] Dave Pengelley: while [39:29] Richard Webbe: mate. [39:29] Dave Pengelley: my brain reconnects. [39:30] Matt: Did I did I tell you guys about what I did for my partner's 30th? [39:34] Richard Webbe: No, but I love to hear [39:36] Matt: Because this was all driven by, you know, AI. This is me just talking to to Jipity on on pro, uh, because [39:42] Dave Pengelley: Yeah. [39:42] Matt: I haven't used up my my usage. Um yeah, [39:45] Richard Webbe: it. [39:46] Matt: so the the idea of like, you know, oh, but these these tools are just there for uh restructuring your email or uh doing a little bit of research or maybe writing code or spinning up a website or something, you know, it it's So limited. Like my dad used to do this physically. He'd look at me, um, not say anything, and he'd he'd go like this. Meaning like blinders [40:09] Dave Pengelley: Huh. [40:09] Matt: on a horse. You know, you're [40:11] Richard Webbe: Yeah. [40:11] Matt: not thinking, you're not like expanding and looking sideways. So yeah, what I did with my partner, it was the 30th, we wanted to do something special. Um, I thought, well, look, I could go and I could just buy something expensive, material approach, wrap it up. Here you go, happy birthday. Um, but no, like it she got a day, a day off work. Um I did my best to take a day off work as well. And we went touring around. Like so I live in Aubrey and you know, nearby we've got towns like Myrtleford, Bright, Beechworth, um, you know, these these really nice hilly country towns, especially in spring, just coming out of winter. And she loves that sort of thing. So I went into just standard Chat GPT, chucked it on pro and I just voice dumped. Now, if if anyone hasn't just voice dumped, like, [40:56] Richard Webbe: Oh, it's [40:57] Matt: do [40:57] Richard Webbe: brilliant. [40:57] Matt: it. [40:58] Richard Webbe: Yeah, [40:58] Matt: Because you [40:58] Richard Webbe: it's [40:58] Matt: need [40:58] Richard Webbe: brilliant. [40:58] Matt: to be able to get everything out. And yeah, it analyzed everything I said. It it it heard the fact that you know she loves wandering around cottage gardens, she likes historical stuff, um, nice cafes, you know, interesting food, places to go look at and like go through like shops and you know, just browse things, um, walking paths, you know, stuff like that. And it it basically wrote out an itinerary for me of where to go and what to do. Now I had one more thing because she loves puzzles. And I was like, okay, I kind of want to have this thing where I want to have handwritten notes, you know, some memories, some like [41:32] Dave Pengelley: Yeah. [41:32] Matt: you know, just nice stuff. [41:34] Richard Webbe: You are so [41:34] Matt: Um [41:34] Richard Webbe: romantic, man. I love it. Good gun, buddy. [41:37] Matt: and I got four cards. So these four little cards, and I will I basically said, okay, I've got these four cards. I want to give these out, you know, during the day, like at different moments and intervals. And in there I want to also have like a cryptic puzzle thing. And you know, it ended up being a ciphertext, you know, numbers relate to the alphabet, but with 30 added on. So it didn't look like one to 26. Anyway, it was really cool. She loved it. And um yeah, everything went really well. So I could have thought all that myself, but it would have taken me probably weeks. And I just did it in a in a couple of days of just chatting to um to GPD [42:17] Dave Pengelley: Yeah. [42:17] Matt: on the site. So yeah, [42:18] Richard Webbe: What [42:18] Matt: it's [42:18] Richard Webbe: I [42:18] Matt: amazing. [42:18] Dave Pengelley: Yeah. [42:18] Richard Webbe: like about the memory function, which a while ago I didn't know existed inside the LLMs, and I think you both told me you select the, you know, connect the memory across your chats. [42:30] Dave Pengelley: Recently turned it on. Claude didn't have it for a long time, the way Chat GPT [42:33] Richard Webbe: Yeah, [42:33] Dave Pengelley: does. [42:33] Richard Webbe: well, I did that, and it's great because similar to you, Matt, you know, I've got I've got some challenging children who I love very much, and I was trying to solve a problem, and the LLM just got that Venn diagram of what I already knew about my child and just reminded me and came back and said, Richard, we could do that, but if we did that, it would breach these things. And I know they're important because we had that conversation a while ago. And I went I it's it is it is a panacea of opportunity for us all make use of to enrich our lives and improve efficiency. [43:10] Matt: I need [43:10] Dave Pengelley: Yeah. [43:11] Matt: to get you uh connected to the core, Richard, so you get actual real long term note taking. [43:18] Dave Pengelley: Okay, [43:18] Richard Webbe: What's [43:18] Dave Pengelley: you guys [43:18] Richard Webbe: the [43:18] Dave Pengelley: say [43:19] Richard Webbe: core? [43:19] Dave Pengelley: that offline's mind is blowing right now. Um I remember what I was gonna say. This is something that came up in the in the interview with uh Tracy yesterday, is like and it comes back. That inspiration, what