Was This Built for a Previous Reality?

[00:00] Dave Pengelley: Most people are getting AI to do the wrong jobs, and then they're wondering, why am I not seeing value from my AI projects? And guess what? Sometimes the AI operators are not immune from that. So today we discussed some of our stories where we perhaps haven't been executing and using the AI in all the places we should have. [00:18] Dave Pengelley: Then how we pivoted and then started using the AI to do things that we probably should have. So join us for some real stories from the field around how we are learning from our own mistakes, and maybe you don't need to make those mistakes and you can get a head start. So join us as we ponder the question, was this made for another reality? [00:38] Richard Webbe: I like the beard though. [00:42] Dave Pengelley: That's a, that's a great, great segue, Richard. [00:44] Richard Webbe: He, he's looking very chiseled and trendy and neat and I've just noticed him grooming it over the months and I think he's doing a good job. [00:51] Dave Pengelley: I think it's a haircut too. Like he, he's looking, he looking, looking sharp on the sides. [00:55] Richard Webbe: He is. [00:56] Richard Webbe: There's [00:56] Matt Slager: been no haircut. [00:59] Marno Brits: He [00:01:00] washed it. [01:00] Matt Slager: I'm very, very due for a haircut. That was actually my second cost savings. Things I was telling you about. Dave. I'm gonna cut back on, uh, haircuts now. I've been cutting my hair since COVID. Um, but yeah, it, thank you, Richard. Honestly, it's like just. Um, it's like shaping a hedge. [01:15] Matt Slager: It grow, it grows back. You trim it back, it grows back. You trim it. [01:18] Dave Pengelley: I, I had a friend and, and, and multiple times over, over months and years knowing him, you'd go, have you, have you lost weight? He's like, no, just trim the beard. It's like, have you trimmed the beard? No, I just lost weight. It was always, always, you'd pick the wrong one. [01:29] Richard Webbe: Yeah. So I, my girlfriend, I was trimming. 'cause my hair grows, as you know, from my age. It's very thick. I have a lot of hair, hairdressers, fawn over my hair. And you've got so much hair. It's so thick. My grandfather had all his hair till he was 88. Uh, 98, sorry. Anyway, I was, I got one of those self trimmer things and I was, you know, and when it gets long, I think like, oh, I waste my time to go to that barber and give him 40, 50 bucks for a five minute haircut. [02:00] Richard Webbe: So I've been doing it myself. And then on this side, I cut it a little too short. And then I was starting to look like Kyle Sandyland, you [02:00] Richard Webbe: know, he's got a shaved on the side and he's got the hair going and, and my girlfriend said that, is it either I cut it or you go to the barber. So as you can see, I went to the barber, but the barber was laughing so much because there was so many short bits here and a long bit there and a bit sticking up there. [02:15] Dave Pengelley: We're using a suck cut. [02:17] Richard Webbe: Sorry, [02:18] Dave Pengelley: we're using a suck cut [02:20] Richard Webbe: from [02:20] Dave Pengelley: Waynes World. It's like, well it certainly does suck. [02:25] Matt Slager: Yeah. It's, it's a technique to learn to do it yourself. It's really easy and in the [02:29] Richard Webbe: mirror with the wrong hand, on the wrong side, you think you're going this way, but you're going now. It's very difficult. [02:35] Dave Pengelley: I, I, I take the, take the punt sometimes and try and, you know, do the back of the neck. And like you're doing, doing these and going, I hope I'm not like ruining my, my hairline. I'm trying to, [02:44] Matt Slager: yeah, [02:44] Dave Pengelley: yeah. [02:45] Matt Slager: Mine's like overdue right now. But yeah, I do the whole thing. I've got clippers, the, you know, different sizes and whatnot and I've got like specific areas that I do certain things and then when I get to the top, I literally take all the combs off the clipper and I like, like [00:03:00] rake my hair like that. [03:00] Matt Slager: And I, I like, I'll cut these pieces off. [03:02] Richard Webbe: Oh, you're getting adv advance. You're like just Mr. Advanced, you're like, you are with your ai you just writing on the advanced stuff. That's scary. [03:09] Matt Slager: My, my hair. I, I gonna get a robot that can do it for me now though. [03:13] Dave Pengelley: That's, I mean, be careful. I mean like, 'cause you're just doing it flat, which my, my hairdresser, she, she sort of holds the thing up and then she does like little angle cuts across it to try and layering. [03:23] Dave Pengelley: Yeah, [03:24] Richard Webbe: I, I ne I needed the short hair anyway 'cause I'm heading off for surfing in Indonesia and when you're in big waves in the middle of the ocean, you don't, the [03:31] Dave Pengelley: hair [03:31] Richard Webbe: and your hair falls down across your eyes and you can't see what you're doing. It's a pain. So, uh, yeah. How to get ready. [03:38] Matt Slager: Have you, have you, um, progressed in your, you mentioned last week, a little bit of stiffness. [03:44] Richard Webbe: Yes, I'm stiff. I've got Celine Dion disease, stiff body disease. Goodness. I had some. It's very sad, but it's like everything we do, when you see yourself doing something, you imagine you're just fantastic and you look like Kelly [00:04:00] Slater. And then I go on the Urban Surf video and I go, oh my God, I look like a retard trying to, you know, an idiot trying to, trying to get up and do that stuff. [04:09] Richard Webbe: So, uh, yeah. So I'm progressing. I'm doing very well. Thank you. [04:12] Matt Slager: You look like one of those, um, the Barbie candles on a surfboard. Just [04:17] Richard Webbe: Yes. [04:17] Matt Slager: Not much [04:18] Richard Webbe: movement. We're like really stiff like this Again, I think I'll do a reentry, but it's really slow and then I fall off. [04:23] Matt Slager: Oh, well you're doing more surfing than any of the rest of us. [04:27] Richard Webbe: Yeah, I probably shouldn't be. I'm just giving surfers a bad name. I'll, uh, uh, yeah. Anyway, so it's all good fun. But thank you for asking. How is it? Sorry. Go on mate. [04:37] Dave Pengelley: I was gonna say we're, we're at time. I mean, that's great to all this catch up, but people are here for a show, so we should probably, uh, do a, do a show. [04:43] Richard Webbe: Okay. Go sir. [04:56] Dave Pengelley: Good to have you all here for this week's show. Um, [00:05:00] there are lots happening in the news, in in ai, in, in the world. I mean, beyond surfing and stiffness. I mean, people, businesses are stiff, right? Let, let's, let's use that for a segue. Businesses are stiff in, in their adhesion to old ways and the inertia of change and movement and doing things, there's a lot going on out there. [05:16] Dave Pengelley: Um, and we're gonna talk about it today and how we're using it, how we're seeing our businesses using it. Customers, clients, everyone's coming at it from different angles, right? So that's what we're here to do. Um, we're gonna, we're gonna do some new segments this week, uh, and I'm gonna have to let the show know, uh, upfront for people that are watching, listening. [05:33] Dave Pengelley: I'm gonna be here for about half the show. Then I'm gonna kind of go remote and try and connect in from my mobile. 'cause I've gotta. I'm doing some volunteering for a not-for-profit to go do some consulting for them around AI and usage. So the way timing works and the size of Sydney, they wanted it at one 30 and I said, no, no, I've got a podcast at one o'clock. [06:00] Dave Pengelley: I can't do that. They said two 30. I was like, I can probably make two 30. And then when I looked at the timeframes of where it is, where I am, travel times, I was like, that's okay [06:00] Dave Pengelley: there. But then I, hi-his, here's where AI failed me completely. Actually it wasn't ai, it was just me. I thought it was school holidays this week, but it's not. [06:08] Dave Pengelley: So I thought my kids were gonna be home and it was cool. I can go out, I can come back. My kids are old enough. That's fine. Didn't realize this is the last week of school. And so I was like, ah, great. I'm not gonna be back in time. That's fine. They can walk home from school then find out. My daughter's got like this musical recital thing they do once a term at 6:00 PM and I'm like, oh my goodness. [06:25] Dave Pengelley: It's like an uh, like it might be an hour drive now for me to get to this place, but coming home in peak hour at four 30. It could be two plus hours. There's no way I'm gonna make it back. So I'm like, great. Now I've gotta flip to the train. So by catching, catching a train means I've gotta leave on being on a train at 1245 during the podcast even I'm though I moved it to two 30 so I could do my podcast because of family commitments that I didn't realize I was gonna have. [07:00] Dave Pengelley: I've gotta catch trains now 'cause train's more reliable than peak hour. And anyway, so I don't know how I, uh, will fix me with that. Um, I do get my daily briefings from Claude Cowork [07:00] Dave Pengelley: every day, but. To [07:03] Marno Brits: your calendar, they're like, Hey, this is everything that's going on. Plug it to everything. Well, [07:07] Dave Pengelley: I mean that, that we talk about context and, and giving all these tools the information they need, right? [07:12] Dave Pengelley: Like I had someone reach out on LinkedIn saying, Hey, I'm would love to connect, talk more about, um, ai and we're putting some Claude into our business, love to have a chat. I'm like, cool. Here's my calendar link. Go book it in. And they booked it for 5:00 PM on Friday, which a, in Australia, you just don't book things at 5:00 PM on a Friday. [07:28] Dave Pengelley: They're in Hamburg. They're, they're not even