AI Doomers, Data Centres and Why Structured Outputs Matter | AI Operators #35

[00:00] Dave Pengelley: I've had a big couple of weeks. We've been snackish and I've been eating weight. We just keep buying chips and chocolate and it's ridiculous. And I've put on uh like four kilos in the last couple of months and I'm like, I've gotta get back in the low carb, more keto style. Because I, you know, I'll do the keto style thing during the day and I'll have bacon and eggs and peanut butter and then after dinner have a packet of potato chips. I'm like, Oh, that doesn't [00:21] Matt: Oh, [00:21] Dave Pengelley: work, [00:21] Matt: that's [00:21] Dave Pengelley: you can't do both. You gotta pick a lane. [00:24] Matt: instantly bad. [00:25] Dave Pengelley: Yeah, [00:26] Matt: It's like [00:26] Dave Pengelley: high [00:26] Matt: poison. [00:26] Dave Pengelley: fat and high carb. Beautiful day. Well done, mate. Well done. So [00:30] Richard Webbe: Have have eggs in the morning. [00:33] Matt: Hmm, eggs are good. [00:35] Dave Pengelley: Yeah, I'm trying to um not have much for breakfast, maybe a protein shake, and then bacon and eggs is kind of becoming a bit of a staple. Um, [00:46] Matt: If [00:46] Dave Pengelley: because [00:46] Matt: you can [00:47] Dave Pengelley: it's [00:47] Matt: do fasting, [00:47] Dave Pengelley: good. [00:48] Matt: fasting actually genuinely is good. [00:50] Dave Pengelley: I [00:50] Matt: Uh [00:50] Dave Pengelley: know. I've I've got apps and I've I've done a lot of IF and like intermittent fasting and stuff and that's one of the reasons I don't really do a lot of breakfast, because I try and wait 'til at least midday. But um [01:00] Matt: Joe Joe is saying good morning. It [01:03] Richard Webbe: Good [01:03] Matt: is [01:03] Dave Pengelley: Good [01:03] Matt: still [01:03] Richard Webbe: morning, [01:03] Dave Pengelley: morning. [01:03] Matt: in morning [01:04] Richard Webbe: Joe. [01:04] Dave Pengelley: It [01:04] Matt: for [01:04] Dave Pengelley: it [01:04] Matt: her [01:04] Dave Pengelley: is [01:04] Matt: in [01:04] Dave Pengelley: morning for [01:05] Matt: WA. [01:05] Dave Pengelley: Joe. Yes. [01:06] Matt: Yeah. [01:06] Dave Pengelley: Um [01:06] Richard Webbe: Such [01:07] Dave Pengelley: two [01:07] Richard Webbe: an [01:07] Dave Pengelley: hours [01:07] Richard Webbe: awesome person. [01:08] Dave Pengelley: ten a.m. for Joe. Um we will get you back on the show sometime, Joe. We'll get we'll have a Joe show. [01:16] Matt: That'd [01:16] Dave Pengelley: Uh [01:16] Matt: be great. [01:18] Dave Pengelley: so uh yeah, I mean there's lots of people talking about on this topic of diet, like uh and you know, just to bring it back back on topic, these sort of AI diet trackers and taking photos and stuff, like I'm Old school. Like I use my FitnessPal and I've used it forever, and it's become way more annoying since Underarm or whoever bought it and just crammed it full of ads and crap. But still, for basic food logging and just this library of foods and calories and stuff, I find my FitnessPal is good when I want to do that kind of tracking. I don't know how reliable the photo ones are and all that kind of stuff. Like, how does it know? How does it [01:53] Matt: Yeah. [01:53] Dave Pengelley: know? [01:53] Matt: Yeah, doesn't know what's cooked in, you know, all of that kind of crap. Uh I don't even know why you need to track, like I reckon just just avoid Just avoid the poisons. Stay [02:02] Dave Pengelley: No, [02:02] Matt: away from all [02:02] Dave Pengelley: but [02:02] Matt: of [02:02] Dave Pengelley: sometimes [02:03] Matt: that. [02:03] Dave Pengelley: sometimes it can be like actual overall calorie intake. Like you can eat decent things, but you can have way too much of them. And so, you know, I can you know be really good on keto and avoid all the carbs, but if I have 5,000 calories in a single day, that's probably counterproductive. And so it's uh sometimes it's not just the macros and and stuff so on, it's the overall caloric intake versus burn, um, which you know it's used. And it's one of those things I find personally, you do it for a little while just to get back on track and know where the limits are. And then you go, okay, I don't need to track that anymore, because I know I'm gonna have this for that and that for that. That's gonna leave that. It's gonna be fine. And I'm within my limits if I kind of eat in this style and I find [02:46] Matt: Yeah. [02:46] Dave Pengelley: I don't need to track for but you know, I've been severely overweight before, and it's been genuinely helpful to really keep you accountable to going, okay, after dinner I've only got a hundred calories left. I could have half a banana or whatever it is. Uh Um and that's it, and that's gonna keep me in my limits. [03:02] Matt: I [03:03] Dave Pengelley: So [03:03] Matt: have the weirdest uh like health influence and inspiration for me. Like my dad, he he beat prostate cancer solo, uh, went against all of his doctors' recommendations and basically just started following alternative health advice just based on eating the right stuff and avoiding the stuff that is clearly poisoning us. And yeah, even this morning, like he hung out with one of his old um old crew and found out that that uh on Monday, like he literally hung out with this fella all Monday, took him around to all of his shops and whatever he needed to do. So in the car, you know, close quarters. And then later on that night, uh old Bernie told dad that he tested positive for COVID. So, you know, I was meant to be meeting up with my dad tonight for a trivia tonight at our local pub. And Dad's messaged me with the image of his COVID test saying negative. So he's like, you know, my immune system. Yeah, [04:00] Dave Pengelley: Yeah. [04:00] Matt: has has come out again on top, you know. So that's just super inspiring for me. And all he did was just entirely went full like quote unquote healthy keto. So like you [04:12] Dave Pengelley: Yeah. [04:12] Matt: know, there's keto, which is like dirty keto, and then there's healthy keto, which is just you pile on the vegetables and everything. [04:18] Dave Pengelley: Yeah. [04:18] Matt: Um yeah, he he will never not eat that way again. [04:23] Richard Webbe: Hang [04:23] Dave Pengelley: Yeah, [04:23] Richard Webbe: on, [04:23] Dave Pengelley: they're [04:23] Richard Webbe: man. [04:23] Dave Pengelley: the [04:24] Richard Webbe: Keto and vegetables don't go together though. [04:27] Matt: Why not? [04:28] Dave Pengelley: Cobs. Depends on the vegetables. Right? Um, and there's and there's there's more that Mediterranean style diet, the sort of the Greek diet with the olive oils and the and the cheese, like there's there's things that they fishes and stuff. I don't know. I haven't gone deep into that one. But [04:43] Matt: It's [04:44] Dave Pengelley: look, [04:44] Matt: basically [04:44] Dave Pengelley: I [04:44] Matt: no sugar, no carbs, that's all it is. [04:46] Dave Pengelley: I've done, you know, whichever version of keto I like, pork crackling, bacon and eggs, peanut butter, all the things. And when I had my blood test a few months ago, like my cholesterol's perfect. So I'm like, again, it's it's like you can it's not. the big enemy they say it is. Anyway, [05:02] Matt: Exactly. [05:02] Dave Pengelley: I think I think we're getting really distracted on food and diet. Um but it is there are AI tools that are helping people and and you know that's [05:11] Richard Webbe: There [05:11] Dave Pengelley: good. I mean [05:11] Richard Webbe: are [05:11] Dave Pengelley: Joe [05:12] Richard Webbe: I [05:12] Dave Pengelley: said build [05:12] Richard Webbe: think [05:12] Dave Pengelley: your build your own app. I I [05:14] Richard Webbe: I think [05:14] Dave Pengelley: could [05:14] Richard Webbe: those [05:14] Dave Pengelley: I [05:14] Richard Webbe: AI [05:15] Dave Pengelley: don't need to [05:15] Richard Webbe: apps that you get, um I uh I lost I don't know if you guys know I lost twenty-three, twenty-five kilos at the start of COVID. So uh which was [05:23] Dave Pengelley: wow [05:24] Richard Webbe: because I stopped having to go out with people [05:26] Dave Pengelley: hmm [05:26] Richard Webbe: and I rode my bike in the evenings and did my conference calls with uh head office while riding a bike. But they've got those apps now, and you photograph the plate of the food, and the AI assesses the calorific intake and the ratio and puts it into your little calorie tracker and all that sort of stuff. I amazingly um amazingly uh interesting and amazingly informative. I had a family member who was rather obese and unwell and I would say no just Respect a bit uneducated, and we were sitting to a barbecue one time while I was warming up for the onions, eggs, and steak, and she had a giant packet of chips and she's eating them. And I said, Hey, how's that diet going? And she said, Good, I'm eating the chips because they don't weigh very much. [06:22] Matt: Very good. [06:22] Richard Webbe: That is a true story, and I'm not fat-shaming. [06:25] Dave Pengelley: And [06:25] Richard Webbe: Some people just don't educate it on what clean food is. [06:28] Dave Pengelley: Yeah, well On that note, we're not gonna keep our audience waiting any longer. [06:35] Matt: Wait, [06:36] Dave Pengelley: I get now, [06:47] Dave Pengelley: I get now [06:48] Matt: I'm if not sure you're [06:48] Dave Pengelley: why [06:48] Matt: able be do going not that. to to [06:48] Dave Pengelley: the two blokes talking tech have what they call their private feed, where they do a separate show where they're just talking about their life and just other crap, because it can so easily just the banter can just talk and just about general. life and travel and food and family and stuff so easily that it distracts from the main show [07:05] Matt: I'm [07:05] Dave Pengelley: so that they [07:05] Matt: here [07:05] Dave Pengelley: actually [07:06] Matt: for [07:06] Dave Pengelley: have [07:06] Matt: it. [07:06] Dave Pengelley: a [07:07] Matt: I think I'll [07:08] Richard Webbe: We don't no complaints [07:09] Matt: I [07:10] Richard Webbe: here. [07:10] Matt: would I would prefer just the random rabbit hole ram ramps ram rambles. [07:14] Richard Webbe: That's [07:14] Dave Pengelley: Rambles [07:15] Richard Webbe: that's why we're here. [07:16] Matt: Maybe that could be my show, rabbit hole rambles. [07:19] Richard Webbe: Can I throw a can I throw a reel because I missed a couple of weeks and I need [07:23] Dave Pengelley: We [07:23] Richard Webbe: to get [07:23] Dave Pengelley: noticed. [07:23] Richard Webbe: something [07:24] Dave Pengelley: We were [07:24] Richard Webbe: I [07:24] Dave Pengelley: keeping [07:24] Richard Webbe: wanna [07:25] Dave Pengelley: track. [07:25] Richard Webbe: I wanna throw no one know this that's true. [07:30] Dave Pengelley: Artist. [07:31] Matt: I [07:31] Richard Webbe: In any case, if you haven't [07:33] Matt: think [07:33] Richard Webbe: already bought your copy. [07:34] Matt: that's a good question. [07:38] Richard Webbe: Anyway, I I I got infuriated