Everyone's Running AI. Nobody's Getting More Done.

[00:00] Dave Pengelley: Yeah. Uh, Matt's got his AirPods. My, my AirPods are out in another room. I can't bring them out and show you mine on the camera. Um- Much distinction ... there's wired ones. I, I've just started going with the, um, the spatial surround, the spatial sound thing. Oh, yeah. Um, I, I didn't like that at first 'cause it was weird. [00:16] Dave Pengelley: It's not too obvious on most of my songs now, or maybe I'm getting used to it. But I did see this thing that talked about the, um, bouncing of left and right and the, the focus, the, the sort of stereoscopic stuff where you sort of have dynamic stereo noises jumping left, right ears is actually good, especially for a scattered brain, um- Dude [00:37] Dave Pengelley: to focus. [00:38] Matt Slager: Isochronic tones. Is that what it is? You should look for binaural [00:42] Dave Pengelley: and isochronic tones. Yeah, I've seen the I- AD, AD they talk about on, on some of the, the, the stuff trying to make you buy their apps. Um, so as a sub for that, for just regular music that's not specially encoded, I've just turned on the, um, the spatial sound thing with the AirPods, um, to try and help. [00:58] Matt Slager: Yep. Yeah. [00:01:00] Isochronic is a, it's a frequency. So in the background- Mm-hmm ... there's a pulse. It's kind of like white noise for your brain or like- Yeah ... brown noise, I guess. And, um, and binaural is when you've got stuff that's out, slightly out of phase between each ear. Yeah. And yeah, I might not be defining that correctly. [01:14] Matt Slager: I'm not an audio- [01:15] Dave Pengelley: No ... like, technical person. I, I remember, like, the first time I, I heard it, it was a song, um, I think it was a Switchfoot song, and it's got this guitar riff, and it's like, da, da, da, da, da, da, da, da, da, da, da. And I was like, sat in my car and I'm like, oh my goodness, and just the sound bouncing. [01:30] Dave Pengelley: Like, I love when a band does that really, really well- Yeah ... and actually uses left and right effectively. [01:35] AI VO: Mm. [01:36] Richard Webbe: It used to happen a lot in the '80s- Mm ... and the '70s on the old records. You play some of your old records with your old stereo, and I used to get excited. It came out from the different speakers. [01:46] Richard Webbe: Nowadays, not so much. [01:49] Matt Slager: Mm. It's still used a lot, and be it, like you said, for your, um, for sort of neurodivergent kind of people, Dave. Like, it's very, very good for focus and for stimulation. Um, it's very cool. I [00:02:00] love it. You get, you get people now, like, discovering it for the first time, and they've got it on Instagram Reels or TikToks, and they say, "Oh yeah, lift your, lift your phone up to your chin like this." [02:08] Matt Slager: Yeah, I've done that. Yeah. And then, yeah, see? And it, it spits it out, like, either, either ear or whatever. It's wild. Um, yeah, it's such a cool thing though. Um, but yeah, the whole idea of the... What I saw just recently, like people, you know, forsaking the wireless ones and going for the wired ones. Some, some interesting concept, um, where they were vir- not virtue signaling, signaling, but, like, showing that, like, that's the new cool thing. [02:35] Matt Slager: It was, like, you know, if you're like Drake, you know- The, the hipster thing, right? It's that, that kind of- Yeah. All of the money in the world, but, like, they wear them because they're, like, signaling something. [02:45] Richard Webbe: You are, you are the coolest, Richard. You are the coolest. All right. You know what's uncool? Trying to untangle these. [02:51] Richard Webbe: That is really uncool and very annoying, even though I use them all the time, except when I'm on a plane, I use my big headphone I [02:59] Dave Pengelley: [00:03:00] bulged a disc in my back with wired headphones [03:03] Richard Webbe: Wow. [03:04] Dave Pengelley: It's self-inflicted and I'm an idiot. Big time. What were you doing? Bowling. I was ten-pin bowling. Oh. So I was doing this thing called Phantom League. [03:13] Dave Pengelley: This is got nothing to do with AI. I was doing this thing called Phantom League where I had no friends that liked bowling, but I really wanted to bowl regularly. So I joined this thing where, you know, there was a dozen other people in this Phantom League, and you'd just turn up any time during the week, you'd bowl your, your two or three, um, games or whatever, and then they'd tally it up and week to week how you'd went against each other and, and, you know, there's a