here. Like they're, they're overseas somewhere. So they're in a completely different time zones. I, I get it, but also I'm like [07:35] Matt Slager: texting in the morning, [07:36] Dave Pengelley: but I'm also like, why didn't my calendar automatically know it's a public holiday and block out good Friday? [07:41] Dave Pengelley: I'm not doing calls on Good Friday. So [07:43] Matt Slager: you've [07:43] Dave Pengelley: gotta start the [07:44] Matt Slager: calendar. [07:45] Dave Pengelley: So, yeah, so I mean, again, I didn't supply the context of the calendar. I thought cal.com automatically brought in public holidays. It doesn't, I thought maybe my own Google Calendar would automatically block out a public holiday. It didn't. [08:00] Dave Pengelley: So as magical as all [08:00] Dave Pengelley: these tools are as humans, we still need to be in the loop and take some ownership to set them up to get the results we want, right? Mm-hmm. [08:07] Richard Webbe: The example of anti AI as a good friend of mine, having done many startups with us or New York based, California based, um, companies, he went on stage to accept a sales award for, for the whole APAC region. [08:20] Richard Webbe: He was a sales director and he took a basketball with him and he got on stage in front of thousands of these people and he held up the basketball. He put an X on one side and X on the other, and Shun shone a light on one of the exes. [08:33] stinger vo: Mm-hmm. [08:34] Richard Webbe: And he said, now if the sun is shining in New York, do not call us or book a meeting 'cause you'll get nothing. [08:41] Richard Webbe: Right. 'cause, you know, training people that we are in an unusual time zone is a challenge and AI, I'm hoping is gonna solve all of those problems. [08:50] Dave Pengelley: I mean people are using voice AI and doing things, sorry, sorry Matt. I [08:54] Matt Slager: I was gonna say something architecturally it needs to be set up in the context management. [09:00] Matt Slager: So it's [09:00] Matt Slager: literally the engineering of the agents in the AI that you're setting up to force it to get the accurate date and time zone. [09:07] Richard Webbe: Exactly. It's like the early days of the internet where, you know, you're logging into a website and they want to know who you are and they asked you to put in a zip code, but they'd only accept a four, or sorry, a six letter zip code 'cause it was a US based written, you know, input. [09:20] Richard Webbe: We've got all of those same things with AI now, like, you know, LT a m and all the directories and, uh, that, that make everything in syntax work compatibility, they're all out the door. Again, until we get these plugins, which is probably my first poised question I had for today. If I'm not derailing us, how are we going in the AI industry with making everything from the orchestration, like open claw and all of that so that the plugins are consistent. [09:49] Richard Webbe: They know how to talk to each other. What's happening there? [09:51] Dave Pengelley: Well, before we do that, can I just do this? [09:57] stinger vo: Welcome to the AI update. [00:10:00] Let's look at what's happening in the news [10:06] Marno Brits: we've not expecting. So [10:09] Dave Pengelley: that felt like a good segue for a news figure around what's happening with development in, in Go. [10:13] Marno Brits: Go. I'll shut up. [10:16] Matt Slager: Uh, I mean, to your question, Richard, that I sort of, for me, I need to sort of understand what you mean by plugins, because if you just think of the way that applications communicate, you need to have that a application or API layer the programmable interface. [10:35] Richard Webbe: Yeah. [10:35] Matt Slager: Um, and people are trying to do things, you know, like MCP is a version of that. At the end of it, the raw root, all it's doing is call requests, essentially web fetch, you know, post get, put, delete, update, like that's all these things are and they're already standardized. [10:52] Dave Pengelley: Although Matt, there, there is a trend happening because, and I think we're gonna see this and, and we'll cover this in, in the next news article a little bit, but [00:11:00] a lot of this AI stuff is moving to the edge and we talk about sovereignty and owning your data, and we talk about local models and owning your models more and more compute is actually pushing out to the edges again, like where this, this, the, the whole pendulum thing, right? [11:11] Dave Pengelley: Everyone got sent, everything got centralized, and then it was like, um, server client, and then the clients became more powerful and we did more stuff on the client and on the phones. And then we moved back to some server stuff with all the AI models being in the cloud and server. And now we're pushing back to the edge. [11:23] Dave Pengelley: We've got this, this, this constantly back and forth of where does compute and knowledge and data live? Is it centralized or is it on the edges? And. Power grids, everything's the same, right? Um, but on that like curl request, that's when everything lives out in the server. But now as people wanna do more things locally, more and more services are producing the, the cli, the command line, um, interfaces and interpreters. [12:00] Dave Pengelley: So that way the locally running models and locally running data on the edge, rather than just like flick stuff out and re authenticating every time to an external thing, they're talking to a local little terminal interface, [12:00] Dave Pengelley: the CLI, which pre authenticates in. And so then they're communicating locally, locally to this local executable. [12:06] Dave Pengelley: It's managing all that external stuff, but it's making it easier for the agents to just sort of have one throat to choke and just talk directly to these CLI tools versus thousands of web requests. I mean, Matt, feel free to argue that point, but I think we're seeing more of that like cracking the trading platform brought out their own CLI this week. [12:23] Matt Slager: Really like you sort of need to just know the primitives of what, what the CLI is and what's actually happening under the scenes. 'cause it's the same thing. Um, like if you interact with, you know, from the terminal or from whatever like program to program on your device. Yeah. There's no web request, there's no h TTP there, of course. [12:41] Matt Slager: 'cause it's, [12:42] Dave Pengelley: yes, [12:42] Matt Slager: same. So at that point it's just a function call. [12:45] Dave Pengelley: It's just running bash commands in the terminal. [12:47] Matt Slager: Yeah. It's literally a function like, you know, uh, the simplest way to put it is it's just, just a function call. Um, so yeah, like the idea of standardizing kind of already is there, but the best question, and this is the way that I design [00:13:00] stuff, is, was this designed for a previous reality? [13:06] Matt Slager: You know, is there a better way now that removes friction and is moreally scalable? And that's where my knowledge core thing, [13:16] Richard Webbe: I like that statement. Remove friction. I had, I, I was playing with David and thank you for my introduction to building my own agents and doing all things like that. I was playing with some ideas during the week and every time I wanted it to go find some data for me, I think, oh, there's a publicly available amount of data, and it would come back and say, you've been blocked. [13:36] Richard Webbe: It doesn't allow you to make these calls into that space [13:39] Dave Pengelley: so that that's your cowork security settings protecting you. [13:43] Richard Webbe: Yeah. But the, the other database was telling me, no, you're not allowed to call with this program. [13:48] Dave Pengelley: Uh, that's the website stopping robots and crawlers from [13:51] Richard Webbe: Yeah, for example, LinkedIn or Pilot or one of those. [14:00] Richard Webbe: So it comes back to the whole scenario, which I used to say a while [14:00] Richard Webbe: ago, who owns the data? And your data's worth a lot of money. [14:03] Dave Pengelley: Yeah, so, so Matt, on your point around the, the interfaces and where they built for the current generation, and so we're seeing this evolution of more providers providing cli so that way you, that, that, that way that provider controls the local authentication on your desktop. [14:18] Dave Pengelley: You run their tool, they, they manage the authentication for you back to their services. So they're happy that that's more secure. And then your agent talks to their tool versus your tool trying to do all the authentication stuff. Um, the video you shared with me this morning around a JavaScript vulnerability actually talked about this in depth. [14:37] Dave Pengelley: There's a, a library that people have been using for some time, apparently not me, um, called Axios. [14:44] Richard Webbe: Cool. [14:45] Dave Pengelley: Uh, and in the video, and Axios has got this massive compromise where people, people have bad people do bad things, and they've found special ways to trick backdoor entries into sharing all your keys. [15:00] Dave Pengelley: And if, like, if you, if you, if you're a developer that's used Axios, [15:00] Dave Pengelley: do some. Check this out. But he made the point that Axios was there to fill a gap in the core technology, and now with Fed request and things in Node, it's actually redundant and you don't really need it, but it's still there. So it's an example of a technology that was built for previous generation [15:15] Marno Brits: mm-hmm. [15:16] Dave Pengelley: Architecture, right? [15:17] Matt Slager: Yeah. Yeah. It was literally for developer experience, you know, when, when we're all writing code, it was a way of abstracting complexity and fetch requests and HTP handling that just made it easier and simpler. So everyone used it and everyone has been using it. And in