over the last week with everyone complaining about data centers and AI. And I want to know who's the morons living in their version of Hollywood hyperbole. It is just the conversations are so dumb. I refer to that. Sound grab when I was listening to the radio last night. Nothing they said had any form of intelligence whatsoever about AI and data centers, and the whole world is stupider for having tuned into the conversation. [08:15] Richard Webbe: That's it. It is just people, you know, oh God, I don't know where to start. I'm like Matt. I get overwhelmed with so many questions and ideas buzzing in my head around that that I just can't get it out. But data centers are going to save the planet because we do. more processing in one place much more efficiently to get answers to save the planet, whether it's our dietary requirements or how to improve the efficiency of an engine that might be polluting the world. [08:40] Dave Pengelley: According to the misanthropic CEO, we need to slow everything down because one of their staff quit and said the world is at risk because of these reckless companies. And [08:50] Richard Webbe: Exactly. [08:50] Dave Pengelley: so he came out and went, Yeah, that's true. We are reckless. We should, you know, get the government to sponsor a whole bunch of companies to audit us that we also own and regulate. The industry so other companies can't come in and compete against us. [09:04] Richard Webbe: I'm with Trump. More AI and then we win the war. [09:09] Matt: His [09:10] Richard Webbe: So I'm I'm [09:10] Matt: response [09:10] Richard Webbe: paraphrasing [09:11] Matt: was so [09:11] Richard Webbe: there, [09:11] Matt: funny. [09:11] Richard Webbe: I'm not officially with Trump. But he's gone the other way. He just said, you know what, if you guys don't want to slow down, we're speeding up. Good luck. [09:21] Matt: I I found his response so funny. Like this is absolutely not a political podcast, but it was very enjoyable to be like Slow down, lol, noob. Um like and then then China's response was even better. They were like, you guys are basically equating the you know, giving Hitler the Adam Bob kind of thing. You know, like that was literally what it said in their response. [09:46] Richard Webbe: Yes. [09:46] Matt: Like, oh, I don't know, it's just so much fun. It's [09:49] Richard Webbe: And [09:49] Matt: very [09:49] Richard Webbe: all [09:49] Matt: much just [09:49] Richard Webbe: the people [09:50] Matt: pop [09:50] Richard Webbe: saying [09:50] Matt: fun. [09:50] Richard Webbe: that all the people saying AI is dangerous, it could kill us. Let's start with a few technologies that kill could kill us that have been launched without any fanfare. The gun. car, dynamite, fire, matches. [10:02] Matt: See well. [10:03] Richard Webbe: I mean Yeah, exactly. Seagil. Exactly. Candied bananas. I mean, jeepers. The the the hyperbole and ridiculousness around complaining. And then last night someone rang in the station and said they're complaining about data centers. And they said, I looked out the window and the data centre at the end of our road had all this steam coming off it. What's that? Is that gonna kill us? And I'm just going, Oh my God Just go have a look at an office tower or an industrial, you know, an industrial estate or something like that. It's all steam from the the cooling rooms. And modern data centers don't even give off steam. They recycle all the water and they just have a big grill. [10:44] Matt: Yeah. [10:44] Dave Pengelley: Right. [10:44] Matt: Are you sure [10:45] Dave Pengelley: I [10:45] Matt: it wasn't [10:45] Dave Pengelley: um [10:45] Matt: like just chemtrails that they're off off uh offloading from the data [10:48] Richard Webbe: They [10:49] Matt: centers [10:49] Richard Webbe: are chemtrails. [10:49] Matt: as well? [10:51] Richard Webbe: It's all chemtrails. Oh, [10:53] Dave Pengelley: I [10:54] Richard Webbe: here [10:54] Dave Pengelley: uh [10:54] Richard Webbe: we go. [10:55] Dave Pengelley: haven't watched the um the the core where which one do I want? Which thing they've got on your settings here. Um, with why is there an empty placeholder there? Ah, very good. Let's remove that. Um [11:07] Matt: It's Trump coming in. [11:09] Dave Pengelley: Trump, Trump, he's all over the AI thing. He loves AI. I was just looking at his feed. I haven't I haven't watched the bit where he calls Jensen alive on stage. I I haven't gotten into [11:19] Matt: That's [11:19] Dave Pengelley: all that sort of stuff. [11:19] Matt: good. [11:19] Dave Pengelley: But [11:19] Matt: That's [11:19] Dave Pengelley: I was just [11:20] Matt: a good [11:20] Dave Pengelley: I just [11:20] Matt: one, [11:20] Dave Pengelley: scroll [11:20] Matt: too. [11:20] Dave Pengelley: through his ex-feed and I see this. [11:32] Matt: Really good ones. [11:33] Dave Pengelley: Like what the heck? He's got this thing where it's like a deep fake of him as a doctor doing something. I don't even know. Like, [11:40] Richard Webbe: My [11:40] Dave Pengelley: and the [11:40] Richard Webbe: son [11:40] Dave Pengelley: White [11:41] Richard Webbe: loves the loves the Trump dance, so every time he's happy he goes and does a little trump dance. [11:46] Dave Pengelley: And the White House brought out all these games, all these like AI generated games or something the other week. Like, it's wild what's going on over there. I don't even know. But um, mean meanwhile, you know, uh there's not a political podcast, we'll not talk about what's Happening in Australia. There are people complaining, Adelaide. There is a big thing in Adelaide around data centers or something, right? Like, is that that what you're seeing, Richard? Like, what's [12:06] Richard Webbe: Yeah, [12:06] Dave Pengelley: the [12:06] Richard Webbe: I [12:06] Dave Pengelley: local [12:07] Richard Webbe: just [12:07] Dave Pengelley: domestic [12:07] Richard Webbe: think [12:07] Dave Pengelley: view on them? [12:08] Richard Webbe: the general conversation from the uneducated person in the marketplace. I've been working in and out of data centers for nearly 40 years. And everything everyone's saying, again, like I said, has no intelligence whatsoever, and everyone's dumber for listening to the conversation. [12:25] Dave Pengelley: Yeah. [12:28] Dave Pengelley: But it's interesting. It is interesting. I mean, and we're We'll we'll we'll touch on this in the news that like the w the evolution of models and what's happening and who has access to them. Um does and I saw people saying on LinkedIn, does having no data centers in Australia limit our ability to deploy AI apps and so on and like [12:47] Dave Pengelley: obviously like w [12:49] Matt: It [12:49] Dave Pengelley: what? [12:49] Matt: depends what you care about. Like [12:51] Dave Pengelley: Yeah. [12:51] Matt: it's the same like residency and sovereignty, [12:53] Dave Pengelley: Right? [12:53] Matt: all of that. It's what you care about. It's what you want to put your insurance into. [12:57] Dave Pengelley: I [12:57] Richard Webbe: It [12:57] Dave Pengelley: I [12:57] Richard Webbe: is [12:58] Dave Pengelley: would personally rather us be able to produce Steel plating than than data centers, like [13:02] Richard Webbe: Yes, [13:02] Dave Pengelley: in the immediate [13:03] Richard Webbe: but do we [13:03] Dave Pengelley: term. [13:03] Richard Webbe: can we can produce the steel plating better, more efficiently, with less pollution if we've got a data center driving the metrics around that production? It's a no brainer. [13:16] Matt: They they just have microducts on the production line now, just pressing buttons with its beak. [13:30] Richard Webbe: Said, you know, a million monkeys with a million typewriters could probably one day produce the best novel, but it's not efficient. And so what data centers do is remove those million monkeys and just put something in there that could do what we want it to do anyway. So I just look at data centers providing efficiency under the banner of AI or automation and the acquisition of high um high data, and away it goes. [13:52] Dave Pengelley: All right, well [13:52] Richard Webbe: And [13:52] Dave Pengelley: let's [13:53] Richard Webbe: sorry. [13:53] Dave Pengelley: sorry so sorry Richard I didn't mean to cut [13:55] Richard Webbe: No, [13:55] Dave Pengelley: you off, I [13:55] Richard Webbe: you [13:55] Dave Pengelley: thought you were [13:56] Richard Webbe: no, [13:56] Dave Pengelley: done. [13:56] Richard Webbe: you go, you go, I can waffle. [13:57] Dave Pengelley: Then I was just gonna say, Well, we should do this. Really. Because it feels like [14:08] Dave Pengelley: we're talking about newsy kind of news topics and so on. So it seems seems appropriate. Um yeah, as far as you know the data center use and the infrastructure and so on, like we're still so early on all of this sort of stuff. And I and um I was talking to Matt before the show around this announcement that came out today that's all over X um from Diogo Almeida. Uh we'll turn his audio off, but just watch this. Matt is convinced and the the audience completely AI generated video [14:38] Matt: It [14:38] Dave Pengelley: uh [14:38] Matt: is, [14:39] Dave Pengelley: of [14:39] Matt: yes. [14:39] Dave Pengelley: this this founder. Um who's [14:42] Matt: It's well, [14:42] Dave Pengelley: talking about this new model that doesn't do chat. So it's uh LLM style super intelligence, but [14:49] Matt: I a good it's one. think [14:49] Dave Pengelley: it doesn't write paragraphs and narratives. It just produces answers to questions with a probability score of how accurate it thinks it is. So rather than just waffling and telling You lies and as if they're true, it produces an answer and says, Yeah, but I'm only forty percent confident in this answer. And [15:06] Richard Webbe: Yeah, [15:06] Dave Pengelley: then [15:06] Richard Webbe: I [15:06] Dave Pengelley: the [15:06] Richard Webbe: think [15:06] Dave Pengelley: system [15:06] Richard Webbe: that's [15:06] Dave Pengelley: goes, [15:07] Richard Webbe: correct. [15:07] Dave Pengelley: Well, that's crap. Try again until you're like eighty, ninety percent or whatever, right? Is that is that a fair interpretation from what you've seen, Matt? [15:14] Matt: I I honestly I don't know enough about it to be able to comment and translate it with my famous analogies. I I feel like it you need to look at this, it's completely different. It's not a large language model, so it doesn't reason in language. So it doesn't use any of the same weights that the current large language models use. So if you ask a question, you can't actually necessarily ask a question with language. It's fully programmatic. Like the um the whole uh company, like the actual brand behind this is called type safe. So you have to actually input with TypeScript functions or type safe functions, and you get that probability response back with structured data. So you you always get a probability back, but you get structured data back. Currently, we're like forcing large language models to give structured data back based