have people seen before and what do they know about? And I I gave her the um Henry Ford quote, which she hadn't heard before, which is if Henry Ford had asked people what they wanted, they would have asked for faster horses. [43:43] Richard Webbe: That is a brilliant quote. Now, did you ever hear what Carl Sagan spoke about the space shuttle and space exploration? Almost the same thing, David. I'm glad you brought that up. So what happened is I believe when they said, let's build a reusable spacecraft to go to the moon, NASA was Went away and gave the stuff to the engineers and said we want a reusable spacecraft. And they came back with this multi-trillion dollar bill of materials. And everyone went, Oh, that's too much. You can't do that. And so Carl Sagan went to the artists and said, Would you please draw for us what you imagine repeatable travel to the moon would look like? And they all came back with shuttles because they were used to flying in planes. They then got that and took it to the engineers and said, How much to build that? And it was, you Much, much cheaper and very viable. And the engineer said, Why didn't you ask us for that in the first place? It is as you said, Dave, the prompt, the question, the objective, necessity is the mother of invention. Build [44:43] Dave Pengelley: Tell [44:43] Richard Webbe: me this. [44:44] Dave Pengelley: tell [44:44] Richard Webbe: I [44:44] Dave Pengelley: that [44:44] Richard Webbe: want [44:44] Dave Pengelley: to [44:44] Richard Webbe: to [44:44] Dave Pengelley: Elon [44:44] Richard Webbe: go [44:45] Dave Pengelley: with [44:45] Richard Webbe: here. [44:45] Dave Pengelley: his chopstick rocket catcher. [44:48] Richard Webbe: Remind [44:48] Dave Pengelley: Obviously. [44:49] Richard Webbe: us of that. Oh, [44:50] Dave Pengelley: Sorry? [44:50] Richard Webbe: yeah, no, sorry, yes, [44:51] Dave Pengelley: Yeah, [44:51] Richard Webbe: no, [44:51] Dave Pengelley: with [44:51] Richard Webbe: yeah, [44:51] Dave Pengelley: his [44:51] Richard Webbe: I don't [44:52] Dave Pengelley: rocket [44:52] Richard Webbe: care. [44:52] Dave Pengelley: return, no no shuttle, no plane takeoff. It's like traditional just sort of like low drag rockets and it comes back down from [44:58] Richard Webbe: Yeah, [44:58] Dave Pengelley: the little chopstick [44:59] Richard Webbe: why did [44:59] Dave Pengelley: goes. [44:59] Richard Webbe: it have To look like a plane. It's like car design. It's like all [45:03] Dave Pengelley: Well [45:03] Richard Webbe: design [45:03] Dave Pengelley: aer [45:03] Richard Webbe: comes [45:04] Dave Pengelley: aerodynamics, [45:04] Richard Webbe: from her history. [45:04] Dave Pengelley: technological capability, ability, all all the things. Alright. [45:17] Dave Pengelley: I'm not sure there's been [45:25] Dave Pengelley: any news. This is a short segment. We can we can move on now. [45:28] Matt: Yeah. [45:31] Dave Pengelley: Where where do we start? I mean, mean model model fest this week, right? Like, um, what have we got? We've got the big one today, GPT 6, Sol, and Luna. Terra is dead. There is no Earth. It's all about the the other celestial bodies. Um, so these are these are competing, like we're supporting Astra now. They've been trained similarly to Astra, but much cheaper. Astra's still the the superior highest level model. So Getting closer and Lunar Max apparently is is up there with Sol when it comes to coding tasks and significantly cheaper. They've halved the price of Sol on Luna with this new release. Wild. [46:12] Matt: Yeah, it's pretty cool. I would argue that um GPT6 Sol is actually better than Astro now as a default choice. The [46:19] Dave Pengelley: Really? [46:19] Matt: only time yeah, the only time you should use Astro is when you need that sort of unbounded creativity. So if you just want to get a job done, use Sol. There's no reason. Like there's no reason. Reason to [46:30] Richard Webbe: Prince. [46:30] Matt: use Astra as your main ever. Um, the way that you should use Astra is to instruct your current agent to use Astra as a sub agent when you need extended creativity. [46:42] Dave Pengelley: Interesting, because I've just I've just set up my um I'm setting up my Hermes with a an updated skill that runs Astra as the the harness handler, but is very clear for most tasks use Lunar Max. If Lunamax comes back and didn't quite get the job done, then give Sol a go. But but sort Of still keeping Astra as the sort of layer above. [47:04] Matt: Yeah, I would I would give it a go swapping that. Because [47:08] Dave Pengelley: Okay. [47:08] Matt: the the new soul is much nicer to speak to. Uh that was a huge problem with the with five-six soul that it would [47:17] Dave Pengelley: Well, [47:17] Matt: just break [47:17] Dave Pengelley: I [47:17] Matt: the [47:17] Dave Pengelley: suppose this this is the thing. I haven't been talking to it. So so and and I'm gonna uh uh wins of the week in my workflow that I've been using. Um I've been getting JPD6 Pro. To actually create the issues in GitHub and then Hermes is picking those up as Astra and then delivering them. So I'm actually not really talking back and forth