All episodes

Your AI Agents Are Overloading Your Laptop. Here Is How Matt Scales.

livestream1:06:1011 Nov 2026

Matt shares how he offloads AI agent workloads from his laptop to a home server and cloud sandboxes. Dave, Richard and Matt discuss AI discovery, model costs, algorithm bubbles and the judgement needed to use AI well.

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

Episode notes

Matt explains how a home server takes agent sessions and heavy development jobs off his laptop. Dave, Richard and Matt then discuss cloud sandboxes, AI business discovery, model costs and the importance of questioning AI output.

In this episode

  • Local servers, cloud compute and the economics of running coding agents.
  • ChatGPT business profiles, local search and platform discovery.
  • Social media algorithms, cognitive security and model bias.
  • Iterating on AI-built games and thumbnails, and applying human judgement to the results.
  • AI regulation, infrastructure and the panel's differing concerns.

Featuring Dave Pengelley, Richard Webbe and Matt Slager.

Transcript
[00:00] Matt: Oh, [00:00] Dave Pengelley: I'm all [00:00] Matt: I [00:00] Dave Pengelley: s [00:00] Matt: can [00:00] Dave Pengelley: I'm [00:00] Matt: see [00:00] Dave Pengelley: not [00:01] Matt: the [00:01] Dave Pengelley: like [00:01] Matt: the top [00:01] Dave Pengelley: um [00:01] Matt: is shorter, [00:02] Dave Pengelley: I'm [00:02] Matt: yeah. [00:02] Dave Pengelley: and I'm way l way less scruffy on the sides. [00:04] Richard Webbe: I'm [00:04] Dave Pengelley: Like it's [00:05] Richard Webbe: just [00:05] Dave Pengelley: all [00:05] Richard Webbe: trying [00:05] Dave Pengelley: been [00:05] Richard Webbe: to [00:05] Dave Pengelley: sort [00:05] Richard Webbe: see [00:06] Dave Pengelley: of [00:06] Richard Webbe: which particular hair [00:06] Dave Pengelley: look [00:07] Richard Webbe: you cut. [00:07] Dave Pengelley: Yeah, no [00:08] Richard Webbe: Did [00:08] Dave Pengelley: no [00:08] Richard Webbe: you [00:08] Dave Pengelley: the the [00:08] Richard Webbe: label [00:08] Dave Pengelley: ones on the side here, [00:09] Richard Webbe: it? [00:09] Dave Pengelley: all the ones on the top on the top I I left pretty long. Um and so still it doesn't actually really necessarily look like I I uh but I had some uh meetings to have earlier in the week, so I thought I'd better get [00:18] Richard Webbe: You [00:19] Dave Pengelley: tidied [00:19] Richard Webbe: know your fans [00:19] Dave Pengelley: up. [00:20] Richard Webbe: will be just twisting in excitement at this conversation about [00:23] Dave Pengelley: I [00:23] Richard Webbe: which [00:23] Dave Pengelley: know [00:23] Richard Webbe: hair [00:23] Dave Pengelley: we [00:23] Richard Webbe: you cut. [00:24] Dave Pengelley: look, I mean I dodged the whole men's grooming thing last week. I thought we'll just we'll just we'll just go straight. in we'll [00:30] Matt: With [00:31] Dave Pengelley: um [00:31] Matt: that. [00:31] Dave Pengelley: dive dive feet first into the grooming [00:34] Matt: Well after this show today, I'm gonna literally be diving into my own haircut. So there we [00:38] Dave Pengelley: oh [00:38] Matt: go. [00:39] Dave Pengelley: very good with your with your suck cut [00:42] Matt: Yeah. Yeah. [00:44] Dave Pengelley: it certainly does suck [00:47] Matt: You know, I've done my own stuff since COVID. I just sort of learnt it and haven't stopped. [00:52] Dave Pengelley: yeah no I we I always think the uh the Ways World thing like oh oh it's suck in my will to live With Garth, just freaking out. The socket. Um, [01:07] Matt: Gotta [01:07] Dave Pengelley: speaking of [01:07] Matt: work [01:07] Dave Pengelley: doing [01:07] Matt: right [01:08] Dave Pengelley: things at home, [01:08] Matt: now. [01:08] Dave Pengelley: Matt, I noticed that you've shared a thing that you're doing at home. [01:15] Matt: Oh. [01:16] Dave Pengelley: Um [01:16] Matt: Damn it, what did I share? [01:20] Dave Pengelley: dun dun dun. [01:22] Matt: Actually [01:22] Dave Pengelley: Um [01:22] Matt: I've been trying to do more content, so finally trying to do [01:25] Dave Pengelley: a [01:25] Matt: some stuff. [01:26] Dave Pengelley: box. You shared a box. [01:30] Dave Pengelley: That that that allows you to go much harder. [01:32] Matt: Literally, that's what that does. It's the go [01:35] Richard Webbe: So [01:35] Matt: hard [01:35] Richard Webbe: what is [01:35] Matt: box. [01:35] Richard Webbe: what is it? Just describe for an idiot like me what it is. [01:38] Matt: Okay, so I've got a fairly old laptop. My my actual laptop that I use daily is an old M1 MacBook. Um it's old now in in like tech standards. So if I do all of the kind of vagentic engineering stuff that I need to do for local development, which for people that don't know, there's a lot of processes that just grind your computer. You know, your CPU spikes, your RAM starts spiking, and all of a sudden you're out of hard drive space. And that's just the the joys of trying to build multiple projects at once. But then you put on top of that your actual agent sessions. So the coding a coding agent sessions like your clawed codes, you know, each one of them will add another chunk of RAM, another chunk of CPU. So basically my laptop would just grind to a halt and it would just panic. Uh Just entirely exhausted. So what that extra computer is, it allows me to run all of that stuff separately. So it's basically like a like a home server. You know, it's like my lone little mini data center. You know, nobody attack me. Um, doesn't use any water. So the it literally allows me to put all of that stuff in there. So at any given time, I'm working on at least like 10 different projects. And each of those projects will have an agent orchestrator. That can spin up and fan out across an entire agent swarm of things, you know, just all running their own processes. And then when it comes time to do those really heavy developer jobs, like, you know, like a compilation or a build or a type check or something like that, instead of it just making my my laptop die, they they join a queue in that server and they each can get processed one at a time with a more powerful machine. [03:27] Richard Webbe: Wow, did you buy it from a Chinese spy that you had an Fair with [03:30] Dave Pengelley: Oh, [03:30] Matt: Yes. [03:31] Dave Pengelley: we we'll get onto that in the news. [03:35] Dave Pengelley: Uh it's it's [03:35] Richard Webbe: sorry, [03:35] Dave Pengelley: funny, you [03:36] Richard Webbe: I [03:36] Dave Pengelley: mentioned [03:36] Richard Webbe: couldn't help [03:37] Dave Pengelley: I [03:37] Richard Webbe: myself. [03:37] Dave Pengelley: know, no, no, we will talk about Mr. Wang. I mean Huang um later on. But the um it's funny, I like you talk about the water thing. We're watching Tulsa King. You guys watch Tulsa King on [03:48] Richard Webbe: Yes, [03:48] Dave Pengelley: Paramount [03:48] Richard Webbe: we [03:49] Dave Pengelley: Plus [03:49] Richard Webbe: love [03:49] Dave Pengelley: with [03:49] Richard Webbe: Tulsa [03:49] Dave Pengelley: Stallone? [03:50] Richard Webbe: King. Yep. [03:51] Dave Pengelley: Data [03:51] Matt: No. [03:51] Dave Pengelley: centers of the bad guy this season, which uh is [03:55] Richard Webbe: Surprise me, some socialist writing for Hollywood [03:58] Dave Pengelley: uh [03:58] Richard Webbe: comes [03:59] Dave Pengelley: no [03:59] Richard Webbe: up [03:59] Dave Pengelley: no [03:59] Richard Webbe: with [03:59] Dave Pengelley: no [03:59] Richard Webbe: a [03:59] Dave Pengelley: but Coming from Sly, Sly, Sly leans to the right. So um he's not in the woke brigade at [04:05] Richard Webbe: No, [04:05] Dave Pengelley: all. [04:06] Richard Webbe: but I [04:06] Dave Pengelley: Like [04:06] Richard Webbe: think he's having I think he's having fun with the concept. [04:09] Dave Pengelley: Yeah, well, I mean, look, it's uh the notorious uh pseudo-European bad guy. It's sort of almost 80s guy with uh let's let's go with the nondescript European dude as the bad guy, the Hans [04:22] Richard Webbe: Yes. [04:22] Dave Pengelley: Gruber. [04:22] Richard Webbe: With a [04:23] Dave Pengelley: But [04:23] Richard Webbe: German [04:23] Dave Pengelley: uh [04:23] Richard Webbe: with a German accent. [04:25] Dave Pengelley: that's right, right? All right, well, um we've I've said the things, we've all talked about the things, so let's Kick off. That's not so bad. Four minutes. [04:43] Dave Pengelley: Four and a half minutes into the stream before I I ran the bumper. Have you ever watched one of those TV shows where you're watching it and watching it, watching it, there's like fifteen minutes in, and then they play the opening title sequence, and you're like, [04:52] Richard Webbe: Happens [04:53] Dave Pengelley: What [04:53] Richard Webbe: a lot [04:53] Dave Pengelley: is [04:53] Richard Webbe: on [04:53] Dave Pengelley: that? [04:53] Richard Webbe: server happens a lot on the streaming channels now, a lot with the big series. [04:58] Dave Pengelley: Yeah. [05:00] Richard Webbe: Straight into it and you watch it, it's like the end of the show because you watch it, it gets to a really critical point, and then they bring up the credits, and you go, What? [05:06] Dave Pengelley: Yeah, like we're like like a quarter of the way through the episode, and then they've dropped the bumper. I'm like, that is wild. And then you go skip intro. [05:14] Matt: I see it with um with the streamers and YouTubers that I watch now where they do their ad segments and you know usually it's it's pretty defined, like similar to how how you try and run this and they'll they'll do that thing and then they'll they'll throw out their little like pun or like little pithy quote. And then that transitions into the ad segment. [05:32] Richard Webbe: Yeah. [05:33] Matt: But yeah, recently I was watching one and it went for ages. I'm like, oh, he's not doing one. And then it jumped in like a little bit [05:38] Dave Pengelley: Yeah. [05:38] Matt: later on. So yeah, that that was my experience of the same. [05:41] Dave Pengelley: I I like the way uh Ryan George does his stuff. He's he's pretty good. I don't know who I was watching Ryan George, he's a pitch meeting guy, does all kinds of just funny sketch stuff. One of my favorite Canadians, named Ryan. And uh because there's [05:55] Richard Webbe: They're [05:55] Dave Pengelley: a few [05:56] Richard Webbe: all [05:56] Dave Pengelley: of [05:56] Richard Webbe: they're all [05:56] Dave Pengelley: them. [05:56] Richard Webbe: called Ryan. [05:57] Dave Pengelley: Yeah. Um very common Canadian. Name. Anyway, he he does uh he'll often say this is a this uh sketch is brought to you by you know whichever company it is. More about them after the sketch or later. So he'll at least give them a like a name, like that [06:11] Richard Webbe: Yes. [06:11] Dave Pengelley: way if you click out straight away, at least they people heard their name. And then later on, he sort of goes into the the ad read and he's like, I'm the astronaut up here in ad space, and it's like a little astronaut guy floating with his face in the helmet and talking about the thing. Like, that's his his uh [06:25] Richard Webbe: Yes, [06:25] Dave Pengelley: his shtick. [06:26] Richard Webbe: very funny. [06:26] Dave Pengelley: He [06:27] Richard Webbe: I [06:27] Dave Pengelley: does [06:27] Richard Webbe: like [06:27] Dave Pengelley: it [06:27] Richard Webbe: that. [06:27] Dave Pengelley: very well. Um [06:30] Matt: Be said about like entertainers like that that do ad segments very well. [06:33] Dave Pengelley: Hmm. [06:34] Matt: Like I know he's probably not everyone's cup of tea, um, but I've always really enjoyed PewDiePie's ad segments. [06:41] Dave Pengelley: I've never watched any PewDiePie anything. [06:43] Matt: There you go. Yeah, it'd be worth checking him out now. He's very, very different now to what he was like 10 years ago when he just