AI Operators
Contact
All episodes

Vibe Coding Removes the Typing. Not the Engineering.

ROI benchmarks66 min29 Apr 2026

608% ROI. 4.2 days down to 6.8 hours. One small manufacturing team. Richard brings the case study that answers the question most operators are afraid to ask.

DP
Dave Pengelley
RW
Richard Webbe
MS
Matt Slager
MB
Marno Brits
YouTube
Show notes

608% ROI. 4.2 days down to 6.8 hours. One small manufacturing team. No groundbreaking technology. Richard brings the case study that answers the question most operators are afraid to ask.

The show opens in unexpected territory — AI in churches, schools, and one particularly impressive letterbox-drop iPad app built by an assistant minister. The tools are already working in places nobody's writing think-pieces about yet.

There's also a harder thread: AI is slowly reshaping the language you use and the frameworks you reach for. Matt calls it cognitive security. Worth knowing which frog you are before the water gets warm.

Dave closes on the monitoring story from last week — Sentry caught a live Stripe misconfiguration inside 24 hours. Matt ends the episode with the line that reframes vibe coding in one breath.

Transcript
[00:00] Speaker: What does Max Walker Taylor Swift and a boiled frog got to do with ai? Well, if you wanna know, you've gotta watch today's show. It was great to have Richard back in the chat this week. Where we looked at some of the news and the headlines releases from Open AI and from Claude, as well as the general state of what is happening and how we're as usual getting our winds of the week. [00:18] Speaker: How are we automating our lives? How are we automating our customers and how are we actually using these tools hands-on to get better outcomes? If you wanna know more, make sure to watch the show like and subscribe and drop a comment so the algorithm knows to share this great content out with your network. [00:35] Dave Pengelley: Richard, you were just saying before the show then that you broke a rib while you were away. Um, for the last few weeks. Yeah. Wanna tell us about that? Breaking a rib, [00:44] Richard Webbe: not fully diagnosed, but yes, I was surfing in, uh, in a, we resort in Indonesia and uh, I fell on my board and, uh, they never know if it's a broken rib or a bro cartilage or whatever, but has the same symptoms and that is, you can't take a deep breath. [01:00] Richard Webbe: You can't sneeze or [01:00] Richard Webbe: cough and rolling over in bed is very horrible. But I'm still alive. I'm still here, sir. [01:05] Dave Pengelley: Not fully diagnosed, just like my A DHD. There we go. [01:08] Matt Slager: I was gonna put my hand up then too, [01:11] Richard Webbe: but Mine's fully diagnosed not about you guys. [01:14] Matt Slager: Awesome. Well, how was, how was last week? I feel like I, I was able to listen to a good half of it or so. [01:20] Dave Pengelley: Yeah, you were definitely like, um, you know, like, um, Waldorf and, and Statler throwing in stuff like from the, from the cheap seats, uh, on the comments. But, uh, it was good. I mean, it was great having Amber and Manjula, um, Manjula didn't wanna talk over people, so we've gotta get her on again. So I wish you can pry out of her shell a bit because I think she has more value to add. [01:38] Matt Slager: Very [01:39] Dave Pengelley: much. [01:39] Matt Slager: There's a lot there. Yeah. [01:41] Dave Pengelley: Um, yeah, so building her up for teachers is awesome. Smart syllabus assistant if you didn't hear about that last week. So that's pretty awesome what she's doing. And we'll continue to bring in different super friends, uh, as we go along and, and have different people on, on different weeks as the, as the roster shifts up and people have, have life things, um, like breaking ribs or cartilage in Indonesia, [00:02:00] [02:01] Matt Slager: I found, um, what, what Manjula was talking about with how the education system, like they don't let you use any of the, the big tools they have like, like their special version of it. [02:11] Dave Pengelley: Yeah, yeah. [02:13] Matt Slager: Forget what it was called. It was like, um, their own like unique chat GBT thing. [02:17] Dave Pengelley: Well, it depends, depends on which part of the education system as well. Right. And we do talk about data security and sovereignty and all that kind of stuff. Like they're worried about teachers putting kids' names in another random crap, I'm sure. [02:28] Dave Pengelley: So they're trying to lock that down. Um, but like, I know like in the Ang of Sydney Anglican education system, they've got their own tool, which understands the specifically more the primary curriculum, um, for helping build out lesson plans and doing all that sort of stuff within their frameworks and what they're teaching. [02:45] Dave Pengelley: So they've got their own systems that they're deployed for that. But I dunno about the state systems and other, other, you know, um, private school entities. What they're doing [02:56] Matt Slager: makes me think about, um, like religious [00:03:00] organizations as well. Like I, I've never really followed any particular. Denomination, but I have many in my family and I've, it's been exposed to a lot of different sources. [03:10] Matt Slager: So the, the, the Jehovah's Witness, like organization, like my, my dad's fully into, um, the JW side of things. They have all of their own jw.org systems and, and applications and all of the various source of information, but I'm pretty sure none of them yet actually have like, like an AI layer in it. And I think it would be really, really useful for, um. [03:36] Matt Slager: Just linking concepts like, you know how if you're speaking to somebody that's really high in like faith and they, they can give you, that reminds me of so-and-so passage from [03:45] Dave Pengelley: Yeah. [03:45] Matt Slager: This book in this chapter. And, you know, there's this story as well that will probably help you. And [03:50] Dave Pengelley: to be fair, the, the, the main models, like the Bible is like the most published book ever. [03:55] Dave Pengelley: Like [03:55] Matt Slager: yeah, it's [03:55] Dave Pengelley: probably, I dunno about, I dunno about like the, the, the J dubs and their specific texts and things, [00:04:00] but, um, just the, the frontier models are pretty good. If you want to say, Hey, help me investigate this passage. I'm gonna run a study on this. I wanna pull these themes out. And it goes, yep, no worries. [04:09] Dave Pengelley: Here, here. And, and you can cross reference off to this one. Like, it actually has a lot of that in the training data and you just, you know, give it an agent prompt that focuses it on, on your guardrails of what you're thinking. Um, like at our church, my, my, my minister, he does, he uses a little bit for some of that prep. [04:26] Dave Pengelley: Um, for studies and stuff, especially pulling the questions and things together. Uh, it's great for research. Obviously it's not writing the finished sermon and then you always double check and, and make sure you, uh, agree with the, the references. I know the Catholic pope came out and said, you can't use AI for anything, don't use it at all, because that's from computers and not from the human heart. [04:47] Dave Pengelley: And I'm like, I can, I can agree with [04:48] Matt Slager: that. I can see the him on his phone just speaking a yeah. [04:53] Dave Pengelley: Um, you got [04:55] Richard Webbe: broken rib. Don't make me laugh.[00:05:00] [05:00] Dave Pengelley: Don't get AI to write you a prayer because then it's not a valid prayer if you say it. Yeah. I'm like, um, that's incorrect. Like if you believe, if that's the words that are like, you pray that like that's, that's valid. Who cares if. AI wrote it. I don't know who wrote every word in the, uh, prayer book 500 years ago, but it doesn't make it a less valid prayer if I'm [05:19] Richard Webbe: Yeah, exactly. [05:20] Dave Pengelley: Wanna say those words to God. Right? [05:22] Richard Webbe: It it is, it is, uh, for anything, religion, learning, knowledge, explanation, consulting, support, love, caring, family, none of it is valid to use AI as a boundary to say it's not valid. And we see that with new technology all the time, right? It comes along, people say, this is really bad, and they're only saying it 'cause they're scared of it, right? [05:43] Richard Webbe: Mm-hmm. At the end of the day, everything comes from the human DNA, everything we use. And that eventually comes from the human DNA. And it can be a spiritual location, it can be a identified patent location, it can be an intelligence location, but it's all bloody human, right? [05:58] Dave Pengelley: Yeah. And, and, and I mean, [00:06:00] beyond, you know, using this for those tools, all the administration and stuff like the. [06:04] Dave Pengelley: Church organizations like any other not-for-profit organization needs to do. I mentioned I'm, I'm building a, an accounting management thing to help me do the annual budgets and reports for the, um, monthly media to, to track actuals versus budgets and, and help us plan all that better than using a Google sheet. [06:22] Dave Pengelley: Um, the one on one of the guys, uh, the, the assistant ministers at the church for doing letterbox dropping to make sure, you know, we don't double up pamphlets when we're announcing the Easter services and saying, Hey, it's Easter. Believe it or not, we have church services at church for Easter. Come, come join us. [06:37] Dave Pengelley: Uh, he normally, they print out maps of the suburb and you go over and you grab your little stack and you highlight the streets you're gonna do, he built a thing using anti-gravity, I think, where he goes in with Google Maps and you tap on the, the streets. That you want. And it tells you how many bundles of pamphlets you're gonna need based on the number of houses in the streets you've selected. [07:00] Dave Pengelley: Then you put your name on it and it locks them out. And so it's now this dynamic iPad app [07:00] Dave Pengelley: for helping manage the um, the letterbox drops. It's awesome. [07:03] Matt Slager: That's so cool. That's really cool. Actually. I think we should have more like little showcase demo, things like that on this show to be like, Lou, this amazing thing that I just found some, I [07:13] Marno Brits: think that'd be fun. [07:13] Marno Brits: Like showcases are so, like it's all, it removes the barrier to entry because they can see a real practical use case. Instead of