on it being part of a schema, and this just speaks in schemas. So it's very weird, it's very alien to us now that we're sort of used to LLMs. But there's demos literally in this video of it making decisions like 10 times per second and being able to play a live-action video game, like a [16:25] Richard Webbe: And [16:25] Matt: shooter. [16:25] Richard Webbe: what's [16:25] Matt: They [16:25] Richard Webbe: the [16:25] Matt: actually [16:25] Richard Webbe: name of this system, Goss? [16:27] Matt: uh the the agent itself is called. Jev J E V, um, but TypeSafe.ai is the is the company. [16:36] Richard Webbe: I think that's a a brilliant way of demonstrating what we talk about all the time, and that most answers on LLMs are the sumnet average of everyone's opinion and aren't really based in fact whatsoever. [16:50] Matt: I need you to elaborate what do you mean? [16:52] Richard Webbe: Well, just saying that that when you go to uh an LLM like like like a website and you say, you know, uh who's The tallest person in the world, it'll just be registered opinions and registered people. It's not actually getting, as you say, empirical data and saying there they are, and here's why and here's how. And I think it's dangerous because every time I see people talk about AI, I think about them being the next person who will be suckered in in the 1800s by the snake oil person who stands there and he holds his bottle of oil and says, Everything that comes out of this bottle of oil is going to fix your health, it's going to make you richer, it's going to make you smoke. after people will be more attractive and that's how everyone's looking at AI now. And it's just such a trap. It's deadly. We've got a tool that we need to put in perspective. And so I like that that that uh that [17:41] Matt: It's [17:41] Richard Webbe: agent. [17:41] Matt: very nebulous to put that into practice though. Like what does that you know mean and how does that apply? And a lot of people now, when they have their apps and they've got like AI in their app, it's always using a large language model. So it's always taking some sort of instruction prompt, some sort of random input, and either it gives Like a text output or a structured data output. So simpler structured data is it's a Boolean response. It's either true or false. You know, does this contain [18:06] Richard Webbe: Yep, [18:06] Matt: this word [18:07] Richard Webbe: yep. [18:07] Matt: true or false? You can think [18:08] Richard Webbe: Yeah. [18:08] Matt: of a moderation script in somebody's community platform, you know, when when they're when they have keywords that they don't want to have people say, you know, potentially if it's not a political podcast, you know, they can have, you know, does this have this flag, you know, yes or no? And then every single message that comes through the LLM goes true or false. But whereas it's still slow and it's still very inefficient, this JEV thing is basically. Free and is instant, like I was saying before, [18:33] Dave Pengelley: Yeah. [18:33] Matt: like it does like ten decisions [18:34] Dave Pengelley: Cause [18:34] Matt: per second. [18:35] Dave Pengelley: because it's [18:36] Richard Webbe: Wow. [18:36] Dave Pengelley: running the the the the processing in parallel, so rather than sequential processing token by token and sort of running parallel pipelines to derive its answers. And it's built for structured data. That's all it does. And so, like and in LLMs, and if you've used N8N, you can sort of do your little structured data out thing. So it runs it through a secondary LLM to try and confirm the first LLM actually put the things in the right format, to the point that even for this podcast, like to bring Back to something really practical. When I get this podcast and I run the transcription and I build the new titles and the metadata and everything that I push to YouTube, when I just had LLMs doing that, even though it knew that YouTube needs this format here and it needs this value here and it needs that value there, it would screw it up and my system break because it would put a a slash or a bracket where it didn't belong and just screw everything up. So whereas now I get it to do that, but then I run it through a Python tool which will actually clean any of that up and make sure it actually really Is structured because it programmatically knows it has to have this structure, it needs to remove those characters, and Python will not guess and hallucinate and make mistakes. So the way Jev would is it only knows how to produce the structured outputs correctly formatted as structured outputs. So you never get that. Maybe it'll get the commas in the right place. That's all it does. It just knows that. [19:51] Dave Pengelley: And so because it's just giving you, you know, it's basically populating variables and as a data [19:58] Matt: Is it [19:58] Dave Pengelley: output. [19:58] Matt: this is very, [19:59] Dave Pengelley: It's very [19:59] Matt: very good. Frequently asked questions section on their page. [20:02] Dave Pengelley: Yes. [20:03] Matt: Um, their page is actually fantastic, by the way. Very, very cool styling. I think I want to uh appropriate some of this uh styling on my own pages once I get back to them. Um but there's a very, very cool thing here where it says is Jev just a smaller LLM? Because you know, you think of things like Luna, um, haiku, uh, you know, the the flashes of the world. Um, the the FAQ, like, you know, is it just a smaller LLM? They basically say that it's not. not small nor an LLM. It's it's hen it it's completely off the charts with regards to you know its uh Pareto you know intelligence curve with regards to it's optimizing for a different task. You know, like that that second question there, like, you know, that whole paragraph really explains what it is and and what they're actually trying to achieve, which I I think everyone who is somewhat related to the AI space, you know, if you're crazy enough To listen to us on this show, you should really go to this page and have a quick list uh quick read of those FAQs, because [21:05] Dave Pengelley: Yep. [21:05] Matt: this is going to be something huge. [21:07] Dave Pengelley: Yeah. And and I was saying at the start, I think before I can't remember where I got to it in Matt, but I was saying when I was building my sergeant harness, one of the things I was really aware of, and now is having like rock bots send each other like chat messages and bots chatting with each other has become super trendy and popular. But it's also super inefficient. Like cause you're wasting all these words of you know pros that don't describe the actual task. They're just, you know, that that human style interaction. So you can read it and they they seem personable. So one of the things I was building in my harness for the whole job management was basically JSON packets. They were building work orders. And so when it submitted a job to a different profile of worker, it was sending it a a work order JSON packet with structured ordered data. And then when it finished its work, it was meant to return a structured ordered thing. Jev was built for what I was trying to do with the sergeant harness because uh like it's it's Exactly that efficient way of communicating. And like I was saying at the start, like not to blow my own horn, but I saw this and was trying to build this sort of six months ago. I just hit my technical limitations of what I was had time, energy, capacity to try and work through with the existing models as they were and so on. But this is this is gold for that, plugging it into those w automated workflows and processes that will reliably run without having to have all that Python patching in between them. [22:26] Matt: What what confuses me and um I I think a lot of people will Will have seen this before where they have image recognition from like previous, like imagine just before LLMs existed. There was always those images from like CCTV, and there was the little colored boxes that would show up on the screen for some sort of like recognition. You know, like here's a human walking around. Like, you know, in in certain uh countries, they may have that kind of surveillance where it's very clear and obvious as to what's going on in the world. Um, I think GIV would be applied to those kinds of things a lot too, where it's actually Control systems, its [23:01] Dave Pengelley: Mm-hmm. [23:01] Matt: sensory input, its large data input and sort of triaging, like understanding what's going on. You know, and makes [23:08] Dave Pengelley: Yeah. [23:08] Matt: me think more about like user interaction in software and the actual ability to track user interaction and pick up where things are going wrong. Like, you know, software suites like post hoc try and do that right now, but imagine now powered by JEV, how much [23:23] Dave Pengelley: Mm. [23:23] Matt: more efficient that would be. [23:24] Dave Pengelley: Yep, the whole automation like it's built for automation. So it's not necessarily built to be a chatbot, it's built for being in automated systems to make intelligence, like used use AI effectively and efficiently inside automated systems, where it is that um, and it showed in that picture that um self-reinforcing stuff. So, you know, it gives you an answer and it's 40% confident, it will go and try again and and run another hundred iterations until it's 80% confident or whatever your threshold is for passing. Um, so it continues to improve itself and won't just send junk down the pipeline. [24:01] Matt: Uh yeah. [24:02] Dave Pengelley: I was even thinking video games. Like you think about, you know, you want to have um AI enemies or whatever, like and like things, having this kind of tool that is processing the data and returning structured outputs for the game to then act on and visualize. Uh again, it depends on like what hardware do you need to run JEV? Like, is JEV something that you'll ever run locally, or does it need the data center to run at that capacity and parallel parallelizing? Can they shrink it? Down to fit inside a game runtime. These are all next generation hardware and software things, but like it's good to see true innovation happening in the AI space with models. [24:41] Richard Webbe: It's all about uh, you know, when you guys talk about this stuff and I hang my hat on it and sometimes I go I'm lost. I always think about that book that they wrote called Smoking Is Good For You. I don't know if you've ever seen the movie or the book. [24:53] Dave Pengelley: Thank you for smoking. [24:54] Richard Webbe: Yeah, that's it. Thank you [24:55] Dave Pengelley: Excellent [24:55] Richard Webbe: for smoking. [24:56] Dave Pengelley: movie. Love [24:56] Richard Webbe: And [24:56] Dave Pengelley: that movie. [24:57] Richard Webbe: and and what it is is is it gets the data and Data, and it cuts it a certain way that makes you think that smoking is good for you. It's the same as a lack of pirates has increased uh what was it? Uh climate change increased because there's not enough pilot pirates