with the model directly. I've actually been talking to 6 Pro, whatever that is. Um, and and that's been my discussion model. [47:48] Matt: That that is a bit of like a conversational hack too. So when you're interfering interfacing with Chat GPT just through chat mode, um whether it's instant or reasoning or even pro, like it it feels better to talk to in those moments than just talking directly to the model. [48:06] Dave Pengelley: Yeah. [48:07] Matt: Um yeah no impressive uh the efficiency as well like the idea of it being half the price like that's [48:13] Dave Pengelley: Well [48:13] Matt: crazy. [48:14] Dave Pengelley: Yeah, that's that's completely wild that they've like taken their old price for five point six, and even five point six Luna, which was already lunacously cheap, and they've half get it lunacously. Uh they've they've dropped it half again, which makes Luna like really Compelling to use a [48:32] Richard Webbe: I mean [48:32] Dave Pengelley: lot. [48:33] Richard Webbe: I'm interested in the marketing behind this and if just between all the LLMs there's a real fight for uh money uh market share usage at the moment. I'm just wondering if there's been a dynamical change in that space. [48:48] Matt: Yeah, yeah, keep keep an eye on China. Like and the that's exactly where all this stuff comes from. So like Deep Seek uh 4.1 flash is now more expensive than Luna and you know. they're [49:00] Dave Pengelley: Wow. [49:00] Matt: they're roughly equivalent. [49:01] Dave Pengelley: Um, [49:02] Richard Webbe: And [49:02] Dave Pengelley: this this [49:02] Richard Webbe: China's [49:03] Dave Pengelley: also [49:03] Richard Webbe: just [49:03] Dave Pengelley: dropped. [49:03] Richard Webbe: released a whole lot of newer AI chips as well, I think. [49:07] Matt: So yeah, this is arguably the the bigger release. I I you know I've I haven't used port a lot recently as much. I've only been on the hundred dollar max as opposed to the 200 end. Um opus 55 is actually huge. Like it it's it's better than Fable in most things now. There's always gonna be that slight you know moment where Fable feels nicer to talk to and like gets you better. know [49:30] Dave Pengelley: Mm-hmm. [49:30] Matt: so fable could continue to be that you know initial ideation phase but then you would just default straight back to opus five five [49:38] Dave Pengelley: I love how love how for computer use they just don't show GPT Six Astra, which is renowned for being very good at computer use. But oh, they didn't use the OS World two point zy must have used another one. So we just can say that they don't have computer use on GPT Six Astra. That's hilarious. That's [49:53] Richard Webbe: Lack of [49:53] Dave Pengelley: marketing [49:54] Richard Webbe: data. [49:54] Dave Pengelley: for you, Richard. [49:55] Richard Webbe: Yeah, it is. Absolutely. What is the net that I throw around the context? Of what I'm saying. [50:01] Matt: The uh the the grok release was pretty funny too though, um where [50:06] Dave Pengelley: What's [50:06] Matt: they they actually compared they compared uh 4.7 extra high to 4.6 high in their chart. Like such a weird like apples to oranges. [50:19] Dave Pengelley: it here's more more models four point seven landed are extra high. [50:26] Matt: Yeah, there you go. There's that that column header. [50:30] Dave Pengelley: Extra high against 4.6 high. So yeah, we're gonna put more effort into it and compare it against the one with less effort. [50:36] Matt: Yeah, [50:37] Dave Pengelley: Um [50:37] Matt: as a [50:37] Dave Pengelley: why [50:38] Matt: I think that's such a weird thing to do from like a just a [50:41] Dave Pengelley: right [50:41] Matt: marketing [50:41] Dave Pengelley: that [50:41] Matt: trust [50:42] Dave Pengelley: it [50:42] Matt: perspective [50:42] Dave Pengelley: should be [50:42] Matt: as [50:42] Dave Pengelley: 4.6 [50:43] Matt: well. [50:43] Dave Pengelley: extra high. Like if the if these two are on max and it's saying that it's better than these ones on max. Um, [50:48] Matt: Yeah. [50:48] Dave Pengelley: and obviously now this pricing is out of date. That's down to 10 for six. But anyway, this was released two days ago, three days ago, whatever. So [50:56] Matt: It was [50:56] Dave Pengelley: um [50:56] Matt: a disappointing one because Elon was promising much more efficient and it's the same price and it uses more tokens so it's actually way more expensive [51:05] Dave Pengelley: Yeah. [51:06] Matt: and then yeah it increases on if on intelligence and stuff like that. No it's basic the same [51:12] Dave Pengelley: I'm finding my I'm on a cheap $30 Grok plan and I would just find that it is just horrendously inefficient. I burn through that insanely quickly in [51:20] Matt: yeah. [51:21] Dave Pengelley: it, my my week's quota. Um Grokbot is more generous. I actually find I'm getting a lot more value out of the Grokbot from my $30 sub than I am from Grok itself. [51:30] Matt: Currently, right now, the best products in that whole ecosystem is Grokbot. But in order [51:35] Dave Pengelley: Yeah. [51:35] Matt: to have that, you need to have Cursor or a Grok sub. In Cursor, which is