us just saying acronyms the whole time, we can be like, oh, well I built this, like showing two people yesterday how to use Codex. And instead of saying, Hey, this is our amazing, this is amazing app. [07:31] Marno Brits: I'll be like, Hey, on your computer you have Chate, download this. It's part of your subscription. There's nothing else that's being paid for. It's just on your pc. And then look what it's capable of doing. And then you connect it to everything. Set up some automations and then mind blown like they made a game for like their kid in like five minutes. [07:47] Matt Slager: Yeah. [07:48] Marno Brits: There's so many opportunities. [07:50] Richard Webbe: Yeah, I bought a, I bought a couple of use cases, business ones today they just mentioned. So I thought to fill in that gap 'cause it's really important at the moment. And I liked what Amber and, and you guys were saying [00:08:00] at last week's podcast about, you know, it's not a technology problem, it's a business problem. [08:04] Richard Webbe: He's, you sure I wrote that, uh, article last night, uh, after getting off the plane with jet lag 'cause I was bored and um, and I think it's really valid. [08:13] Dave Pengelley: Alright, well let's, let's, let's do a show about AI and business then. Let's do this. [08:29] Dave Pengelley: So we are doing a show about AI and business. We are AI operators. It's back to the, the, the, the regular roster for those that are joining us, that are regulars in our audience. And if you are joining us, jump in the chat. Say hello. We'd love to hear from you. Uh, gentlemen. Um, I'm still working out with the new, where we do the banter and then we do the show logo, and then we do a bit more bantering and then we do news, or do I just jump straight into the news or, or headlines like, uh, the whole show flow we're working on. [09:00] Dave Pengelley: I'm, I'm working with my AI partners to help me with the show [09:00] Dave Pengelley: flow. 'cause uh, they, they have the, they, they have the full like transcript of all 13 episodes so far. So they can see, they can see where we're going. They can see how we're doing it. They can help us bring some of the news in. So we are trying to self apply the technologies from that we talk about into even producing the show. [09:18] Dave Pengelley: I mean, a few weeks ago, Natalia, when I was complaining about the content production pipelines, that we could automate that for you, Dave. I'm like. W we should do more of that. Um, this isn't a win of the week. This is a story about automating stuff. Um, for my chart reporter app that I talk about a few a little bit, my journaling app for FX traders that I've launched, I've started as of in the last 24 hours doing live streams for that as well because I just, I love live streams. [10:00] Dave Pengelley: I just changed the color of the lights behind me and away we go showing people how the app works in real life. So actually placing trades, seeing the trades appear. And so I was sort of a, a live show and tell, walk and talk, doing some trading, like, you know, chatting with the audience. And so I did [10:00] Dave Pengelley: that for half an hour last night until like 1:00 AM and pulled the transcript out and dropped it in my thing. [10:05] Dave Pengelley: And, and I've worked with my team of agents so they know, Hey, Dave does the AO operators show, but he also does this chart reporter show as well. That's gonna have a different spin when we, when we analyze his transcripts and do all that. So I said, okay, cool. Here's what you should retitle that, that, that live stream with, here's a new subject, here's all the new descriptions and everything. [10:22] Dave Pengelley: And I was like. Can you just do this for me? Like I don't want to be copying and pasting this stuff into YouTube? Yeah. Uh, I said I've got N eight n let's let, that's already linked into my YouTube account 'cause it pulls my analytics and things. Can we just do that? And because N eight n in their latest MCP release can push and write workflows directly into N eight n more easily than ever before. [11:00] Dave Pengelley: So right there in my, in my IDE codec, not the Claude, whatever it was, I can't remember, went and wrote me a new little module, stuck it up there in N eight N and then could take all my detailed transcript analysis and just push it up and update my YouTube [11:00] Dave Pengelley: records for me without any of the copy and pasting. [11:02] Dave Pengelley: And so now you spend a little bit of time now that's gonna pay dividends and, and as I do, like dozens or more live streams for, for chart report. And even as I start doing some of the work for this, now I can help it with the AI operators. Pushing changes up, which is just, [11:16] Matt Slager: that's where, that's where it starts. [11:18] Matt Slager: It's, it's like you, you notice a friction, you notice something that you don't want to do. Yeah. Um, you know, some sort of process that is repeated or, I don't know, has a little bit of, in like intelligence, like a little bit of a decision enrichment, whatever. Like how can we make this semi-automated? Yeah. [11:35] Dave Pengelley: You [11:35] Matt Slager: know, like what components need to exist for this to actually start to come together. You know, especially like in different industries, you don't want to have the whole thing automated. You still want to have that like, human in the loop and whatnot. But yeah, that's how it starts, dude. [11:49] Dave Pengelley: Yeah, I mean, I've been, I've been watching a few, um, of, of Nate on um, U YouTube's videos. [12:00] Dave Pengelley: Nate Herk, he's does some good content and he's been talking about using hyper frames for video editing and putting all the [12:00] Dave Pengelley: animations and stuff. So more the talking head stuff versus editing a live stream. But in there there is video cut and edit skills that he is been using. So I was like, I do use the script and I still like the script and there's a place for descrip in the workflow, but I can't get my AI to easily send just like cut here, cut here, commands in a descrip, their API makes you talk to their AI to then make the edits. [12:21] Dave Pengelley: And I'm like, I don't wanna do that leapfrogging. So the next step is I pull these downloads down, um, and then I'll get my AI. To find the best clips and then use those video skills to actually clip those out and put some animations and transitions on automatically. So yeah, we will get there. Like today, I was like, oh, I didn't clip anything from last week's episode. [12:44] Dave Pengelley: We go live in an hour and a half. Gimme one clip I can put on LinkedIn and he goes, okay, grab this, the soundbite from Amber, use that. And I was like, okay, I can quickly cut one thing and push that out. That was a [12:54] Richard Webbe: good one. I like that one. [12:56] Dave Pengelley: Um, [12:56] Matt Slager: I think I was talking to Mano about this previously. Um, [00:13:00] I, I've sort of hit this weird turning point with when it comes to stuff that is generated, controlled or like decided upon by ai, so the, the quote unquote slop, you know, um, where I think I'm, I'm almost getting to the point now where I'm, I'm tolerating AI generated stuff a lot more as if it's just the norm. [13:21] Matt Slager: You know? So Dave, like if you just went nuts and you kind of just let it go and create a bunch of clips and all these clips just started to get updated, uploaded, and they weren't necessarily like. You know, upper echelon, um, [13:34] Dave Pengelley: yeah, [13:35] Matt Slager: really curated clips. I, I almost would find the novelty in them being kind of shit that they're fun. [13:42] Dave Pengelley: I think, I think there's two sides to that. There, there is like if, like, and, and we're gonna play our, our, uh, winds of the week bumper in a second and so on, which are imperfect AI generations. Um, but I suppose the difference is when we're talking about clipping this podcast, if it just gets half sentences and half ideas that make no sense. [14:00] Dave Pengelley: 'cause it thinks, oh, [14:00] Dave Pengelley: that sounds really good. And it gets like this weird fragment of our conversation that you just push up and it's like, that's crap. Why would I watch that? Those people are idiots. They make no sense at all. I'm like, may, maybe people watch us because we're idiots and we make no sense at all. [14:12] Dave Pengelley: Maybe that's half the, half the, uh, the gig. Um, maybe just, I should experiment with that for a week and just say, go nuts. Clip post. Eat your heart out. Um, it can't hurt as far as like the vending [14:25] Matt Slager: machine bench, you know, because like they, they have to earn, they keep, you know, [14:30] Dave Pengelley: doing things. Get them use. I mean, at the moment with what we're doing, we're not growing particularly fast because I'm not doing enough. [14:36] Dave Pengelley: Maybe just doing more of something, anything is the answer. As long as we don't burn people, I don't know, we, we will, we will try that. [14:41] Marno Brits: Yeah. I feel like I'm curious on that. 