leaderboard. [03:33] Dave Pengelley: And it got to the last week, and you could be a week behind, and then you go, "I'll just catch up next week," or whatever. But I went up there and went, "I just wanna play one week." And they're like, "No, no, we're wrapping the thing up. You have to play both weeks now." So I had to play, like, six games back to back. [04:00] Dave Pengelley: Like, I had to bowl, you know, 60 frames. That's a, that's a lot of bowling. Um, and I nor- I normally went with headphones in. This was, you know, this was 15 years ago, before there was a lot of wireless [04:00] Dave Pengelley: headphone stuff or before Bluetooth was really minimized to the in-ears and stuff that we have now. So then I had my, uh, my, my wired headphones in and I had them running down my shirt into my pocket and stuff. [04:08] Dave Pengelley: S- But the cable wasn't quite long enough, and so I was hunching a little bit, I think. And so the, being slightly hunched and, like, throwing a 16 pound ball, and I was a spin bowler, down the lane over and over again, being in that slightly hunched position and throwing that weight, all of a sudden my back just went bang, and I was like - Yep [04:27] Dave Pengelley: "Oh, my goodness." Um- Yeah. Yeah ... and, and for, like, for it was a year or more of physio and recovery before I could actually take proper strides again. I was like peng- walking like a penguin for ages. I will say, though, that happened in, like, the second or third of my six games. I finished my six games, bowling, like, just, just walking up and just, just, then just, sort of, just doing the, [04:49] AI VO: uh, [04:49] Dave Pengelley: like, 'cause I couldn't swing. [05:00] Dave Pengelley: I couldn't bowl properly anymore, but I persisted through and finished my six games, like, like a legend, and won the Phantom [05:00] Dave Pengelley: League. Wow. Yes. [05:02] Richard Webbe: Did you get the little rolling ramp thing? The ramp? Did you? [05:06] Dave Pengelley: No, I don't, I don't know that I did, because, I mean, especially 15 years ago, that, like, not everyone, like... [05:12] Dave Pengelley: It's socially acceptable to use that. Everyone uses that now. It used to be that was for the wheelchair people. [05:17] Dave Pengelley: I [05:18] Dave Pengelley: find it, I find it very helpful. Like, yeah. Um, I'm, I'm that old- Using that ramp ... when it comes to my bowling. No ramps, no bumpers. I feel [05:23] Matt Slager: like that's the same as using the, uh, the, the reaching stick on a, on a pool table. [05:29] Richard Webbe: I... It's called the rest, but, you know, some of us call it a reaching stick. The rest [05:34] Matt Slager: kind of makes sense, too, yeah. [05:38] Dave Pengelley: Ah, brilliant. Brilliant. Uh, well, I suppose we should probably do one of these. [06:00] Dave Pengelley: Welcome everyone to the AI Operators show, where we are [06:00] Dave Pengelley: building with AI in the wild, learning, sharing, talking, not hype bro-ing, hype sistering anything. We, uh, don't have any of the ladies on the show this week. It's back to just three amigos. Mano could not join us, and, uh, we'll continue to have different people on through the weeks to weeks. [06:18] Dave Pengelley: Uh, if you're well joining us on the chat, welcome. Say hello as we, uh, don't just talk about headphones and bowling, but talk about technology and AI and business and where all those things converge. To jump straight into [06:32] Matt Slager: something really hectic- Oh ... I feel like this is where my brain has been at recently a lot, and it's the concept of, you know, Richard, you might like this first statement, like we've spent 20 years or more designing user interface and user experience for humans. [07:00] Matt Slager: Now we're in the process of rediscovering what that means and focusing on the actual agent experience, [07:00] Matt Slager: so how it feels and the friction and the general smoothness of an agent working through your software. [07:06] AI VO: Yeah. [07:07] Matt Slager: Um, I've literally... My, my core knowledge thing that I've mentioned a few times, um, I've now absorbed the entire functionality that I was using Linear for, so project and task management, as well as, um, any CRM related workflows. [07:22] Matt Slager: That all lives in my core thing now too, and trying to get my agents to understand what that means and what it is and how to use it and all of the things is, is, is crazy. Like, yeah. That's just something that I've... It's been on my mind a lot. What do you think about that? [07:40] Richard Webbe: Well, it's interesting you should bring that