the main interface layer between Grokbot to actually spawn cloud agents and that sort of thing. If you use Grok 4.7 in Cursor, it's actually way worse than using it in the Grok Build CLI itself. And but you can't spawn anything with Grok Build by default. You have to do special. actual things like use GrockBot to either in its own machine or in a remote [52:03] Dave Pengelley: Yeah. [52:03] Matt: machine or spawn the cloud [52:04] Dave Pengelley: It's [52:04] Matt: cursor agent [52:05] Dave Pengelley: it's [52:05] Matt: and [52:05] Dave Pengelley: so [52:06] Matt: then [52:06] Dave Pengelley: so [52:06] Matt: off [52:06] Dave Pengelley: weird, [52:06] Matt: the machine. [52:06] Dave Pengelley: right? Because because Grokbot, I wanted to generate video and Grok Imagine exists, and that's a thing. Um Grokbot has built in image gen, but if you want to do video gen, the only way I could do that was to sign into its web browser with my Grok account to give it access to the web browser interface for Grok Imagine to generate videos. And I'm like, are you kidding me? Like, what what is that Grok Family? I think it's because they're Differentiation between Cursor and Grok and the fact that Grokbot has actually spawned out of the Cursor family and came with the cursor acquisition. It looks like, by the way, you log in and where things go, it seems that Cursor was building Grokbot before they got acquired. And so then when it got acquired, it got launched as Grokbot, not CursorBot or whatever. And [52:48] Matt: Yeah. [52:48] Dave Pengelley: so it's more from the Cursor side than the Grok side, which is why it doesn't have the video gen properly integrated until we get to you know Grok 5 or um whatever it is. [53:00] Matt: According to [53:00] Dave Pengelley: Apparently. [53:00] Matt: the uh so [53:01] Dave Pengelley: Apparently. [53:02] Matt: I [53:02] Dave Pengelley: But anyway, [53:03] Matt: I'm [53:03] Dave Pengelley: lots and and and and that that that's the the the the old school boring LLM news, let alone the the explosion of Jev. And we were we were ahead of the curve, guys. We were talking about Jev last week before we even knew what it was. This week, I'm still not even sure anyone actually knows what it is because the way that everyone's talking about how to use it is just insane. Like you can't [53:24] Matt: cases [53:24] Dave Pengelley: be on the internet without hearing about Jev this [53:26] Matt: yeah, [53:26] Dave Pengelley: week. [53:27] Matt: use cases are so much difficult to define. than actually what JEV is. I think it's so much easier to define what it is now, now that everything's settled. [53:35] Richard Webbe: Yeah, [53:35] Matt: It's a [53:36] Richard Webbe: because [53:36] Matt: classifier. [53:36] Richard Webbe: the abstraction layer of an LLM is a lot easier to understand for people than what Jev really does. Sorry, go on, Matt. [53:42] Matt: Yeah, it's literally a classifier. So if you think like a perfect example is in your email inbox, if you've got automated labels, you know, an email comes in and it's it's a it's a label that attaches to this particular, you know, folder, label, whatever you want to call it, depending on your email inbox. Jev can do that kind of thing. That's a use case. You're classifying something instantly. Now think about a different one, like moderating a chat channel. You know, a message comes through, you know, is this offensive? You know, like whatever the moderation settings are. You you have those kinds of things. It's it's classifying. So it's really a smart if statement. It's a it's a router. And those those routing moments, those little, those little decisions, those classifications, you have to pre-define the output. So the idea of Jev is you tell it what the outputs are with regards to the schema that's going to be an input and an output, and then you pump it at it, and it'll give you an instant result based on probability. [54:43] Richard Webbe: What a great explanation, Matt, because I've been struggling. So it's like a a super advanced filter, but [54:48] Matt: Yep. [54:49] Richard Webbe: you need to know what you're looking [54:50] Matt: Yeah, [54:50] Richard Webbe: for. [54:50] Matt: you can't get an output until you tell it what the outputs are. [54:54] Dave Pengelley: Yeah. The the the their sandbox is really good. It sort of says like, what is is food a sandwich, and then you tell her what food is But you've also got to define what a sandwich is. And so you say, well, it is a sandwich if it's, you know, two wheat-based layers with fillings in the middle, and then you s pass it a description of a food, and then it goes, Oh, it does food fit that classification of what you've said sandwiches, yes or no, with a percentage chance. And so it it's very [55:20] Richard Webbe: That [55:20] Dave Pengelley: much [55:20] Richard Webbe: would have that would have huge applications in the medical [55:23] Dave Pengelley: well, [55:24] Richard Webbe: industry. [55:24] Dave Pengelley: pe people are using it for things like, you know, um case triage and uh email, email sorting. And so on and things like that. Whereas before you might have sort