'cause I feel like there's just value in the consistency of showing up rather than Okay. Constantly having to edit, like where you making the content. [15:00] Marno Brits: So some grace for yourself there. But I'm curious, and I support Matt's idea. There was like, yeah, just embrace the AI slop in that. Like, you don't wanna put out terrible content, but if it cuts off half [15:00] Marno Brits: sentences, then. People might be curious like, what's the rest of the sentence Now? Maybe bloody find the podcast and go watch it. [15:05] Richard Webbe: You just described, you just described YouTube reels, right? [15:09] Marno Brits: Yeah, exactly. [15:10] Richard Webbe: They show you something and you're going, Hey, that's really interesting. It stops you go, what the bloody [15:15] Dave Pengelley: Well that, that, that's where we'd be posting these too. That'd be going straight onto on shorts and reels. I would [15:19] Marno Brits: be curious. [15:20] Marno Brits: Yeah, I mean, I mean, you go first, Dave. [15:22] Dave Pengelley: I was gonna say just, just from a point of view of that, that, you know, just not overthink it with the chart reporter stuff it, and analyze my transcript and I did not review it. I just went, you know what, just push the update. Just, just, just give me a description. [15:33] Dave Pengelley: Don't just leave my generic Dave's live streaming tonight as the subject. Just, just populate that so I'll appear in the algorithm more. Just, just do something. So think LL like what does a, what did Lord Aaron Bert say? Smile more, talk less. I just need to think, think less and push more. [15:48] Marno Brits: There's a great one on the topic of that. [15:50] Marno Brits: Um, I'm not sure if any of you, any of you follow Modern Wisdom by Chris Williamson. [15:54] Matt Slager: Yes. Podcast legend. [15:55] Marno Brits: So I just saw a snippet last night where he is interviewing someone. Uh, [00:16:00] he's a. Um, he, I think he's an author and a linguist or whatever he's talking about, but the, the. Topical conversation was the impact that AI is having on us. [16:10] Marno Brits: Evolving in that, there was a, the example he gave us, has it been a clear marker of, before CHATT came out, everyone spoke in a certain way in the U, the language that we used, and then Chatt came out. So since 2022, more and more people have been using the word delve. It's such a like Latin derived word, and that's a funny one, but it's something where like AI is trying to like sound sophisticated. [16:36] Marno Brits: So it starts using that. And then people started talking about that way and started using those words. And even looking at, um, lawyers and like ministers and stuff, you can see the language that they're using has increased and is a lot more Latin like, because that's what Chate communicates in. And it's almost like we've trained this thing, but now it's training us in the way that we think and the way we communicate. [17:00] Marno Brits: That's, so I'm curious to [17:00] Marno Brits: see how much of that is our own fork, but how much of that is in influenced by ai? [17:03] Richard Webbe: Look, it, it's happening everywhere and it didn't need technology to happen. You've just described fashion trends. Humans' ability to want to belong and follow yet remain as individuals. And so they'll take a percentage that they'll copy so they feel like they're in the club, but then they wanna put a different spin on it so they feel like they're individual. [17:22] Richard Webbe: And you know, you get the four personality types, which I laugh at 'cause I'm looking at a scream of four people and I think I can see all four of them. And, uh, and, and from that point of view know it, it's kind of. The Americanization of our language right around the world accelerated with Word and uh, with all the, uh, all the new office tools and the online tools and the computer and Zeds instead of ses. [18:00] Richard Webbe: And of course now in language, it is outrageously becoming very common. And, uh, I, you know, I think I told the story of someone recently who was selling in the, uh, in the IT space and they were using AI heavily to try and differentiate themselves. And I reminded this person, you know, you're [18:00] Richard Webbe: using the sum net average of everyone's language. [18:03] Richard Webbe: Yeah. And you're not differentiating yourself at all. Get this neck as Max Walker. My friend Max Walker used to say, use your neck top computer for the creative stuff and use everything else for the automation. And that's it. That's it. That's exactly what you guys are saying. And it, it is happening more and more and more. [18:21] Richard Webbe: Um, I have a friend of mine who's on online dating. And, uh, she was telling me that, uh, all of the, uh, profiles from the men are all the same. I go, of course, men are lazy and go straight into ai. Gimme a good profile. That's [18:34] Marno Brits: great. [18:35] Richard Webbe: Very much [18:36] Matt Slager: that, um, what you were saying there, mano about that shift of language and, um, the gentle exposure. [18:42] Matt Slager: It reminds me of the saying, like, how to boil a frog. Like, oh [18:46] Marno Brits: yeah, that's terrifying. [18:48] Matt Slager: Instantly though, instantly my, my alarm was going off, and I hope everyone who potentially watches us and listens to us has heard me ranting about in the past is cognitive [00:19:00] security. [19:01] Richard Webbe: Oh, yeah. Versus cognitive bias. Very true. [19:03] Matt Slager: That's literally what I've been talking about and like, you know, trying to build awareness about is, is that exposure to language change, you know, thinking patterns change. Similar to what Richard said, you know, it, it's existed in the past already, like of course it has. You know, this, the, the access that we have now and the access that they have now to us and our brains, and to slowly boil that frog. [19:25] Dave Pengelley: Well, even, even you saying that you're kind of embracing the junk AI stuff and going, well, that's, that's kind of what it is, and I don't mind it so much anymore. You're like, oh, that frog's getting warm. But that's good segue into some news. [19:40] AI VO: Welcome to the AI update. Let's look at what's happening in the news. [19:46] Dave Pengelley: All right, let's look at what's happening in the news where Dave forgot to pre share his screen. So we'll do that now. All right, look at that Taylor Swift files to [00:20:00] trademark voice and image after AI concern. [20:03] Marno Brits: Nice, because we're not expecting that headline at all. [20:06] Marno Brits: Like when you were sharing your screen, the last thing I was gonna see is Taylor Swift. [20:10] Matt Slager: My, my, our news things are so different, [20:12] Marno Brits: completely different, [20:15] Dave Pengelley: but we're just talking about, you know, that, that, that popularization and ideas and, and AI eating itself a little bit. And where does it learn, where does it get its ideas from? [20:24] Dave Pengelley: Apparently it's not gonna get them from Taylor Swift anymore. She is saying No. [20:29] Matt Slager: It's an interesting thing though, like if you, if you were like heavily, you know, like her in the entertainment industry, your whole brand and everything is revolving around your ability to create music or, you know, some sort of thing. [20:43] Matt Slager: I, I, I can totally empathize with the artists, [20:45] Dave Pengelley: her likeness. She doesn't want to turn up in, in random ads and things. And, uh, [20:50] Richard Webbe: can I just say something nothing against Taylor Swift. She's talented, very wealthy. She's reasonably attractive. But she looks like any girl next door. I mean, if she tries to [00:21:00] trademark her, her image, she's gonna rule out about a billion people around the world. [21:04] Richard Webbe: I mean, how do you do that? Where's the differentiation on the image? You're gonna need AI to say yes. No, that was a general one. No, that is you. No, that's a general one. [21:16] Dave Pengelley: All, all, all. She's got generic lady face, all, all, all mousey blonde girls look the same. [21:20] Matt Slager: Yeah. Pretty much. Makes me think of what Richard just said before, like the sum net average. [21:25] Matt Slager: You know, if, if, if one individual is saying, Hey, you can't use my voice. Um, yeah, why not? Like literally, why not? What's the problem? Like if everyone can use each other's voices and we're just fully aware of the fact that, you know, it could be used. For like harmful reasons, um, I don't really understand 'cause it just becomes that summate average thing again. [21:48] Dave Pengelley: Do Doppel Gang exists? People with similar sounding voices exist? It, it is a tricky one. I mean, I, I, um, went to a funeral, uh, last week. And the guy that was [00:22:00] a, it was an older bloke and they put up a photo on the screen of him from when he was in the seventies when he was young. Like he was twenties or thirties in the seventies. [22:07] Dave Pengelley: And he had that seventies look with the sort of the metal rim glasses and the, the, the big red beard and so on. I'm looking at that going, that could have been my dad. That was generic seventies guy. Like they all looked like that. They all kind of with a big beard and the metal rim glasses. There's this generic look of, of dudes from the seventies where they all kind look the same. [22:26] Dave Pengelley: I mean, Matt, you are not too far away from it. If you wanted to, you can get that. I don't have one [22:29] Matt Slager: of those like furry green [22:31] Dave Pengelley: jacket. Yep. Yeah. The, the, the, the furry sort of check flannel type jacket or something and put, put the little metal rim, the, the, the gold metal rim glasses on. Grow your hair a little bit flop, fuzzier. [22:40] Dave Pengelley: It's [22:40] Richard Webbe: got really good teeth. They didn't have good teeths back in those days. [22:44] Matt Slager: Oh, that's just my, um, that's my AI filter. I need to get one of them. [22:50] Richard Webbe: Oh. [22:50] Dave Pengelley: But like likenesses are a tricky one. Anyway. I, I thought I'd just open up with that. Um, it is interesting, this whole like. Disney did a big deal to license all the Pixar stuff over to Sawa, [00:23:00] and then Sawa shut down. [23:02] Dave Pengelley: So OpenAI had this goldmine of Disney IP that they were licensed to generate what people like to monetize, that [23:09] Richard Webbe: they're unique. You see animations, artwork, style that's kind of unique and short of serendipity occurring, which it does and needs to be taken into account. I understand why someone like, uh, sorry, excuse me. [23:22] Richard Webbe: I understand, uh, why someone like that would be, um, [23:25] Dave Pengelley: you, you mean when, when a studio releases ants and another studio releases a bug's life, like less than 12 months later? [23:30] Richard Webbe: Yeah, exactly. Exactly right. So, you know, we're gonna have to make a decision whether someone was plagiarizing, supporting izing, following, you know, but it, you know, we'll let some courts solve that. [23:43] Richard Webbe: I'm sure it'll be obvious to us as humans when someone copies 'em and someone doesn't. [23:47] Dave Pengelley: Alright, well let's, let's, let's get back into something that is a bit more in your swim lane, guys. Uh, this is, uh, back on the Sovereign thing. Open AI's latest models, codex now available on Amazon Bedrock. Uh, so [00:24:00] open AI is now letting you use your models, uh, and keep all the data residing here on Australian shores using Amazon Bedrock's platform. [24:10] Dave Pengelley: So, um, [24:12] Matt Slager: what, what's the news though? Like, I feel like that's been a, a case for a very long time. [24:18] Dave Pengelley: No, apparently it hasn't been. Like that's, that's New Codex wasn't on