yelled at computers while he was playing games. Um but now he's literally he's doing what we do. He's building his own like AI stuff. So um yeah, [06:59] Richard Webbe: Yeah, [06:59] Matt: it's [06:59] Richard Webbe: just [06:59] Matt: fascinating. [06:59] Richard Webbe: uh It's PewDiePie. [07:02] Matt: Compu [07:02] Dave Pengelley: So Pew PewD, like P E W like a church, and then die and then Pie like a thing you eat. [07:10] Richard Webbe: PewDiePie. [07:11] Matt: D M A. Yeah. [07:12] Dave Pengelley: Pew. [07:13] Richard Webbe: PewDiePie. [07:14] Matt: There you go. That's [07:15] Dave Pengelley: Pew [07:15] Matt: your last [07:15] Dave Pengelley: die [07:16] Matt: question. [07:16] Dave Pengelley: pie. [07:16] Richard Webbe: Pew Die Pie. And it's it's we're looking for him on streaming, YouTube, TikTok. [07:25] Matt: Yeah. [07:26] Dave Pengelley: He he was Mr. Beast before Mr Beast was Mr. Beast. [07:30] Richard Webbe: Ah, [07:30] Matt: Yeah, pretty [07:30] Richard Webbe: thank [07:30] Matt: much. [07:30] Richard Webbe: you. [07:31] Matt: Yeah, that's [07:31] Richard Webbe: Mr. [07:31] Matt: probably a good [07:32] Richard Webbe: Based [07:32] Matt: way to put it. [07:32] Richard Webbe: on the world. [07:34] Dave Pengelley: Yes, I've also watched no no I've watched No Beast. My kids have done a lot of uh what are the guys? I can't remember the names. They guys that do they used to just do like trick shots and now they do a lot more. [07:47] Matt: Dude, perfect. [07:47] Dave Pengelley: Dude, perfect, that's it. They they they just went crazy. I mean good on them. I mean they they were just a couple of guys just doing random trick shot videos and they've just they've become bigger than Ben Hur, their stuff, and it's it's crazy. The amount of like hyped up energy they put in their videos with the fast cuts and the music and drives me insane to listen to it from the room next door. [08:08] Matt: There's a weird sort of transition, but what like what we're talking about here is exactly what people are freaking out about with regards to like software and the ability to have a moat in in industries is vanishing. And it's getting replaced with literally just that. It's the attention economy still, like the ability to drive markets, build up. Your own following for your own thing, regardless of if how many different versions of the same thing already exists. You know, it's just that that following, that kind of brand. And yeah. [08:40] Richard Webbe: The amount of AI attractive women promoting things that aren't real but are hanging themselves out as a middle-aged single or a young single or something like that to capture a male market or a male doing the same for the female market because my algorithms don't play the same way my partners. Too. It's just astoundingly obvious. [09:03] Dave Pengelley: I don't [09:03] Richard Webbe: It's [09:03] Dave Pengelley: see any of [09:03] Richard Webbe: just [09:04] Dave Pengelley: that content in my algorithm, uh, Richard. I don't know what you're I don't know what you're perusing and what your dwell time sits, but I don't I don't get that in mind. [09:12] Matt: My my algorithm just tells me about new cloud compute services. [09:17] Richard Webbe: you two [09:17] Dave Pengelley: Uh [09:17] Richard Webbe: are as pure as the driven snow. [09:20] Dave Pengelley: it's it's weird. I mean, my in my Instagram, you know, it's a bit of F1 memes and and ADHD memes and other funny stuff and music stuff that I like, um, Creed and Monster Mm. like like m man meme type stuff, not all the half naked lady stuff. But it's funny, like I've been setting up new X accounts for different things. Like AI operators are getting a little bit established now, but Tin Mud Games um I I hasn't done much there. And when I go to X, you do start getting just all the random different clickbait stuff across all of that spectrum. And I know it's like don't fall into the trap of dwelling on any of the stuff you don't like because that's what your feed will become. Like 'cause [09:58] Richard Webbe: And [09:58] Dave Pengelley: it you [09:58] Richard Webbe: so [09:59] Dave Pengelley: do just [09:59] Richard Webbe: it's like [10:00] Dave Pengelley: Off in your feed, and it's like, no no, flick through, don't don't fall for the honeypot, keep going, because I don't want this to become what this feed [10:07] Matt: And there's [10:07] Dave Pengelley: is. [10:08] Matt: temporary adjustments too that they'll do. So, you know, if you start binging a particular thing, your feed will be full of that particular thing. But then if you don't touch it again until tomorrow, those things won't be on your feed again. It's [10:19] Richard Webbe: Oh, [10:20] Matt: fascinating. I I actually, because I I do a lot of like X stuff now and yeah, they have their algorithm fully open, [10:27] Richard Webbe: yeah. [10:28] Matt: fully [10:28] Richard Webbe: Well, [10:28] Matt: publicized. [10:28] Richard Webbe: we should put that in. I mean, it's it's AI algorithms that are feeding on the data you give it to make decisions. And the objective of TikTok and YouTube and all of that is for you to stay looking at it. Because the longer you look at it, the [10:42] Dave Pengelley: Yeah. [10:42] Richard Webbe: more money they get. And this is where it becomes dangerous because the original AI algorithm supported you looking at the same thing and then them feeding it to you. And of course, it created it creates confirmation bias for [10:55] Dave Pengelley: That's [10:55] Richard Webbe: the morons [10:56] Dave Pengelley: the Flat Earth's [10:56] Richard Webbe: out of there. [10:57] Dave Pengelley: problem. [10:57] Richard Webbe: Yeah, it is. It's it's like so if you've got a flat earth or an Looks at more flat earth stuff, YouTube says, Well, I'm gonna send you more flat earth stuff. And then all of a sudden, in their eyes, the whole world is talking flat earth when none of us are in their bubble. They're in a little world over there that we don't even know, right? [11:14] Matt: Yeah, [11:14] Richard Webbe: Like [11:14] Matt: the [11:14] Richard Webbe: your [11:15] Matt: bubble [11:15] Richard Webbe: teenagers. [11:15] Matt: is the right one. [11:16] Dave Pengelley: I mean [11:16] Richard Webbe: Sorry? [11:17] Dave Pengelley: this this [11:18] Richard Webbe: Yeah, [11:18] Dave Pengelley: this is [11:18] Richard Webbe: bubble. [11:19] Dave Pengelley: this is the issue with some of these platforms. I I don't know where X is nowadays, but it's certainly there was you were penalized for sharing external links in your main post for some time. LinkedIn I think still penalizes you because they don't want you leaving. The platform. They want you to stay on the platform. So, like, oh, here I am, especially LinkedIn. That's ridiculous. Like, here I am promoting a business or a service, and I want people to go and find out more about my business or service. So I want to put a hyperlink out to my offer and things. But as soon as you do that, you don't get shown to anyone. No one can see your business or service or offer or anything else that you put on LinkedIn. So it's this weird thing where, in order to drive that sort of um dwell time and engagement, they hobble the core thing that people would want to use. The uh the platform for in LinkedIn is another one that like if I'm posting about AI stuff, it's not because I want to see a bunch of other AI stuff, it's because I'm trying to promote that to people outside the AI bubble. I don't know how you do that because of the way audienc things work, you just see what you are in common with. So I get all the other AI consultancies and freelancers and influencers and journalists and stuff, all about AI, AI, AI. I'm like, how do I get out of that to find the small medium business owners that need some uh Some direction or insight, and [12:31] Richard Webbe: And [12:31] Dave Pengelley: maybe that [12:31] Richard Webbe: and [12:32] Dave Pengelley: is not [12:32] Richard Webbe: and [12:32] Dave Pengelley: on LinkedIn. [12:32] Richard Webbe: and the AI people, I think, you know, the LLM platform are going to overtake TikTok. Look, Facebook says they're both, right? But they're all going to be overtaken by clever advertising inside the LLM. So I go to my LLM and I search for information now, and I get some really pointed, clever advertising appears underneath that's right on the money for what I'm looking at, almost like it knows who I am. [13:00] Matt: It's a good uh it's a good uh transition into the Chat GPT business registration. [13:06] Dave Pengelley: Wait, wait, yes, I'm looking that up. And before we do [13:10] Matt: Oh, [13:10] Dave Pengelley: that, [13:10] Matt: I have the links if you want [13:11] Dave Pengelley: that feels like news. [13:19] Dave Pengelley: We're running the news bumper. We're doing it this week. We're [13:22] Matt: to [13:22] Dave Pengelley: having a news bumper. And I believe this is what you're talking about, Matt. Introducing businesses in Chat GPT. Claim your business. So this is going. out for the whole Google business profile stuff or the whole local search thing, right? [13:34] Matt: Yeah, [13:34] Richard Webbe: Absolutely. [13:34] Matt: I mean while while Google continues to fumble, like I guess we'll talk about carbon in a sec, but um why not jump on it? [13:41] Dave Pengelley: Yeah. [13:41] Matt: Like it's already a huge user base. [13:46] Richard Webbe: This is not paid advertising, by the way. [13:49] Dave Pengelley: No, no, this is the news. [13:53] Dave Pengelley: Well, [13:53] Richard Webbe: Oh [13:53] Dave Pengelley: what [13:53] Richard Webbe: yes. [13:54] Dave Pengelley: open [13:54] Richard Webbe: I [13:54] Dave Pengelley: AI [13:54] Richard Webbe: I [13:54] Dave Pengelley: are [13:54] Richard Webbe: am [13:55] Dave Pengelley: not [13:55] Richard Webbe: actually [13:55] Dave Pengelley: sponsoring [13:55] Richard Webbe: talking [13:55] Dave Pengelley: us [13:55] Richard Webbe: to [13:55] Dave Pengelley: yet? [13:55] Richard Webbe: a I'm talking to a sponsor, so [14:00] Matt: Yeah, I've actually just been in Tebow's DMs. [14:03] Richard Webbe: Hello s hello sponsor. Your name could be here right now, so hurry [14:07] Matt: So [14:07] Richard Webbe: up. [14:08] Matt: with with this, like what how do you guys read this? Because I see it pretty much exactly as what you were just saying, Richard, when you are in your chats and then, you know, all of a sudden it says, Well, you know, if you go to this business, which is local, like they do these things. [14:23] Richard Webbe: And and I think the fine line is what you guys were alluding to earlier. You want to get people to your Platform like TV, but you want them to see the advertising, but you don't want the advertising to detract from the platform, and this is where a balancing act comes along, right? So no one's going to watch a TV station that is all advertising. Well, TVSN [14:46] Dave Pengelley: No, [14:47] Richard Webbe: proved [14:47] Dave Pengelley: but [14:47] Richard Webbe: me wrong on that, but [14:50] Dave Pengelley: uh as a as a guy who once upon a time put in a uh uh a showreel to try and get a job on T V S N um didn't didn't go that well. Um, [14:58] Richard Webbe: no [14:58] Dave Pengelley: didn't didn't go anywhere. Uh, I mean, that was before I'd spend like a decade doing actual demonstrations and pre-sales and stuff. I probably have a better chance nowadays, maybe, maybe [15:06] Richard Webbe: Yes. [15:06] Dave Pengelley: not. Uh [15:07] Richard Webbe: Yes. [15:07] Dave Pengelley: anyway, I think where this is, this is less about advertising, and this is around becoming the go-to. This is trying to become the new Google, right? Uh Google do very well with their Google business listings and stuff. And you talk about talk about localized SEO is a whole subcategory than SEO. So SEO is your general web search keyword, engine optimization. But then local SEO is I want a plumber in