the past, you know, with any supply chain, uh, vulnerability, you know, the, the security advice has always been make sure you're using the most UpToDate stable releases of packages so that you, you get those security patches, you know? [15:51] Dave Pengelley: Mm-hmm. [15:51] Matt Slager: Um. This particular hack came from the most UpToDate versions, you know, like, so there's, there's two specific versions where [00:16:00] it actually went through and Yeah, you're right Dave. Like if you, it it, it's coming to the point that the most insecure command that you can run on your system is NPM install. [16:09] Matt Slager: 'cause if there is some sort of supply chain issue that hasn't previously been detected, like this one was nuts. Like you, you install it, it runs its things and then it completely clears its history so that you can't see that it existed, but it put its thing on your computer already. [16:24] Richard Webbe: Wow. [16:24] Matt Slager: Um, yeah. Intense. [16:27] Matt Slager: So [16:27] Richard Webbe: when I hear about, uh, uh, you know, hacking on scale, uh, I think about SQL injection. That hung around for a fair while before someone solved that as a problem, to be able to hack a website for those not technical like me and the audience, SQL injection was a way of using the internet to find databases that you could crack the password, and you ran the application code inside the web browser and it went in and hacked inside everyone's password with a plus or minus high low and worked out passwords based on the, the error code that came back from the server. [17:00] Richard Webbe: I remember a few [17:00] Richard Webbe: years ago, um, I, I got shown that and I type, and it came back with 44 million vulnerable servers, which was hilarious. Yeah, [17:07] Dave Pengelley: I mean that, [17:07] Marno Brits: I imagine there's still a amount out there. Like I worked at a company last week where they were using a custom PHP program and that custom PHP program had a, it was an info php, which shows you what the system details are and also where it's hosted, and sometimes even the username and password, what's going on. [17:26] Marno Brits: Like, there's so much where people just have not done the fundamentals. But a quick fun one to answer, Joanna. Also I think addressing Sure. Just saying good morning. But you were talking before about getting stuff on LinkedIn, Richard. And a cheeky way I found is you can't necessarily get it to use API request or HGTP, even if you get like a developer app that you have to go through the process of. [18:00] Marno Brits: But if you go into your browser, you press F 12, go into developer tools, you can get AI to write your little script that'll scrape that entire page. [18:00] Marno Brits: So what I've done is I've gone to like a couple people that are quite influential, like Alex and Mercy for example. Um, and he's got a bunch of posts on there. [18:08] Marno Brits: I'll look at the link for all these posts, I'll give that to ai. And then it developed a little scraper that you can do inside your console and it just scrapes all of it within a second. [18:18] Richard Webbe: And [18:18] Marno Brits: then I use that to now make a playbook based on all the chat that he's had rather than his books. It's now pulling from like these little nuggets of information. [18:26] Marno Brits: Then it made it into a, a play by play playbook [18:30] Richard Webbe: that, you know, I hadn't actually done that, but it occurred to me while I was doing it, based on the stuff David showed me, uh, earlier in the week, could I go, how do I hack LinkedIn? [18:42] Marno Brits: Yeah, it's, if you can talk to, like, like about, said, if you talk to like a person, if you can talk to people, you can talk to it. [19:00] Marno Brits: It's just how do you communicate and what instructions do you give? Like how many touch points. There's a great video about someone teaching their kids how to make peanut butter sandwiches and it's like, okay, write the instructions and then he goes to make it and it said, okay, [19:00] Marno Brits: take the peanut butter, put on the bread, and then he grabs the whole jar. [19:03] Marno Brits: Like he has to be specific with your instructions. [19:06] Richard Webbe: Yes. That's the old, that's the, I used to run that in, uh, executive thought leadership workshops, uh, and we'd blindfold someone and tell someone else to describe what they should do and they ended up doing pretty stupid things. And that's it. I mean, this clarity of communication is, is at the core of it all, isn't it? [19:23] Matt Slager: You know, like that sort of logic too, where you can just spin up your local agent in whatever systems you're using, start learning express a need, a desire or some sort of situation that you need help with. And you can not only like come away from that conversation with so much more learning yourself, so much more awareness of what you can do in those scenarios too. [20:00] Matt Slager: Like, one thing that I love doing now is, you know, if you have your agents on your system, they have access to your system, right? So if you think of it from like a, a security hacker perspective, 'cause we've spoken about that a little bit so [20:00] Matt Slager: far, that's about keeping people out of your system. And it's not only before stealing information, but it's for controlling your system too, right? [20:08] Matt Slager: Like you don't want them to be able to, you know, hack your webcam and you know, take pictures and stuff like that. So I had a scenario this morning where, where this particular camera wasn't working. And, you know, I was on a Zoom call and I had my laptop camera chosen and I couldn't choose the other one. [20:23] Matt Slager: And, and I was like, you know, flabbergasted. But I was like, let me just quickly open up, you know, my terminal agent and I'm gonna say, Hey, I've got this camera, I use it every day, what's going on? And I got it to go through, find the device, find all the device settings, all of that stuff that you know, you don't want anyone else to see. [21:00] Matt Slager: And I said, okay, cool. Run some diagnostics. Like, let's figure it out. It activated the camera. I saw the, the permission request pop up on my screen from Mac Os saying, Hey, something's trying to access. So I was like, yeah, allow. And then it took a, a, still it used FFM Peg to take a still [21:00] Matt Slager: from the camera. And I was like, okay, you know, just to prove that you actually did take that picture. [21:04] Matt Slager: What color shirt am I wearing? And I was like, well, it's a, it's a burgundy, maroon kind of shirt. No, sure. I'm like, okay, well clearly you got it to work, so the camera works fine, so what the hell zoom? Um, [21:15] Richard Webbe: ah. [21:16] Matt Slager: But yeah, like just another example of use the, use the tool because it knows more than you do in a way. [21:22] Matt Slager: And like [21:22] Matt Slager: I [21:22] Richard Webbe: got, I saw a very scary video of a guy turning up to an Airbnb place and not a very technical person like you guys, just a bit of a, a moron like me. And they sat down and they were worried about hidden cameras in this Airbnb, and he'd got into this habit. So he used AI to hack the router and find out what devices were connected. [21:46] Richard Webbe: And of course it showed all the devices on the router and a webcam. Yep. So he clicked on that and found. The camera was behind him on the light and he could see himself at his computer hacking the router. There [21:59] Matt Slager: you go. [00:22:00] Another [22:00] Richard Webbe: perfect [22:00] Matt Slager: example, [22:02] Richard Webbe: by the way, 1200 that do that when you go there and that hack the router using ai, it'll help you get in there, find the IP addresses, and then click on it and find out [22:13] Dave Pengelley: that's, um, I mean that, that, that's what I talked about last week where I, I salvaged the recording of a USB recorder. [22:19] Dave Pengelley: Yeah. Like the AI was able to help me image the file system and then use the right tools to interrogate it, find the unallocated file sitting in the, in the back end of the, the file table that just hadn't been registered as a file name, and then bring it to the surface like that. That is specialist expertise knowledge that humans can do, and humans could use the same tools to do, but it would probably take them a lot longer. [23:00] Dave Pengelley: It would take me an incredible long amount of time to work out what tools to use, where to point them how. Configure them what to look for, whereas the AI was preloaded with that pattern knowledge and able to, to just do that. On the LinkedIn thing though, the other hack is, um, [23:00] Dave Pengelley: if you're comfortable doing this and, and it's, I can't say it's a hundred percent safe, but you can give it a go, is it's got the, if you installed the Chrome, like using Claude Cowork, if you're using, um, the Chrome, the Claude for Chrome for Claude, Claude for Chrome plugin. [23:14] Dave Pengelley: Yeah. We [23:14] stinger vo: still that [23:14] Dave Pengelley: You can actually just say, you can just say Open Chrome and go look at these pages in LinkedIn using your Chrome browser. As long as you're logged in with Richard, it'll, it'll. So, but it's just built into the workflow. [23:26] Matt Slager: I did that with, um, a whole bunch of X articles. So I use the playwright CLI 'cause I just find it more token efficient and faster. [23:33] Matt Slager: But yeah, with playwright, CLI, you can get it to start a persistent session, which just basically means that every, every call it does, it's like having your browser open and then you can also get it to quickly open up headed mode, which brings up the browser for the screen log in, and then it will just shut headed mode again for you and keep going on that session. [24:00] Matt Slager: And I got it to go through all of my bookmarks on X like anyone who's a, an X user out there that just constantly saves to bookmark but [24:00] Matt Slager: never does anything with them. Yeah, that I, I've put all those inside my knowledge core now so that I can reference them and, and use them. [24:08] Richard Webbe: Mano, you look like you had something to say there, sir. [24:11] Marno Brits: No, just, it's so much fun, man. I, I, I keep thinking about how much we're able to achieve and although we're using simple tools. I, especially the last few days, going back to first principles and also not taking for granted what we have. I'm saying that to say that tool that Matt has built for himself, that would be an industry that would be a massive business. [25:00] Marno Brits: Like why are we skipping over these opportunities? Because we take it for granted. 