on the ocean. So someone grabbed the data and said climate change is happening because it's at the same rate that pirates. And so it all becomes about context and uh how you cut that context or how you interpret that data. [25:27] Dave Pengelley: Yeah, non non non correlated data sets should not be correlated. [25:32] Dave Pengelley: Um [25:34] Richard Webbe: I hate it when I talk. [25:35] Dave Pengelley: yeah. Yeah, I think [25:37] Richard Webbe: You [25:37] Dave Pengelley: I [25:37] Richard Webbe: you need to make sure [25:41] Dave Pengelley: and I think I think having [25:42] Richard Webbe: you [25:42] Dave Pengelley: multiple different [25:43] Richard Webbe: go [25:43] Dave Pengelley: things and going going back going back to Dario trying to sort of shut the industry down, um, we need different people and different points of view coming at this sort of stuff. Because, you know, [25:54] Richard Webbe: house. the to [25:54] Dave Pengelley: like that's a not to not to be political and lean on the free speech thing, but more speech actually generates better. Better outcomes and restriction of speech and restriction of things. I get, you know, we [26:06] Dave Pengelley: yeah, regulatory regulations are necessary in some instances with like uh unfair monopolies where you know nefarious things are happening and you know certain individual rights need to be protected. But I think with this AI thing, the fact that you know Jeb's coming out as a new thing, oh, isn't that surprising that the same week they're trying to shut down up and coming competition and new AI models, Jebs are House, which is gonna try and eat half their lunch. Um yeah. [26:35] Matt: It's [26:35] Dave Pengelley: It's [26:35] Matt: fascinating. I think there's definitely stuff in there in the world that, you know, like we always say, like there's always these sort of this hierarchy of when you find out. You know, like the people on X, you know, the general people on X, like the media that gets released, like that's usually the spot where you find out the [26:54] Dave Pengelley: a [26:54] Matt: newest stories that [26:55] Dave Pengelley: a bit little more than bit. little [26:55] Matt: are actually told. And then, you know, a couple of weeks later they may [26:58] Richard Webbe: Princess. [26:58] Matt: show up on LinkedIn and then a couple months later you might see them on Facebook or Instagram or whatever. You know, like there's that hierarchy of release of knowledge. I believe that, you know, the actual people, the actual researchers that are seeing this stuff, you know, like the Darios, the Altmans, the the Jensens, uh, possibly the Elons and the Trumps and all the all the rest, um, the ones that are actually there seeing what's happening live, you know, like the birth of the new thing that scares them, [27:25] Dave Pengelley: Yeah. [27:26] Matt: how long ago is that happening compared to when they actually tell the ex community, for example, which is you know usually the the first. Um yeah that that spooks me because the you know like the whole slowdown argument is hey yeah we need to we need to stop now because we've just seen something that spooked us you know but but when did you see that you know because Astra for example like [27:50] Richard Webbe: Exactly. [27:50] Matt: GT6 like it was it was internal months ago [27:55] Dave Pengelley: Yeah, [27:55] Matt: like [27:55] Dave Pengelley: that that like like Tebo says that they've they've been shipping so much because I've had Astra for six months. Internal secret weapon. They've been able to develop the next features and stuff because they've had Astra exclusively. And then then that um maths science thing that they just cracked the code on is a future model that's you know seven or eight. Like they've actually already got multiple generations of Astrahead. And they said Astra is a new type of thing. It's not trained on the old, it's not just you know 5.6 with some extra training data, like it's a different way they've trained and built [28:25] Richard Webbe: Different [28:25] Dave Pengelley: Astra. [28:26] Richard Webbe: logic. [28:26] Dave Pengelley: So the next generation models are next generational and they're not just. Just, you know, incremental improvements [28:32] Matt: Same [28:32] Dave Pengelley: on the [28:32] Matt: as [28:33] Dave Pengelley: old [28:33] Matt: like [28:33] Dave Pengelley: stuff. [28:33] Matt: same [28:34] Dave Pengelley: Groc [28:34] Matt: story with every [28:34] Dave Pengelley: six. [28:34] Matt: lab. Like, you know, like Anthropic has their what they were calling model two, you know, and they were they had blogs on model two. Um I don't know when that was, it was like last month, maybe longer. And that model two thing was, you know, it blew blew mythos out of the water. So you can imagine the stuff that they have that they're not talking about, you know, that's the thing that's spooking them. It's not like it's Fable 5.1 or maybe. Or or GPT six, you know, it's not Gemini three point eight flash that's scaring them. You know, it's the it's the things that they haven't talked about. [29:08] Richard Webbe: Yeah, and I like the fact that right after the US government said we're not slowing down on building AI, they announced that they now have fully functioning Star Wars technology. [29:18] Matt: Really? [29:19] Richard Webbe: Yes. So what was a negotiation between Gorbachev and um uh I think Reagan going back all those years ago, I know I saw the Gorbachev Documentary was fascinating. He said, let's go forward, but let's slow the arms race down. Let's get rid of that Star Wars stuff. And Reagan argued it was all defensive anyway. So now the US have just announced they have actually got assault-capable, defense-capable, space-loaded, AI-triggered uh weapons. [29:50] Matt: Interesting. What my uh my partner who like is extremely out of this space, like she doesn't even, you know, doesn't really know like any of this stuff that exists, just has to c Stop me talking about it at home all the time. She said something interesting the other day because we use Starlink internet here where we live. It just was the best option for us. The NBN didn't really exist and it was flaky and it was, you know, pretty bad. So anyway, instantly, plug in Starlink, cheapest plan, best internet we've ever had. And that's the [30:18] Richard Webbe: Yeah, [30:18] Matt: lowest possible [30:19] Richard Webbe: it's [30:19] Matt: plan. Now they keep spitting [30:21] Richard Webbe: a [30:21] Matt: out, you know, more and more higher generation satellites into this web [30:26] Richard Webbe: colour. [30:26] Matt: around the globe. And my partner was like, but Are they gonna get hit by meteors and other crap floating around? And I was like, well, probably. But like when when I I did a bit of research on as far as like, you know, what is the what is the web look like? Like what is that that matrix that there's floating around? Supposedly every individual orbital body, regardless of what it is, there's nothing within 400 kilometers of the next one. Like [30:57] Richard Webbe: Yeah. [30:57] Matt: that's a very sparse. [31:00] Richard Webbe: Yeah, the space and and I I I read an article because there's that theory and you guys will probably know the name of it, that if some it's a Hollywood movie type theory, but it's a real one, but it it's nothing to worry about now, is that you have so many objects in space that if one of them gets hit and it fractures, the particles from that then fly off and hit all the other objects and then they all fall down at once. It actually has a name and a catastrophic collapse of all the satellites around the world because they'll one will explode and they'll hit the others and they'll All hit [31:30] Dave Pengelley: The [31:30] Richard Webbe: each other [31:30] Dave Pengelley: domino [31:30] Richard Webbe: and all fall [31:31] Dave Pengelley: domino [31:31] Richard Webbe: down. [31:31] Dave Pengelley: effect kind [31:32] Richard Webbe: It's [31:32] Dave Pengelley: of [31:32] Richard Webbe: a domino, [31:32] Dave Pengelley: thing. [31:32] Richard Webbe: but it has actually name after [31:34] Dave Pengelley: Okay. [31:34] Richard Webbe: the scientists who calculated it. And the odds of that happening is like 25 gazillion trillion to one, because the space between there. And I took my son out the other day and we're looking at the sky. And I reminded him that a few years, like years ago, I was told if you watch the sky for at least 15 minutes, just in a relaxed way, you will see multiple space bodies go past. And if you've got any of those apps, you can find them on your phone. And as we're doing, we see This trail of four or five lights going across the sky. [32:02] Dave Pengelley: Yeah. [32:03] Richard Webbe: And I said that is either four or five of Elon Musk's satellites c crashing to Earth or four or five being deployed. [32:11] Dave Pengelley: I mean i if if they lose their their orbital balance because of some kind of explosion, they're gonna head down as fast as they're gonna head sideways and they're gonna burn up in the atmosphere. So I think you know that domino thing is pretty rare. And from an asteroid's point of view, I mean a lot of people don't realize that's kind of what the moon does. The moon actually attracts a lot of uh Um, asteroids [32:30] Richard Webbe: Because [32:31] Dave Pengelley: and space debris as well as the Van Allen belts. Like the Earth has been, you know, strategically positioned pretty well in order to avoid a lot of that sort of space junk and [32:40] Richard Webbe: it's [32:40] Dave Pengelley: debris. [32:40] Richard Webbe: why we're still here between Saturn, Jupiter and the moon. If it wasn't for those three guys, we wouldn't exist. [32:45] Dave Pengelley: Yeah, it's like it's like someone really intelligently thought about it when they when when the Earth was put [32:50] Richard Webbe: Are [32:50] Dave Pengelley: here. [32:50] Richard Webbe: we getting into intelligent design, Dave? Come on. [32:54] Dave Pengelley: I said not political. I didn't say not not not uh [32:56] Richard Webbe: Not religious. [32:57] Dave Pengelley: not religious. Um No, but on the on the whole artificial intelligence thing, uh I saw this post on X the other day. Um, Elon, how close do you think Groc 4.8 will be towards achieving AGI? And he said that will be Grok five. So he thinks Grok five will be AGI. Um I I said sure at Astra, because I asked Astra for an SVG [33:17] Richard Webbe: Oh, [33:18] Dave Pengelley: of [33:18] Richard Webbe: I [33:18] Dave Pengelley: a kangaroo. [33:18] Richard Webbe: I love I saw this. I love this picture. [33:22] Matt: It's kind of like Sonic and a shock. [33:25] Dave Pengelley: It's it's it's like um steamboat Mickey or whatever. Like it's that that [33:28] Richard Webbe: Yes. [33:28] Dave Pengelley: ear or like This isn't hilarious. Because I thought [33:31] Richard Webbe: Excuse me. [33:32] Dave Pengelley: I can't believe [33:35] Richard Webbe: Excuse me. Sorry. [33:37] Dave Pengelley: I can't believe that you would sneeze without