up and, um, I, I think David and I had a chat earlier today, uh, about some of these things and how business is using AI. [08:00] Richard Webbe: I caught up with a friend of mine, and she's running a small business. It's a services business. She's a psychologist. She wants to do invoicing and get calls come in and [08:00] Richard Webbe: greet and... But she's got these disparate systems, right? And she's had to go through a couple of practice managers, and they're disparate, and all of them say we've got AI, and they all automate some part of what they're already doing this much, but none of it's singing together, and it's doing her head in. [08:18] Richard Webbe: And she's so bad, she's thinking, "Why don't I just back... go back to being a single sole trader and, you know, get rid of my other 12 service providers and, and just do it myself?" So there's a big gap in getting everything to coordinate and sing together and make sense for a business. [08:35] AI VO: It's interesting. [08:36] Richard Webbe: And, and, and I think... [09:00] Richard Webbe: I, I, I said to her, "So, you know, u- using AI," and she goes, "Oh, Rich, I know you and David got an AI consultancy and you work with..." She goes, "Well, I am using AI." And that's meant to be the end of the discussion. I said, "No, you're not." Yeah. You've got some software And they sticky tape the letters A and I on the end, and someone told you about Canva or ChatGPT, so you've [09:00] Richard Webbe: accelerated your text messages to your family so they're written better, and your emails, you are not using AI. [09:07] Richard Webbe: AI, as we all know, and I know you know this, I'm preaching to the converted, is about being the conductor of complex processes that you can't really add value to, and you need them to just happen in the background. And that may be inside someone's individual software they're selling you, whether it's an invoicing software or a CRM, but unless it's across your whole business and saving you time and money to do what you do best and push that admin into an automated space, it is really everyone's just getting ripped off. [09:39] Matt Slager: Hmm. Makes you, um, curious about the, the concept of productivity in general in a way. [09:45] Richard Webbe: Absolutely. Absolutely. You just- And I want, I want you to come down here and to my place and get all your stuff and load up, because Dave's done some fantastic work with his stuff. I've had a play with it, but I'm not as smart as you two. [09:59] Richard Webbe: So I, uh, from a technical point of view, it takes me a lot longer, even after a conversation with you guys, to even get something started. But here's the funny thing. I know where the gaps are, I know what I want, and I know I need you guys to help me. How many businesses out there are stumbling along, and it's getting worse because as you introduce AI into disparate systems, your world becomes more complicated? [10:26] Matt Slager: Hmm. [10:27] Richard Webbe: It makes it worse. [10:28] Dave Pengelley: This is, this is, yeah, we, we were touching on this this morning, Richard, when we, when we had a chat while I was out for my walk, huffing and puffing on my AirPods. Um, but I've been thinking about this a- as sort of approaching and, and working with some different businesses and looking at their needs and going, it's one thing to have the business in a box, the AI operating system for a single operator, and have that individualized per desktop, per person, with their mail, their inbox, their so on. [11:00] Dave Pengelley: But when you wanna start scaling that to the organizational context, that becomes a whole different thing. It's one thing to have, you know, Claude Cowork and some markdown [11:00] Dave Pengelley: files and stuff on an individual PC. But as an organization, if you want people to be sort of singing from the same songbook and having those commonalities in there, we are seeing some of the frontier providers having their cloud managed agents and cloud workflows and things like that to try and do some of that. [11:18] Dave Pengelley: Um, and I think ChatGPT and their business plan now have shared agents, so within the ChatGPT interface, you can have, create these sort of shared, uh, I suppose it's the next evolution of their all custom GPTs, but now within a business workspace. I don't know if it's just a rebrand. But yeah, having... And then to your point, Matt, having the key corporate systems, your CRM and stuff, I s- suppose that's where- my, uh, previous employer, Salesforce, is going with things like their Slack bot and their agent force is saying, "Look, we've