'cause it's so easy. Yeah. Like there's a couple applications from the last week that I didn't build anything unique. It took me five minutes, but it made my life easier. And if all we're trying to do is to help people live better lives using ai, I feel like I'm taking [25:00] Marno Brits: this mindset at least towards myself, that I'm doing a disservice by not creating that product for them, for them to use. [25:07] Marno Brits: Even if it's a free app or if it's just, hey, install guide. But there's so many people that I'm meeting that are so stressed and they're working 60 hour weeks and with Little to show for it. Because they just dunno how to use the right tools. [25:21] Richard Webbe: Yeah, [25:21] Marno Brits: yeah, yeah. I'm constantly playing in our mindset of like, there's so many things that we could be building or packaging and then shipping and sure, open AI might do an update tonight that makes it all redundant. [25:33] Marno Brits: But we could at least be trying to like build up this catalog of toolkits that people could be using. [25:39] Richard Webbe: I find watching you guys like watching a locksmith, you ever get locked outta your house and you go, oh my God, I can't have the keys. I don't get in. You're thinking of smashing a window or kicking the door down. [25:49] Richard Webbe: The kids might be inside or the dog's inside, you can't get in. And then the locksmith turns up and goes done and that's, and I bought a watching. You guys like that. [25:59] Dave Pengelley: I bought [00:26:00] a lock picking set off Amazon or something just to, just to play around with it and practice it. It is insanely difficult, even with the see through lock and you can see what's happened with the pins and you're like, this is how do, like, it's, it's actually really, really tricky to pick a lock if [26:15] Matt Slager: you, I have, I've got five test locks here, or four Oh yeah. [26:18] Matt Slager: Um, that I can literally pull apart and reap in and, and they come with like a little key. Yeah. Nice. But yeah, I've got those like sets too, because that was just another random, like a DD um, hobby, like mini obsession. [26:34] Dave Pengelley: A million years ago I bought a set of bump keys that was sort of precut and like, I think most of 'em for American style locks, none of 'em actually like fit their sort of like Lockwood locks here in Australia and stuff. [26:44] Dave Pengelley: So it was a complete waste of money. But I know with this little key ring of like bump keys that I've never, ever used, I think it's, 'cause you know, this classic story you get locked out of somewhere once that you really should be able to get into and you're like, I'm never gonna get locked outta this place again. [26:55] Dave Pengelley: And so. Do these things. [26:57] Matt Slager: No, I, with, when that, with the lock picking stuff, [00:27:00] I, I cracked every single lock that I ever like, came across except for this one that is for the power meter. Like it was in one of our old rentals. Um, and it's, I guess to like lock the power meter shut or something, but it's for the electrical services to use. [27:15] Dave Pengelley: Yeah. [27:15] Matt Slager: So the key in that and like the pins and whatever in that was, I don't know, just on the concept of hard locks, that was like the most intense one. [27:23] Dave Pengelley: We, we, we just, we just finished watching the last Mission Impossible film where they've got the, the cruciform key where it's like these two interlocking cross keys to make this more special cross lock to unlock the AI or the submarine thing or whatever. [27:34] Dave Pengelley: It might [27:35] Richard Webbe: bomb outta the sub. [27:37] Dave Pengelley: Yes. Anyway, uh, let's, let's do this short segment and then we'll jump into our last segment and then I'm gonna sort of step away and let you guys discuss and I'll try and jump in maybe audio only, but, uh. [27:57] Dave Pengelley: Love AI wins of, I'm [27:58] Richard Webbe: Loving the, [00:28:00] [28:00] Dave Pengelley: the fir, the first one I said make wins of the week for a podcast called AI Operators. And it said, wins of the week, AI ought ride, like the, the second word was gibberish. I'm like, drop the AI operators word, make it more celebratory. And I was like, wins of the week, the week. [28:14] Dave Pengelley: I'm like, I'm just running with that. That's, that's fine. Thank you. Google VO for your generation. Um, so I, I am part-time am a treasurer at my church for volunteer, uh, roles and we had our a GM on the weekend and I faced the issue of pulling data out of zero and reconciling files and trying to map our sort of budget draft and a spreadsheet to this and bring all the totals in. [29:00] Dave Pengelley: And inevitably all this back and forth, um, you forget to carry something over from one spreadsheet to the other, or you've summarized something there and one number just doesn't fit and someone calls you out on it. When you sort of. Discussing the numbers. Um, and, and it's no, not a big deal, but it's just like, come on, you work so hard trying to make sure all this thing fixes and [29:00] Dave Pengelley: someone spots that one little typo in your numbers and you're like, come on. [29:03] Dave Pengelley: You just feel like a, like an idiot. So I came home from that and went, that's it. I'm building a tool now. So I've um, spent the week building a church budget finance operating dashboard that integrates to zero. So I can pull the data in and I'm gonna get all my live reports. I'm gonna get better charting and reporting than I've ever had before. [29:21] Dave Pengelley: And we're gonna be able to draft our new budgets, seeing the actual numbers from the previous year and, and review all of that and do what if analysis. And it's just incredible. These tools are there and if only I'd used them 48 hours earlier would've really helped me with my a GM report. But you know that that's how we learn. [29:37] Dave Pengelley: You trial by fire and you go, that didn't work well. What is a better way of doing that? Uh, and. [29:43] Richard Webbe: I, I read it. David liked that. I read a great article from a very experienced lady who was a CFO for, uh, a, a, a regional business. Uh, and she had an understudy coming under her from about 24 years old, and the board's sitting there, okay, give us the news, the [00:30:00] reports. [30:00] Richard Webbe: And it was, uh, totally true. And I, I can reference that. So she'd worked for a week on her, all her reports and bits and pieces and stuff like that. And she had the, the paper document ready to flip and go through for everyone. And the young understudy, uh, obviously turned up to hijack her, sadly, uh, turned up with an interactive display [30:19] Dave Pengelley: Hmm. [30:19] Richard Webbe: All written in ai, which he'd done, not a programmer, but leveraged on some people like us. That gave him the impetus and the knowledge to do it. [30:28] Dave Pengelley: Mm-hmm. [30:28] Richard Webbe: And, uh, within five minutes, he'd communicated everything they wanted to know. Live in. And, uh, she was sitting there with a two week prepared paper document going and they said, yeah, you've put that on the table. [30:39] Richard Webbe: We'll look at it next week. And of course that means we will never look at it. Mm-hmm. And she was talking about how someone knew in the game who, who who'd worked with people like us to get across the AI had just g uped her. And she left the room thinking, I think my job's over because I've been doing this for ages. [31:00] Richard Webbe: And he just turned up, he hasn't been [31:00] Richard Webbe: here long, he's 24 and he's gone bang, bang, bang. And they've got everything they needed. Yeah. And, and this, this acceleration of personal efficiency, uh, married with key subject matter expert objectives required. Is where things are really moving fast and happening and, uh, yeah. [31:19] Richard Webbe: So, uh, like with your church or any, that's happening everywhere. [31:22] Dave Pengelley: But in my example, like I'm in this space, I'm building these solutions. I have the tools and equipment, but you just have it. I, we've all, well, we've always done the budget this way. We've got our Google sheet and we copy these numbers and we paste that there. [31:33] Dave Pengelley: That's, that's how we do it. Until you realize the way we're doing it is not working and it's not good enough. A lot of, a lot of people don't realize that there's, there's something else they should be doing instead. [31:43] Richard Webbe: Yeah. And, and I, I tried to explain to a friend of mine, um, who runs a very successful business. [32:00] Richard Webbe: He's been doing it for 16 years, um, really well. And I said, Hey mate, just step over the fence. Come over here and look at the AI thing. Yeah. And he is a fabulous human being. [32:00] Richard Webbe: One of the best process driven, articulate workers I've ever seen. And he loves to get this from his accounting system. And then he translates