Bankstown. Right. I I want the specific person with a suburb and then it Google will leverage its Google business listings and whether they're paying for the promoted spot or not, you're gonna get those business listings and the ones that are populated and have reviews and stuff are gonna rise higher just in their rankings because people have just fed Google the information. Like restaurants, like often you'll click in and you can see menus and opening times and things if people have populated their Google business listings. So you don't have to go to their website, you just see it from inside the Google Maps integration or inside just the Google Web Search. I see this as ChatGPT going, sure, we can run around and we can index and turn get the LLM to try and read 55,000 pages on your website to try and interpret what you do. Or how about you just tell us? And that makes it really easy when someone wants a plumber in Banksdown for us to say, Jim Bob's Banksdown Plumbing offer these services and these prices, and this is their call sheet. [16:23] Matt: It's um it's yeah, I think it's two other things as well, which is like one consent. You know, consent to actually promote your thing and to have your information there indexed, plus getting more stuff that they couldn't get from their own scraping, which is obviously then more stuff that they can, you know, sell and distribute. But then the second one is it's kind of like MCP. So it it means that there is this now structured data set that is super easily reasoned across and provided to responses, you know, in the sense that they can instantly find out and from their own Categorization rather than just interpreting what they scrape. [17:03] Dave Pengelley: Yeah, and you know, this whole reassure customers, your profile is managed by the actual business. The whole check mark verification, not hallucinations. Uh, I think that's important as well. And it's, you know, they they get looks from that video ad that we watched. Looks like that you can then chat with the information about that business. [17:19] Matt: Yeah, you can you [17:20] Dave Pengelley: And [17:20] Matt: can have your own agent set up as to answer your own business [17:22] Dave Pengelley: yeah, [17:22] Matt: questions. [17:23] Dave Pengelley: yeah, which I mean that's that's that whole democratization of chatbots and things, which [17:27] Matt: Yep. [17:28] Dave Pengelley: um is massive. Like [17:30] Matt: And [17:30] Dave Pengelley: like [17:31] Matt: then you also can just like if you have a Shopify store, that this is one of the first integrations they have, you just connect it. So if you're you're chatting and you're like, oh, where can I buy this thing? And all of a sudden they find some Shopify results, you can just purchase it there in the thread, like Richard's saying, stay on the platform. [17:48] Richard Webbe: Yeah, [17:48] Dave Pengelley: Yeah, [17:48] Richard Webbe: yeah, [17:49] Dave Pengelley: here [17:49] Richard Webbe: yeah. [17:49] Dave Pengelley: we go. [17:49] Richard Webbe: It's it's [17:49] Dave Pengelley: Best best [17:50] Richard Webbe: it's [17:50] Dave Pengelley: toy [17:50] Richard Webbe: it [17:50] Dave Pengelley: stores [17:50] Richard Webbe: was [17:51] Dave Pengelley: for [17:51] Richard Webbe: successful. [17:51] Dave Pengelley: creative play. Colorful clutter full of making. Colorful clutter's got an integration here, so you can start a chat directly in here with colorful clutter. And then it's like What's a good gift for my six year old niece? Here's the gifts, here's the price sets. Um, yeah. [18:05] Matt: Yeah, it's pretty good. That that same Shopify integration has been in Grockbot as well. Uh they put that in the last couple of weeks, and it it does work very good. Like I haven't purchased anything through it, but in all the demos that I've seen and the you know anecdotal evidence of people on X, like it does look really nice. [18:21] Richard Webbe: Did you guys upgrade your iPhones this week? [18:25] Matt: To what? [18:28] Richard Webbe: Yes, the [18:28] Matt: Oh [18:28] Richard Webbe: twenty seven point Two or whatever [18:30] Matt: Not [18:30] Richard Webbe: it is [18:31] Matt: not [18:31] Richard Webbe: or [18:31] Dave Pengelley: I [18:31] Matt: the [18:31] Richard Webbe: underneath [18:31] Matt: jeweler. [18:31] Dave Pengelley: I I don't know. Yeah. Uh [18:35] Matt: Oh, [18:35] Dave Pengelley: not [18:35] Matt: I didn't [18:35] Dave Pengelley: physically. [18:36] Matt: I didn't [18:36] Dave Pengelley: Um [18:36] Matt: intentionally do it. I'll see [18:38] Richard Webbe: no, [18:38] Matt: if it's [18:38] Richard Webbe: it's [18:38] Matt: done [18:38] Richard Webbe: a [18:38] Matt: it. [18:38] Richard Webbe: software update, and um I got it, and it's changing the way it's more closely bringing AI into Siri with a much more functional integration and a lot of the features we're talking about of banner ads, quicker connection to transactions, quicker connection to people and and businesses. [18:58] Dave Pengelley: I haven't, but I mean I was hearing about some of the new Siri stuff, Richard. And you, I remember you asked about this weeks ago, and I wasn't quick enough at the time to mention this because I had heard about it, but my brain didn't connect the dots. You were asking, like, when are we going to get an AI that's an aggregator across all my different accounts and calendars and emails [19:15] Richard Webbe: Yes. [19:16] Dave Pengelley: and messages? And I think I think Billy sent me an email about an event, but I can't find it. Uh, where was that? Siri wants like Apple Intelligence wants to be that, where it's plugged into your messaging and your WhatsApp. WhatsApp and your emails, and we'll be able to search across all your messaging surfaces [19:34] Richard Webbe: I [19:34] Dave Pengelley: to [19:35] Richard Webbe: agree. [19:36] Dave Pengelley: bring that up. Mind you, Grokbot and Dot and everyone else is trying to do that as well. Like Grockbot lets you um connect straight to iOS uh applications now as well. Um they're [19:48] Matt: Yeah. [19:49] Dave Pengelley: all heading into that territory. [19:51] Matt: The the that kind of like level of just internec interconnectivity and um just clean, smooth integration is such a Crazy quality of life improvement when you when you realize. And when you sort of go back to a scenario without it, all of a sudden you're you're sort of stuck with this weird situation where you're like, how did we survive with this before? Yeah. Like, I'm not sure if you guys have experienced this yet, or if you sort of have like consciously been aware of any of that stuff yet. Like, Richard, [20:19] Richard Webbe: I [20:19] Matt: when you [20:19] Richard Webbe: have. [20:19] Matt: came across yeah, with this update, Richard, what like what have you found so far? [20:23] Richard Webbe: Well, [20:24] Matt: Like [20:24] Richard Webbe: um, [20:24] Matt: what's [20:24] Richard Webbe: Siri's completely changed now and is a lot more AI L L M pointed. Um it doesn't there's those and here's the user interface, they change it a lot, which could be a win and a problem. So Siri now acts differently. I couldn't work out how to get out of one part so I could just ring someone. Uh how do I connect to someone? Am I clicking on this thing here? Why does that? So there'll be a a period of oh bugger, I hate this now, but that'll that but that'll go away. Um and it's really important, but I'm I am enjoying Both within the application and within the phone. And it's the whole story. Who owns the customer? Is it Facebook? Is it Apple? Is it LinkedIn? Do you know what they're all trying to do that? But I'm enjoying that macro AI that's automatically applied on the user interface. So I can now look at my phone when I wake up in the morning and just the notifications has already abbreviated everything and say, Oh, Richard, you got an email from Dave. He said, you know, you've got to get on early to the podcast today. Matt said he's coming down in a few months. And if just so [21:31] Dave Pengelley: Yeah. [21:31] Richard Webbe: productivity-wise I'm I'm really enjoying the journey but it's like we're saying who's going to own that journey is it Facebook Apple LinkedIn Google AI podcasters group operators group [21:44] Dave Pengelley: And it's it's also the the cross plot platform capability as well. Uh where Apple is very Apple and if you're all in there on the Apple ecosystem, that's really good. But like even like iMessage, for instance, right? My daughter need I had to sign up for some fundraiser thing for her girl guides. today and I did that on my Windows PC. And then I had a hyperlink. And she's like, how can I get that so I can send it to people on iMessage that she's got like a lockdown iMessage on her iPad? And I'm like, okay, remote desktop into my Mac Mini, which has iMessage, which I can then directly message her iMessage through there. Like whereas you know WhatsApp is very easily cross-platform. Even from the and then if you go back to the AI and the LLM point of view, the GrocBots and the chat GPT dots and etc are all becoming very cross-platform friendly. Matt, you talked about having your remote server and being able to offload jobs to that. At the moment, from my Chat GPT app on my phone, I've got codec sessions on my PC and on my Mac that I can run different codec sessions easily from. You've just been able to give your dot specifically access to use your PC for computer use. [22:50] Matt: Yeah. [22:51] Dave Pengelley: And so I can actually get Hamilton to load up and use a PC separately from the actual codec stuff. [22:55] Matt: Remotch [22:55] Dave Pengelley: Grockbots [22:56] Matt: two. [22:56] Dave Pengelley: across all my different devices, so I can easily be messy. Grockbot on my phone and then jump onto either of my different desktops and then click on the link to test the thing that are built for me. It's just not caring where I am, but having access to everything, no matter which thing I'm talking to, it is a productivity unlock. [23:14] Richard Webbe: I [23:15] Matt: Yeah, [23:15] Richard Webbe: I [23:15] Matt: she's [23:15] Richard Webbe: actually wrote [23:15] Matt: a good idea. [23:16] Richard Webbe: a white paper a while ago, not one that was published, but it was a strategy that your face becomes a critical component to your security, and that everyone's computer screen is available for you to. access all your cloud messages, services, and email. And the concept for me was I'm over at Matt's place. He's got a computer. Oh Matt, I got an email. He goes, yeah, just look at my computer and a window opens up with all my world. Just because it's his computer, but I can access it. Do you know what I mean? Although I I used to think about the um the internet connected fridge, which probably got more attention than it needed all over the world, the fact that someone released a fridge with an internet browser on it. But again, imagine if Multiple devices across the world just recognized your face, that's your security, and you're immediately into your email, your messages, your emergency notes, everything. It's going to change the way we communicate. Communicate profoundly and make it a lot quicker. [24:16] Matt: How like it's terrifying though because it makes me think of every sci-fi movie ever where they copy someone's likeness. [24:21] Richard Webbe: Yes. [24:22] Matt: And like, yeah, there there needs to be some sort of like crazy unique bio or like identifier that can't just be face. Um I I feel like even that that biomarker could probably even be spoofed. Like uh it it blows my mind, especially when we get like full on compute uh quantum compute stuff. [24:42] Richard Webbe: I agree. [24:44] Dave Pengelley: Oh, [24:44] Richard Webbe: I [24:44] Dave Pengelley: I'm [24:44] Richard Webbe: agree. [24:44] Dave Pengelley: not gonna [24:45] Richard Webbe: And [24:45] Dave Pengelley: get [24:45] Richard Webbe: that's [24:45] Dave Pengelley: it. [24:45] Richard Webbe: what made the Mission Impossible series that came out, believe it in the late [24:49] Dave Pengelley: Sixties. [24:49] Richard Webbe: sixties, early seventies, yeah. About changing the face, it changed the way spy shows were seen, because they had this weird technology where they could change their face. I even even uh Tom Cruise used it even again [25:02] Dave Pengelley: Oh, [25:03] Richard Webbe: initially. [25:03] Dave Pengelley: a lot. They used it a lot in some of those movies. Almost too much was like every person had a robber face. Like you never knew who was who. Which is great. Great, great, great for story too. I love the mission to possible films. Um, this was a video from 15 years ago, 2011. I don't know if you guys ever saw this video that talks about this particular concept world. I don't know if you ever saw this one. [25:28] Dave Pengelley: Is that too loud? [25:32] Dave Pengelley: This is before Alexa devices. This is before AI. [25:42] Richard Webbe: Pointing obviously the very large glass company. [25:50] Richard Webbe: I know the Australian MD. [25:53] Matt: Yeah, is that a little bit news? [25:55] Richard Webbe: Ah, all interactive touch screen. Okay. [26:00] Richard Webbe: Yeah, that's it. That and that's the concept. That's exactly what it is. [26:04] Dave Pengelley: Like wherever you are, every surface can be an interface. Um [26:13] Richard Webbe: Yeah, [26:13] Dave Pengelley: bench [26:13] Richard Webbe: that's [26:13] Dave Pengelley: tops. [26:14] Richard Webbe: it. [26:14] Dave Pengelley: Like look at that, isn't that? [26:17] Dave Pengelley: But they did a follow up one as well. Here's the kids. [26:22] Dave Pengelley: I don't know, I don't want to be looking down at tables the whole time. Oh, but interactive car dashboards very cool. I think Polestar actually have some of their Interfaces are a bit like this. [26:31] Richard Webbe: Yeah. [26:32] Dave Pengelley: Like Pulsar release skin kits where you can make your um your car look like kit uh or back to the futures DeLorean with all those sort of displays on it. [26:41] Richard Webbe: Yes, yes. And I saw someone demonstrating the video pass-through on the other side of a truck or a bus, so that as the bus has cameras on the other side, it projects the image so the car looks like it's see-through. [26:58] Dave Pengelley: Yeah, all that the NFC type stuff. Devices to things and pulling the data off one thing. This I love this. Oh, put your phone down and becomes a display. [27:11] Dave Pengelley: Yeah. [27:12] Matt: Do you think that like that is more of a likely situation than the augmented reality glasses? [27:20] Dave Pengelley: Yeah, this will take off because but no one wants to wear things on their faces all the time and have little things and worry about headaches and visual focal links and everything. Like [27:30] Richard Webbe: Want privacy. [27:31] Dave Pengelley: Yeah. [27:33] Matt: Your algorithm. [27:35] Dave Pengelley: I mean, did [27:35] Richard Webbe: Someone. [27:36] Dave Pengelley: you guys ever see The Jerk with Steve Martin? [27:38] Richard Webbe: Yeah. Greatest [27:39] Dave Pengelley: Have you [27:40] Richard Webbe: film [27:40] Dave Pengelley: seen that [27:40] Richard Webbe: ever. [27:40] Dave Pengelley: movie, Matt? [27:41] Matt: Oh no, I thought you were [27:42] Richard Webbe: Matt's [27:42] Matt: referring to someone. [27:43] Richard Webbe: way too young. You thought he thought you were talking to me. [27:45] Dave Pengelley: I've heard of this cat juggling. [27:51] Dave Pengelley: My wife and I quote that movie all the time. [27:53] Richard Webbe: Stay away [27:53] Dave Pengelley: Uh [27:53] Richard Webbe: from those [27:53] Dave Pengelley: anyway, [27:54] Richard Webbe: petrol cans. [27:54] Dave Pengelley: but [27:54] Richard Webbe: The petrol cans are trying to kill me. [27:56] Dave Pengelley: Yeah, it's like when he goes up to the water cooler of wine and he goes In just a few months, we've achieved a level of class. Some people don't get it in a lifetime. [28:06] Richard Webbe: Sorry guys, we're [28:07] Dave Pengelley: But [28:07] Richard Webbe: off [28:07] Dave Pengelley: anyway, [28:07] Richard Webbe: tracking. [28:08] Dave Pengelley: no, no, the whole point of that is that the he got rich from these glasses that um had a little nose attachment or something, and it made people go cross-eyed because they were staring at the thing in front of them so closely that it gave everyone these eye defects and he all got sued and lost his money. Um, sorry, spoilers, but it's a four forty-something year old movie. So I think the the statute's gone. Um But yeah, I worry about that with these augmented reality things. If all of a sudden you've got this thing this far from your face and your eyes are our our brains, even if it's clever projections and optics, I don't see that our uh we were wired for looking at something this close to our eye all [28:44] Richard Webbe: No, [28:44] Dave Pengelley: the [28:44] Richard Webbe: and [28:44] Dave Pengelley: time. [28:44] Richard Webbe: and some of them I tried on one the other day um uh from my friend Malcolm Atlanta's McKinsey's consultant, if you're listening, Malcolm. And um he has it so it's like you're looking at a 120-inch screen. So even though it's this far, you think it's this big screen, and as you move your head. It gets it there. And you can put your hand up and start to in front of it. I mean, the the the I think that's the next big change for humanity is how we interface with our information. It started off as the written word, then it became the typed word and the printed word. Now it's going to be just objects and our hands and our eyes interacting with these environments to try and get the answers or be productive as we can be. It's gonna be very exciting. [29:30] Dave Pengelley: It's yeah, I think I think as they get better at like photonic physics um and you know like small scale holography and other other things, we'll see just more of that ambient stuff, a bit like this this world, this day of glass stuff versus putting giant chunky things on our faces. Because I think you know the glassware where it's this close, I think is you know potentially dangerous. I think the quote the oculuses and the the Apple vision stuff where it's you know at least sort of five four inches from your face where the little screens are, but then you've got the whole neck and Weight issues of this whole apparatus on [30:02] Richard Webbe: I'm [30:02] Dave Pengelley: your face. [30:03] Richard Webbe: I'm a little bit older than you guys. So I had my lenses in my eyes replaced a couple of years ago. And uh [30:10] Dave Pengelley: LASIK? [30:11] Richard Webbe: sorry? [30:11] Dave Pengelley: Let [30:12] Richard Webbe: No, [30:12] Dave Pengelley: little the [30:12] Richard Webbe: no, [30:12] Dave Pengelley: LASIK. [30:12] Richard Webbe: no. This is completely different. So a lot of people, I didn't know about it at the time. So when you get older and your eyes are playing up, um, if you've got cataracts, and that is that uh sepia tone color that goes across your your lens of your eye, sun exposure or whatever, um you start to lose the ability. To focus and colour and everything goes a bit so literally, and it sounds horrible, and I I but it is the simplest surgery I've ever had. They give you a bit of relaxation and they put a little tube in the corner of your eye and they vacuum out your lens. And you can actually see the little pole as it and it doesn't hurt. And then in inside another little tube is a wrapped-up new lens, and they slide it in, and as they pull it out, it's got little springy wings on it, and it slots into space and it can. It cures your entire vision issue. Now, I still have to have some reading glasses for up close. The new ones, you don't need that. They're concentric. So your brain adjusts when you go forward. Now, they're talking about the new glasses with AI, you put them on and they adjust to where you are, right? Imagine that photoelectric behavior in my eye, in that lens, which is totally conceivable with today's technology. [31:30] Dave Pengelley: Yeah. [31:31] Richard Webbe: And imagine the possibilities then as my eyes age, I can't see it, or I want to be the six million dollar man. I'm looking for something I can't see it, and I just tell my brain, would you do a super thousand-bit, you know? Oh, look, I can see man on the moon. You know, [31:46] Dave Pengelley: Yeah. [31:46] Richard Webbe: it's just going [31:47] Dave Pengelley: And [31:47] Richard Webbe: to [31:47] Dave Pengelley: I [31:48] Richard Webbe: be [31:48] Dave Pengelley: think [31:48] Richard Webbe: amazing. [31:48] Dave Pengelley: that's different. I think I think if you're changing the optics of the light into the into the cornea is different than focusing on something an inch from [31:55] Richard Webbe: It's [31:55] Dave Pengelley: your [31:56] Richard Webbe: true. [31:56] Dave Pengelley: face. [31:56] Richard Webbe: That's [31:56] Dave Pengelley: And I think [31:57] Richard Webbe: true. [31:57] Dave Pengelley: I think that's the difference because if you're trying to look at something close to your face. Just the eye muscles and the way they're stressed. Whereas if it's just lights coming in from all the same distant bouncing sources ambiently everywhere, but you know, you've got different mechanics, which is what eyeglasses do, right? You're not staring at the glass surface [32:12] Richard Webbe: No, [32:12] Dave Pengelley: on your reading [32:13] Richard Webbe: no, [32:13] Dave Pengelley: glasses, [32:13] Richard Webbe: no. [32:13] Dave Pengelley: you're [32:14] Richard Webbe: But [32:14] Dave Pengelley: staring [32:14] Richard Webbe: what I'm [32:14] Dave Pengelley: through [32:14] Richard Webbe: saying [32:14] Dave Pengelley: that. [32:15] Richard Webbe: is Elon Musk style, Elon's [32:17] Dave Pengelley: Yeah. [32:17] Richard Webbe: got people now who are moving parts of their bodies they couldn't because he put things in their brain, he's accelerating. What if everyone at a very young age gets an electronic lens put in both [32:28] Dave Pengelley: No, [32:28] Richard Webbe: eyes? [32:29] Dave Pengelley: but now we're going to transfer. Humanism. [32:30] Richard Webbe: They'll never [32:30] Dave Pengelley: This is this [32:31] Richard Webbe: need [32:31] Dave Pengelley: is [32:31] Richard Webbe: glasses. Um, but what will be dangerous is you know some governments will start transmitting messages to our lenses to convince us of certain things. [32:41] Dave Pengelley: Yeah, I saw I saw the episode of Doctor Who where everyone had the little Bluetooth headset and then that transformed more into Cybermen. [32:47] Richard Webbe: Yep. [32:48] Dave Pengelley: Bad. [32:48] Richard Webbe: Now very importantly, um Anthony Albanese in our news thing, and this [32:52] Dave Pengelley: This [32:53] Richard Webbe: is [32:53] Dave Pengelley: is [32:53] Richard Webbe: not [32:53] Dave Pengelley: not [32:53] Richard Webbe: political. [32:53] Dave Pengelley: a political podcast. [32:54] Richard Webbe: It's not a political, he has announced that they're going government's going to get on the front foot in Australia. And start to regulate, control, and comply AI to protect us. Now, no matter what side of politics you're on, that's probably a good thing. But what are your thoughts on that, guys? [33:10] Dave Pengelley: This is not a political podcast. [33:12] Richard Webbe: Yeah, well it's not about the government. It's [33:14] Dave Pengelley: No, no, I mean [33:14] Richard Webbe: about. [33:15] Dave Pengelley: I th I think I think we've got to be wary of of government regulation and and [33:18] Richard Webbe: Yeah. [33:18] Dave Pengelley: I think there are things that they should have some, you know, say in, but there's a lot of things they shouldn't, and some of the things they're saying they should have a say