got all your, all your business context, and you can chat to it through Slack and share that and have that organizational thing." [12:00] Dave Pengelley: But that's great at the enterprise level if you're running Salesforce and paying a squillion dollars for it. What about the, the [12:00] Dave Pengelley: small, medium businesses, the 5 to 50 to 500 employees that may not wanna spend that? And just something I've been working through and trying to work out, what does that solution look like? [12:08] Dave Pengelley: And can you use the off-the-shelf tools? Do you need to bring in extra infrastructure? Where do you start decisioning out, okay, this is stuff we store locally in context on the individual computers, versus here's where we need to, you know, go and pull from SharePoint repositories, from knowledge cores, et cetera, and, and CRMs. [12:27] Dave Pengelley: And, and where is that differentiation when you set up the agents and their systems of local context, shared context, corporate context, et, et cetera? [12:35] Richard Webbe: Hmm. That, that is, that is so true. I said to this, uh, person, who's very smart and a great service provider and has been a friend for 20 years, and I said to her, um, "Why can't your agents talk to, you know, this, that, and the other?" [13:00] Richard Webbe: And she was smart enough to know what the agents were. And she said, "Oh, it's not out in that version of their software yet." And I'm going, "Oh my God. All these s- software providers holding all their [13:00] Richard Webbe: customers to ransom because..." And, and, and you know, sometimes with some of the agents we use, if, uh, we can get around it of course, they go to talk to one of your databases and it's, uh, the database says, "You're not allowed to do that. [13:13] Richard Webbe: Go away. Come back another way." So this data engineering for the SMB as much as the enterprise, it's a really, it's a real thing, and it needs help. Hmm. It needs a lot of help. And you've got some companies doing it at the top end of town, but the biggest automation, the biggest savings and advancement will be for the SMB in having the data ready to be accessed and all the agents singing together. [13:37] Richard Webbe: And we can do that. We can help people do that, and they should reach out to us. But I just find so many people suffering in silence. Yeah. Now I'm going to one of the big AI conferences this week, and I'm gonna be canvassing a lot of people, and if you see me there, come up and say g'day, because I'm gonna be canvassing a lot of people. [13:54] Richard Webbe: Whereabouts? [13:55] Matt Slager: What event? [13:55] Richard Webbe: Uh, am I allowed to tell? Open [13:59] Dave Pengelley: Summit- Yep ... [00:14:00] [14:00] Richard Webbe: I think. That's it. It's, I think it's... Is it an Anthropic thing or? I don't [14:03] Dave Pengelley: know. I think that's a... No, just, just, um, some other AI guy running an AI practice who's, you know, pulled together a big event. Yeah. Good on him. Um- I gave him money. I'm gonna [14:12] Richard Webbe: turn up. [14:12] Richard Webbe: Yeah. And, uh, I think it was sold out in Sydney, and it looks like it might be sold out in Melbourne. We will be running a series of our own AI updates soon, 'cause Dave and I have been talking about doing that- Yeah ... beyond just doing a podcast, so we'll keep an eye out for that. [14:25] Dave Pengelley: We'll, um... Yeah, we'll, we'll talk about that a bit- Yeah [14:27] Dave Pengelley: in a bit. Um- Yeah. Well, Matt, you, you said you're, you're integrating CRM stuff, so you're not using a third party CRM, you're just building your own bespoke thing for you and your team? Yeah, 100%. [14:36] Matt Slager: Like, I, I started just with the idea- Yeah ... of, like, what, what is a CRM? Like, what's it for? And, you know, what- Yeah [14:41] Matt Slager: what are the questions? You know- Mm ... if you know you need a CRM, it's because you have a question and you need an answer for it. So therefore, you know, if you look at it like a chessboard, you know at a chessboard you've got all those little pieces that you put on the board. What are the pieces? Mm. Yeah. [15:00] Matt Slager: You know? That, that's when I use the word primitive. It's like the pieces, the building [15:00] Matt Slager: blocks. You know, what are the things that the system actually needs to answer those questions for you? And then Richard, you just said data engineering. Like, ugh, the, like, the whole reason why my business name is Slaytech Systems is because systems is, like, my love language. [15:17] Matt Slager: Mm. Um, as