of fuzzy logic type stuff. Now you can actually get this to do the classification. So rather than looking for keywords with a Python script or something, this is a little bit more not LLM, but gives you that like that ability to just have a looser definition than specific keywords. [55:48] Richard Webbe: To give people a context, I used to sell supercomputers many years ago out of a supercomputer company called Control Data, and we sold the ECA and the cyber supercomputers. And we sold them to the Walter and Eliza Hall Institute, which is one of the world's leading medical research organizations. And they were running classification filter searches across DNA of certain cells and viruses and things like that to try and qualify what the AIDS virus was. And they used to run it, they used to run it on, I think, a VAX, a very large digital VAX, and the answer would take them four months. They then used what we call parallel processing. Which is a sequencing where you set up almost similar to what we'd remember, Jeff, where you set up a series of calculations, but they run in parallel in a stream, but they're very narrow. They're not scalar. They can't change as they go and you just run it really fast. And as you were talking about that, because when they got the cyber onto it, they could do the answer in a day. Now, of course, with the AI we're using and things like Jeb, I'm guessing those answers will come in a few seconds. Actually, you know what? There was an example of this. You triggered my mind. Solving endodometriosis for young ladies [57:03] Dave Pengelley: Yeah. [57:04] Richard Webbe: was originally a 12-month minimum exercise. They would take a scan and then they would wait 12 months and take another scan, and they could clearly define if the, and then I could be wrong, but there was a doctor talking about it in an article I read the other day. It now takes one scan and 0.018 seconds for them to work out if someone's got endodometriosis. [57:28] Dave Pengelley: That's that's crazy, and as things speed up you can you can do more experiments you can try more things I watched an interview with a guy talking about the installation of a crazy supercomputer at one of the big research places in Europe and he's saying because like in old like traditionally with research you'd have to spend a long long time making sure your parameters for running a simulation were right because it was going to run for like six twelve eight hours days whatever but as these things run so much faster you can try things much quickly you can actually burn through options and just quickly experiment because you know you can get them back quick And so you're not wasting a whole week waiting to see. So you can try five things instead of one in the same period of time. And so you can, you know, try different outcomes, get better, better ideas and better things. Um [58:10] Richard Webbe: Some history on what you just said. Dr. Cray, who I met many, many years ago, was the architect of the original control data supercomputers. [58:20] Dave Pengelley: yeah. Yeah. Yeah. Crazy, crazy, crazy, crazy [58:25] Richard Webbe: Smart [58:25] Dave Pengelley: hardware. [58:25] Richard Webbe: people. Yeah, crazy. [58:26] Dave Pengelley: Um [58:26] Richard Webbe: Go for it, dad [58:27] Dave Pengelley: we [58:27] Richard Webbe: joke, [58:28] Dave Pengelley: we we're [58:28] Richard Webbe: man. [58:28] Dave Pengelley: we're on the hour. I I Pushing, I'm trying to cram, I wanna wrap this up soon because we we do like to try and keep the show about an hour reliably for our uh our viewers. Um but yeah, Jev, Jev was crazy. I saw this guy, uh Jack Rudenko, Aussie guy, talking about how he couldn't see it. And he actually makes reference to back in twenty eighteen Google, who also made the GPT, released Bert, a model that reads the sentence and decides which bucket it belongs in. A model like that's been hiding spam from your Gmail inbox for years. [58:56] Richard Webbe: Exactly. [58:56] Dave Pengelley: Um and so like this Jev thing, it's it's It's great that people are getting awareness of this sort of thing, but these sort of classification models aren't necessarily new either. And I think some of these are not necessarily helpful, these videos and diagrams, right? Because it can't read the picture. You actually have to describe the picture and tell it what its next moves are. These are all screenshot images that another model or something is describing to Jev to then choose what to do. Like so all these fancy visuals is playing a game doing that, it's getting that pre-processed by another model or something that's describing what it's seeing [59:30] Matt: It's based on it's based on state. So like game board's state can be can be structured data. So it's you just pump in the structured data. [59:38] Dave Pengelley: yeah [59:39] Matt: So [59:39] Dave Pengelley: yep [59:39] Matt: like position on the board, you know, i i is a value. [59:43] Dave Pengelley: and you know you meant you mentioned China Matt that within sort of 48 hours this uh layer model is being talked about as a local model that does the same thing the difference is layer