in, I don't think they should. I mean a lot of the Judicare stuff is massive overreach around freedom of ability to communicate, but [33:34] Matt: So [33:35] Dave Pengelley: that's I'm wary because it's very easy to become politically sided on this topic. Um I don't like I [33:43] Richard Webbe: It's [33:43] Dave Pengelley: and [33:43] Richard Webbe: bizarre. [33:43] Dave Pengelley: I think I [33:44] Richard Webbe: It's bizarre. [33:44] Dave Pengelley: I think on [33:44] Richard Webbe: I [33:45] Dave Pengelley: the top [33:45] Richard Webbe: find [33:45] Dave Pengelley: on the [33:45] Richard Webbe: it. [33:45] Dave Pengelley: idea of government regulation, I think where it comes into play to try and segue you into something a little bit more that we like less contentious [33:53] Matt: we can [33:53] Dave Pengelley: than [33:54] Matt: just [33:54] Dave Pengelley: Australian politics right now is [33:56] Matt: do that. [33:56] Dave Pengelley: American politics. Um I mean we we talked About um uh Jensen's Huangate, um or [34:03] Richard Webbe: Yeah. [34:04] Dave Pengelley: Jensen Hu Huang's Huang's Wangate, whatever you want to call that. Um, [34:09] Richard Webbe: Cave place, [34:10] Dave Pengelley: which [34:10] Richard Webbe: I like [34:10] Dave Pengelley: the [34:10] Richard Webbe: going. [34:11] Dave Pengelley: and the issue is national security in that instance. Is he selling the super fast black world chips or whatever to a foreign nation state that doesn't necessarily have, like in that instance, the United States best interests at heart? That becomes a discussion for you know that kind of regulatory. [34:30] Richard Webbe: Even [34:30] Dave Pengelley: Stuff. [34:30] Richard Webbe: worse, is she stealing secrets over pillow talk? I mean, it's rather dangerous because they're at the forefront. Um, the interesting thing about the regulation side of things, and it's it's not a political angle, is you know, you look at the forces, we've got the opportunity with AI to improve the lives of humanity worldwide. But of course, humans are defaulting to the arguments rather than the positives. And that's upsetting me at the moment, right? Oh, not enough water. What a load of rubbish, right? We've Got technology that can super I I heard them on the radio you say talking about and just splinter groups. We don't have enough water in Australia for data. That is a load of pigwassel rubbish, right? There's new types of cooling systems that we can use that will solve this problem easily. It's not about water. And then of course, we don't have enough power. Well, you know, go out in the desert and put up a lot of solar panels or a nuclear. I mean, private companies around the world are building nuclear power stations. So we shouldn't be pushed. So What does worry me, it's it's like humans seem to snatch uh defeat from the drawers of victory when they get all worried about these things by putting too much now. They [35:40] Dave Pengelley: Well, [35:41] Richard Webbe: should have some compliance in, right? [35:42] Dave Pengelley: yeah. [35:43] Richard Webbe: Deep [35:43] Dave Pengelley: Look. [35:43] Richard Webbe: fakes and all that sort of stuff, and cyber is important, but it does worry me the way that's all going. [35:48] Dave Pengelley: I mean, you you said you're you're pro the regulation stuff. I think one of the issues is we have a lot of short term political wrangling here in this country and not long [35:57] Richard Webbe: Yes. [35:57] Dave Pengelley: term vision, and it's all based [35:58] Richard Webbe: Yep. [35:58] Dave Pengelley: around fear. Everything [36:00] Richard Webbe: Yeah. [36:00] Dave Pengelley: we're seeing is very fear-based. Let us do this for your safety. We've been through that numerous times in the last half a decade alone. Didn't always work out particularly well because they went for the short term, quick fix, band-aid. We're here for your safety. Like I'm from the government, I'm here to help. And I think some of this AI fear mongering is exactly the same. [36:20] Richard Webbe: I [36:20] Dave Pengelley: Um, [36:20] Richard Webbe: agree. [36:21] Dave Pengelley: and and you know, I was having this chat with someone yesterday, and the whole water thing came up. And uh the biggest issue is NIMBYism. No one wants it in their bank. Backyard. Um, I don't want a data center in my backyard. And because of the makes the power prices go up. And I was like, well, maybe we shouldn't have shut down all that coal power. Uh [36:38] Richard Webbe: Yes. [36:38] Dave Pengelley: and then that person, that person, that same person then said, Why don't we just have nuclear? I'm like, why don't we just have nuclear? Exactly. We could um make coal is even cheaper. We [36:48] Richard Webbe: Europe [36:48] Dave Pengelley: could just [36:49] Richard Webbe: Europe has reclassified nuclear as clean energy. [36:53] Dave Pengelley: yeah. So I mean that and that eliminates the the whole power cost issue. Like, oh residents don't want their Power costs going up because of data centers. Okay, well, let's build better power infrastructure. That's not a hard problem. [37:06] Richard Webbe: Yeah, [37:06] Dave Pengelley: And it doesn't [37:07] Richard Webbe: yeah, [37:07] Dave Pengelley: need [37:07] Richard Webbe: yeah. [37:07] Dave Pengelley: a billion solar panels crushing our farmland. Um, or you know, Mr. Cannon Brooks, who wanted to put a billion kilojoules of lithium ion batteries that are known for thermal runaway in the middle of Alice Springs. [37:20] Richard Webbe: And it's not a political pod pass, but that was a very good point there, Dave. [37:25] Dave Pengelley: Not a political podcast. [37:29] Dave Pengelley: Um Unfortunately, all these things have everyone who wants to have opinions on politics. I mean, the for the all even, you know, the whole trying to rename it superintelligence is just ridiculous. Like just politicians want to have a say on things. And whether you agree with or disagree with them, um, what is their motive? What is their intention? Is it actual for the good of the people or is it the good of getting re-elected? [37:51] Richard Webbe: Can I ask Matt? Matt, what do you think of runaway AI? [37:56] Matt: What do you mean by that? [37:57] Richard Webbe: Well, all of these technologies and Infrastructures to deliver us groundbreaking improvements. Like, you know, I use the example, and someone someone else told it back to me the other day that we can now isolate uh interimetriosis in about 14 seconds when it used to be 14 months. I mean, these things are starting to happen daily. That's the positive side, but we have so much just what do you think? [38:20] Matt: What what do you actually mean by runaway [38:22] Dave Pengelley: Did [38:22] Matt: AI [38:22] Dave Pengelley: you mean [38:23] Matt: though? [38:23] Dave Pengelley: like the recursive self learning stuff? [38:25] Richard Webbe: Yeah, [38:25] Dave Pengelley: So, like [38:25] Richard Webbe: the fact that AI is getting smarter and moving ahead and everyone's trying to slow it. Damn. [38:32] Matt: Um [38:34] Matt: so I don't think it really matters what I think. Um but [38:39] Richard Webbe: Yeah. [38:41] Matt: I okay, so obviously technology increases and uh benefits to humanity and all of those things are the positives. They are obviously the the the marketed reasons why we need it, why we should have it. You know, the the best way to analyze any individual. thing is to look at something else in parallel, you know, an analogy. Like try and tell somebody that's never had or owned a phone before why they need to have an iPhone. You know, should [39:10] Richard Webbe: Right [39:10] Matt: why why [39:10] Richard Webbe: comment. [39:11] Matt: should we why should we sell this thing? Why should we manufacture it? Like [39:14] Richard Webbe: Yeah. [39:14] Matt: what's the point? It's useless. Like that [39:17] Richard Webbe: Very good. [39:17] Matt: that [39:18] Richard Webbe: No, I like that. [39:19] Matt: that's my argument. And all and all the all the political talk like uh I don't I don't care about any of it because nobody ever actually speaks the truth. Nobody understands [39:29] Richard Webbe: Very good. [39:30] Matt: Things to the same degree that they should understand it in order to actually have an applicable opinion. Um [39:37] Richard Webbe: Yeah. [39:38] Matt: I I don't know, don't meet don't put me in power because I'd instantly become some sort of tyrant. Um [39:43] Richard Webbe: No, I like [39:44] Matt: but [39:44] Richard Webbe: the way you're talking. You like the benevolent dictator who's got [39:47] Matt: I [39:47] Richard Webbe: some [39:47] Matt: I [39:47] Richard Webbe: common [39:47] Matt: just [39:47] Richard Webbe: sense. [39:47] Matt: don't think anyone ever actually truly understands anything that they say ever. And it's very frustrating. Like, you know, yeah, I I try not to get involved in any of that discourse because it just makes me angry. [40:00] Richard Webbe: AI company that I've been doing some work with on the way down here before I got back in for the podcast. And they were being asked for their um uh the essential age, you know, the security tick box that you get through yourself, or you can buy a package with Microsoft. And there's a new Microsoft security score that comes in relating to AI and stuff. I think it's and the government was saying, We need you before you can provide us, we need you to have a score above 85. And they're saying, but the funny thing is the government's only Score was forty-two. Which is bizarre, isn't it? [40:34] Matt: Yeah. [40:34] Richard Webbe: That they ask more of us than they do of themselves. [40:37] Matt: So like Dave, you said before, like what's the motive? Like everything is data and everything is control. Everything is bending your own reality to the version that they want you to experience. That that is the case. And you know, that could be both a positive and a negative. You think about a therapist, you know, a therapist is trying to bend your own reality into the better version. That you need to realize that's all they're doing. [41:03] Richard Webbe: What [41:03] Matt: You [41:03] Richard Webbe: a [41:03] Matt: know [41:03] Richard Webbe: great statement. [41:04] Matt: that [41:04] Richard Webbe: I like [41:05] Matt: so [41:05] Richard Webbe: that. [41:05] Matt: and that's all marketing is. That's all propag propaganda is. So [41:09] Richard Webbe: Yeah. [41:09] Matt: the only thing I will say here is the same thing that I've said since the very beginning of this show is think about your cognitive security. Where are you being too gullible? [41:18] Richard Webbe: Yeah, I like that. [41:21] Richard Webbe: As you two know, I'm very gullible, but I like that. I'm gonna hang [41:24] Dave Pengelley: It's [41:24] Richard Webbe: on to that. [41:25] Dave Pengelley: I've um as you guys know, I've been been building games and so I've been watching a bit of game uh content on On YouTube and its algorithm suggests things. I saw this guy who is a bit like me. He's like similar age and he's using AI to build games and he just had a massive rant because people are like railing on him going, AI and game can't use AI to make games, it's a games. And there there is a lot of contention around, you know, a bunch of crap slop games being published, the whole one shot thing, every AI demo we talked about. This AI demo uh the dev day thing, open AI. Let's make a little rocket ship game. Like that's a core demo. And I'm like, is does everyone make game demos because we're all Nerds and we like video games or just because it looks flashy and it's interesting. But from a gameplay point of view, and I can t attest to this, even I'm just trying to build like a simple little casual game, not even my more complex strategy games. And I think I'm up to version twenty two in forty eight hours with Grockbot. Just iterating, changing, [42:17] Richard Webbe: Wow. [42:18] Dave