soon as, like, anyone says the word engineering or systems, I instantly just get chills. Um- Yeah ... and yeah, Dave. So I have... If you think of, like, a standard CRM, you've got people. Those people might be associated with an organization or company. You may have notes about said people or, or organizations. [15:36] Matt Slager: You may have opportunities. It could be a lead, it could be a client. There's, there's various stateful things there, um, including then projects and tasks and whatever else. So the, like, the disparate stuff, Richard, that you said that most people struggle with, where if their tool integrates AI or not, they're still across all of these different random tools. [16:00] Matt Slager: What I tried to do is be like, "Well, [16:00] Matt Slager: why can't they all just be in the same system?" [16:02] Richard Webbe: Yeah. Yeah. And I get... I, exactly. Export... And so a couple of points on that, and I think that's a really good point. Um, years ago, and, you know, I've been in the technology industry for 40 years, and every time I watch a software company, a technology company try and lock everyone into their technology and not have it open, eventually they die. [16:21] Richard Webbe: Mm. Whether it's Token Ring or you name it, you try and lock it in, and you're gone, right? People won't stand for it, and they'll do exactly what you're doing. They'll build a better fly trap or mouse trap somewhere else, and your mouse trap will be left on the wall 'cause no one could work with it. And it is quite astounding at how difficult our industry makes it for people, right? [16:45] Richard Webbe: We used to have the ISO standards and they'd all come out. All, all of that seems to have gone by the by. No one's making an effort to connect things together and make it connectable. Mm. [16:54] AI VO: And [16:54] Richard Webbe: I think there's a real problem there. [16:56] Matt Slager: You said the word standard. I, I love standards. Um, standards is [00:17:00] like, you know, you talk about trying to keep people trapped in a box or whatever. [17:04] Matt Slager: Um, they, that box will change. You know, the reason for being in that box changes, but the standard is usually the same pain, the same questions, the same reason for doing anything, you know, across any of these av- um, avenues. So- Yeah, I'm trying to build this standard where it's just this collective set of primitives to do work. [17:27] Richard Webbe: So- Yeah ... yeah. I, I, you know, on top of that, Matt, not to interrupt you, I call it the Microsoft paradox, and someone taught me this many years ago. We go out and buy Word or Excel or some of these standard tools or software that we think everyone needs 'cause they're all compatible. Trouble is, people only use 3 to 4% of the features of those software. [18:00] Richard Webbe: They don't want or need anything else. Yeah. And that's why I like your approach with the chessboard in your CRM. What do I need? Good. That'll be fine. [18:00] Richard Webbe: I'll go with that, thanks. [18:01] Dave Pengelley: Right? That, that, that's what Google did when they tried to take on Office. They didn't try to recreate everything in Word and Excel. [18:05] Dave Pengelley: They went with what are the core things that people do, and they've expanded over time, but their Docs and Sheets were trimmed down versions 'cause they're like, "We don't need all that advanced stuff-" Absolutely ... "for the majority of users." Yeah. And then, then that's evolved further with, like, Notion. For a lot of what we're doing, like, you don't need Word documents. [18:20] Dave Pengelley: You're not meant to be emailing Word documents back and forth. We just want shared documents, um, with interactivity and collaboration. So then that's another layer which becomes a real pain when you actually do wanna print something, 'cause it's like, ah, where's the page layout? So how do I, how do I sort of- Split by page [18:34] Dave Pengelley: split by page and A4 and pagination and stuff. It's like, ah, this is not the right tool for that. Uh, I- But- ... haven't printed anything for a very long time. No, but even, like, you know, um, a simple example is something like you wanna make a PDF of something to send to someone, and all of a sudden you, you're in Notion and you say, "Okay, export to PDF," and the pagination's all, like, stuffed, and you've got weird breaks through paragraphs and things, and you're like weird, um, margins that you can't control. [19:00] Dave Pengelley: And it's like, [19:00] Dave Pengelley: ah, come on. Yeah. I'm used to, used to my Word document tools