um doesn't have as much base data like if there's tread there's model training data is is more Limited and so out of the box, you can't zero shot as much stuff with Layer. But Layer, you can program your own weights into it. So you can actually train your own weights based on your own data sets to then make Layer reply better, um, have better accuracy, and then it's configured specifically to your needs, and you run it locally, so it costs you nothing. JEV's super cheap, but running it on your own hardware for nothing can be faster and cheaper than even running JEV. Um, so I'm I'm actually using that on Grokbot at the moment. I've got Grokbot running an experiment. Where it installed Layer for me on the Grokbots computer, and then it had went and got a training data set and it used this virtual GPU service called Kaggle, which I'd never heard of, but basically you're renting GPU off people for free. So it went and used someone else's GPUs to train the weights to then plug in the layer to give me this classification model for free on my cloud grokbot. So it's not using a single [01:00:55] Dave Pengelley: percent of megahertz off my off my home laptop. Anyway, crazy. Crazy stuff that's going on with all this stuff, and um no doubt we're just scratching the surface in the time we have. Back to real workflows, and we're gonna finish. I I what I'm I've been talking about all show, so we're doing it. Damn it. [01:01:15] Dave Pengelley: Windsor. [01:01:26] Dave Pengelley: Um we talked about Jipity and Web Gipity quite a bit this. Today. And one of my big unlocks is I am trying to build a video game. And I do have my Hermes harness running Jipity to do the actual coding and execution. And we talked last week about the whole dark factory thing. And do you give things a giant big spec, or do you, you know, do small blocks and you iterate and iterate and iterate? And it's funny, again, Sid Meyer was saying, you've got to be testing and playtesting your game over and over and over again. Like this sort of structured little micro building is a way better way of building actually good finished products than this sort of dark factory loaded. In and seeing a few points of view on that. But anyway, I've been using Jipity as my voice sounding board as I've been out walking around. And on the weekend, I was like, oh, I've got all these ideas, all this information. I've had all these conversations. I'm making a lot of decisions. Now, this could be applied to any business thing that you're doing, not just a video game, any birthday planning, whatever it is. And I was like, oh, how do I get this out into some kind of structured place where I'm going to need it at some point? Is there some kind of game, narrative design database system? There is something open source. What are the costs if I want to pay for someone else's web service? Maybe it's an obsidian vault. Maybe it's with WikiLinks. Maybe it's mind mapping. And I was trying all these different things and trying to work out what it to do. And [01:02:42] Matt: Maybe [01:02:42] Dave Pengelley: then [01:02:43] Matt: the core. [01:02:44] Dave Pengelley: sorry, Matt. [01:02:45] Matt: Maybe the core. [01:02:46] Dave Pengelley: Yeah, maybe the core. Um, but then I stumbled across the fact that, you know, I was looking for something and in the just the mobile app. I was on my phone a lot because it was a weekend. I wasn't sitting at my desktop, and I saw the GPD plugins. And I I know they'd been talking about plugins upgrading from the old custom GPTs, they're going away, plugins is the new meta. And I saw all this sort of stuff and went, hang on, here, I'm not even in Codex. I'm in my GPT app that does not use any codex tokens. And they're telling me I can connect to services like Airtable and GitHub. And so I did. And then I was able to say to ChatGPT, hey, I've good just created a new Airtable blank workspace. Can you go and create a whole data model? And save all of our decisions into Airtable. And I went away and thought about it and connected the services, asked for some permissions, and then bang, created, you know, eight tables, entire data models with different columns, linking decisions about narrative or mechanics or game engines and all this kind of stuff. Extracted all the knowledge from our sort of days of chats and loaded it all into Airtable. No coding, all just using my mobile app. [01:03:56] Richard Webbe: That is brilliant. [01:03:58] Dave Pengelley: And then, and then the good thing is. And then I'm like, well, how do I pass this information now? I want to start building the prototype. How do I actually get this information over to the the thing? Because at the moment you've been giving me some prompts and I've been copying and pasting prompts into it to get it running. And it goes, well, if you use GitHub issues, I can create issues. And then when it commits, it can go and check the issues. It can get work as these sort of GitHub issues [01:04:21] Richard Webbe: Resolve [01:04:21] Dave Pengelley: and [01:04:22] Richard Webbe: the [01:04:22] Dave Pengelley: then [01:04:22] Richard Webbe: problems. [01:04:22] Dave Pengelley: do the work. And so