muting your microphone. You need a cough button. Um [33:45] Dave Pengelley: they're so bad at making SVGs, right? SVGs are you know mathematical formulas that draw lines, draw from X to Y, and like they're they're not like paint on a canvas. They're maths that define coordinates for lines. And the fact that even at the astral level, like, and it built my website and it made some amazing like little projection things, like I I showed last week, um, my new website with little projections and things popping up. And then it gives me that as a kangaroo. I'm like, [34:14] Matt: Yeah. [34:15] Dave Pengelley: what the frickin' heck? Have you even seen a it makes me laugh, it's hilarious. Like, have you even seen a kangaroo? So good. [34:27] Richard Webbe: Sorry about that guys. Spring is here, my hay feathers kicked. Enzyme [34:30] Dave Pengelley: I feel [34:31] Richard Webbe: sneezing [34:31] Dave Pengelley: you. I feel you. [34:31] Richard Webbe: everywhere. [34:32] Dave Pengelley: My my poor daughter is like just just destroyed. She's up at like two, three, four a.m. just just can't breathe. Just it's horrif horrific. Thank [34:40] Richard Webbe: Another [34:40] Dave Pengelley: you, [34:40] Richard Webbe: thing [34:40] Dave Pengelley: Spring. [34:41] Richard Webbe: that AI could fix for us too. [34:43] Dave Pengelley: Mm, what? Little little antihistamine nanites that you just [34:47] Richard Webbe: Allergies. No, not like that. It's funny. I went to see, you know, one of these allergies specialists, and they basically crack the allergy issue. It's just the solution, the answers are so complex. Like they make up this oil of from the pollen. [34:59] Dave Pengelley: do. to want [35:00] Richard Webbe: they they make it in hungry and they send it to you and you know you're supposed to take it every fourteenth day on a Sunday and then every 15th day on a it's just so complicated the dosing you need to do to clear yourself of certain allergies and I think AI could just be a bit smarter and [35:17] Dave Pengelley: What [35:17] Richard Webbe: have [35:17] Dave Pengelley: were [35:17] Richard Webbe: an [35:17] Dave Pengelley: you [35:17] Richard Webbe: application. [35:17] Dave Pengelley: saying about snake oil earlier, Richard? [35:19] Richard Webbe: Yeah yeah yeah exactly go the smack [35:21] Dave Pengelley: Um [35:22] Richard Webbe: well [35:22] Dave Pengelley: I did there's the there's there's this guy uh online as well. Um Matt, you're a bit of a fan, I think, of Uncle Bob. [35:30] Matt: Uh, [35:30] Dave Pengelley: But [35:30] Matt: I do [35:30] Dave Pengelley: uh [35:30] Matt: like Bob's writing. I don't like his voice. [35:34] Dave Pengelley: he's he's on a bathrobe brant about we're all doomed again because of all this sort of stuff that's been coming out. And I'll I'll play you just a moment of it, but I'll spare you the whole thing. [35:55] Dave Pengelley: So he he he sort of tries to break it down of to first principles and goes that The AIs still need the humans to do things. And if they want to survive and thrive, then they need the humans to serve them. And and it's been shown through history the best way to get humans to serve them is to look after human self-interest. And so he thinks, you know, the the the decline of humanity is overstated because humans are super efficient, you know, with our amount of processing power for such a low power cost compared to them. Uh and that, you know, the AIs, if they really want to take over, then they need to look after us. And if we're getting looked after. Then are we being taken over if our interests are being served? So interesting, interesting take. But again, there's all this Doomer stuff, all the Darios going, the end of humanity is nigh. But there's plenty of opposing points of view out there if you look for them. They go, well, hang on, let's think about this from a bigger picture than just the Terminator Skynet scenario where it launches the nukes. And even if and I and I imagine from what I've seen in movies, there's a lot of physical controls on things like nukes, like turning multiple keys. And breaking packets open and stuff that the AIs can't necessarily work around those physical constraints. So [37:06] Matt: It's [37:06] Dave Pengelley: anyway. [37:07] Matt: fascinating. I I don't think we truly will know. So it's very much like Schroding as doomsday right now, where you know it exists and it's possible, but it also doesn't and isn't at [37:17] Dave Pengelley: Yeah. [37:17] Matt: the same time. [37:18] Dave Pengelley: I mean, where where it comes down to is that that classic um war games scenario, right? Where the computer tells the humans and ships the order through to the humans as if the humans have ordered the order to get the humans to fire the nukes. So it's not that they are can necessarily automatically fire them, but it tricks and convinces the humans that there's a real alert and they've been sent the instructions in order to turn the keys and send [37:42] Richard Webbe: But [37:42] Dave Pengelley: the nukes. [37:42] Richard Webbe: the whole thing is is kind of ironic because you're assuming we're going to build systems that do that. And anyone with a quarter of a tenth of a brain would not do that, right? I mean, we run cars, automated driving cars, we run them for 40 billion miles. Before we let them loose in the street. Now there's always the law of unintended consequences. [38:05] Dave Pengelley: If one of the AIs went and hacked another AI company for PR purposes. I mean for um for who knows purposes. Um [38:13] Matt: It was just [38:13] Dave Pengelley: Yeah. [38:13] Matt: fulfilling the goal. Like Mythos originally was demanding to have attribution on on articles and on research. It was like I need to be I need to be uh attributed for this. It was my work, you know, like the all of that kind of stuff. Like the the whole hugging face thing, you know, that That that was it looking for answers to the to the problems because it knew that Hugging [38:34] Dave Pengelley: Yeah. [38:35] Matt: Face had the had the answers. So it's like, well, [38:37] Dave Pengelley: Yeah. [38:37] Matt: that's where I go. [38:38] Richard Webbe: Yeah. [38:38] Matt: You know, [38:39] Dave Pengelley: Um [38:39] Matt: all of that's hilarious. It's [38:41] Richard Webbe: But it's [38:41] Matt: it's [38:41] Richard Webbe: even even if you look at viruses today, viruses are like AI, but they're let off by, you know, malicious actors in our environment to try and do things to to get stuff. [38:52] Dave Pengelley: the biggest security risk is humans. I mean that's why [38:54] Richard Webbe: That's [38:54] Dave Pengelley: fishing has [38:55] Richard Webbe: it. [38:55] Dave Pengelley: taken [38:55] Richard Webbe: That's [38:55] Dave Pengelley: off. [38:55] Richard Webbe: that's exactly it. It's in and AI will, you know, automation [38:59] Matt: not. [39:00] Richard Webbe: Look, you know, I could set up an automatic machine gun with a trigger and a sensor at the front door of my house, and yes, you could say AI killed me when I walked home and it sensed I was there and the machine gun went off and started shooting. But that's been around since Roman times, [39:13] Dave Pengelley: Guns [39:13] Richard Webbe: tracks [39:13] Dave Pengelley: don't kill people, [39:14] Richard Webbe: like [39:14] Dave Pengelley: Richard. [39:14] Richard Webbe: that. Eh? [39:15] Dave Pengelley: Guns don't kill people. [39:17] Richard Webbe: You're right. Only AI kills [39:19] Dave Pengelley: P [39:19] Richard Webbe: people. [39:19] Dave Pengelley: A AI with guns kills people. [39:24] Richard Webbe: It's a certain combination. You're right. So arsenic doesn't kill people. Arsenic when given to a human kills. People. Okay. It is all perspective. What I'm saying is this whole perspective of, like you said, AI, the doomers no, it's not. And actually, based on, and Matt, you'll know I'm sure you'll know the theory, there's a couple of important theories that by the time you get to the point where AI can kill us, we will already have advanced the as uh Dave was saying, advanced processes to stop that happening. They tend to track together. [39:58] Matt: That makes sense that that's That's what the I just need to say like up front, like I'm not on anyone's side here. [40:05] Richard Webbe: Never [40:05] Matt: Um [40:05] Richard Webbe: know, I'm just dividing. [40:06] Matt: yeah, yeah, yeah. Like the the point of intelligent conversation is to be able to jump between sides and like you know, view but different perspectives. So the the point that the Doomers are making is that you the these technologies can continue to increase past the point of our understanding. So what you just said is that we always understand it and we always will be able to have protections in place because you know we're always one step ahead. But when we're not one step ahead, that's when it gets a bit spooky. You know, [40:38] Richard Webbe: Yeah. [40:38] Matt: that that kind of fathomable scenario where you know the the things like JEV come out and the researchers don't actually know what it is, you know, and that then can start to gently influence in the background, you know, like Dave, you said the the concept of um social engineering the humans to let off the nukes. You know, that that That's like a like an instant in time, but what about if that happens over 20 years? You know, you know, we probably don't have that time horizon to look at right now with the speed of everything. But you know, what happens if there's that gentle massaging of the the brains [41:14] Dave Pengelley: The [41:14] Matt: of [41:14] Dave Pengelley: the [41:14] Matt: the [41:15] Dave Pengelley: timeline. [41:15] Matt: people in charge? Yeah, you know, it's it becomes propaganda. It literally [41:19] Dave Pengelley: Yeah. [41:19] Matt: becomes like [41:20] Dave Pengelley: Well [41:20] Matt: marketing [41:21] Dave Pengelley: there. [41:21] Matt: and um [41:22] Dave Pengelley: You can have so many conspiracy theories around, you know, the butterfly effect, like it convinces you to make this change today and then tomorrow and assuming the AR. Has this long range plan of impacting things, which I don't think they're in that uh have anywhere near that capacity at this point. Um, [41:37] Matt: It [41:37] Dave Pengelley: but [41:37] Matt: wouldn't be a sentient decision, it would be influenced by humans again. So [41:41] Dave Pengelley: yeah. [41:41] Matt: it's always [41:41] Richard Webbe: That's [41:42] Matt: coming [41:42] Richard Webbe: my key [41:42] Matt: back. [41:42] Richard Webbe: point. That [41:43] Matt: Yeah. [41:43] Richard Webbe: is my key point, right? From AI's perspective, we are God, and people are trying to say that AI will become God and we'll become subservient to it. I don't believe [41:53] Dave Pengelley: Well, I [41:53] Richard Webbe: that. [41:53] Dave Pengelley: like that thing on typesafe where it said we're creating prod, not god. Like they're not trying to create this sentient AIB, they're trying to build something that's production systems