it here, and then he puts it in there and it goes into his MYOB and it goes off to the accountant. [32:12] Richard Webbe: And I said to him, you don't have to do that. But he's not comfortable. He, he just will not step outta that because to his mind, he needs to check all those steps as they go forward. A bit like a person who likes to drive a car with a manual versus an automatic, I mean, the computers in automatics now will change the gears in whatever mode you want much better than a human could. [32:32] Dave Pengelley: Hmm. Once, once it's established and, you know, it's reliable. The issue with AI is that non-deterministic issue where it's not always A equals B. Um, yeah. But anyway, that was, that was, that was my, that that's a half win. Like it was a, it was a, uh, that I'm moving in that direction now. I'm fighting zero authentication and trying to work out how do all this, that sort of stuff. [32:48] Dave Pengelley: So the app's nearly ready, but it's, it's looking, looking great. Um, any other wins for me? We've, I, we've discussed this in the first half, but any, any key things where you're gone. [32:59] Marno Brits: [00:33:00] But I'd like to just add a few things there and also read some of the comments. Um, said, I'm wondering if Mano knows how to scale the roof and find his way from ceiling. [33:09] Marno Brits: Um, maybe a couple times. And then Joanna's one. Yeah, he has everything except the wisdom and the experience. I think that's a comment to the, both that you're talking about, Richard. And I'd say this is something that's probably getting a bit like, it's, it's getting on my nerves. Like it's annoying me where there's people that are so qualified and so great at what they do and they see someone else that's younger than them come in or just someone that has a different approach to it. [33:34] Marno Brits: Come in with AI and say, oh, my job's over, what am I gonna do? Like Alex Moey had this guy on the podcast or whatever talks he did, and they do websites and the guy's like, well, I don't know what to do anymore because everyone can make a website now and it's super, super easy and it costs nothing. And then I'm worried on business is gonna die. [34:00] Marno Brits: And Alex Mosey basically turns around and says, well. There's this new great tool called ai, have you [34:00] Marno Brits: heard of it? And he's so scared of it. He is like, oh, it's gonna replace me. It's like no, just tool. And then benefiting from it being easy and like it, it's your capped that your experience, it amplifies your existing skill. [34:14] Marno Brits: If you more, then you're gonna be able to produce more. If I try and do a presentation on sales versus Richard, doing implementation on sales would look vastly different because Richard knows more promptly, he understands the language. If I have to do one about ai IT or ai, I can do that. 'cause I know all the acronyms, but it's scales, your knowledge. [34:35] Marno Brits: And I think that's just a pain point where people need to stop giving themselves the quick out of like, oh no, I'm gonna be replaced. It's like if a kid can learn it, then you can learn it. [34:46] Richard Webbe: It's, you're, you're exactly right. And this particular CFO, which was for a multinational organization, reset her perspective. [35:00] Richard Webbe: And it is subject matter experts. Mm-hmm. With the right tools and the right resources and data [35:00] Richard Webbe: available. And it's like I say, if you know the question, you've got access to the data [35:04] Marno Brits: mm-hmm. [35:05] Richard Webbe: And you've got the skill to use the tools, then you're gonna lead the world. Right. So no one's gonna be made redundant and everyone will just be more efficient. [35:12] Richard Webbe: And I laughed at, um, a friend of mine recently who was, you know, well, I've got a lot of friends. And they go, oh, it's great. When I go to put in my proposal for my clients, I cut and paste all the information about who they are. I cut and paste all the information about my products on the web, and I merge 'em together and I send a proposal in. [35:29] Richard Webbe: That's a good way to lose, because all you're doing, as we always say on this podcast, all you're doing is grabbing the sum, net average [35:37] Dave Pengelley: mm-hmm. [35:37] Richard Webbe: Of information available in the public marketplace. Right. And as we know, Dr. Google is not always right. As a subject matter expert, you know what to look for and you know the unique questions that will isolate and competitively excel you into a zone that no one can get to. [36:00] Richard Webbe: And uh, yeah, if you get lazy and think, oh, this is easy, I'll just cut pastes, [36:00] Richard Webbe: LLMs off I go, you're gonna lose always, every time. Yeah. [36:02] Marno Brits: It's, it's the, it's always been humans versus humans are better tools. Like we're, we're just, as we, as the tools evolve, we evolve, and it's probably not the best analogy, but like farming. [36:14] Marno Brits: Like you could give a farmer a tractor and then he's gonna build a crop and he's just gonna do very love it, lavish things. Or you can give me a tractor and I'm gonna try and go through a wall so I can like, or like do some reconstruction. 'cause if you don't know how to use the tool set. Then it's pointless. [36:32] Marno Brits: Like there's different use of different ways of using the tool, but, uh, if it's purpose built, and, um, another analogy is probably like when you are going to someone's house and they say they, they want a drill, they actually know you just want a hole in the wall, don't you? So then what's the best tool to get you there? [36:47] Marno Brits: Maybe it's a drill, maybe it's a hammer. Um, what are we trying to accomplish? You know, and getting really clear on the outcome. [36:53] Richard Webbe: And that, and that's the best metaphor, right? If all I've got is a hammer, I'll see every problem as a nail. If you ask me [00:37:00] to build the back fence and I'm a plumber, I'm gonna use pipes. [37:03] Richard Webbe: That'll look normal too, when I probably should use, that's [37:05] Marno Brits: beautifully put. That's going in my little diary, man. That's [37:08] Richard Webbe: great. I've got a book coming out soon. Uh, anyway, so metaphors to Utes. So exactly as you're saying, mano and David was alluding to, and Matt was describing earlier, that your knowledge now as a subject matter expert, or an inquiring mind as I like to refer to it, is much more valuable with AI than it was in the past. [38:00] Richard Webbe: Mm-hmm. Excel did not make us all accountants. Okay. Excel made us a bit more efficient, no matter who we were, but the accountant still knows the rules. And, uh, uh, um, in a recent litigation matter, a friend of mine said, oh, I'm gonna do this myself. I'm not gonna use the lawyer. Right. I said, okay. So they go into, you know, using all the, the information they could on an LLM to draw, draw a, and it even wrote it, wrote the request letter, it did the brief for the court, it did everything, and they used it all, and it [38:00] Richard Webbe: achieved nothing. [38:01] Richard Webbe: Because the legal ability to leverage and negotiate was not there. Just pure blank regurgitating legal, uh, rhetoric. A good example to a lawyer. They won't. Sorry. [38:15] Matt Slager: A good example of that sort of thing is when you see some sort of politician or like, you know, it could be the prime Minister or whatever, um, come up and they, they, they start saying a speech, you know, and they're talking away and they're saying a lot of words, but nothing is being said. [38:29] Richard Webbe: Yes. [38:30] Matt Slager: It it, it's the kind same sort of like soul soulless information that you can be resultant with if you don't put your own spin on it by your own taste and your own experiences. [38:40] Richard Webbe: Yeah. That's where the, the competitive advantage. My father who was an inventor, uh, an Asperger, a pretty interesting guy and ran an ad agency for 50 years and did all these things, he said, Richard, the person that invented the hair clip for women did not make money and was not successful, but the person who invented the hair clip for [00:39:00] women. [39:00] Richard Webbe: Then put those little kinks in it so that it stayed in the hair. That person became very successful. And I think about that when I think about ai. So information is now becoming democratized on a mass scale that is, getting information is not hard now and it's gonna be even hard, easier, and easier. And I say that information will become a free, totally democratized available resource in the next few years for everyone. [39:27] Matt Slager: Yeah, [39:28] Richard Webbe: right. Just a complete like, I want to know. Yes, there's the answer. What do you do with it? And without the skill of understanding, uh, um, based, yeah. Yeah, exactly. Who said what are the other weekly wins? We [39:42] Marno Brits: stopped after, after Dave's. We stopped. [39:44] Richard Webbe: We did, we should get back to the weekly wins. Any more weekly wins. [39:47] Richard Webbe: Come on Matt, you've got something for us. I'll put my glasses on. So I read the chat. [39:51] Matt Slager: It, it's very, very hard to summarize my wins this week because there's been [39:55] Richard Webbe: a lot, [39:56] Matt Slager: um. But I'll just, I'll, I'll say one small sort of [00:40:00] story for you, Joe. This one might be interesting and also thank you for, uh, tuning into the idea of copy and pasting. [40:07] Matt Slager: 'cause as soon as I heard that as well, I was like, hmm. Um, so yeah, my win this week is actually along the same lines. So I, I'm not sure if you remember last week I mentioned I had a couple of really big, um, client meetings coming up and, um, the, the one for the local IT service provider, the managed service provider, um, that