now my planning database is not only integrated with decisions I've made, but then it goes and creates jobs for the worker. As I continue to iterate decisions, it can out go and refresh and update jobs that haven't been done yet with new information as I as I pivot things. And then when the worker goes and does the coding and puts it in GitHub, my mobile app can go and check the code and review it. And when I ask questions about implementation stuff, it goes, yeah, I can see that in the code. It's done that. Oh, that makes sense because the code defaults to this. So it's actually a code reviewer as well, but it's not a coding tool. It's the Chat GPT app on my mobile. And it's just been a huge phenomenal unlock. And All these other third party systems can plug into it, but this is kind of the model that I've been running with. I put this up on LinkedIn and Grokbot made this for me. Grokbot's been great for a bit of like on-brand content creation. Um, but I [01:05:12] Richard Webbe: When [01:05:12] Dave Pengelley: just thought [01:05:13] Richard Webbe: when are [01:05:13] Dave Pengelley: this [01:05:13] Richard Webbe: you gonna [01:05:13] Dave Pengelley: was [01:05:14] Richard Webbe: when are you gonna publish your results, Doug? That sounds cool. [01:05:17] Dave Pengelley: like games take a long time to make, mate. That's the problem. I'm working on it. I just run it run out of tokens. But that was my big win of the week, actually, leveraging, you know, the subscription half of my subscription, not just for the codec. But for the chat GPT stuff. And I've just this morning hit my limits on using 6 Pro um through the mobile app from doing all this sort of stuff. But huge unlock. And and whatever project you're working on, you can connect your chat GPT to some other back end, even if it's a Google sheet or something, and you can get it to start saving data. Which way back when 12 months ago, when I was still when I was first using this, I was trying to work out how to can get notes saved into Notion and I had to set up my own GPT and I was trying to send JSON packets to an N8N server. To then create pages and put stuff and bookmark things into an in the Notion. It was a nightmare. But twelve months on, it's all integrated and amazing. [01:06:08] Matt: So I've got a question. [01:06:10] Dave Pengelley: Yeah. [01:06:11] Matt: Um [01:06:14] Matt: it's not really a question, but [01:06:15] Dave Pengelley: Okay. [01:06:16] Matt: like [01:06:16] Dave Pengelley: Yeah. [01:06:16] Matt: this this unlock that you're feeling that you're feeling is is literally the reason why I built the core. Like is genuinely the reason. Like [01:06:25] Dave Pengelley: Yeah, [01:06:25] Matt: not to [01:06:25] Dave Pengelley: but [01:06:25] Matt: pitch like that, that's why it needed to exist. [01:06:28] Dave Pengelley: the core the core doesn't And give me like I get what you say. The core is the memory part and the memory and the storage part, but it's not, you know, the whole connected loop of interface of the mobile app and the talking and stuff, right? But you're saying your core [01:06:40] Matt: Yes. [01:06:40] Dave Pengelley: could plug in to the ChatGPT mobile app to replace the air table bit. I mean that's that's what core would replace. It wouldn't replace the GitHub and the Hermes and the Chat [01:06:48] Matt: It [01:06:48] Dave Pengelley: GPT [01:06:48] Matt: would [01:06:48] Dave Pengelley: mobile. It's just the air table piece, right? Primarily. [01:06:51] Matt: No, it's all of that. It's the operational layout. It's but yeah, that [01:06:54] Dave Pengelley: Yeah, [01:06:55] Matt: that [01:06:55] Dave Pengelley: but [01:06:55] Matt: that [01:06:55] Dave Pengelley: but [01:06:55] Matt: unlock [01:06:55] Dave Pengelley: it but but it'll burn tokens, right? Like if I plug Hermes into your core, then I'm burning tokens on. every single soundboarding conversation. [01:07:03] Matt: You can't speak in absolutes like that. Like the I'm just all I'm saying is the the unlock [01:07:08] Dave Pengelley: I love [01:07:08] Matt: that [01:07:08] Dave Pengelley: I love [01:07:08] Matt: you're [01:07:09] Dave Pengelley: your core, [01:07:09] Matt: feeling. [01:07:09] Dave Pengelley: Matt. I'm not trying to diss your core, but I'm also saying like in this instance, like part of the value for me is the Chat GPT mobile experience, which you've said is great for talking to and great for having that back and forth with. I wouldn't get that if I used any other model and had it plugged into the core. It might keep the conversation history. It might be able to pull memories and do those things, but I'm not going to get that whole Circle experience [01:07:31] Matt: You [01:07:31] Dave Pengelley: necessarily. [01:07:32] Matt: you can stop saying you're not. You don't actually know what it is. You don't know what a [01:07:36] Dave Pengelley: Okay. [01:07:37] Matt: building. [01:07:37] Dave Pengelley: All right. Fair. Fair. Yep. I I haven't experimented with it, but I also it's it's part of the, it's not the whole wheel, like because it's not a model. I'm not talking directly to the core as a model, right? You have you built an LLM? Is