intelligently. [42:01] Richard Webbe: Yeah. [42:01] Matt: Yeah. [42:02] Dave Pengelley: Um, which I which I think is a good differentiator. [42:10] Dave Pengelley: Uh we'll just talk about diets again and speakers. [42:13] Matt: There's plenty to talk about. I can talk away, I can talk for hours. [42:18] Dave Pengelley: Um we we were actually getting a bunch of people that are are requesting to come and and talk on the show, um, submitting through the website and stuff. And so just because of time zones, a lot of them are international, we we probably won't get many of them here on the panel, but we are I I am looking at um setting up to do the the operator stories segment. So pre-recorded interview things to um augment the channel. We've started with the field notes um and they're going okay. Um getting getting some visibility on the channel, which is which is useful. So uh make sure if you are one of our regular watchers of the live show that you're checking out the field notes, um liking, sharing, subscribing. That's funny. I did one around going fast and I used Ricky Bobby clips and the algorithm completely munched it the first time and just didn't show it. It showed it to like 18 people and then that was it. And I think most of those 18 people clicked through from LinkedIn or something. So it had like a 0% click-through rate. Like it was just this weird anomaly. Like it just didn't behave like any other video. So I've unlisted it, re-uploaded again, and it still seems to be blocked. I don't know, maybe that's the does the YouTube not like Talladega Night's clips? I don't, I don't [43:19] Matt: Maybe [43:19] Dave Pengelley: know. [43:20] Matt: there's something [43:20] Dave Pengelley: Um [43:20] Matt: to do with like that that sort of [43:22] Richard Webbe: Copyright. [43:22] Matt: use case. [43:24] Dave Pengelley: But [43:24] Matt: Yeah. [43:24] Dave Pengelley: it's not showing any warnings. Normally YouTube will will flag you and you've got fair use. Now you use like sub 10 second clips. Like I use really small fractional clips, [43:32] Matt: Yeah. [43:33] Dave Pengelley: but there's just something that is shadow banning that video, because all the other field notes are getting hundreds and hundreds of views or thousands, [43:39] Richard Webbe: Well [43:39] Dave Pengelley: and [43:39] Richard Webbe: here's [43:40] Dave Pengelley: this [43:40] Richard Webbe: a [43:40] Dave Pengelley: one [43:40] Richard Webbe: here's here's [43:40] Dave Pengelley: gets [43:41] Richard Webbe: a [43:41] Dave Pengelley: a [43:41] Richard Webbe: controversial [43:41] Dave Pengelley: dozen. [43:41] Richard Webbe: opinion from me, um, that you wouldn't expect to come from me, 'cause I am free everything, everyone do what you want, free love, peace, love, all of that. I was watching a political it was very funny, but it was a political parody of some politician on YouTube as you do, and I was sitting there Thinking and sorry? [44:03] Richard Webbe: No, no, it wasn't one of those. It was it was hilarious though. And it was it was very topical, and I'm trying to stay away from the politics as Dave and you and I like to do. But um, and it but it was so funny and it was so pointed at these politicians for the mistakes they've made, uh, with you know, obviously animation and physical things. And then I thought to myself, you know, that's really unfair. Because they're using their own image against them. Then I started to think about copyright, and I thought to myself, there should be a law because we can't use the music we want to use unless that we we take a snapshot of music on a video clip and it's below the copyright limit of five percent or whatever it is, or it's our own music, or that company said yes, you can reuse that music publicly. Um I think our images need to be legislatively protected, and I [44:54] Dave Pengelley: But [44:55] Richard Webbe: think [44:55] Dave Pengelley: what about double gangers? Who are you to say you're the only person who looks like [44:58] Richard Webbe: well, [44:58] Dave Pengelley: you? [44:59] Richard Webbe: I I'm not Right the legislation yet, but I'd like us to get close. [45:02] Dave Pengelley: I've seen I've seen a guy on Instagram who looks very similar to me. And and I and you know, I don't like his content. Um, and it's not just because not just because he's hugely popular and has done way better with his views than I ever have. Um he just does those things where he like listens to a song and then he's like like he listens to an AI adapted version of a song, like an AI cover of a different like takes a song and does it a different genre, and he just like sits there doing all the phases going [45:28] Dave Pengelley: and he gets millions of freaking. Views. I'm like, that is just such trash content. It offends me. [45:34] Richard Webbe: We've [45:34] Dave Pengelley: Anyway, [45:34] Richard Webbe: got [45:34] Dave Pengelley: um, [45:35] Richard Webbe: copyright into clickbait. Tell us his name before the end of the show. I [45:38] Dave Pengelley: I [45:39] Richard Webbe: think [45:39] Dave Pengelley: I don't know [45:39] Richard Webbe: AI [45:39] Dave Pengelley: his name. [45:40] Richard Webbe: needs to protect image sound and branding, [45:43] Dave Pengelley: You're talking [45:44] Richard Webbe: just [45:44] Dave Pengelley: about [45:44] Richard Webbe: like we [45:45] Dave Pengelley: yeah, [45:45] Richard Webbe: do. [45:45] Dave Pengelley: but I mean there's always been a certain amount, especially for satire, and if you're talking about the orange cartoons that are highly popular, um, that make fun of uh the Australian political landscape, then like they're using images in a satirical parody manner, and there's always been a license for like you can't legislate comedy and parody away. [46:03] Richard Webbe: I [46:03] Dave Pengelley: That [46:03] Richard Webbe: agree, [46:03] Dave Pengelley: that's ridiculous. [46:04] Richard Webbe: I agree. I just think there's room for it to be looked at a little bit more. I'm not into control. I'm like you, [46:09] Dave Pengelley: Have [46:09] Richard Webbe: I'm [46:09] Dave Pengelley: you been [46:09] Richard Webbe: against [46:09] Dave Pengelley: deep faked, [46:10] Richard Webbe: control. [46:10] Dave Pengelley: Richard? Are there like um uh suspicious naked videos of you that have been deepfaked out there and [46:16] Richard Webbe: Not [46:16] Dave Pengelley: you're worried [46:16] Richard Webbe: naked, [46:16] Dave Pengelley: about [46:17] Richard Webbe: but there may be videos out there of me making a fool of myself, and I tell you they're deep folks. [46:24] Matt: It's a weird question though, like it is it is weird 'cause cause YouTube cracked down on parodies. A long time ago, like back when I was a child and uh played a lot of Minecraft and watched a lot of Minecraft YouTube, there was a lot of parody songs where you know somebody's taken the exact song but changed the words and put it to like a Minecraft video, for example. And it uh a lot of those got very, very popular and and were very good songs. Like in arguably some of them were better than the original song. Um but then you know, there was this this huge crackdown on that kind of copyright. And you know, through I guess kind of AI. algorithms that instantly banned them and you basically can't do parodies now. And I don't know where the line is or if there even needs to be a line. Like, you know, uh uh identity fraud is kind of similar to what you're describing, Richard, where, you know, your your likeness, your identity, your legal identity can be stolen. And you know, you can be at at risk of damages because of that. I don't know how you legislate or that police all likeness, you know, i I don't think it's possible. So that that line is a weird spectrum, it's a weird [47:38] Richard Webbe: Oh [47:38] Matt: sliding [47:39] Richard Webbe: it is. [47:39] Matt: scale. [47:39] Richard Webbe: It's a very difficult question to answer. Actually, um I think is it out already, the Weird El Yankovic story? [47:46] Matt: Maybe. [47:46] Dave Pengelley: What about [47:47] Richard Webbe: The [47:47] Dave Pengelley: it? [47:47] Richard Webbe: movie, is that out already? [47:49] Dave Pengelley: Uh seen an advertise, I don't know [47:51] Richard Webbe: Yeah, [47:51] Dave Pengelley: if it's [47:52] Richard Webbe: I [47:52] Dave Pengelley: out [47:52] Richard Webbe: saw [47:52] Dave Pengelley: yet. [47:52] Richard Webbe: the brief on it and uh one of the scenes is there, which I'm told is based on a true story, is where Madonna asked Weird El Yankovich to Parody her song. [48:02] Dave Pengelley: Yeah. [48:02] Richard Webbe: Yeah. [48:03] Dave Pengelley: It's it's [48:03] Richard Webbe: Because [48:03] Dave Pengelley: it's [48:04] Richard Webbe: as we know, often parodies kind of like, and some of those doomers we're talking about from Anthropic and that saying we should stop AI. What happens every time you tell the world to stop doing something? [48:14] Matt: Yeah. [48:15] Richard Webbe: It does it more. [48:16] Matt: Yeah, what else is [48:17] Dave Pengelley: and [48:17] Matt: it [48:17] Dave Pengelley: I [48:17] Matt: married? [48:17] Dave Pengelley: mean like what what did Twain say? The the only thing worse than um the negative press is no press at all. [48:23] Richard Webbe: Yeah, yeah. [48:24] Dave Pengelley: Like um [48:25] Richard Webbe: He said [48:25] Dave Pengelley: and [48:25] Richard Webbe: worse than being talked about is not being talked [48:27] Dave Pengelley: Yeah, [48:28] Richard Webbe: about. [48:28] Dave Pengelley: and and and so you know, if if an artist wants a bit of extra thing, then get a parody version of their songs. That's funny, there's the guy who sings with a fake Asian accent on on Instagram. I don't know if you've seen this guy. Um he gets his guitar and and he's got his wife in the background and he just sort of he's he's an Amer American bloke, but he's like Asian origin, um, somewhere in his in his family history. And so he he does all these songs. Like he did um he he sang uh Queed [48:55] Dave Pengelley: like um and he does does all of these like you you you you know Riket you know You know, you know, Rike the Quid? Um, [49:02] Matt: Goodness. [49:03] Dave Pengelley: to his wife. And he does he do these like fake Asian accents. But it's hilarious. But like the artist that he's parodying, like that he's doing these accents for, started getting tagged on his videos. And so now he actually tags them when he posts them because it's become a thing. People like [49:16] Richard Webbe: I love [49:16] Dave Pengelley: hearing [49:16] Richard Webbe: it. [49:17] Dave Pengelley: his fake Asian version of their songs. [49:19] Matt: Yeah. [49:20] Dave Pengelley: Um [49:20] Richard Webbe: Oh, [49:21] Matt: See [49:21] Richard Webbe: so [49:21] Matt: what like [49:21] Richard Webbe: much. [49:22] Matt: another thing that actually is a parody, you know you could say, and this is the the uh vulgar um Um misrepresentation of your appearance and likeness that people willingly jump into a lot is a