was intense. [40:31] Matt Slager: Like when I really think about it, it was really intense for somebody who's never been involved really in any kind of, um, executive corporate communication and sitting in a boardroom with the, you know, the big dudes like looking down at you. Um, yeah, that, it was, it was huge. So like for me, getting in there and. [41:00] Matt Slager: Uh, getting out of that meeting successfully, and I think doing a fairly decent job and like talking through them and meeting these guys, um, it [41:00] Matt Slager: was, it was really cool. It was really exciting and it was fun. And they actually said a lot, a lot of really heavy hitting and impact impactful statements during that meeting where one of the guys literally said in relation to, you know, me and, and others in this space, you know, like all of you listening as well. [41:20] Matt Slager: Um, he said, we don't have someone like you, and if we don't find someone like you, we won't be around in the next five years. [41:30] Richard Webbe: Mm. That is very true. [41:32] Matt Slager: It was huge. And when they said that, I was like, whoa. Okay. Fair enough. Um, so yeah, [41:39] Richard Webbe: as long as they gave you a blank check after they said that, Matt. Of course. [41:42] Matt Slager: Yeah. Well, well, I, I, I basically recorded the whole meeting with their permission of course. And. Um, went home and started just digging into the transcript, you know, finding out exactly what it is that they were asking for, what they needed, what their first steps would be, what that would look [00:42:00] like for me to do it for them. [42:01] Matt Slager: You know, I put out a couple of different pathways with different stacks, different methodology, but in order for me to generate that proposal, put that together, you know, one of their things that they needed was reporting and regular structured reporting. And like you guys were saying before, the non-deterministic nature of the output that you get from an AI system means you need to have certain things in place to make it more deterministic, probabilistically better, you know, at each time. [42:32] Richard Webbe: Yeah. [42:32] Matt Slager: So in the process of creating this proposal, I created the system that would create their reports for my own proposal. [42:40] Richard Webbe: Because it's a look, it's a very close parallel, right? Again, it comes back to what is the question? And the best question I got given in my career very early, and I used to as a young, you know, shoot my mouth off Phil, when I was very young as an engineer, I got a lot of those meetings. [43:00] Richard Webbe: Matt, I know exactly how you feel and you know it's not your world and it's their business. And they, [43:00] Richard Webbe: and the best one is whenever I used to get lost, I just stop and say, what is the problem we are trying to solve? And I think when we asked that question more and more, uh, an example was I worked at IBM for many years and someone came out with a report and said that IBM produced over 40,000 individual reports that no one ever read anymore. [43:24] Richard Webbe: Yeah, right. And we are heading into, I think that territory. With ai, right? So someone says, I need reports. And you say to them, well, what's the problem you're trying to solve? Oh, I need to know when the price gets over this. Or a customer feels like that. Okay, let's have it so you don't have to read those reports. [44:00] Richard Webbe: 'cause that's just a waste of time. How about if my tool takes you from there to there? Right. I'll give you, I'll give you continuance so you can check the data as it flows past. But why keep reading the reports? Why not just have it come out here and then your experience, [44:00] Richard Webbe: your subject matter experience will go, now I know what to do with that situation. [44:04] Richard Webbe: Yeah. And that's what we're doing. And that's why I say information's gonna be democratized. It's gonna be open to everyone from every corner of the planet. [44:10] Matt Slager: A hundred percent. Like for these guys, the way that I see their reports and like what they wanted, um, is kind of similar to if you are in a, in a shopping mall and you purchase something and you have that option on the, on the, let's be real, the self scan checkouts where you, you can say, yes, I want a receipt, or No, I don't want a receipt. [44:29] Matt Slager: And you know, the, the, the, the green or like, you know, the save the paper version is the, no, I don't want it. But then you are now walking through the shopping center with an item that you have no proof of purchase for necessarily. So. [44:44] Richard Webbe: Yes, [44:44] Matt Slager: the, um, the reports that these guys are sending out are to each of their clients that they manage their hardware, their software, their systems, their cloud solutions. [45:00] Matt Slager: Like all of the things for these individual businesses to run. These guys are sending reports of like, you know, [45:00] Matt Slager: here's all the information that we have of the stuff we've done for you and the statuses over the last week. So it's kind of like a receipt. So there, I agree that the, the point of the, the delivery and the reading of the reports is kind of a little bit erroneous, but the actual data that exists there is the useful thing. [45:19] Matt Slager: And it's more so that these companies understand that these guys are doing something for them and there's evidence of that. So, yeah, [45:30] Richard Webbe: no, no, you're, you're exactly right. So the, the example of that is where are you adding value once you know the question. Once you know the problem you're trying to solve, everyone's gonna be a little more exposed in where are you adding value right now? [46:00] Richard Webbe: Here's the data, what do I do with it? What's the next decision with your experience? And it comes down to, I think, the human element. People understanding how humans react or need and what their needs are. You know, you can, you can personalize that, right? I was having a conversation with a [46:00] Richard Webbe: a, you know, a friend recently and we're gonna go overseas on holidays and I was talking about the fundamentals of the sort of footwear you should wear in the environment we're going to, it's not a pair of thongs 'cause that doesn't work. [46:11] Richard Webbe: And you're walking on coral and you're walking on some rocks under and, and thongs come off in the work. So there's a particular, and they said, oh great. Let's go look around the shops. I said, no, let's get on the weapon. Look it up, but I wasn't solving the problem they wanted to solve. Right. The problem they wanted to solve is actually wanted to go to a shopping center and walk around. [46:30] Richard Webbe: That was their objective. Mm-hmm. It wasn't the shoes at all, but my objective was, well how do I, you know, like a bloke, how do we get straight to getting that thing? I mean, yeah, stuff, that shopping experience, but the shopping experience to some people is an aphrodisiac. They love it. It makes 'em feel good. [47:00] Richard Webbe: It pumps their serotonin and their dopamine. So I wasn't solving the right problem. And while what I use is a personal extreme version of what's happening in business and in supply with humans, those [47:00] Richard Webbe: things are gonna become a lot more important. You know, do you go to the movie theater or just sit in your own home and do your streaming? [47:07] Richard Webbe: Right. [47:07] Marno Brits: Yeah. [47:08] Richard Webbe: It's not the same thing. It's, it's not the same problem solved. Even though the movie's the same [47:14] Marno Brits: different experience, [47:15] Richard Webbe: the [47:15] Dave Pengelley: humans. I, uh, I watched it and Alex, can you guys hear me? Sorry, [47:21] Richard Webbe: we can David. [47:22] Dave Pengelley: Hello. Excellent. [47:24] Richard Webbe: Yes. [47:24] Dave Pengelley: I watched to Alex video this morning talking about ai and he was saying like, and all these things, you're not replacing a person, you're replacing a workflow. [47:33] Dave Pengelley: And so we think of these things architecturally, like, I need a marketing agent, but what does the marketer do? What are their actual breakdown of their tasks? And within those tasks, what are the subtasks and the workflows? And then being able to establish the ins and the outs and sort of assign these agents and things to solve workflow problems, not entire role problems, uh, because they're much better when you get more specific. [48:00] Dave Pengelley: And so back to what you were saying around the problem, what problem do you need to solve? It's not, we need to buy shoes. [48:00] Dave Pengelley: It's like we need to have social time. We need to do that. And so the friend, friend bucket, if you just make an agent friend that's not necessarily gonna solve the right problems because the agent friend is too broad. [48:09] Dave Pengelley: What are the, the sub workflows that that friend agent really needs to be doing? And he, he's sort of likened it almost less back to a, um, an org chart and more to a manufacturing line. [48:20] Richard Webbe: Yeah. Oh look, Joanna just, uh, popped in the chat and made a very good point. What does, uh, evidence and value mean customers like to be kept in the loop, which is what Joanna just said, right? [48:30] Richard Webbe: So what does being kept in the loop mean? It means they believe what is happening. They have evidence and a crumb trail to follow and say, I agree with that. I, I, I do a, uh, fun card trick, which I'll do for you guys one day. Uh, and it is a series of things that I do. And then I get the person to tell me what the card is on the top of the deck, right? [49:00] Richard Webbe: Mm-hmm. Now, if I start that card trick and just say, Hey, um, you know, say this name and I lift