from the '90s. Yeah. True. True, though. It's interesting. Um- It's funny ... whether you print it or not. Um, it's funny how PDF just became the sort of de facto standard for document sharing. [19:15] Richard Webbe: So- Yeah ... it, exactly, and, and, um, it's funny how you say the usage of things and the practicality and simplicity of things is where people rally around, and that becomes the core of what we use. [19:28] Richard Webbe: And we can see with AI at the moment that data, data format and, to get Matt excited, data engineering is the cornerstone, right, of AI. AI is easy. Yeah. AI is automation and algorithms. How do I get that data? How do I get it over here, treat it, treat it here, treat it here, and have the outcomes that I want? [19:47] Dave Pengelley: Any business I've spoken to, it comes back to, like, where are their sources of truth? Like, what is their sort of system of record? And system of record is probably arguably strong language for what most people are using. Um, it's- [00:20:00] Yeah. It's a scattering of- Very true ... spreadsheets and then a few different online systems that they update by hand, and it's all very manual. [20:07] Dave Pengelley: And to maximize your ability to do anything, it comes down to aggregating, actually standardizing the systems of record and knowing what you've got and where you've got it, so then you can- Put workflows around it and take actions on it. Matt, my question for you, because your system's all very AI-driven, does that mean you've got to burn tokens to interact with anything in your entire stack? [20:27] Dave Pengelley: Like, if I just want to go and update a contact and, and log a phone call, can I do that or I've got to burn tokens to interact with that? [20:34] Matt Slager: That's a fantastic question, and it comes down to that first statement that I said to do with user experience versus agent experience. Yes. So my user experience right now is interaction with the agent. [21:00] Matt Slager: Yeah. But if you think of the actual agent experience, the, the, the UI, the APIs that exist here, the actual interaction method that the agents use, it's all just simple CLI [21:00] Matt Slager: commands. If you know how to traverse a CLI interface, look up help commands, look up the command syntax, you can run the exact same commands that the agent can. [21:10] Matt Slager: Costs you zero tokens. But let me ask you this. If I wanted to jump in my car and go down to the shop to have a bit of a look around or to do some sort of work that my car needs me to do, I have to pay for fuel. It's, for me personally, it's part of the game. I don't chase zero cost infrastructure anymore because there's kind of no point. [21:31] Matt Slager: You know, if you want to be part of the game, you have to burn tokens. [21:36] Dave Pengelley: Interesting. That's it. I mean, and, and the conversation becomes then at scale and h- and how do you control the tokens? And I've been thinking about this, talking to some people around custom harnesses that they're writing and different systems, and thinking about my own harnesses that I've been building. [22:00] Dave Pengelley: And even with my, my app, my, um, FX Traders app that I've built, Chart Reporter, I'm not putting AI into it other than MCP so people can, you know, attach [22:00] Dave Pengelley: their own thing and, and query the data. That's easy. That costs me nothing except a few API calls. But if I wanted to go, "Oh, well, what if I did put an agent in the system but it was like BYOK?" [22:09] Dave Pengelley: Then all of a sudden I'm on the hook for how efficient is my agent at, uh, leveraging their tokens, because now they're trusting me to burn their tokens with their key and it, it gets really curly and who's on the hook for the responsibility for token burn? And if you're building a system and saying, "Plug your key in"... [22:24] Dave Pengelley: Um, and we're seeing, I suppose that's true of Hermes, of OpenClore, of anything, if you're plugging a key in and expecting magic, like how efficient are these systems and where are the benchmarks and how do people know they're getting value and that you haven't built this amazing thing that does great work, but it's just ridiculously inefficient on tokens. [22:41] Dave Pengelley: And I say that from my own personal experience with my system, given that I'm- You burnt them all the [22:46] AI VO: other day, didn't you? You, [22:46] Dave Pengelley: you know, you know, I'm, I'm doing everything in Coworker at the moment. That's my primary sort of automation platform that I'm using. And I hit my 100% of Claude tokens for