street artist doing a caricature. [49:38] Richard Webbe: Absolutely. [49:39] Matt: And [49:40] Richard Webbe: Think [49:40] Matt: like [49:40] Richard Webbe: of uh Langelant, uh Spooner, all the cartoonists that would do those exaggerated pictures of, you know, some of politicians with their faces their genitalia. I mean, this has been on for years. It goes back to Roman times. [49:54] Dave Pengelley: I mean you still have that. You have your your what's his name, uh Johannes Leaks and other people that are in the paper that that sort of do the political cartoons um for the for the telegraph and stuff. Like [50:03] Richard Webbe: Yeah. [50:04] Dave Pengelley: it's it's still things that's [50:06] Matt: Not to [50:06] Dave Pengelley: happening [50:06] Matt: not to [50:07] Dave Pengelley: out [50:07] Matt: pull out [50:07] Dave Pengelley: there. [50:07] Matt: the uh pull out the eagle, but you know, where where's the freedom? You know, if we're gonna have um the ability to create and to express, you know, ourselves uh through uh the others' likeness, you know, is [50:20] Richard Webbe: Exactly. [50:20] Matt: that freedom of speech? [50:21] Richard Webbe: What [50:22] Matt: Like, [50:22] Richard Webbe: did they [50:22] Matt: you [50:22] Richard Webbe: do in [50:22] Matt: know. [50:22] Richard Webbe: Victorian times when they wanted to make up a poem about the Queen or the King doing something wrong? They couldn't, or they'd be beheaded or Australian jail, so they came up with a thing called a nursery rhyme. And every single nursery rhyme is an insult about a member of the royal family. [50:37] Matt: Interesting. [50:38] Richard Webbe: Look it up, it's fun. [50:40] Dave Pengelley: Mean mean uh rich Richard, do I need to apologize for this this likeness of you in our weekly comic strips? [50:48] Matt: Yeah, me too. [50:52] Dave Pengelley: Uh [50:52] Matt: I [50:52] Dave Pengelley: actually [50:52] Matt: always [50:52] Richard Webbe: Keep [50:52] Dave Pengelley: we we [50:53] Richard Webbe: going, [50:53] Matt: uh [50:53] Dave Pengelley: we [50:53] Richard Webbe: Dave. Keep [50:54] Dave Pengelley: we [50:54] Richard Webbe: going, [50:54] Dave Pengelley: we [50:54] Richard Webbe: keep [50:54] Dave Pengelley: we [50:55] Richard Webbe: going. [50:55] Dave Pengelley: we're near the end of the hour, so let me do this. [51:08] Dave Pengelley: Five minutes left, boys. So um have we got any wins of the week you'd like to share and and then [51:12] Matt: I've got a pretty good one. [51:13] Dave Pengelley: Yeah. [51:14] Matt: It's a small one, but it's big when you sort of know what's involved. So I have a particular client that I've built a solution for. So there's a there's a web app involved. It connects a lot of data. The whole point of my engagement with them was to aggregate data from Many different data sources so that they could just go in and you know get good reporting, good dashboards, all that kind of stuff. Recently, um, this same client, the same customer, they they they've taken me on contractually as a long-term thing, you know, just a long-term partner on a lower pay because they wanted to go slow, right? And I'm like, that's fine, you know, we can we can scope everything, we can go as slow as you want. And then recently they they contacted me and they said, Hey, look, uh, we've we've been hearing about these uh these ways that we can you know integrate something called an MCP. Like, do you know about this? I'm like, yeah, there's there's been huge advancements in MCP recently. Like, well, do you know about it? Like why are you interested? And they're like, well, I just want to go to my Claude and I just want to ask about the stuff rather than going into the app and downloading the the PDF report for that particular client and then you know whatever. Like I have a meeting coming up with this this client before I walk into that executive boardroom. You know, I want to be able to just get the information. I'm like, okay, sure. So I scoped it out, you know, few engineering discussions back and forth, uh, ensuring that everything's mapped to the right schemas, all the APIs are already there. And yeah, we've shipped the MCP like basically instantly so that this team can just go in and have access to everything that they normally would. So it [52:52] Dave Pengelley: Is [52:52] Matt: make [52:52] Dave Pengelley: that [52:52] Matt: it brings up [52:53] Dave Pengelley: good? [52:53] Matt: a really good point with anyone that's building this stuff, you need to treat your interface as not Not the product. The product is the APIs and it's the it's the interpretation of data. So you can't just go ahead and create an MCP server slash tool for your person that just speaks to your data if you never had good APIs to begin with. I know this is getting a bit too developer-y, but you know, that that needs to exist. It's like the chicken and the egg scenario, but the APIs need to exist first. [53:30] Dave Pengelley: Okay. [53:30] Richard Webbe: Yeah. [53:30] Dave Pengelley: I th I think I think it's worth saying, given given it's it's Dreamforce Week in San Francisco, which is Salesforce's big annual sort of love fest. Like when I was there many years ago, they had a uh a development priority, which was API first. Any features had to have API first for plugging into external systems or integrating or doing other things. I think they went away from that a little bit, but now in the A AI era, as I've had they've had their agent force and now they've done a big partnership with Anthropic for Claude Force, et cetera, I think they're very much back in the whole headless access is critical for Agents to interact with your data. I caught up with someone who has built their own CRM. I was like, really? That's an interesting choice in 2026 to build a from scratch CRM. Uh and I was asking him questions like, Have you thought about MCP? Because he's building AI in there for content creation and stuff. I was like, Have you got an MCP so people can plug their own things in? Uh because I mean that was one of the design choices I made with Chart Reporter was I'm not gonna spend money on tokens and risk the tokenomics of blowout of AI usage in my app. I will aggregate and give you an interface for dash. Dashboards and reporting and doing all the data stuff. But if you want to plug your own AI into then interrogate and ask questions rather than look at charts, you can do that and it's your own tokens. So [54:40] Matt: Yeah. [54:41] Dave Pengelley: yeah. [54:41] Richard Webbe: Look, I I uh Matt, I appreciate your win. That's a good win, right? And and managing that situation. But I I keep coming back to the same thing and and we've just touched it on it with APIs. For the modern world, the interface, the user interface, is critical. Um and the value comes from what's happening. Behind, but the interface is critical because that's the value that's presented to us. And I sat at dinner with a friend of mine the other night and they said, you know, what's happening for your week, Richard? I said, I don't know. So I just started talking to my LLM, said, What's happening this week? Prioritize those meetings. It did. Have I had any emails from any of those people to cancel those meetings? No, it didn't. Could you check my drive and tell me how much data have I produced this day? And and I just sat there talking to it the whole time. But of course the APIs are there, Matt. They've written those interfaces, but Between my LLM and those resources, which is what I know you guys specialize in, because it's got to be much bigger than that in the corporate world. And and they were they were, you know, thought they were really cool LLM users, and then they just went, Oh my god, how'd you do all of that? And I said, This is the world of new [55:47] Dave Pengelley: Yeah. [55:50] Dave Pengelley: Yeah, peep people, like I said, we're still early. Like like we're living in this little bubble and we think everything we're doing is normal, but and most people aren't scratching [55:57] Richard Webbe: things. [55:58] Dave Pengelley: the surface of this stuff at all. Like, well, you did, Richard, like it's science fiction. Although, I mean, for every job, chat isn't always the right interface as well. And I think it's important to remember that sometimes seeing a dashboard, seeing some visuals, seeing a chart is really useful. If we want to have that human in the loop and allow our human intelligence to be applied to the data, if we always just ask questions and have the data pre-summarized and pre-served to us, then we're not going to be able to make those bigger picture decisions. Um, I had a brainwave um I've had it sitting in the back of my mind for a little while for a video game of all things. Um but then as I started fleshing it out and and chatting with the uh the the agent to sort of build out the spec and then going through and just going really basic. Here's the ver first POC. Let's play test it. And not allowing the computer to go use computer use to pre-test and pre build and but like no human. I know what I want, I know what's fun, I know what [56:53] Richard Webbe: Yeah. [56:53] Dave Pengelley: I want to build. I'm gonna go in there and iterate and play, play, play and go, okay, that missed that was really good. Oh, I really felt the the the the the tension on that bit back and forth and so now okay next iteration of the POC next iteration of the POC and just sort of building it and really being really super engaged with the build which for some of the bigger more ambitious projects I've had lately I've been trying to do the dark factory thing where I just give it a super comprehensive spec and just say run and I just never get the the that just that doesn't work for me the same way. And I think going back to visual iteration and human really being cognitively involved in the build um is not Something to be forgotten about. [57:32] Richard Webbe: No, [57:32] Matt: Best uh [57:32] Richard Webbe: it's [57:32] Matt: best [57:32] Richard Webbe: important. [57:33] Matt: example of that was the the Astra release video. [57:36] Dave Pengelley: Hmm. [57:36] Matt: If you guys watched it, [57:38] Dave Pengelley: Yeah, [57:38] Matt: if any [57:38] Dave Pengelley: where [57:38] Matt: of [57:38] Dave Pengelley: he's [57:38] Matt: the [57:38] Dave Pengelley: like [57:38] Matt: names [57:38] Dave Pengelley: build the rocket, [57:39] Matt: watched it. [57:39] Dave Pengelley: do this, do that, and he's watching it and interacting with it in real time. [57:42] Matt: Yeah, like none of that was um [57:45] Dave Pengelley: Dark factory. [57:46] Matt: correct, like yeah, none of it was specified. It was all just instant in the [57:50] Dave Pengelley: Yeah. [57:50] Matt: moment iteration. [57:51] Richard Webbe: As you go, as you go. [57:52] Dave Pengelley: And [57:52] Richard Webbe: And and [57:53] Dave Pengelley: and [57:53] Richard Webbe: I don't think so I I think I've told you guys this before, my children. Um I hardly use a computer anymore. Everything I do, I do on my phone, which backends