Pengelley: fitting things, changing things, getting it right, making it [42:20] Richard Webbe: Wow. [42:20] Dave Pengelley: work, making it feel good, adding features 'cause I go actually that's not really a good game, or would be better if we did this. And like if you just one shot a game, you're getting really poor results. Anyway. This guy was going, people that are saying like that AI and the data centers and you're killing the others like, do you use a cell phone? Do you drive a car? Where did the [42:36] Richard Webbe: Yeah. [42:36] Dave Pengelley: fuel come from? Where did the rubber come from? Where do the materials come from? Is like, unless you're like an Amish person living completely off the land and you hand build everything, then I don't need to listen to your feedback about AI is killing the earth because you are doing so much destruction with everything you use and do in your life as well. I am no different. [42:55] Richard Webbe: Yeah, it's [42:55] Dave Pengelley: Um [42:55] Richard Webbe: true. [42:56] Dave Pengelley: it's [42:56] Richard Webbe: It's [42:57] Dave Pengelley: like [42:57] Richard Webbe: true. [42:57] Dave Pengelley: did you ever watch the good place with um Kristen [42:59] Richard Webbe: Yes. [42:59] Dave Pengelley: Bell? And [43:00] Richard Webbe: Yes. [43:00] Dave Pengelley: uh Ted Danson. [43:02] Richard Webbe: Yep. I like that [43:03] Dave Pengelley: Um [43:03] Richard Webbe: shot. [43:04] Dave Pengelley: yeah, there was there was a bit, and hopefully this again, it's very old and um it's a good show, and I don't want to spoiler it too heavily, but in trying to find new good people to go in the good place that were trying to find who the best people on earth were, and the and it was almost impossible because the scoring system had all these second, third, fourth order effects. And so picking a flower had like this this fourth, fifth order consequence that the person had no idea about, but that then would Mark them as having done a bad thing for picking a flower out of their own garden or something. Like they just become this overly convoluted [43:36] Richard Webbe: The butterfly [43:36] Dave Pengelley: mess of things. [43:37] Richard Webbe: effects in [43:38] Dave Pengelley: Yeah. [43:38] Richard Webbe: real [43:38] Dave Pengelley: And [43:38] Richard Webbe: time. [43:38] Dave Pengelley: and so because our world is is so interconnected and complex in the modern world, and everything we have is built on the previous things that the previous people built, that the previous people built, and this sort of industrialized revolution world that we live in, there is so much going on to say that AI is the bad guy, and it's funny, you're like all the narratives change. Again, not a political podcast. But we've gone in the last few years from the whole it was all climate change and AI and carbon and reducing carbon usage to now it's uh the AI data centers and the water usage. And there's always a new bad guy, there's always a new gremlin, there's always a new bogey to chase, to blame for why things are bad, to you know, set up that class division of the haves and the have-nots. And it's all political messaging for some individuals to get what they want over the top of other people. [44:28] Richard Webbe: Okay, [44:28] Matt: I don't [44:28] Richard Webbe: I [44:28] Matt: understand. [44:28] Richard Webbe: have an economic question. Right? But I think you two could answer quite well. So when search came out, um people paid for search inside their offices, mainly because they wanted to search for their own information, right? Inside it. And then it it expanded. And of course, we get search for free today, but big search engines are still paid for by large corporates. But now, of course, that's changing because AI is the new search, right? It's the digital transformation search on steroids. And I notice just with my Apple phone, I'm getting free AI now that I've upgraded my phone through Siri. It's it's going in. How long before we all stop as individuals playing for LLM and the corporate world is financing our AI stuff and that model changes? What do you guys think about [45:21] Matt: You [45:21] Richard Webbe: that? [45:21] Matt: you can entirely use free AI forever right now. [45:25] Dave Pengelley: Yeah. [45:26] Matt: There's [45:26] Dave Pengelley: But most [45:26] Matt: nothing. [45:26] Dave Pengelley: people [45:26] Richard Webbe: But [45:27] Dave Pengelley: most people are. We [45:27] Richard Webbe: there's [45:28] Dave Pengelley: we're actually [45:28] Richard Webbe: still [45:28] Dave Pengelley: in the in [45:29] Richard Webbe: restrictions. [45:29] Dave Pengelley: the [45:30] Richard Webbe: Are we paying for [45:31] Dave Pengelley: We [45:31] Richard Webbe: it? [45:31] Dave Pengelley: we are, we are. Um, we we're complete weirdos for paying for it. Most people are not paying for it. Um, they're using the free version. Chat GPT gives you a lot of usage for free [45:40] Richard Webbe: It [45:40] Dave Pengelley: uh [45:40] Richard Webbe: does. [45:41] Dave Pengelley: without paying for it. I think Claude gives you some, Perplexity gives you some, Gemini gives you a bunch for free. And they change it and they tweak it. And depending on what the training data they want on any given day, changes how much they do and don't give you. [45:54] Richard Webbe: Yep. [45:54] Dave Pengelley: Uh, I've I've got friends that are building and coding and and doing stuff using open routers free. models and and they you know they just switch between the free models and when they do hit a limit they go all right oh this is not my core job anyway I'm gonna go do some other tasks and I'll come back when I've got some limits back and they're just hopping between free models to build stuff because that works for them. [46:15] Matt: Richard, your question, like I hundred percent agree with the premise of what you're actually asking there, and that's when does the technology become efficient enough to just not be relevant with the unit economics. Like that is very Very true, and it's very, very important for people to realize because yeah, at this at this point in time, you can't just operate Opus 5.5 for free. Like it's not possible. But you know, how many years away is it going to be where we walk up to our browser on the fridge and we have Opus 5.5 for free? Like that, that I believe is your question. And yeah, like the free models right now, certain companies, like if you go and get an open code go subscription, um, that's 10 bucks. And it's not now free because you've paid $10, but you can through that use their free service of the Deep Seek 4.1 flash, which is actually [47:09] Richard Webbe: Yeah. [47:10] Matt: an extremely capable model. And like that's insane. Like and and to anyone who thinks that like they need to go and buy the the $200 cord subscriptions to go and actually build something, you don't. Like you just need to put a little bit more effort into it. And you can absolutely get almost anything done with those those models. And people freak out. about like oh but it's it's the Chinese one you know that those kinds of comments but the problem that they they forget is that because it's open source you can have anyone actually host this for you and [47:42] Richard Webbe: Yeah. [47:42] Matt: including yourself if you want to purchase the hardware so you can find inference in your own country on a ZDR endpoint where you're using a free model like that's [47:51] Richard Webbe: Yep. [47:51] Matt: what we have [47:52] Dave Pengelley: Yeah. [47:52] Matt: right now. [47:53] Dave Pengelley: And [47:53] Richard Webbe: And [47:53] Dave Pengelley: and [47:53] Richard Webbe: that [47:53] Dave Pengelley: I mean [47:54] Richard Webbe: makes [47:54] Dave Pengelley: some some [47:54] Richard Webbe: you [47:54] Dave Pengelley: of [47:54] Richard Webbe: two. [47:54] Dave Pengelley: the concern is is what are the values underlying on the Chinese models and what do they put them in? I saw, I think, Noose published. uh like [48:00] Matt: Yeah. [48:00] Dave Pengelley: a quen that they retrained on western values [48:03] Matt: Yeah, yeah, [48:03] Dave Pengelley: so [48:03] Matt: yeah. [48:04] Dave Pengelley: they [48:04] Matt: That's [48:04] Dave Pengelley: they re-weighted [48:04] Matt: insane. [48:05] Dave Pengelley: it [48:06] Matt: Yeah, I actually [48:07] Richard Webbe: Wow [48:07] Matt: I bookmarked that. I don't know if I still [48:09] Richard Webbe: Wow. [48:09] Matt: have [48:09] Dave Pengelley: um [48:09] Matt: that [48:10] Dave Pengelley: but [48:10] Matt: answer. [48:10] Dave Pengelley: but [48:10] Richard Webbe: That [48:10] Dave Pengelley: also [48:10] Richard Webbe: brings [48:10] Dave Pengelley: remember [48:10] Richard Webbe: a new that brings a new inference to cognitive bias. [48:14] Dave Pengelley: but remember also these models all trained on English so regardless which country they come from they're actually absorbing all the English knowledge and other like primarily if you're asking English questions it's probably going to be feeding from an English training set. Not [48:25] Richard Webbe: Wow. [48:25] Dave Pengelley: that these models care. I mean If fed a model French I'm sure it could work it out but [48:30] Matt: Yeah, I've got that link, Dave. Um [48:32] Dave Pengelley: Yeah. [48:33] Matt: I I I find that fascinating. And and to f well, it it's even though it asks, like, okay, so along with cognitive security, you know, this is something that I've learned from Lauren Potato like P Stack, attack the premise. Like try and figure out the premise and attack it always. Like contrarian beliefs. Like so what what is Noose actually checking here? You know, they're saying, oh, okay, so it's um yeah, yeah, what does that mean? [48:58] Dave Pengelley: Well, [48:58] Matt: And and [48:59] Dave Pengelley: C C P [49:00] Matt: But like what is the actual thing that they're checking? [49:02] Dave Pengelley: Yeah, [49:02] Matt: So [49:02] Dave Pengelley: yeah. [49:03] Matt: I kind [49:03] Dave Pengelley: Oh [49:03] Matt: of want to go have a look into that and dig through it more because you know that [49:07] Dave Pengelley: well they [49:07] Matt: that's [49:08] Dave Pengelley: they they say here the the seven value categories including freedom of expression, democracy, and privacy. Imagine how often a model adopts China's party line. And [49:16] Matt: yeah, [49:16] Dave Pengelley: so [49:16] Matt: I [49:16] Dave Pengelley: then [49:16] Matt: mean, [49:17] Dave Pengelley: they [49:17] Matt: which is a nice sentence, but yeah, like actually concrete mechanically, what does that mean? So [49:22] Richard Webbe: Yeah. [49:23] Matt: that's the kind of stuff that um I think is worth knowing because yeah, you read that and you're like, oh wow, well I'm never gonna use the thing that's that's not. Not aligned to Western values, because I'm a Westerner. [49:37] Dave Pengelley: But [49:37] Richard Webbe: Not [49:37] Dave Pengelley: but but [49:37] Richard Webbe: a [49:37] Dave Pengelley: how [49:38] Richard Webbe: political [49:38] Dave Pengelley: much values [49:38] Richard Webbe: popcorn. [49:38] Dave Pengelley: are in your C code or your JavaScript? Like [49:41] Matt: Right. [49:41] Dave Pengelley: that that's the other thing, right? Like horses for courses. If you're getting this to write you a research paper, then that's probably more relevant than if you're getting it to do highly technical, discrete coding tasks. [49:52] Richard Webbe: And that is uh and it is wherever humans program or set something up, there's room for bias. [50:01] Dave Pengelley: I mean bat back on the efficiency and the the general like availability of these things at scale. I mean, that's what the labs are working on. That's why OpenAI keeps going, Yeah, we've been able to re-release, you know, six Sol or Terra, and it's half the cost because we've got more model efficiency now, and we can use