Bedrock before, [24:23] Matt Slager: as in the model. [24:24] Dave Pengelley: Yeah. ' [24:26] Matt Slager: cause they, they've done that really fun thing where they've named everything Codex, um, [24:30] Dave Pengelley: everything has moved that, that's the new brand, right? [24:33] Dave Pengelley: They're moving away from chat GPT to Codex as their main thing, like their desktop app is all revolving around being the Codex app now. [24:40] Matt Slager: Mm-hmm. Yeah. So like when I see that story, like previously it was always, if you wanted. ZD endpoints in Australia, you had to go through Azure and they had all of the 5.4 series in there. [24:56] Matt Slager: Um, like I currently use that for a few clients. [24:58] Dave Pengelley: That's, that's what this headline's saying [00:25:00] then, because the, the sub headline is the day after the chat, GPT creator loosened its ties with Microsoft. So yeah, so maybe it was on the Azure data centers, now it's on the Amazon ones as well as, as they sort of, um, do that. [25:11] Marno Brits: It was a big problem because it was like what Microsoft gave them over $50 billion and then all of a sudden it came out that, oh yeah, they also went and basically window shopping and they went to Amazon and got a big deal there and it makes sense for, it makes sense, but it's also just bad business, like the company that's funded you for so much, but then it could just be all publicity. [25:33] Marno Brits: Like there's so much smoke and mirrors. A [25:35] Matt Slager: hundred [25:35] Marno Brits: percent. Um, but I mean the money that's being thrown around is just tremendous. [25:39] Matt Slager: Similar to, um, you know, like a number of months ago we talked about meta buying. Manus. [25:45] Dave Pengelley: Yeah. [25:45] Matt Slager: And now the, the, the CCP is trying to stop that. Um, but Manus, since that, uh, initial deal, they've actually moved their operations to Singapore. [26:00] Matt Slager: Um, so they're no longer in c [26:00] Matt Slager: ccp, you know, borders jurisdiction. Yeah. So it's this real weird, like vibes, but I, I totally, totally put my tinfoil hat on with you, mano that I think these headlines and the, the news stories and the stuff that they do to create a reaction [26:17] Marno Brits: mm-hmm. [26:18] Matt Slager: Is part of the game. You know, all of the stuff with, um, Elon and Sam at the moment, you know, oh, Sam's a scam. [26:25] Matt Slager: Altman, you know, he's, he's stealing all of the, the profits, which not, was not meant to be a for-profit and all this kind of stuff. We will never know what the actual game is here. [26:36] Marno Brits: It also doesn't matter, like at the end of the day, I mean, I'm a big proponent of. The news stays the same. Just the names change only the names change. [26:44] Marno Brits: Mm-hmm. So I just, I don't care for the news at all unless it's impacting my immediate family or my immediate day-to-day. I can't do anything about it. I can make my corner of the earth better and that's all I can focus on. Right. [26:55] Matt Slager: A hundred percent. [26:55] Marno Brits: Um, but yeah, there's a lot of stuff where it's just, it's high level [00:27:00] plays from people that have way too much money and they have a lot of resources. [27:03] Marno Brits: And what's the incentive? I always ask the question, what's the incentive for doing this action or making this post? Yeah. It's never just, okay, I'm doing it for the good of the greater good. It's almost always this an incentive. So what they get incentivized from making the deal of AWS or making this article or getting news stories. [27:21] Marno Brits: It's like there's more business. So [27:23] Richard Webbe: it's always, it's like you said, there is a, I think someone said once famously might have been Walter Re, that there's only two pure things on the news sport and the weather. Everything else has cognitive confirmation, bias and advertising stuck in it. And it's true, right? [27:39] Richard Webbe: Mm-hmm. Uh, if you think about it, every little piece of advertising is either political, nominal, advertising, you know, I always laugh, um, uh, and, and say to friends when I'm watching the news, there'll be some, I dunno if, you know, some obscure news article comes up and you go, why the hell did that? It was so out of place. [28:00] Richard Webbe: And then you think about, ah, they're advertising a new, [28:00] Richard Webbe: a new, you know, bit of food. [28:01] Dave Pengelley: Yeah. So some of the puff pieces that appear on, on the news or on the morning on the television shows. I mean, we, we, the other night we were, um, my wife and I were flicking around and found Anchor Man Two was on, have you seen Anchor Man Two? [28:12] Richard Webbe: That's good. [28:14] Dave Pengelley: Um, which really details like, you know, like in a parody comedy way, the story of how the news became audience captured and just started chasing eyeballs and became very sensationalistic. And there is a whole subplot in there around commercial capture as well, and not reporting on things because of commercial interests and stuff. [28:32] Dave Pengelley: For a, for a funny movie. It's actually got like this sort of thread of [28:35] Richard Webbe: truth. It does have. I agree. David and Command. Very clever, very funny. [28:41] Dave Pengelley: What [28:41] Matt Slager: um, what else is in the news though, fellas? I feel like there's a few [28:45] Dave Pengelley: other, I mean just, just, just on a product update back to just sort of AI operations. Codex is getting a lot new feat and a lot of new stuff, and some of that's pretty fun, like in the Codex app, like the integrated browser use. [29:00] Dave Pengelley: So yes, a lot of them have browser use so you can have your playwrights and so on. [29:00] Dave Pengelley: But one of the things that's nice is the fact that, and I, I wanted this when I was first. Developing stuff in Antigravity that had a Chrome plugin, which was great. But there's like an annotate button in the built-in browser with Codex. [29:11] Dave Pengelley: So you can actually select elements on the screen and comment. So the, the AI isn't trying to guess from a screenshot what you're referencing, it actually knows directly. 'cause you clicked on this element that I need to look at that element. It makes some changes to that specific thing. I was doing a demo of it yesterday at this, uh, build day. [29:28] Dave Pengelley: I was at showing someone how these new tools can work. 'cause, 'cause she was like, what is all this stuff? Uh, and I brought up an app that I've been building and it took me to my AU zero page, which I had no control over. It's the old zero page. But just to show that in principle, I was like, annotate the blue login button, make this red. [29:47] Dave Pengelley: And they came back and said, I can't control that page. I can't make that red for you. That's on the author platform. And I said, by the way, the system you're building, the brand color is orange. Why are you asking for red? And I was like. [29:58] Matt Slager: Yeah. [29:58] Dave Pengelley: Brilliant. [29:59] Matt Slager: [00:30:00] Contextual knowledge. I love it. Yeah. Codex is gonna be insane. [30:03] Matt Slager: Like, I genuinely am so excited to see what Tebow and his team just keep shipping like they're achieving. I, I think he, he tweeted something this morning when I woke up. I saw his words. Were we achieving escape velocity? Like, I, I just love that kind of wording Wow. Where everything is speeding up and accelerating to some sort of insanity. [30:24] Matt Slager: Um, [30:24] Dave Pengelley: yeah. Which we, we, we discussed this a few weeks ago where we were showing how like the philanthropic were trying to eat their lunch and like had news release after news release, after news release. And now the coins flipped and open ai, uh, just sort of pumping out all the new updates and releases. [30:38] Dave Pengelley: They're both fighting each other on things like computer use, uh, which is super fun. 'cause I, I think I mentioned last week I've started building a video game as well, just for funsies and it's using a thing called the Go dot engine. Mm-hmm. And so what I do is I say. Play test it and it goes and spins up this video game in its own go dot engine window. [31:00] Dave Pengelley: And then it can actually click around and test whether, you [31:00] Dave Pengelley: know, the resource production is fast enough or too slow, or whether the enemies come too fast or too slow. And it can do this play balancing for me by just physically playing the game with computer use. It's awesome. [31:11] Matt Slager: Yeah. There's, um, there's been a couple of, like, I, I know you're not on, not next day, so you don't see like the, the latest news. [31:18] Matt Slager: Mm-hmm. Um, but the anthropic put a few new things out today that I saw that it's the same story, like they've packaged and productized something that has existed for months already. And they've made it just look super clean and, and which it, it's fantastic. Like if you can find the post stage, 'cause I can't share, but, um, there's stuff to do with Claude can connect straight to Blender. [31:44] Matt Slager: So if you do 3D modeling in Blender Ooh, nice. It [31:47] Marno Brits: can [31:47] Matt Slager: completely control everything in there for you. Yeah. [31:49] Marno Brits: Yep. [31:49] Matt Slager: Um, as well as, uh, some sort of audio platform for creating music. I can't remember what it's called, um, as well as. Touch designer, I believe is another one. [00:32:00] Um, and yeah, like that, those kind of interactions and integrations. [32:04] Matt Slager: Oh, also with Affinity, um, with like the Canva Affinity whole setup. [32:10] Marno Brits: Okay. [32:11] Matt Slager: Uh, [32:11] Marno Brits: I don't know about the Canva Affinity. I've not touched Canva much. I feel like they're so like primitive, like it's just so much easier to use anything else. [32:20] Matt Slager: It'd probably on the Claude page rather than a traffic. But anyway, um, I think that's awesome for people that are in that ecosystem and they use those tools. [32:28] Matt Slager: That's really, really cool. Like, there you go. Is the Blender one Oh yeah. And Fusion. So I used to use Fusion 360 a lot when I was doing 3D. Um, [32:37] Richard Webbe: wow. [32:37] Matt Slager: Really, really simple, easy, quick integrations that [32:40] Richard Webbe: we're seeing. We're seeing the democratization not only of information, but also the applications. Is that a fair assessment? [32:47] Matt Slager: Yeah. The, did you guys see, and Dave, I'd be curious if you saw like Salesforce is kind of moving away from their, their entire GUI [32:56] Dave Pengelley: it, it, it's funny, I mean this comes full 360 a little [00:33:00] bit. Uh, when I was there, uh, over a decade ago as they release things, the policy at the time was API first, everything was meant to have a fully functional API. [33:09] Dave Pengelley: So that way when they released their various mobile apps and things, you could get the full functionality, not just on your desktop web, but they could push that out and have that in the mobile apps. Uh, that as people built things on the app exchange and extensions, it was made have full API functionality. [33:24] Dave Pengelley: Now, as time went by and acquisitions happened, I think that that fell. Fell away a little bit and there wasn't as much of a focus on all products releasing API first all the time. But it sounds like they've gone back to that. 