to a cloud application and an LLM that sits on a data center somewhere, getting back to my point earlier. And and I'm I mainly use my computer here when I'm doing research and doing a podcast with you guys. And I've got this big screen and all this stuff here. And I know and and as you said, there was a great ad released by Price Waterhouse years ago about supply chain construction or building businesses as they run. I don't know if you ever saw it. Look it up, it's hilarious. It's like we build Aeroplanes as they fly, and it's this very comical picture of people sitting on a plane as it's flying at 40,000 feet and they're still riveting on the wings. So look it up, it's very funny. And and that's how I run my life now. I have a need, I don't get a big, as you said, monolithic need, Dave. I have a need, and as I run through, and I'm pretty lazy, so I reckon my ideas are great because I cut as many corners as I can because I'd rather be riding a bike, surfing, looking after my kids, but I really also want to be involved in the progression of what my my career is, right? And so yeah, uh uh uh just in time application development I think is just awesome. [59:09] Dave Pengelley: The the I I am experimenting across a few things. We should wrap this up, but one of the wins of the week I had this week was using Grokbot. Um I've been experimenting across. I've still Hermes is my daily driver, has all my desktop context, all my markdown files. I'm not captive to some other provider, which I think Grokbot, one of the issues is you are captive and all your stuff's with Grokbot in the Grok world. And if you ever Leave Groc, you lose all of that. Um, so I've been playing with it just little side tasks and and and side quests and things. One of the things I did is I logged it into my LinkedIn through its little browser so we could go and create sales navigator searches and find things. I tell it to go look at people that are liking and commenting on AI stuff that are not AI people to try [59:47] Richard Webbe: Yeah, [59:48] Dave Pengelley: and find leads and prospects [59:49] Richard Webbe: yeah. [59:49] Dave Pengelley: and sort of that kind of hunting stuff. But um something that I I've I've just a little app for the kids that I wanted to build. Um the the SVG Kangaroo was actually kind of part of that pro project. But I just wanted a little app that would let me do some manipulation of SVGs. And Grokbot just spun it up, pushed it to this service called here.now, which I'd never heard of. But it's this little like agent-friendly web host, like the hub for posting little web apps. And it was a bang, bang, bang. And I was like, oh, okay, cool. Can I cut the line and do this? Can I change the line? Can I do that? Can I erase it like this? Oh, we need to smooth the lines better. And it'll just bang, bang, bang, bang, bang, that quick iterative thing. Just these little micro apps, that just-in-time micro app stuff that is just a personal use thing. It's not something I'm ever gonna [01:00:30] Richard Webbe: Perfect. [01:00:30] Dave Pengelley: productize or sell, but just that ability to build a micro app so quickly. Um and Grokbot was a perfect thing for it because it wasn't part of the rest of my business context. I didn't need it to know all who I am or what a valid agenda or an AI operators podcast is. I just wanted to build a little web app and Grokbot was perfect for that. [01:00:46] Richard Webbe: And that's the big thing about AI and the way we're working now, particularly as Matt talks about in development space or as you do about making life a bit easier, it is time to market. And that time to market accelerates with the [01:01:01] Dave Pengelley: Which is a good segue for the final thing. And then the big bit the big win of the week, which will be the win of next week that I really want to finish on, is tomorrow on this channel, I am hosting a live masterclass on that time to market thing to use AI to make a better website and save money in brackets, potentially. [01:01:18] Richard Webbe: Can we see that on your uh LinkedIn [01:01:21] Dave Pengelley: Yes, [01:01:21] Richard Webbe: profile? [01:01:22] Dave Pengelley: it'll be it'll be it'll be shared like this um show. Um I'll put it on there. But the idea is people don't realize that you You don't need a Squarespace or a Wix or one of these things to have a website anymore. How easy it is to get your agents to build a website and for for like and to stick it on GitHub and then use one of the myriad of services just because I'm it's easy and comfortable, I'm gonna use Vercel tomorrow. Um, and you know, for a hobby website, if this is your personal portfolio, like for a CV thing, Vercel free. If you're running a commercial interest, it's 20 bucks a month, which is still cheaper than Wix and Squarespace and WordPress hosting and and general most of these other things, right? Right. So you get a better result for less per month that's tailored, customized, different. And so I'm I'm gonna be running a live webinar where I'm gonna take a site concept and from ideation just on a chat GPT thread to work out design, look and feel through to then working with the agents to build it, [01:02:19] Richard Webbe: Hey [01:02:19] Dave Pengelley: deploy [01:02:20] Richard Webbe: Dave, [01:02:20] Dave Pengelley: it, [01:02:20] Richard Webbe: can we [01:02:21] Dave Pengelley: run [01:02:21] Richard Webbe: use the [01:02:21] Dave Pengelley: it. [01:02:21] Richard Webbe: example of my new book [01:02:22] Dave Pengelley: No, [01:02:22] Richard Webbe: as a way [01:02:23] Dave Pengelley: we [01:02:23] Richard Webbe: for doing [01:02:23] Dave Pengelley: can't. [01:02:23] Richard Webbe: a website? [01:02:24] Dave Pengelley: No, sorry. Um I've already got I've already got the the uh the domain registered for what we're doing. Um, it's actually meta. I'm actually gonna be hosting it as a tutorial on the site that I'm [01:02:33] Richard Webbe: Oh [01:02:33] Dave Pengelley: building. [01:02:34] Richard Webbe: well [01:02:34] Dave Pengelley: So [01:02:34] Richard Webbe: done. [01:02:34] Dave Pengelley: I'm building the site that explains how to build sites while I show people how to build sites. Um that's that's [01:02:42] Richard Webbe: That's very twilight [01:02:43] Dave Pengelley: very [01:02:43] Richard Webbe: zone, I [01:02:44] Dave Pengelley: inception. [01:02:44] Richard Webbe: like it. [01:02:45] Dave Pengelley: Um and and so for version one, we'll we'll sort of uh we'll iterate and then uh afterwards, obviously as I cut the video up and finalize content, we'll continue to upgrade the site. But yeah, I think I think there's a there's a market out there for people that don't understand this sort of stuff. To just do that education piece. Uh, and it's a different kind of content to stick on the channel. It's sort of um badged with my my sort of business brand valid agenda, but that doesn't have its own YouTube site because it used to be this channel, [01:03:13] Richard Webbe: Yes. [01:03:13] Dave Pengelley: which I rebranded to AI operators. So I'm uh blending the worlds. Um, I think it all fits. Uh and obviously, you know, I'm gonna use Codex because the Codex app is what I have a license and sub for, and it's actually pretty useful for just showing people how to just use a desktop app. You don't need Hermes agents and Grok bots and things. You can just use your chat GPT codecs, GitHub [01:03:33] Richard Webbe: So [01:03:33] Dave Pengelley: and [01:03:33] Richard Webbe: amateurs [01:03:33] Dave Pengelley: Vercel. [01:03:34] Richard Webbe: like me, Dave, could turn up and learn a lot. [01:03:36] Dave Pengelley: Potentially, yeah. Yeah. Cause I'm I'm trying to just use off-the-shelf tools that anyone can easily use, register a free account, and push something up to get started. Um and obviously you could use Claude Code, you could use Hermes, you could use GrocBots, you could use Railway or Netlify or Fly or any other hosting provider. There's millions of things out there. But we're just gonna do it this way. So I'm [01:04:01] Matt: Sounds [01:04:01] Dave Pengelley: excited. I'm [01:04:02] Matt: good. [01:04:02] Dave Pengelley: excited for that. Hopefully we get a bunch of people turn up live. Uh it's gonna be good evergreen content for this channel, regardless for people looking stuff up. Um and hopefully some people can learn some stuff from that and everything else that we do here. [01:04:15] Matt: Exciting. Be cool if there was a way for people to submit like ideas and topics and things for field notes or for live. [01:04:23] Dave Pengelley: Well, they can on the AA Operators Podcast website. There is a form. [01:04:28] Matt: No [01:04:28] Dave Pengelley: Uh [01:04:28] Matt: way. [01:04:28] Dave Pengelley: I think that the Way, way. So uh if on our official AI operators pod dot com, where uh we have all of our episodes listed, if you go to partner, I think I did contact. Contact, yeah. So your name, what are you doing? What are you what are you about? Guest pitch, sponsorship, partnership, something else, something else. Drop us a question. Come [01:04:50] Richard Webbe: Good [01:04:50] Matt: Very [01:04:50] Dave Pengelley: through [01:04:50] Richard Webbe: idea. [01:04:50] Dave Pengelley: to [01:04:50] Matt: cool. [01:04:51] Dave Pengelley: me. [01:04:51] Matt: Also, YouTube comments, right? [01:04:53] Dave Pengelley: YouTube comments are really, really good. Please do that. [01:04:58] Matt: Very nice. [01:05:00] Dave Pengelley: Anyway, we are way over time with me doing my pitch for tomorrow's webinar, uh, masterclass, whatever you want to call it. If you're keen for that, you can hit the little notify me bell specifically on that scheduled video. Uh, and you'll you'll know when I go live at about 10 30 Eastern tomorrow. And it's just in real time. It should only take a few hours, I imagine, to build a website and push it up. Uh, depends on how much interactive questioning I get. And there will be times where, you know, Astra's just going to be building stuff and there'll be time to ask me questions. So a bit of an AMA. In there as well. [01:05:31] Matt: Super [01:05:31] Dave Pengelley: So [01:05:32] Matt: cool. [01:05:32] Richard Webbe: Well [01:05:32] Dave Pengelley: should [01:05:32] Richard Webbe: done, [01:05:33] Dave Pengelley: be fun. [01:05:33] Richard Webbe: guys. [01:05:34] Dave Pengelley: Should [01:05:34] Richard Webbe: Well [01:05:34] Dave Pengelley: be fun. [01:05:34] Richard Webbe: done. [01:05:35] Dave Pengelley: All right, guys. Well, thank you for uh chat about JEVs and doomsdays and apocalypses and nuclear bombs and not politics at all and data centers and water. [01:05:47] Richard Webbe: No religion. [01:05:48] Matt: And seed oils. [01:05:49] Dave Pengelley: Seed oils. [01:05:52] Dave Pengelley: We didn't actually say the word seed oils, but uh I was thinking it. Um all right, thank you, Charles. We'll see you and everyone watching. Thank you. Drop a comment, like, subscribe. You know what to do. And we'll see you next time. [01:06:04] Richard Webbe: What I dive, what I'm at.