the same number of CPUs to run double the number of tokens, so we can half the price and provide more power to everyone. And Elon's just come out with saying the same similar about Grok, like that they've rewritten the entire thing. In C C because it's more directly closer to the metal efficient than using some of the more abstracted um languages. And so it can run faster. They can run more efficiently using these more traditional, sort of like I said, more direct closer to the hardware languages. And so then they can they can do more on their existing hardware and provide better things for cheaper. [50:56] Matt: The the like with without going too deep into technical crap, the um The amount of languages that are coming out now and frameworks and people actually redesigning things from the ground up because they [51:06] Dave Pengelley: Yeah. [51:06] Matt: can, um, I think is crazy. And like this is probably like a follow-on answer, Richard, to what you were saying before about runaway AI. Like without looking into the darker end of things, which is like hackers and terrorism via Yeah, via that. Well, yeah, and that. Um I I think the ability, like the actual renaissance, like the actual, [51:27] Richard Webbe: Yeah. [51:27] Matt: I think one of you guys said Renaissance last week from [51:30] Dave Pengelley: Yeah. [51:30] Matt: from the buttons thing. Um I truly believe in that. Like the idea of well we never we never touched this before because it's just too complicated, but now we can. You know, and a a lot of people freak out about AI written code because it's not as good, that's entirely wrong now. And there's there's literally, I mean citation needed. There's literally experts out there now that I I see through it throughout my feeds saying if you believe human written code is more secure, you're actually wrong. Now [52:00] Richard Webbe: Yeah. [52:00] Matt: like that that kind of blows my mind. So, you know, all of these old systems that no humans wanted to touch before, because they're just archaic and painful, you can now get that thing refactored [52:12] Dave Pengelley: Yeah. [52:12] Matt: and all of [52:13] Richard Webbe: Fix [52:13] Matt: a sudden [52:13] Dave Pengelley: And [52:13] Matt: it's [52:13] Richard Webbe: him [52:13] Matt: better. [52:13] Dave Pengelley: and [52:14] Richard Webbe: fixable. [52:14] Dave Pengelley: what we're what for for those that are less familiar with the whole difference of languages and why I talk about closer the the metal, so to speak, is you know, all these languages are abstractions, like, oh, what if we package these things up and give you a single command instead of having to write those 10 lines manually to control that part of the thing? And so languages. Have sort of been built and stacked on the previous generation to have these frameworks. And the frameworks are sort of models that sit on top of models on top of underlying languages. [52:37] Richard Webbe: It's removing the always [52:38] Dave Pengelley: Um, [52:38] Richard Webbe: hypervisor. [52:38] Dave Pengelley: to the point that vibe coding itself and then talking to an LLM means I'm not directly writing, you know, functions and code, I'm describing what I want, and then it does the underlying processing to build the thing I want. All these other frameworks have been abstractions doing the same thing in different different levels of order. And when you get back to sort of basic old school languages like C, where you are controlling memory. Memory pointers, they call them, and you're actually allocating direct memory space where you're gonna store data versus just saying store data. Like, and so when you're getting to that and that level of precision of control of the code and and how the bits and bytes are moving around, [53:13] Richard Webbe: She's the [53:14] Dave Pengelley: there's [53:14] Richard Webbe: worst. [53:14] Dave Pengelley: a whole different level of optimization you can put in at that level. And so that's why they're saying we are getting rid of some of these higher level abstraction things and we're getting back to these core basics from you know the the 80s, because that talks more directly to the hardware. And runs faster and more efficient [53:31] Richard Webbe: And the [53:31] Dave Pengelley: if [53:31] Richard Webbe: bare [53:31] Dave Pengelley: we get it [53:31] Richard Webbe: metal [53:32] Dave Pengelley: right. [53:32] Richard Webbe: becomes a lot more efficient because there's less overheads because we've removed operating systems and put in [53:38] Dave Pengelley: Yeah. [53:38] Richard Webbe: very thin hypervisors that allow us to go straight from the CPU straight into the application. [53:44] Dave Pengelley: Yeah. [53:46] Matt: I [53:46] Dave Pengelley: So [53:46] Matt: see many more things, but it's getting too technical [53:48] Dave Pengelley: all right, [53:48] Matt: probably. [53:48] Dave Pengelley: let's let's do this then. That better map. [54:00] Dave Pengelley: And no boing. [54:02] Matt: Oh, [54:02] Dave Pengelley: No [54:02] Matt: I [54:03] Dave Pengelley: swing. [54:03] Matt: can't remember what it was now. [54:06] Dave Pengelley: I've deleted the last version, so I can't show you the swing version. Um, [54:10] Matt: Oh [54:11] Dave Pengelley: but [54:11] Matt: schwing, wow. [54:12] Dave Pengelley: yeah, [54:12] Matt: Um [54:12] Dave Pengelley: it was [54:13] Matt: yeah, what's your wins? Uh I I have like a few things that I can talk about. [54:18] Dave Pengelley: I think just on the on the streamlining and stuff of AI and I've been I've been testing things. I I did cover this off I think in Friday's field note, last Friday. But um for the channel itself. Just inside baseball on the operator stories series that we do. So I'd been building these thumbnails in Canva manually with sort of me as a consistent sort of anchor across them all and the sort of various guests and their business logo or whatnot there to do that. But I was thinking seeing other people's thumbnails and saw some of the Dara with CEO type thumbnails. I was like, no one's going to click on any of these. Why what why would anyone click on any of these unless they know the person or they recognize me over time and go, Oh, yeah, he's he's an interesting interviewer. Well, click on. That. Um, and so I said to Grokbot, I want to make these more like the art of a CEO. And now on our channel, we have these style [55:09] Richard Webbe: I [55:09] Dave Pengelley: ones [55:10] Richard Webbe: love [55:10] Dave Pengelley: like [55:10] Richard Webbe: it. [55:13] Dave Pengelley: um much better [55:14] Richard Webbe: I think [55:14] Dave Pengelley: with the sort of the the short quote with the highlight with the big face of the guest, and people don't even know who the guest is, but they see the little quote and they go, Well, that sounds interesting. They look like an interesting person. It adds more of that style. But Grokbot built all of the the and the then you got my field note ones mixed. [55:29] Richard Webbe: it's a good thing. very [55:30] Dave Pengelley: in here. But all the operator story ones, they um were built in like minutes while I sat on the couch by Grockbot and pushed up into the channel as well. [55:39] Richard Webbe: And the [55:40] Dave Pengelley: Um [55:40] Richard Webbe: subject matters I if I can add, I know this is a bit of shameless self-promotion, but the the subject matters that you cover are are are fabulous. [55:48] Dave Pengelley: these are all people reaching out [55:51] Dave Pengelley: um and so just just having the chats with them. But I mean this this one was for the field note Canva for thumbnails and Hamilton built that one for me and I thought that was that was Great. I loved Hamilton's work on that one. Um, that's not actually me. That's an AI version of me, but it's it's not bad. Um, [56:08] Richard Webbe: It's [56:08] Matt: That's [56:08] Dave Pengelley: anyway. [56:08] Richard Webbe: got [56:08] Matt: not [56:08] Richard Webbe: better [56:08] Matt: terrible. [56:09] Richard Webbe: hair than you [56:11] Dave Pengelley: That that's shut up. What's wrong with my hair? Why why are you being [56:16] Richard Webbe: Nothing. [56:16] Dave Pengelley: that way? [56:17] Richard Webbe: No, no, I think I love your everyone loves your hair. [56:20] Dave Pengelley: That's it. No more book [56:22] Richard Webbe: Ella [56:22] Dave Pengelley: publishing [56:22] Richard Webbe: Max [56:22] Dave Pengelley: for you. [56:23] Richard Webbe: Beard, and everyone loves your hair. [56:24] Dave Pengelley: Anyway, uh, that was that was one of my many wins of the week was just uh redoing the thumbnails. Uh And just uh when I got Grockbot to try and make one for the field note, it was rubbish. Um, Hamilton smashed it. But, you know, people wonder why Canva. My my my wife actually said this. She uh at her work, they've just been given bigger access to copilot. And someone dropped in a map of the facility and copilot made this beautiful brand new map. And my wife's like, and that's why Canva just fired a thousand people, because people just get these AIs to build these things and they're better than most people would ever build in a design tool like Canva. Like it's Sad for [57:00] Richard Webbe: I [57:00] Dave Pengelley: all the Canva [57:00] Richard Webbe: think [57:01] Dave Pengelley: people, and I feel [57:01] Richard Webbe: I [57:02] Dave Pengelley: for them, [57:02] Richard Webbe: yeah. [57:03] Dave Pengelley: but [57:03] Richard Webbe: I [57:03] Dave Pengelley: it's [57:03] Richard Webbe: think at [57:03] Dave Pengelley: tough. [57:03] Richard Webbe: Lessian, Canva, lawyers, some other people, they're all headed for the trash heap. [57:13] Matt: Well, I I think that they're headed for better realities. Like you don't need to keep doing the same thing, go do something better. [57:20] Dave Pengelley: But [57:20] Richard Webbe: Very [57:20] Dave Pengelley: it's also [57:20] Richard Webbe: good [57:20] Dave Pengelley: a [57:21] Richard Webbe: point. [57:21] Dave Pengelley: shakeup, and there's a place for Canva. There's a place for one to manually move and tweak designs and stuff and um not always get the AI to do it. Sometimes [57:28] Richard Webbe: There's [57:28] Dave Pengelley: you want to have that [57:28] Richard Webbe: no [57:28] Dave Pengelley: humor [57:28] Richard Webbe: place for [57:29] Dave Pengelley: corrupting. [57:29] Richard Webbe: lawyers, I'm Sorry. [57:34] Dave Pengelley: Um I mean human-to-human communication, negotiation, interaction. Uh you're the one that told the story about the how the the lawyer knew what the other person was likely to want versus what the AI was recommending, was able to negotiate a better deal than the AI would have gone [57:49] Richard Webbe: Yes, [57:49] Dave Pengelley: with. So [57:50] Richard Webbe: very true, very true. [57:51] Dave Pengelley: um I had an AI red line uh an offer and it just gets overly minutiae and just like wants to change. Change every single paragraph, like arbitrarily sometimes. Like, that's too much. We did need to change those paragraphs. [58:06] Richard Webbe: Yeah. [58:07] Matt: Yeah. [58:07] Dave Pengelley: So um, no, I think I think it is interesting time, and this it gets back to people being more creative with it and not just asking to rewrite emails, but but you know, be ambitious, ask it to solve problems and see how it goes. Anyway, [58:18] Richard Webbe: It's it's [58:18] Dave Pengelley: some [58:18] Richard Webbe: about the question [58:19] Dave Pengelley: thumbnails. [58:19] Richard Webbe: because I dropped a contract into AI the other day, and instead of saying, could you check this for me, I was a bit more circumspect, and I said, is Is there any financial risks for me in this contract after signing? And it very [58:35] Dave Pengelley: Yeah. [58:35] Richard Webbe: quickly came back and said no. [58:38] Dave Pengelley: That's good. Yeah. Uh knowing what to ask for. And and I'm watching