'cause they realize, oh, now we just wanna plug agents in. And the agents need APIs to talk to all the things and then access all the functionality. [33:43] Dave Pengelley: So they're announcing, you know, headless 360 or whatever because they love sticking extra taglines, like 360 and Force and 2.0 and Cloud and Einstein. And [33:53] Matt Slager: it's like Apple, [33:55] Dave Pengelley: uh, it's, it's a running gag like in the whole Salesforce community that they rebrand their products every 12 [00:34:00] months at their annual conference. [34:01] Dave Pengelley: And it's the same product, um, with a new badge on it and the few upgrades and it's like, makes things so hard for everyone. Yeah. Just remembering what are we calling it this year? Um, but yeah, I mean, again that's, and I saw, um, Google have released some new A two A spec as well. They're, they're sort of trying to push the into agent thing. [34:20] Dave Pengelley: There's, there's so much news every week and trying to work out, you know, what to cover. I decided we should talk about Taylor Swift. [34:26] Matt Slager: I wanna talk about 5.5 quickly for, for those of the people that are actually out there with OpenAI subscriptions still. Yes. Um, whether they've maintained it the entire time, or they've just come back since 5.4 and 5.5, um, I was one of them. [35:00] Matt Slager: I, I have to say that 5.5 on low reasoning in my pa in my pie set up. So in my agent harnesses that I've built is phenomenal. Like if people throw around the term a GI and all that sort of thing, um, I genuinely don't need [35:00] Matt Slager: anything smarter or anything more upgraded at this point. I can do everything I need from start to finish. [35:06] Matt Slager: Or without any issue. [35:08] Dave Pengelley: That's, that, that, that's a good warning. 'cause I was, uh, I went out in the city yesterday and I took my laptop with me and I upgraded Codex to make sure I had a local copy of the, of the tool. Um, and when I logged into Codex, it defaulted to 5.5 extra high fast. [35:24] Matt Slager: Oh wow. [35:25] Dave Pengelley: Like that was, that was the default setting was burn as many tokens as fast as you freaking can. [35:31] Dave Pengelley: And I was like, whoa, okay, we'll turn that off because I've been caught by the fast trap before. And you do go through tokens at two x the speed on fast mode, so get rid of fast mode. [35:40] Matt Slager: Did um, did you guys ever use 5.3 Codex Spark? Did you ever use the Spark model? [35:46] Dave Pengelley: Not really, no. [35:47] Matt Slager: So Spark like it sounds was fast. [36:00] Matt Slager: It was when, um, it was when they did a, a deal with cereus. So Spark actually runs on those big Cereus chips and it is [36:00] Matt Slager: intense. Yeah, yeah, yeah. I totally, totally agree with Sean. A hundred percent, like 5.2 was a waste of time. It was definitely the inflection point. But um, yeah, it was pretty scary. And then [36:12] Dave Pengelley: I think, I think the Codex version, so 5.2 chat might have, like I, I think I was started using Codex heavily for building some rust stuff and some, some structured things. [36:19] Dave Pengelley: And the Codex addition around that 5.2 time, I think it was 5.3 codex. Yeah, there was no regular 5.3. I think they went from 5.2 to 5.4 in chat and then they dropped the Codex moniker off the model. And they're not putting Codex on the models anymore. They're putting Codex on the harness. And I think kinda like Anthropic have gone with clauses, they go to market title. [36:40] Dave Pengelley: It feels like OpenAI are trying to move away from chat GPT and move to Codex as their go to market. [36:46] Matt Slager: Yeah. [36:46] Dave Pengelley: Product. [36:47] Matt Slager: I think it, the, the branding of, of chat GPT, like, you know, I, I adopted the Primos, um, moniker of, of Gyp, uh, which was from Boris Johnson, um, where he sort of pronounced it wrong. [00:37:00] And yeah, so like I always say gyp pity, and I, I say it tongue in cheek, but it's kind of like a bit of a tainted brand now. [37:05] Matt Slager: Like when you think of people that don't really know what exists in the world, they're like, oh yeah, I've heard of chat, GPT. Like, but you know, so you, you're, you're targeting the, the, the small, like, you know, not very highly adopted user base if you're using that term, but if you're saying Codex, codex are the builders. [37:22] Matt Slager: So I think that kind of makes sense for them to try. And we [37:26] Dave Pengelley: we're gonna, we're gonna move to some wins and discuss how we've been, uh, hands on with the tools this week. Uh, but I just wanna on the whole Codex thing. One last thing that they pointed out yesterday. So I was at these events and OpenAI were there doing some presentations, and one of the, um, the technical guys, he said, look, we know that the GPT models are not the best at doing front end ui. [38:00] Dave Pengelley: Like he, he just said that and, but he said, but now, especially like in the Codex app, you can do image gen and we've got the new image gen two, which is really good at generating anything. And so the trick is you use the image gen to create, to mock up the interface you want. [38:00] Dave Pengelley: And then Codex is really good at list reading that interface and then building off that mockup so that that would saying, look out of the box. [38:07] Dave Pengelley: If you tell us you want an interface, we're not gonna do the best job. But if you tell our image gen you want an interface, then you'll get something really good and then tell our model to use that image to build the front end and it will be really good. [38:17] Matt Slager: Well, yeah, if you're one shotting, like if you're just trying to one shot from the scratch, definitely a hundred percent like that makes that workflow is actually super, super effective. [38:26] Dave Pengelley: Yeah. And, and you know, j having the, the image gen built in, like I'm building a game and I've been doing it a bit in Codex, a bit in in Claude, but I need to move more into Codex I think. 'cause it can image gen the pixel assets and things to actually use as elements and sprites all in one tool. 'cause Claude doesn't have image gen at all. [38:45] Matt Slager: Yeah, well you can wire it in, you can literally, [38:47] Dave Pengelley: well you can, you've gotta go nano banana and you gotta do all other Yeah, but all in one. If it's there and I'm paying for it already and I don't need x extra API keys, then would Well [38:55] Matt Slager: you do, you, do you need to use your not pap and AI key in order to do that? [38:59] Matt Slager: Well, [38:59] Dave Pengelley: I've got a [00:39:00] Codex login. I'm using Codex with my sub the image gen. Right. Cool. It's all built in. You just, it's, it's part of the license I'm paying them for. [39:10] Matt Slager: So you've already had some successful gens on non API, key gen. [39:18] Dave Pengelley: I assume so. I haven't tried. I'm gonna try that in the background while we talk. [39:21] Dave Pengelley: I'm gonna bring up, uh, wins of the week and then we'll talk about that. And while you guys talk about your wins, I'm gonna load Codex in the background and see whether I can generate an image. Let's do that. [39:29] Matt Slager: Good. Because yeah, last time I checked it didn't work. I'd be stoked to see if that does work now. [39:33] Dave Pengelley: All right. [39:34] Richard Webbe: Can I bring a. [39:45] Matt Slager: Oh no, you [39:46] Richard Webbe: can't. I was gonna say that was timing, wasn't it? I forgot about the titling. Uh, so Mano and I were talking the other day about, um, about, uh, you know, how we position all this great technology [00:40:00] that we use for the. Technophobe that are there. And I checked with some of our clients and customers and some of our partners and that, and it really is, um, a group of haves and have-nots. [40:10] Richard Webbe: And I'm sure we all agree with that, right? So, you know, I I, I, I think I used to example, my ex-business partner from very recently is, is is still one of my closest friends. I said, Hey, you gotta look at this AI staff and we'll build some agents for you. And he's one of those people that loves to do his processing and loves to know where his stuff is and goes to his accountant with his perfect records and going, look, I'm, I've been a good boy. [40:31] Richard Webbe: Do I get an a and a gold staff at it's, I'm saying to him, can you just cut that crap out and put it to the back? Uh, it's like someone who likes to change gears on a car versus an automatic, of course they get emotionally attached to it. So I looked at a couple of business cases that we were involved in and just to give the headlines generic generically, one of these was a, uh, manufacturing company and they produced about, I know, a thousand to 2000 invoices a month for small manufacturing. [41:00] Richard Webbe: They did in the mid market. What was astounding [41:00] Richard Webbe: is that the most invoices were completely touched by human hand and checked. And of course I think, uh, mana or Matt might've written on my article, you know, keep the human involved now was another guy wrote, keep the human involved in the process. Like you've always looking at it and you know what's going on 'cause you can't use your own brain for patent collection. [41:17] Richard Webbe: But one of these ones that was all about, um, uh, yeah, agent orchestration across invoices going out, received