up the card, look, you said [49:00] Richard Webbe: the name on the top, top, it has no value at all. But if we go through the card trick process of questions and answers and we walk around this, then they see, ah, I understand how we got here and why we got there. [49:12] Richard Webbe: People dunno why they got there. It's not worth it. The journey, as Joanne's pointed out. Is as important as the answer. And if you can't justify that journey with value and evidence, right? Humans don't like it. [49:23] Marno Brits: Mm. [49:25] Matt Slager: A good, uh, example of that is, um, notifications. So in a lot of the systems that I build, especially for some of my, my bigger long-term clients, I have specific Slack channels that are dedicated for giving system updates. [50:00] Matt Slager: You know, things that are happening. So one guy in particular, he does, uh, a regular webinar that's just like the main funnel loop that he does. And I have these specific notifications come through saying, webinars started, you know, webinars finishing. You're putting together the, the participant [50:00] Matt Slager: lists, okay, the list is loaded. [50:01] Matt Slager: Here's your stats from today's webinar. And like those things happen just on auto autopilot. And, um, it's evidence of a system that I've built for him, and he gets to see it every single time it happens. So, like, you [50:13] Richard Webbe: know. [50:14] Matt Slager: That is very much being kept in the loop of the system that's actively running and working. [50:20] Matt Slager: And um, another one just really quick is the same company. I have set up a different channel for actual development updates. So again, for them I'm using linear just as like a task management, project management thing. And every time a new issue is created or a status has changed, I that, that goes into that same Slack channel so he can actually see me working and, you know, see all the updates live. [50:43] Matt Slager: Plus when I've actually merged a PR back to Maine and we've got new updates on the app, I've got another message that goes, you know, hey, we've got a new version, here's the new changes. Um, and it's cool 'cause like I don't have to get on the phone. I don't have to book a call. I don't have to do any of that stuff to make sure that they know that I'm there doing work. [51:00] Matt Slager: Um, [51:00] Matt Slager: they just see it happening live. [51:01] Richard Webbe: Yeah, an answer without evidence or, uh, proof of how we got there is not really a very good answer, particularly with clients and us and customers because if we get an outrageous answer or a really big one. We go, why? How did we get here? And then we have to lay out that whole, if it's Angen ai, we've gotta lay out exactly how we got this so people understand. [51:24] Richard Webbe: Uh, um, and I, I, I honestly think it's gonna have a weird impact on humans decision processes as we go into the front, because there's certain things we trust and there's certain things we don't trust. [51:37] Matt Slager: Yeah, a hundred percent. There's a quote from one of my favorite, um, agent engineers that I follow and I get a lot of my information and inspiration from. [51:46] Matt Slager: He specifically says, do not defer your thinking to your agents. [51:55] Richard Webbe: Of course, defer the process, not the thinking. Defer the data access, not the [00:52:00] thinking. So true. [52:01] Matt Slager: Yeah. Yeah, it's a really, really interesting point. 'cause you start just getting lulled into a false sense of reality when you start just never thinking for yourself. [52:08] Richard Webbe: Well, we all know the conspiracy theorists who get on the web and look, Google something and think that's the answer, and enough they go down a rabbit hole. [52:14] Matt Slager: Yeah. [52:14] Richard Webbe: God save us from AI doing that for us. [52:17] Matt Slager: Yeah. Another good one. Back in the day and another life I used to work at Bunnings and my favorite situation was somebody coming up to you with their phone in their hand, like shoving it in your face saying I Googled this and it says Bunnings has it. [52:31] Matt Slager: And you know, it wasn't even saying anything, it was just Google giving a sponsored result saying Bunnings. And yeah, that was, that was [52:39] Richard Webbe: my, my doctor has in Google font on a sign behind him, um, my degree. A little bit smarter than your five minutes Googling. [52:48] Matt Slager: Yeah, yeah. Literally [52:52] Richard Webbe: six years. All this study, cutting up bodies, learning things, and you're Googling. [53:00] Richard Webbe: Yeah. Oh, well, there we go. [53:00] Richard Webbe: And are you still with us, Dave? [53:01] Dave Pengelley: I, I am. I am indeed. Um, I think that the, the third thinking is interesting because you do, especially once you start building some agents and you've got some good system prompts, and you can get a bit lazy with your prompting because you've set up the tools and the systems to just work for you. [53:17] Dave Pengelley: Um mm-hmm. I've, I've certainly fallen into this habit where I just, and, and with Open Chlor, I can't remember if I was talking about this on this call last week or another call. I was, I was on, but saying, I tried setting up open core and it hasn't really done much for me because I kind of expected it to be super magical and everyone talked about how great open core was, so I was pretty vague with it and just expected it to set itself up and do things. [53:38] Dave Pengelley: And it doesn't, it needs more work and more effort than that. And so it's a reminder that. These things aren't magical black boxes. You do need to be more specific. Even that, um, in the news thing where I said it came back with rubbish results from my video, I was really slack. I gave it a really short prompt that wasn't highly descriptive. [54:00] Dave Pengelley: I didn't get the AI to write the prompt that is full [54:00] Dave Pengelley: of all the detail about cinematic angles and do this and use this kinda lighting, which gives you a much better output. But I'm just, you know, sometimes you're in a hurry and you just get lazy and then it's a good reminder that these things are highly fallible. [54:13] Dave Pengelley: Um, I think we're out time to really dive deep into it, but maybe we'll talk about about it more next week after everyone's had a bit of a chance to read through it. But Matt Barry, the, um, CEO of freelancer.com, Australian entrepreneur, businessman, he does very thoughtful, very long essays. And his latest one is around AI and token burn and the sustainability of the subscription models, like, like the core plans and the open AI plans. [54:39] Dave Pengelley: And he does talk about the fact that, you know. Compared to a human where you see a problem and you analyze it, you assess it, you don't try 50 things that are all stepping over itself and retrying the same thing you tried three turns ago. 'cause you've forgotten that you tried it three turns ago, just burning tokens. [55:00] Dave Pengelley: But that's what these AI do. You hit. And if you doing any kind of coding, anyone who's used these things will notice [55:00] Dave Pengelley: sometimes it makes mistakes. And when you try to get it to debug itself, it goes down all these rabbit holes that are just completely irrelevant and are just burning tokens for the sake of burning tokens. [55:10] Dave Pengelley: Uh, you've really gotta go, hang on, stop. No, and that, that's, that's where we talk about the engineering side of engineering versus vibe coding. Yeah. And having those humans and those experts back in the loop. Right. [55:23] Richard Webbe: Oh, look, and, and it's so true. Like, I, I, I think of the clown at the, uh, you know, at the show with its Ted going back and forth and everyone putting the ping pong balls in hoping they're gonna win. [55:33] Richard Webbe: Uh, they're not. Um, and if you think about, um, again, getting back to AI being the subnet average. Every time I look at social media, there's another revelation, snake oil sales person like, and it says, Hey, I've got the key prompts for everyone to be able to win the stock market. And it is just like the stock market. [56:00] Richard Webbe: It's gotta be the stupidest comment ever. Whenever our friends come to me and go, oh, gotta get on this stock, [56:00] Richard Webbe: it's gonna go really, really well. By the time they're telling me it's too late, the stock's gone up and I'm pretty sure that it's gonna crash As people pull out, you know, if you're gonna run with the pack, you're gonna get no value. [56:11] Richard Webbe: If you're gonna use your noggin and your AI architectural engineering and not rely on someone else and get lazy, you are gonna do really well. If you're gonna get lazy, then you'll just be another me too. Good luck with that. You'll hover in just below the sum net average of earnings and success in your life, and that's not what we're about. [56:30] Marno Brits: I'd love to get Blake Sims on the podcast. He's been instrumental. Um, I know he's influenced Stephanie. And then as a byproduct, Matt told me about it and I've just been obsessed with his teachings. Um, the way that he approaches it is very analytical, but his reasoning is sound. And it's very much the 80 20 rule of if I do 80% of my time in planning, I don't have errors. [57:00] Marno Brits: I don't have this random thing where it just gets lost for half an hour and wipes, [57:00] Marno Brits: tokens. Um, yes, it's not fun work, but it's the work that it's doing and the work, the work works more on you than it works on it. So just putting in that 80% to do the planning, like I spent probably about two and a half hours just working off it to, to develop the um, PRD and then I gave it to Claude, and the output is significantly better. [57:20] Marno Brits: But it's taking the time. It's