the core an LLM now? [01:07:52] Matt: No, no, no. I'm just saying that that that [01:07:53] Dave Pengelley: Yeah. [01:07:53] Matt: unlock, like if you if you connected your chat GPT to the core, you've got the same thing instantly. [01:08:03] Dave Pengelley: Okay, we'll take that offline. Because I'm confused. I I don't I'm trying to understand which part of the wheel or how much of the wheel is the core. Because the core's come a long way since last time I looked at it. [01:08:13] Richard Webbe: You know [01:08:13] Dave Pengelley: So [01:08:13] Richard Webbe: what I love about this conversation is the amount of tools that AI is introducing for multiple people who approach problems and solving problems in different ways, and we can combine them together. And the stronger MCP and APIs and all of that come together, the more that we'll be able to connect and make things happen in a more automated and efficient fashion. It really says a lot for how fast this is all going. [01:08:43] Matt: You just gotta ask the right questions. [01:08:45] Richard Webbe: That's [01:08:45] Dave Pengelley: Yeah. [01:08:45] Richard Webbe: it. It's the prompt. What is the problem I'm trying to solve? And [01:08:49] Dave Pengelley: And [01:08:49] Richard Webbe: it's been [01:08:49] Dave Pengelley: and [01:08:49] Richard Webbe: my [01:08:50] Dave Pengelley: what [01:08:50] Richard Webbe: favorite [01:08:50] Dave Pengelley: is the best [01:08:50] Richard Webbe: saying. [01:08:50] Dave Pengelley: way of solving it today, which might be different than it was yesterday, right? [01:08:54] Richard Webbe: Yeah, [01:08:54] Dave Pengelley: That that that's [01:08:54] Richard Webbe: exactly. [01:08:55] Dave Pengelley: a thing. [01:08:56] Richard Webbe: Exactly. [01:08:56] Dave Pengelley: Because it is all moving so fast and so Quickly and so um yes, maybe maybe I need to rip this apart and plug a core in. I don't know. I'll need to do some more homework with Matt on [01:09:08] Matt: Richard's [01:09:09] Dave Pengelley: that. [01:09:09] Matt: point there is that there's too many options. It doesn't matter. [01:09:12] Dave Pengelley: Yeah, [01:09:12] Matt: There's no there's no right [01:09:13] Dave Pengelley: no, [01:09:13] Matt: option [01:09:13] Dave Pengelley: but I'm [01:09:13] Matt: per [01:09:14] Dave Pengelley: curious [01:09:14] Matt: se. [01:09:14] Dave Pengelley: I'm curious to learn what value add I would get from using the core versus airtable, and I'm I'm happy with I'm already paying for the the Chad GBD mobile bit. So if I plug something else into the core, like anyway, there's a holistic you know thing where my brain hurts going, but this works, why would I change? It which is a good thing, actually, it's a good good point to end on for businesses. Sometimes if what you're doing works, you don't need to change it [01:09:37] Richard Webbe: Don't [01:09:37] Dave Pengelley: for another shiny [01:09:37] Richard Webbe: touch it. [01:09:38] Dave Pengelley: toy, [01:09:38] Matt: Yep. [01:09:39] Dave Pengelley: right? And that's that's worth considering. [01:09:44] Dave Pengelley: Yeah. [01:09:45] Richard Webbe: Wasn't [01:09:45] Dave Pengelley: Like [01:09:46] Richard Webbe: a good idea. Wasn't not a good idea. [01:09:48] Dave Pengelley: what what what what is the value add you get from change? If there's no value add and it's just changed for change's sake to feel productive, then it's probably a waste of time and money. Um, and so in this instance, maybe, and that's why I need to understand. Stand from Matt, like where the core fits, maybe there is a real value add and it will enhance this and make it better and add real value to me. Or maybe it's as a re-engineering exercise that you know I'm too far down the track and it is just expensive and hard to do. Um, but maybe good to know for the next project. So um keep that in mind as people want to throw all these solutions at you. Is it solving an actual problem that will drive value versus doing what you're doing? Because that's always the thing with AI. Everyone's to pitch you the AI solution. Solution, sometimes it's just more work and sometimes it's not. And that's where people like us will help have that conversation to help you work out which one's which. [01:10:39] Matt: Make the mistakes for you. [01:10:41] Dave Pengelley: Yeah, we've already made them. So all right. Thanks, gents. We are way over time. I'm gonna wrap it there. [01:10:46] Richard Webbe: Thanks, Doug. Doug [01:10:47] Dave Pengelley: Awesome [01:10:48] Richard Webbe: threw in a lot of [01:10:48] Dave Pengelley: chat. [01:10:48] Richard Webbe: great comments there. [01:10:49] Dave Pengelley: Yeah, thank thank you, Doug, for uh for chiming in for us. Um, love having you along uh for the ride, and thank you for the positive feedback about the show. That's good to See. Um, if you are joining us online on the replay, please make sure to hit the like icon, drop a comment, ask us questions. What do you think of Jev, the new models, and uh my win? Is it a win, or am I just behind the curve and you knew all this anyway? Let us know, and we'll see you next time. Thanks, boys. [01:11:21] Matt: So yeah.