double checking payment optimization. The results were just astounding. I mean, these guys avoided late penalties of $180,000 a year. Their discounts captured eight to 70, uh, early payment discounts for them paying out, saved them up to 71% of their costs. [42:00] Richard Webbe: Their processing time went for, from 4.2 days, uh, down to 6.8 hours. This is a small mid-market company that got a couple of buffoons like us to sit down and help them handle their invoicing processing in a very simple way. It wasn't updating data, it was no groundbreaking [42:00] Richard Webbe: technology. Um, their ROI, the implementation cost for them was 54,000. [42:05] Richard Webbe: Um, uh, but, uh, they saved, uh, an ROI of 608% Wow. On that. So, you know, don't hire us, but hire some buffoons like us. Right. And you might spend 50 grand with us, but you're gonna save all this money. And I remember when I was talking to years ago, Peter Svic was head of, was CIO of Cisco. I met him many years ago. [42:31] Richard Webbe: And I was the Cisco spokesperson for Europe and Asia at one stage about the new stuff, right? Going out and giving these talks. And the worst thing we could do is talk about an intranet because all the CIOs thought they were gonna have to sack all their staff. And I remember I created this graphic and I said, and it's kinda like my article said, no, you've got these people doing very basic shit, sorry stuff and you should, sorry, David, I did it again. [43:00] Richard Webbe: You just keep those people 'cause they know your [43:00] Richard Webbe: business, they're loyal, they love you, you love them. And just move that idiotic stuff out of their brain and use them in a more creative way. And this great graphic, so it made sure people didn't think you're gonna sack anyone. And so the key message here is, I've got example. [43:14] Richard Webbe: Of even small businesses growing their businesses substantially and saving hundreds and hundreds of thousands, if not millions of dollars, just by a very short engagement of a few months. They spend $50,000 with some nerds like us, and all of a sudden they don't have to. And it's so obvious it's happening so much. [43:33] Richard Webbe: I've got other examples that are just outrageous. I had another salesman where these people grew their business by 300% in six months, and it's because they were mining the data that they already knew, right? But they didn't have people assigned to looking at the moving market they had assigned to processing their market. [44:00] Richard Webbe: Anyway, that's my view of a win. And I think the ROIs is so big, but no one's, everyone. I, I think Mano you might have told me or [44:00] Richard Webbe: someone told me and, and, and other friends. People go, oh, how does it cost to get my AI stuff in? And it's a whole lot of tools and a waste of time and they're scared. No. How much money do you want me to print for you? [44:10] Richard Webbe: That's what I wanna tell my next, our next client. How much money would you like us to print for you? Because we will print it for you anyway. That's my view of a win this week. [44:19] Matt Slager: That's like borderline sort of like Gen Z, gen alpha lingo there. Is it? [44:23] Richard Webbe: Yeah. Oh good. I'm getting younger. I hate getting that paper. [44:28] Matt Slager: Well that's awesome. That's really cool. That's super nice wins. I, I, I've been stupid with like all of my client engagements. I haven't really captured any metrics for like before and after case study stuff, so, um, yeah, I wish I could say like 600% improvements and stuff like that. That'd be fantastic. [44:48] Richard Webbe: You know, people like yourself and Mano and David to get under the weeds and do the dirty work, like the mechanic on our car that we don't wanna see. [45:00] Richard Webbe: And it is really important that we acknowledge [45:00] Richard Webbe: that and then take a step back and say, you know, my mum used to always say, which is in my new book, coming out. Uh, forget the crap in the middle. Look at the start, look at the end. And that's the reality of what you're doing. Yeah. [45:12] Matt Slager: That's, that's how I describe systems to people. [45:14] Matt Slager: Don't worry about the middle part. [45:16] Richard Webbe: Yeah. [45:17] Matt Slager: I'll handle what comes out. [45:19] Richard Webbe: Exactly. So, uh, inputs, outputs. [45:21] Matt Slager: Yeah, [45:22] Richard Webbe: exactly. [45:23] Matt Slager: That's awesome. How did you go with your image, Jen? Dave? [45:27] Dave Pengelley: Uh, yeah, it totally worked. I'm just, I'm just tweaking it, uh, before I brought it up. Uh, but, uh. [45:32] Matt Slager: That's awesome. [45:33] Dave Pengelley: It's, it's, it's definitely working just outta the box. [45:36] Dave Pengelley: Um, just gimme gi gimme a sec. Just, just, just someone else told me win of the week and then, uh, I'll fill the image in a sec. [45:42] Matt Slager: I can, I can explain a, a win of the week. I've, um, I've managed to go about two weeks of pretty much almost 14 hour days, uh, without burnout. So yeah, that's, um, [45:54] Marno Brits: that's, I'm happy for scared at the same time. [45:58] Marno Brits: I'm [45:58] Richard Webbe: done, man. I'm [45:58] Marno Brits: sorry. I said I'm [00:46:00] happy for you, but I'm scared at the same time. You gotta get that sleep. [46:03] Matt Slager: Oh, I've just got the stuff that I need to deliver. Uh, I had a, I had a call with somebody last night and if you're watching, you know who you are. You know, you're a legend. Thank you very much for the call. [46:11] Matt Slager: We'll, we'll keep chatting. Um, then I, I told them there's, there's zero chance that I can commit to anything in the short term 'cause I just have these things I need to deliver. So like, literally after I hop off these, like this call today, I'm gonna just get back to the grindstone and keep thrashing 5.5. [46:29] Matt Slager: Um, but, uh, I would say a win honestly is the ability to deliver two massive projects in parallel. Um, [46:36] Marno Brits: amen. [46:37] Matt Slager: And, and both of those together have, you know, been on top of what I would normally be doing every month. And so this month would be, uh, almost triple what my normal monthly revenue would be based on. [46:54] Marno Brits: Amazing. [46:55] Matt Slager: Yeah, it's been insane. So I'm just trying my absolute best to serve [00:47:00] and to deliver. Um. You know, as this one man show, uh, with my, my, my pack of racing dogs in front of me. [47:09] Richard Webbe: Well, Matt, you know, the best thing about, um, about what you're saying from a, a, a market and a, a concept point of view, you've got a lot of people that wanna support you, of course, but the same token, you say, I'm triple my revenue. [47:21] Richard Webbe: What I wanna hear from you next week is how much money have you generated for your customers? [47:26] Matt Slager: Yeah. [47:26] Richard Webbe: Because that will be bigger by a magnitude. And that's the bit that we get blind to sometime we talk about we did this and we delivered that, and wasn't it great? But, hey, what'd you do for the community? [47:35] Matt Slager: Well, what [47:36] Richard Webbe: customer? [47:37] Matt Slager: I can say one, I think in, um, in, in careful words without reaching NDAs. Um. One of my clients turns over in the business that I am the sole systems engineer for. If, if I didn't have the stuff running for them, this business wouldn't exist. So I think that that's enough for me to say that I somewhat am responsible for it. [47:59] Matt Slager: Um, they, they turn over about three to 400 K US per month. Um, that's, that's top line. So, you know, before costs and everything, but that's a crazy, crazy amount of like, pressure that I feel like I literally was on a call earlier today and this, the, the CEO founder of this company was messaging me saying, is something broken? [48:22] Matt Slager: I'm like, oh no. What do you mean? So yeah, that's my one that gives me the most stress and, um, doesn't necessarily let me sleep. But [48:30] Dave Pengelley: this, this is where we sound a few, few weeks ago, Matt, that there's a point where maybe having a second human that can step in. Is useful. And so it's not all on you. And if you get hit by a bus and a lying in a hospital, your customer can still have a human helping make sure things are happening. [48:46] Richard Webbe: Yeah. Well he is. [48:49] Dave Pengelley: Yeah. No, no. Matt, Matt, Matt, Matt, uh, he, uh, he wants to build agent to never work with humans. He said that on the show, he is like, he is never hiring. He can't work with any of us. I, [48:59] Matt Slager: I actually had this [00:49:00] chat last night to one of, with one of my agents, you know, in deep, um, no, I, I was like, everything that I'm doing is, is to provide, means to do more cool stuff with cool people. [49:11] Matt Slager: So like yeah. The, the whole point of it is to continue to do it and if it means working with others to mean achieve that end, then it's only a benefit. Like, [49:20] Dave Pengelley: I'm just teasing man. [49:21] Richard Webbe: You mentioned that earlier. You mentioned it earlier. Um, when you're doing something, it's the big wide world, particularly as we get these tools and you dunno if you're heading off down a weird path of discovery. [49:31] Richard Webbe: And so using AI and your agents to reflect back to you. Where you are and what you're doing as well as humans. So you've got two. Now you still only have humans and they're all biased as we spoke about earlier. But if you apply that human reflection, which you need, we all need, uh, and then you apply that to some AI reflection coming back to you, you get an even better perspective. [49:51] Richard Webbe: One's pragmatic and one's got cognitive bites, and you know what to do between the two. [49:56] Matt Slager: You know what I did last night, Richard? I built, I built a framework, [00:50:00] um, to picture my life as a tabletop game. You know, those insane nerdy tabletop games where they and pizzas and weird resources and a rule book that's like a Bible and nobody wants to play it. [50:17] Matt Slager: It's just reserved for those people that know it. I, I said, okay, look, let's look at my life and, and business and, and every decision that can possibly be made as. Something that can be boiled down to these resources and, and where you apply your resources and focus, and how do you achieve a net benefit at the end? [50:35] Matt Slager: You