not, yeah, not delegating the thinking in the actual, um, I said, I think you said it best before the subject matter expert. Like, I know what I want, I know what I need, I know what frameworks to use. I just needed to put into a word or put into a, a document that can easily describe and structure the play by play. [57:42] Marno Brits: Almost like a project manager for the ai. [57:46] Richard Webbe: Yeah. [57:47] Dave Pengelley: Yeah. I took a, i, I took a leap outta Matt's book with the linear stuff. So I've been using linear ad hoc a little bit. I started using it when I was building something else and wanted just to go through and do user testing and catalog all the bugs that needed fixing. [57:59] Dave Pengelley: But, uh, since then I've evolved it and now across, you know, six different projects that I'm running that got their own, like, linear codes before. But when I wanted to do my last dev project, I did what you said. I was Claude and, and, and, and brought out a whole full spec, PRD. Then I went into Claude Code and said, Hey, review the PRD. [58:21] Dave Pengelley: Then let's develop the build plan. And it broke it up into milestones with issues and waves for development into linear. And so then it used linear as its checkpoint and then it worked through the linear tasks. And when it finished a milestone, it goes, cool, all right, we've done this. I'd do some testing. [59:00] Dave Pengelley: I'm happy with milestone one. All right, set up the task list for for milestone two, and then it would go and build out all the issues ready to run through. And so it was using that as that sort of system of record of what to do next and building out all the issues. It's been been phenomenal. And then because that's tied back in with all my other cowork and stuff, cowork, when I've got my sort of daily rhythms and my pulses going on, it reviews and tells me where all my projects are. [59:01] Richard Webbe: Yeah. David, can I just change tack just for a bit? We've only got a couple of minutes left. Um, something you texted all us this morning, uh, which you spotted in the media, which I missed, and it's about the rise and fall of AI similar to telco and software and cloud and everything else. You know, you know, the, the, what did we talk about? [59:19] Richard Webbe: The trough of excitement of adoption, the trough of disillusionment, and then we settled with a reality somewhere and, um, kinda like, uh, video streaming. And when I first got my first streaming TV stations, I was wrapped. And then of course I find I'm streaming 30 different TV stations and it's costing me a fortune and the quality of my TV has not gone up. [59:39] Richard Webbe: Um, so the, the interesting one is with AI and AI subscriptions we're, we're getting so many of them, and as you said, they're asking us to use more and more token, and it's becoming a new sinkhole for money. [59:52] Matt Slager: Mm-hmm. [59:52] Richard Webbe: What, what does everyone think about that? [59:56] Matt Slager: That is a big topic and [59:58] Richard Webbe: just throw it in. We'll [01:00:00] come back to it next week. [01:00:00] Richard Webbe: It's only one minute to go. [01:00:02] Matt Slager: There's actually one line from that thing that I think is worth mentioning. It's the, it's the one that I, um, made reference to in that message that Dave sent us. And it was in the summary of, uh, of old mates full talk thing. And basically it said, you know, the closing line of this author, he basically was summarizing that AI is game changing, but the current crop of companies funding it won't be the ones who profit from it. [01:00:35] Richard Webbe: Yeah, good comment. [01:00:37] Matt Slager: It's the ones that, you know, build systems around it and that sort of thing. Mm-hmm. The, the whole token use thing. You know, if, if the, if we are fully usage based on API costs and those API costs rise and fall, we have no control over that. There's no unit economics there to like scale off. [01:01:00] Matt Slager: But for the ones that are using subscription costs, like what you just said, Richard, the subscription costs can be insanely [01:01:00] Matt Slager: exploited right now. Yeah. I use that word. [01:01:02] Richard Webbe: I think, I think Kat just mentioned absolute tokens are going to be the new commodity or the new monetary. Yeah, yeah, [01:01:10] Matt Slager: yeah. Like the, the, the max plans that you get from these different companies, they lose money on those plans like they a hundred percent do. [01:01:18] Matt Slager: And if you can commercialize your value you get from that subscription, it's in, you've got multipliers, better margins than if you did it directly through API costs. [01:01:28] Richard Webbe: Yeah, actually I saw a video, um, over the weekend and the. It had a, a particular role, a processing role that was quite clear and it had, you know, this much money we'd play an employee to do that. [01:01:42] Richard Webbe: And then it spoke about an LLM doing it and how many tokens it used to do that role. And, uh, the cost of the human doing it was like $150,000 a year. And the cost of the LLM and the tokens doing it was $335 for the year. [01:01:57] Matt Slager: Yeah. [01:01:57] Richard Webbe: So the revolution's here, but we've gotta be [01:02:00] careful. We don't, like with cloud, we all end up with build shock and we flip over the edge into some of these crazy subscriptions. [01:02:06] Richard Webbe: And like Kat said, someone with a smart brain, maybe me, is going to do a new financial transaction engine like Bitcoin. But it relates to AI tokens. [01:02:18] Dave Pengelley: I think the thing is just, and, and yeah, we can dive, Matt wrote a lot of like a 10,000 word essay. It's massive. Uh, and we can come back to it maybe in more depth next week. [01:02:26] Dave Pengelley: But one of the things he made, the point is these subscriptions aren't sustainable. And so there will be a point where we get the sticker shock where they're gonna start charging us per token, API pricing. But, and you know, he's, he's, he's called it, um, uh, like insert coins and pre PRAI like, he just like that, that luck. [01:03:00] Dave Pengelley: Luck of the draw. Will you get the result you need? Or are you gonna be spitting the wheel five times to get the right output? You want like the video generation and so on. But at the end of the day though, like you said, like if you've got a fully loaded cost of an employee, even if you do have to burn some tokens, is there a [01:03:00] Dave Pengelley: relative benefit still to ai? [01:03:02] Dave Pengelley: We've been spoiled by going, yeah, these things are 20 bucks. They're a hundred bucks for a month. I can do unlimited things. But even if, you know, it becomes $5,000 a month, is that still gonna be cheaper than, you know, having five people on an engineering. Yes, [01:03:14] Richard Webbe: absolutely. And so yeah, the benefit, it's [01:03:16] Dave Pengelley: gonna be that cost benefit analysis. [01:03:19] Richard Webbe: And we'd love, we'd love to work with some of our, well, we'd like to work with some of our customers on a real ROI, because guess what, we're all gonna start with very basic processes. Some of the things we talk about here are very cerebral and the high end of, of thinking and planning and changing for competitive advantage. [01:03:35] Richard Webbe: But the general, let's reduce the cost of my business. Uh, ROI against any of these tokens is still thousands of percent. [01:03:43] Marno Brits: Mm-hmm. [01:03:45] Matt Slager: A hundred percent. That actually wraps into, I don't know if you guys wanna wrap it here, but there's a, like, I talk about habits a lot and I've got a little, like stinger quote here. [01:04:00] Matt Slager: Um, a bit of a spicy line, like stop, stop designing AI systems. Just in calling it that, [01:04:00] Matt Slager: you need to start designing business habits. AI is working. [01:04:05] Richard Webbe: Yeah. New ways of working. I love it. [01:04:07] Matt Slager: Yeah, yeah, yeah. [01:04:09] Richard Webbe: Okay. I like that a lot, man. Let's do it. [01:04:11] Dave Pengelley: Yeah. It's, it's, you're solving a problem. AI is a tool, and as we know, a lot of the workflow stuff isn't even AI based. [01:04:18] Dave Pengelley: How many, I think someone did a study of all the NAN, like thousands and thousands of NAN instances, and, you know, 90% of NAN workflows had no AI in them. They're just deterministic. If this, then that, if, if, if nodes had a much higher, uh, usage rate, JavaScript code nodes had a much higher usage rate than any AI nodes in NAN. [01:04:38] Dave Pengelley: So [01:04:38] Richard Webbe: yeah. [01:04:39] Dave Pengelley: And understand the workflows. Maybe AI has a place, maybe it doesn't, but it comes back to let's solve the business problem and go from there. [01:04:49] Richard Webbe: Yeah. Yeah, yeah, exactly. Well, do you wanna dance this out, sir? Leader Dave? [01:04:57] Dave Pengelley: Yes. Um, yes. Uh, sorry, I'm just [01:05:00] dodging train notifications that are coming in through my speakers hall. [01:05:03] Dave Pengelley: That's a weird angle. There we go. Hello. This is, this is me on a train. Hello. Um, all right, well, uh, thank you everyone. Apologies for my, uh, hit and missing, but I am working with a not-for-profit, so it's for a good cause that you've had to deal with weird travel. Dave. Um, you, you're good, man. Thank you guys. [01:05:18] Dave Pengelley: Good, good, good chats. I hope, uh, audience got some value from it. Thank you, uh, for all in the live chat, Sean, Catherine, Joe, for joining us and, and chatting and anyone else who joined us and it was too shy to chat. You can jump in next time, but, uh, thank you gentlemen. We'll wrap it there and we will see you all next week. [01:05:35] Richard Webbe: Thank you. Thanks. Good luck. See you guys. See you, everyone. [01:05:40] Marno Brits: Yeah.