some of the other um game devs that are the established game devs that are using AI and stuff. They were talking about this, um, you know, knowing the things, and Matt, you talk about this as well, like knowing what to ask. So actually having some software engineering background so you know the language to ask the AI to do the things matters, you know. [59:00] Matt: That [59:00] Dave Pengelley: Use [59:00] Matt: was [59:00] Dave Pengelley: words [59:00] Matt: one of [59:00] Dave Pengelley: like [59:01] Matt: my uh my posts, I think like quote tweeted um the one of the creators of Pi, Mario. Um, because yeah, uh it's up to you if you want to pull that one up or not. I can give you the link. But yeah, this is uh there's a few choice words in there from Mario. [59:15] Dave Pengelley: we're running running the time. But I mean, yeah, using words like [59:17] Richard Webbe: Yes. [59:18] Dave Pengelley: a like affordance and diegetic and things like that that the people don't necessarily use all the time, but you give that to the AI and it knows what to do with those sorts of words and how to to do it. So the this guy said he I want to describe for a character who's building a lady's hairstyle and he didn't know what to call it. So he got AI to generate a palette of different ladies' hairstyles with descriptions. So then he could go, oh okay, it's that one and that one and that one. So he could learn the language that he could use to then ask the AI to build the things he wants. [59:45] Matt: Hmm. Yeah, that's that's clever. My my main point was just to be skeptical, just to push back, stop being so nice, just say no, not like that. [59:54] Dave Pengelley: Oh yeah, I saw that post. Is that the post you're talking about? You're bringing up, yeah, where like saying if you're not telling your AI to F off. This is a piece of then you're not doing it right. [01:00:05] Matt: Yeah. [01:00:05] Dave Pengelley: Yeah, [01:00:05] Matt: Yeah, [01:00:06] Richard Webbe: I [01:00:06] Dave Pengelley: I [01:00:06] Matt: that's [01:00:06] Dave Pengelley: tell [01:00:06] Richard Webbe: had [01:00:06] Dave Pengelley: my [01:00:06] Matt: one [01:00:06] Richard Webbe: your [01:00:06] Matt: more. [01:00:06] Dave Pengelley: AI [01:00:06] Richard Webbe: own problem. [01:00:07] Dave Pengelley: things look trash all the time. This looks [01:00:08] Richard Webbe: Yeah. [01:00:08] Dave Pengelley: trash, that's janky, that doesn't work. Why are you doing that? [01:00:11] Richard Webbe: Yeah, exactly. [01:00:13] Dave Pengelley: Any other wins, guys? [01:00:15] Matt: Um my mine is just that infrastructure win again. So like making that machine that you talked [01:00:20] Dave Pengelley: The [01:00:20] Matt: about at [01:00:20] Dave Pengelley: box. [01:00:20] Matt: the beginning better and better. I I actually do the majority of my work now in cloud boxes. I've I've been using boat, if you guys haven't come across boat yet. Um and then And also lizard. Boat and lizard are, you know, like what that box is is like for fractional compared to what I can actually do on boat and lizard. Yeah, which is crazy. [01:00:43] Dave Pengelley: What [01:00:43] Matt: They're [01:00:43] Dave Pengelley: a world [01:00:43] Matt: funny [01:00:43] Richard Webbe: So [01:00:43] Dave Pengelley: we [01:00:43] Matt: names, [01:00:44] Richard Webbe: just [01:00:44] Dave Pengelley: live [01:00:44] Richard Webbe: quickly [01:00:44] Dave Pengelley: in. [01:00:44] Matt: hey. [01:00:44] Richard Webbe: for my weary mind, what is a boat and what is a lizard? [01:00:49] Dave Pengelley: Cloud [01:00:50] Matt: So [01:00:50] Dave Pengelley: service. [01:00:51] Matt: yeah, they're they're literally cloud boxes. So like that that physical box that's on my desk, um it takes it takes processes. But if I want To go and ship changes live to my production software, there's a whole bunch of front-loaded verification that needs to happen that I've baked into my CI pipeline. Those things I just send to the cloud now, and boat will literally just chuck them into individual sandboxes, runs, I get the result, and then I put it live. [01:01:17] Richard Webbe: And we've got security that is tenamount to sovereign AI for us, hasn't it, hasn't it? [01:01:24] Matt: What? [01:01:25] Richard Webbe: So you you know, the concept that my my my data is protected and sovereign and not accessible by someone else. [01:01:35] Matt: Yeah, yeah, that is the case. You can turn on all of those protections if you wish. Um, and you can just run your own boxes locally as well. Like [01:01:43] Richard Webbe: Yeah, [01:01:43] Matt: you don't need to use a [01:01:44] Richard Webbe: that's [01:01:44] Matt: cloud [01:01:44] Richard Webbe: what I [01:01:44] Matt: thing. [01:01:44] Richard Webbe: meant. Yeah, yeah. [01:01:45] Matt: The [01:01:45] Richard Webbe: Throwing [01:01:45] Matt: the reason [01:01:46] Richard Webbe: a [01:01:46] Matt: why [01:01:46] Richard Webbe: net [01:01:46] Matt: I'm [01:01:46] Richard Webbe: around. [01:01:46] Matt: using I'm using a cloud thing because it's more efficient, it's cheaper [01:01:49] Richard Webbe: Yeah. [01:01:50] Matt: to run, and it allows me to scale infinitely. [01:01:53] Richard Webbe: So that's exciting. So then i i if I think about it for a second, I'll shut up Dave and I will wrap it up. now get a monetary program that will work out what is cheaper whether it runs on your box or in the cloud and make a decision on where that data and and processing goes. Is that correct? [01:02:10] Matt: 100%. Yeah. [01:02:11] Richard Webbe: That is [01:02:11] Matt: So [01:02:11] Richard Webbe: exciting. [01:02:11] Matt: there's there's quite literally arbitrage right now to just run things from providers that actually provide you better infrastructure than what you can provide for yourself. That's [01:02:23] Richard Webbe: Yeah [01:02:23] Matt: the whole [01:02:24] Richard Webbe: I [01:02:24] Matt: point of [01:02:24] Richard Webbe: like [01:02:24] Matt: a data [01:02:24] Richard Webbe: that. [01:02:24] Matt: center. [01:02:25] Dave Pengelley: And that that that's the whole point of like that's why AWS is a thing, right? With EC two. The whole elastic compute concept started with them twenty plus years ago, where they went, Oh, what if we just sell CPU and RAM by the by the minute? [01:02:40] Richard Webbe: Yeah, but it's but it's dynamic, right? So Bill Shock comes to everyone with AWS and others. And I'm just thinking, by the minute, if you know what the current rate of your cloud is and you know what the rate of how much it costs you to run that box, you can immediately move that processing back on in-house. [01:02:57] Dave Pengelley: Yeah, which I think and then people are doing that across Different services and distributed platforms and even AI with open router, people are looking at which is providing the cheapest rate on a particular model at any given time and moving moving models and all that dynamic [01:03:10] Matt: So [01:03:10] Dave Pengelley: costing [01:03:11] Matt: I [01:03:11] Dave Pengelley: stuff. [01:03:11] Matt: can give you an example, like with boat, I'm on the 100 USD plan for boat, which gives me 840 starts per day. So that's literally firing up a container is one start. [01:03:22] Richard Webbe: Wow. [01:03:22] Matt: And so that's a 24-hour window that's constantly sliding, right? And I'm basically capped constantly. And The actual monetary usage of those box, so other than that starts per day being a quota, the actual like cost per minute, you know, I I've spent this month so far. So this last billing period, which was since the start of November basically, um, I've spent uh $15. [01:03:52] Matt: And [01:03:52] Richard Webbe: Wow, [01:03:52] Matt: I've put a lot through there. So yeah, the the economics of like if I was to go and buy the same machines that boat is using to run. This for if I would was to buy that and chuck it next to that box on my desk, I would be spending a lot more than fifteen dollars. [01:04:06] Richard Webbe: So the economies of scale work in a very dynamic manner. [01:04:09] Matt: Yep. [01:04:10] Richard Webbe: I like that. [01:04:13] Matt: Anyway, [01:04:13] Dave Pengelley: All [01:04:13] Matt: it's [01:04:13] Dave Pengelley: right. [01:04:14] Matt: a good place to end it. [01:04:16] Richard Webbe: Dave, [01:04:17] Dave Pengelley: Yes. [01:04:17] Richard Webbe: there's one hair sticking out. [01:04:19] Dave Pengelley: Shush. Uh and uh Joe, our friend Joanna Jones has been in the comments. Thank you, Joe, for for participating. Sorry, I just saw them. We would miss you in real time. Uh but back on that topic earlier, she did write. AI is ultimately a utility like electricity. [01:04:35] Richard Webbe: Oh, [01:04:35] Dave Pengelley: Um, [01:04:36] Richard Webbe: so true. [01:04:36] Dave Pengelley: will [01:04:37] Richard Webbe: So [01:04:37] Dave Pengelley: become [01:04:37] Richard Webbe: true. [01:04:37] Dave Pengelley: somewhat ubiquitous. [01:04:40] Matt: Hundred [01:04:40] Dave Pengelley: So [01:04:40] Matt: percent. She's actually met with Mano today. I got some [01:04:43] Dave Pengelley: oh [01:04:43] Matt: insider knowledge there. [01:04:44] Richard Webbe: Oh, [01:04:45] Dave Pengelley: wow. [01:04:45] Richard Webbe: wow. [01:04:45] Matt: She was in Cal. [01:04:46] Dave Pengelley: Ugh. [01:04:47] Richard Webbe: Nurse Joe hanging out with the mechanic Mano [01:04:51] Dave Pengelley: In the gold fields. [01:04:53] Richard Webbe: in [01:04:53] Dave Pengelley: Well, [01:04:53] Richard Webbe: the [01:04:53] Dave Pengelley: uh, [01:04:53] Richard Webbe: cold. [01:04:53] Dave Pengelley: thank thank you all. Thank you to Joe. Thank you to the others that have joined us live. Thank you if you watch us on the replay. Make sure you like and subscribe. That way you will catch all of oh, we've got Kate, friend, our friend Kate as well. Um, great to see you there, Kate. Uh, thank you for joining us. Uh, make sure if you like this sort of if you like our live chats, if you like the uh the one-on-one interviews and the operator stories, if you like the little hot tips and fragments from my life in field notes, make sure you're subscribed and hit the bell icon so you get those notifications. Uh, we have hit two milestones, guys, and I'm sure I'm I'm appalled at myself for not raising this last week. Last week was episode 42. We crossed the answer. Last week was 42. I had no fanfare. I had no bumpers or stingers. And like afterwards, when I was processing the episode, I was like, okay, and process the transcript for episode 42. [01:05:42] Dave Pengelley: Um, [01:05:42] Richard Webbe: Well, [01:05:42] Dave Pengelley: so [01:05:43] Richard Webbe: well [01:05:43] Dave Pengelley: we've [01:05:43] Richard Webbe: done, Dave. [01:05:43] Dave Pengelley: crossed [01:05:44] Richard Webbe: It's all your logistics and hard [01:05:45] Dave Pengelley: we've [01:05:45] Richard Webbe: work [01:05:45] Dave Pengelley: crossed [01:05:45] Richard Webbe: that gets [01:05:46] Dave Pengelley: 42 [01:05:46] Richard Webbe: used. [01:05:46] Dave Pengelley: and we have crossed the ton. We've got um a hundred followers on the channel now. So uh it's all up and up from here. So thank you everyone for your faith and confidence and voyeurship, voyeur uh viewish viewership, not voyeurship. Maybe a bit of both, and we'll see you all next time. Thank you, gentlemen. [01:06:07] Richard Webbe: Thanks, guys.
Related episodes

Keep going

Chapters

Jump to a section

0:00The home server behind Matt's AI agents
8:08Software moats, attention and algorithm bubbles
13:19ChatGPT business profiles and local discovery
18:21AI assistants, cross-device access and security
32:48AI regulation and infrastructure debate
41:09Cognitive security and building games with AI
44:28Who pays for AI? Free models and model bias
50:01Efficiency, old code and AI-assisted development
54:18AI-generated thumbnails and human judgement
1:00:15Local servers, cloud sandboxes and compute costs