know, surely this is a universal thing that every single person thinks about and feels. Um, there's gotta be some pattern here, right? So. We built this framework where it's like a full scoring system metric stuff. And I was like, okay, cool. I know where I can make a decision now. And it, it just was so clarity providing, um, yeah, I don't know. [50:58] Matt Slager: So yeah. That's [50:59] Richard Webbe: [00:51:00] awesome. I'm so jealous. So [51:01] Dave Pengelley: Matt Matt's investing wheat into the mech factory so that way he can cross the river and deploy the oil. [51:08] Matt Slager: Yep. And avoid the trap cards and all the things. [51:11] Richard Webbe: You know, it's, it's funny 'cause I'm not really like you guys in that I was a nerd, but I've, I've forgotten more than I could ever imagine. [51:20] Richard Webbe: And I remember I did a leadership course when I was at BHP and what I liked about that, le and, and, and, you know, I don't mind saying it was fabulous. Right. And there was 62 of us over three days, all senior executives at BHP trying to get through a bunch of these, you know. Pretend situations like, you know, here's a rope, you're all blind. [52:00] Richard Webbe: How do you stand in a perfect circle? Or weird crap like that designed by some weird psychologists from all over the world. And, and, and, and we, you know, our team won it. We're very successful and it was really great. What I loved about that course wasn't the problem solving that's always fun. But it was watching the, the, the cognitive bias and confirmation bias from humans [52:00] Richard Webbe: coming from different backgrounds. [52:02] Richard Webbe: And, and, and that's, that's where you get your creativity from, right? But what we get from ai, if we get a gen AI and agency into that, it brings us back to a pragmatic reality. Like when you are laying out your life as a game and the two is where it comes, it's the human element with the AI that makes it so exciting. [52:22] Richard Webbe: Sorry, end of speech. Mano. You haven't said much. I'll shut up. [52:25] Matt Slager: No, [52:25] Marno Brits: that's fine, mate. That's fine. I'm learning. [52:27] Matt Slager: Stop there. [52:28] Marno Brits: Do you wanna go, Matt? Yeah. [52:30] Matt Slager: I'm always like, I'm captivated whenever you speak, Richard. I think it's because you speak my language. And [52:35] Richard Webbe: I think we're both mental and I agree. Yes. [52:37] Matt Slager: Yeah, there's like a wavelength thing there that it just tunes in. [52:41] Matt Slager: Nice. If anyone else resonates with what Richard says and they happen to, um, still be listening or watching this in replay, put comments and like just explain whether or not you agree with it or not, because like that, those, those things always resonate with me so much. Anyway, back to Mano. [52:57] Richard Webbe: No, no, thanks Matt. [53:00] Richard Webbe: I appreciate you saying that and [53:00] Richard Webbe: I, I, I agree. I, I trade in that space. Mano [53:03] Marno Brits: easy. Um, yeah, once for the week, uh, talking with Matt and I've implemented my homies agent, um, I've, now, I've, I've started it by putting it in a docking container, so I wanted to make it as secure as possible because the first time I installed it, it was just locally on the pc. [53:19] Marno Brits: And although it was incredible to watch it work for about 25 minutes at a time. It scared the living crap outta me because it had complete control of anything and everything, and it was cool. Like it did a lot. But at the same time, I'm like, okay, well let's just, let's tone this back a bit. Let's put it into a docker container. [53:37] Marno Brits: Set that up. Spent about probably four hours setting it up to make sure it's safe and secure. It has a dedicated folder. This is how I communicate with it. Um, learning how to use WSL correctly. Mm-hmm. The windows subsystem for Linux. And then on Sunday, Matt sent me a message, Hey, do the Hermes update and then do the um, TUI as well. [54:00] Marno Brits: So I was like, cool, I'll do that. Long story [54:00] Marno Brits: short, seven hours later, her still isn't updating. I get furious. I was like, what the F is going on? I then pull it out of the docking container. It's now running locally. It just has a bunch of restrictions where like before I can do anything, it has to ask me for my permission. [54:15] Marno Brits: But at least then I now, now now know I can interact with it in a normal way instead of jumping through all these hoops and it still has some level of security where I have to still approve every little task. But I'd rather that because then I know it has the capabilities, I just have to monitor it. Um, so long story short, Hermis agent installed, working, running the business, help me build my website. [54:37] Marno Brits: Um, it's been extraordinary so far. [54:39] Richard Webbe: You just described a teenager getting your teenager to do gardening in the backyard. Oh, mate. [54:45] Dave Pengelley: I mean that, that, that was my open claw experience. I tried setting it up all my nemo claw. I tried setting up in Dockers and Wsls and just completely gave up. And then I ended up. [55:00] Dave Pengelley: Getting open, kind of running in WSL without Docker, but it never, I [55:00] Dave Pengelley: just was like, you know, I don't need it. With what I'm doing in Claude and Cowork and, and other things is not adding sufficient value to my life to spend, you know, 7, 4, 11, 12, 18 hours configuring yet another agent system. Mm-hmm. Um, because mano, a few weeks ago you'd built your own thing and you had your little pixel workers and you're doing all this sort of stuff and then you saw some stuff from Blake and you rewrote everything and it seems like you're just like agent hopping between which, which, which agent assistant am I gonna run this week? [55:29] Dave Pengelley: None of them seem to be quite landing for you, like, I dunno, is [55:32] Marno Brits: that right? Yeah. Well I started the, the process with that one where I had like pixel agents, which was cool to see. And it, it had some functionality, but the constraint there was, it was burning tokens too quickly. And then that's when I saw Blake's one of like, okay, cool, if you wanna build it. [56:00] Marno Brits: With first principles, then I took that approach, built a task management platform. So it's a lot more effective. It's efficient, and it is token efficient as well, which is perfect. So I like that. And then Claude Code decided to ban anything related to Open [56:00] Marno Brits: Cloud or Hermes or anything. I was like, well go Claude Code now sucks. [56:04] Marno Brits: So then moves it over to, um. Not moved it over yet, but then did a, an basically analysis between Open claw versus Hermes. Yeah. What my goal is, like I created a PRD for what my end product will look like of me as a business, and then provided that to perplexity to help me do a review of which agent makes more sense, my long-term business. [56:24] Marno Brits: Like I don't need this to be a Tony Stark level where it does everything. I just needed to support me and help me operate like a team of 20. So I want this chief of command and this command center that I can run the business from by talking to this one thing. And Hermes was the best result and it just, it won by a clear mile. [56:42] Marno Brits: So now this will be me for the foreseeable future until I hope so open AI cancels it. [56:48] Matt Slager: What model are you running it on at the moment? [56:50] Marno Brits: 5.5. [56:51] Matt Slager: In low, [56:52] Marno Brits: no, you don't have the option. In Hermes, you can only choose 5.5 [56:56] Matt Slager: slash reasoning. [56:56] Marno Brits: 5.4. No chant, buddy. That's how [56:59] Matt Slager: Yeah. Slash reasoning [00:57:00] low [57:01] Marno Brits: slash reasoning. Hey, you're too good. [57:04] Matt Slager: You, you need low, low, low in your life. Like, low is the new high. [57:09] Marno Brits: It's now set the low. 'cause it doesn't have to do very difficult things. I just had to digest all of my Alex or Moey playbooks as well as some mentoring courses that I paid for. [57:20] Matt Slager: Yep. [57:20] Marno Brits: And now it's just my day-to-day, my accountability coach, my strategist, um, and then hopefully doing marketing for me soon as well, which will be nice. [57:29] Matt Slager: Have you, have you gotten all around, um, for those of you that don't know about Hermes that I happen to be listening or, um, still here, Sean. Hello? Um, the, there's this really cool system with Hermes where it, it's it, it has like a recursive pedagogy loop where it teaches itself new skills. Um, have you gone around that, like with the HO stuff and those mentoring things, did you get it to set up frameworks for you as repeatable skills or did it [57:54] Marno Brits: already do? [58:00] Marno Brits: Yeah, so the way I've done it, I'm not sure if it's done properly, but what I, my practice that I followed [58:00] Marno Brits: was I got antigravity to go through and create the playbooks. So digest the books, create playbook, MDs for me, gave those playbooks to Hermes and then said, Hey, digest this. And it said it's created, um, references or skills to pull from that. [58:16] Marno Brits: But I've not tested it yet. This is in the past week. Yeah. Yeah. [58:19] Matt Slager: It'd be really, really cool in the future if you see it, pull that out and be like, well actually, you know, like that would be a really fun one. [58:25] Marno Brits: Yeah. It's been cool so far for the website copy, like writing the copy for that and doing my, um, outreach emails and stuff. [58:32] Marno Brits: It is pulling from Hormo without me telling it, which is nice. [58:35] Matt Slager: Nice. [58:35] Marno Brits: Um, so I'm curious to see now that it has about probably 12 books that it has to go through. How will that work? 'cause that's a lot of information. [58:43] Matt Slager: Yeah. Yeah. I, I'm feeling so for anyone curious, I think Hermes is the great local agent and you do exactly what Mano did, install it locally on your system, you know, and then you secure it in place. [59:00] Matt Slager: You know, you give it certain restrictions so that it can't go outside [59:00] Matt Slager: of boundaries, but without containerizing it. 'cause containerizing, it just ns it and you need an external agent in order to run it and update and do all things. So my mental model I have now is, is Hermes is the best sort of like local thing. [59:13] Matt Slager: And then you obviously can still do open call. And Open call is increasing in its functionality and competition with the same sort of setup here. But I love Open Claw as a containerized hosted deployment for production reasons. So, and I have a number of them now that I have set up for myself that I access through various means and I pipe data in and out of, and also for clients. [59:37] Matt Slager: Um, I know I said I was gonna make you guys one for, for the podcast, and I still need to do that. Um, but yeah, uh, the, the, the containerized option is for production deployment for sure. Um, yeah. And the local option for just that extensibility and sense of evolution is, um, unparalleled. [59:54] Dave Pengelley: Hmm. Yeah. I'm loving it. [59:56] Dave Pengelley: Nice. [59:56] Matt Slager: So cool. [59:58] Dave Pengelley: Moving so fast, all these different agents, like I [01:00:00] said, I, I, I kind of, unless I've got a good reason and we come back to, you know, talk about the ROI of implementing AI for businesses, I'm the same for myself. Like what I'm doing with co-work is working really, really well for me. And until Claude really decide to change things up and force my hand and given the evolution of Codex and, and some of the routines and, and stuff that I could do there, I could probably move a lot of my co-work infrastructure if I need to with all my schedules and my updates and so on, across to, to Codex order, Hermes or anything in the future. [01:00:27] Dave Pengelley: But at the moment I'm like, it's not broken. In fact, it's working really well. I am not gonna spend two days like mano rebuilding stuff at the moment. I'm just gonna use what's working for me. [01:00:37] Matt Slager: Thanks for grounding that. 'cause I, I totally agree. I'm just a spastic and I do all of them all at once. Um, so yeah, if, yeah, you, you literally use one thing and master it. [01:00:47] Matt Slager: People always say, should I change subscription? Should I change companies? Should I use that agent or look at this new tool, you know, just keep using the same ones. Yeah. And just get really good at it. [01:00:56] Dave Pengelley: Like, I moved from Antigravity to Claude because Antigravity usage [01:01:00] licensing just got crushed heavily. [01:01:02] Matt Slager: Yeah. [01:01:02] Dave Pengelley: And Claude gave me upgrades, like having scheduled things, which I couldn't do in Antigravity at all. Um, yeah. So now I've locked in on there and that's doing really well. But like I said, I, I bounced back and forth. I've got the Codex app here. You asked about image gen, let's see. Boom. Draw me a logo for the AI Operators podcast that captures the idea of a panel of people talking about the state of ai. [01:01:22] Dave Pengelley: And here we go. Use the image Generation skill. It's a podcast logo. There we go. Skills System slash image, gen Skill md. [01:01:30] Richard Webbe: Why do I have a ballet bun? [01:01:34] Dave Pengelley: What maybe. [01:01:34] Matt Slager: Interesting. [01:01:35] Dave Pengelley: And, and so then, then, then, then I gave it a little c nodes and said, integrate this concept. And it said, boom. There you go. [01:01:43] Matt Slager: Nice. [01:01:44] Richard Webbe: Nice. [01:01:45] Matt Slager: Is that, um, um, is that Amber and Mandu? [01:01:49] Dave Pengelley: You saw the prompt man. You saw the prompt. Uh, 5.5 low. Yeah. But this was a fresh install and it literally came in at extra [01:02:00] high speed, like it came in like that. That's [01:02:03] Marno Brits: crazy. Yeah. [01:02:04] Dave Pengelley: Um, which like when you go to speed, it says. Uh, so I was saying one and a half now I was saying two X at one point. [01:02:11] Matt Slager: Yeah, it was [01:02:12] Dave Pengelley: two x increased. [01:02:13] Dave Pengelley: Uh, yeah. So one and a half times faster with increased plan usage. The plan usage is two x, so you get one and a half times the speed for double the burn. It's like, ah, that maths doesn't math for me. [01:02:22] Matt Slager: That's the, uh, that's the Gary Tan mode that boil the ocean. [01:02:24] Dave Pengelley: Yeah. Yeah. So, um, but anyway, that's, that's image gen built into the Codex app. [01:02:29] Dave Pengelley: So that's [01:02:30] Matt Slager: Yeah, [01:02:30] Dave Pengelley: using, [01:02:31] Matt Slager: I don't know if that's using image gen two though, 'cause it'd be curious to see what that skill actually describes. It looks like you have a scroll bar there. Um, [01:02:41] Dave Pengelley: yeah, I do. All right, here we go. Um, skill use image gen tool, image Gen PI for CLI mode generate edit. Who knows? [01:02:53] Marno Brits: Oh, there we go. [01:02:54] Dave Pengelley: Uh, PT [01:02:55] Marno Brits: image two. [01:02:56] Dave Pengelley: Yes. Never silent switch from built in image. Gen C-L-I-G-P-T image two to [01:03:00] CLIG PT image, gen 1.5. [01:03:04] Matt Slager: Oh, um, I was speaking to someone recently about like how to get used to communicating with, with agents and, um, they were like, how do you, how do you talk to them so well, like you've got such like a good grasp of like wording. [01:03:16] Matt Slager: And I go, I'm just skeptical. I just, I don't trust them and I don't like believe what they say the first time. I have to force 'em to prove it. So like in that situation there, Dave, I would just be interrogating being like, which model did you use? Did you, you run this through an API key? You know, like all of the things, show me the exact payload that you used to do that thing. [01:03:34] Matt Slager: And yeah, again, [01:03:36] Richard Webbe: managing a teenager. Where did you get the money? What did you do to get it and where have you been? Where [01:03:41] Matt Slager: you with, you know, what's their phone number? Tell me what time. Yeah, [01:03:49] Dave Pengelley: yeah, yeah. I, I don't know. There is lots of mentions of G pt image two in here, but without spending the time to read through it or get it to reread itself and explain it to me, which we're not gonna do now 'cause we [01:04:00] are out of time. [01:04:01] Dave Pengelley: Um, I've already shared lots of things I've done this week, but one little win of the week that ties back and close the loop on. Last week we were talking about vibe, coding apps and hacks and security holes and so on. One of the things when I was prepping my app for launch, I, uh, stay what I should be building in to help me as, as sort of a founder launch infrastructure. [01:04:19] Dave Pengelley: And it told me to put in a few, um, hooks in the backend for a tools called Sentry and Post Hog. And so I put those in and sent, went, Hey, there's a problem with the way you've set up Stripe, where that's the wrong product code and that's not gonna bill and you can't actually get charges and stuff like that. [01:04:36] Dave Pengelley: So. Uh, again, like this is something I would've known, like you said, interrogate, ask questions. If I'd just gone, here's my app and here, here it is, it's working. And then you've got no monitoring infra, you wouldn't know these things. And so again, just asking more questions and going, what are the tools I need to manage this as an ongoing SaaS app? [01:05:00] Dave Pengelley: And doing those extra monitoring tools in is is really important. That was a bit of a win for me getting that. And [01:05:00] Dave Pengelley: because I've got post hog in, which is like a analytics kind of monitoring tool in some sense it can map and show me when people come to my marketing website, but then transition into customers and can maintain the ID 'cause they've got post hog across both different platforms And yeah, some, some cool stuff there. [01:05:16] Matt Slager: I love OG and Sentry's amazing as well. [01:05:18] Dave Pengelley: Yeah, this is amazing how it just, what, how it's just monitoring and going, Hey, you did a lot of database calls there. That's weird. Why, why would I be doing that? And just proactively just says, maybe go check that out. And I'm like, oh, okay. I'll go check that out. [01:05:30] Matt Slager: Yep. [01:05:31] Dave Pengelley: Nice. That's, that's a bit of a win. Um, I installed it monitoring infrastructure that I had no idea about, and when I went and went, what is this stuff that I installed and went logged in and was like, oh my goodness, it's actually got news and events. That's, that's very cool. It just, the system works. [01:05:44] Matt Slager: It brings back the point, like, as a final note. [01:05:47] Matt Slager: To everyone, like vibe coating removes the typing. It doesn't remove the engineering. [01:05:54] Richard Webbe: Exactly, exactly. Wise words. [01:05:57] Matt Slager: Hmm. [01:05:58] Dave Pengelley: Yeah, yeah, yeah. Which [01:05:59] Matt Slager: [01:06:00] one? [01:06:03] Richard Webbe: We're done. [01:06:04] Dave Pengelley: Alright, well thank you all. Thank you gentlemen. Thanks. Thank you. Australia. Thank the world. I'm sure we're global. I, I'm pretty sure global viewership of this show. And so, uh, we'll automate some pipelines. We'll spam you with half sentence clips of this show on your feeds from now on and on that [01:06:23] Richard Webbe: I might, I might publish a couple of those business cases too. [01:06:25] Richard Webbe: Anyway, go on, Dave. Sorry. No, that's it. [01:06:27] Dave Pengelley: We're good. Goodbye. Goodbye Australia. Goodbye, gentlemen. Yeah. We'll see you all next week. [01:06:32] Matt Slager: Thanks for listening.
Related episodes

Keep going

Chapters

Jump to a section

00:00Cold Open: Richard's Indonesia Rib, ADHD, Regular Roster
01:00Ep 13 Recap: Manjula's Smart Syllabus, Super Friends Roster
01:37AI in Education: Data Sovereignty, Curriculum-Specific Tools
02:21AI in Religious Orgs: Sermon Prep, Church Admin, Papal Ban
05:22Live Build Showcase: Letterbox Drop iPad App (Church)
07:53Dave's Win: n8n to YouTube Metadata Pipeline Automated
11:14AI Video Clipping: Descript, Hyperframes, the Slop Debate
19:04AI Update: Taylor Swift Trademarks Voice and Image
23:12AI Update: OpenAI Codex on Amazon Bedrock
26:00AI Language Drift: The Delve Effect
28:07Cognitive Security: Boiling the Frog with AI
39:00Wins of the Week: Richard's 608% ROI Manufacturing Case
44:00Wins: Matt — 3x Revenue Month, Two Projects in Parallel
49:20Matt's Tabletop Life Framework
51:50Wins: Marno — Hermes Agent Live, Docker Saga
59:18Wins: Dave — Sentry + PostHog Caught Stripe Misconfiguration
1:03:43Codex Image Gen Live Demo (On-Air)
1:05:09Closing: Vibe Coding Removes the Typing. Not the Engineering.