my 5X plan yesterday and got locked out for 24 hours. [23:00] Dave Pengelley: Except when [23:00] Dave Pengelley: Anthropic turned off the OpenClore tap, they gave everyone, you know, $150 of like a month worth of free credits of your plan as extra usage, which you can't use via the API. You can only use essentially by burning over your quota. Um, so I don't know how that was really meant to help the OpenClaw users. [23:17] Dave Pengelley: But anyway, I had 150 bucks of free extra usage, and I turned that on when I hit my quota, and within half a day of my usual workflows, my hourly heartbeat pulse, generating my morning email. And obviously when it does these things, it's like, ooh, I found a task, or I've got to go grab that YouTube video, I've got to summarize, I've got... [23:36] Dave Pengelley: So I was doing work, but I burnt like $50 of credits in half a day. And I'm like, that is insane. Like, my system is not optimized for token efficiency, clearly. Um- I-- You've [23:49] Matt Slager: got a little bit of token optimization possibly as part of the variables of that equation, but the other part of the equation is anthropic, and that cannot be, like, overlooked. [24:00] Matt Slager: You know- [24:00] Matt Slager: Yeah ... the fact that if you look at any of the cost to value ratio, anthropics models are the lowest as far as the s- the money that you spend compared to the token economics of value that you get out of that. Like, it's insane. And, like, I haven't, I haven't completely put my... and I don't think I ever will, put my eggs in any one basket, but I keep jumping because I just keep seeing the shifts. [24:25] Matt Slager: I'm trying to be as flexible and pliable- Yeah ... uh, to the market that exists. [24:30] Dave Pengelley: And- I, I installed, I installed LM Studio yesterday 'cause I had no tokens left, and I was like, "Oh, I'm burning all this extra usage. What can I do locally?" And I installed LM Studio and downloaded Gemma 4B that would fit on my laptop, and I got a Nemotron build that would fit on my laptop, and then tried hooking up Hermes again, 'cause, you know, Mano's been, been bullish on Hermes. [25:00] Dave Pengelley: So I was like, "Ah, I'll try Hermes," and got it to CLM Studio, and I got it to populate the models, and I picked the model I wanted, and then every time I tried to send the thing, it came back going, "Chat response not validly [25:00] Dave Pengelley: formatted." And I'm like, "Ugh, I don't even have a good AI that I can ask questions to help me fix this." [25:03] Dave Pengelley: So- [25:05] Richard Webbe: Quick question for you nerds. Can you transfer tokens between your accounts? What do you mean? Well, I, I'm, I'm not fully utilizing my- [25:15] Dave Pengelley: No, like within your plan, you'd have to give over your whole subscription login to share tokens. Like, with my ChatGPT business account, which, um, we've got two... you gotta have two users, so it's got me and my wife. [25:26] Dave Pengelley: Their Codex tokens are separated, but if I run out of Codex and wanna borrow hers, I've actually gotta log into my Codex account, like log into my ID as my wife- Got it ... in order to use her Codex tokens. Got it. But no. The other thing- It's not like frequent flyer points ... [25:40] Matt Slager: is the, like the price of certain models or certain providers, the actual inference that they're using and the efficiency gains that they're getting during their actual inference, um, is, is insane. [26:00] Matt Slager: And a lot of people probably don't realize this. Like, people who have been or are used to working with, you know, the top GPT, the top Claude [26:00] Matt Slager: models, you know, burning through millions of tokens on that expense. Um, when, when the recent DeepSeek V4 stuff came out for s- both the Flash and the Pro model Yes, it's Chinese. [26:13] Matt Slager: Yes, they're probably mining all of your data at the same time. But if you just look at the actual efficiency of it, these, these people pumped the same sort of workflows through Pro, which is their expensive one, and they couldn't even crack five bucks worth of usage, which is... It's just nuts. And the... I don't know. [26:31] Matt Slager: I'm so excited to see what happens with all of the, all of the infrastructure com- concerns and demands because a lot of people are saying that we're gonna hit this, this plateau where everyone's burning so much tokens and all of a sudden all of the, the VC funding and the subsidization will, will burn up, and you'll actually start to experience the real cost of these things and- Mm-hmm [27:00] Matt Slager: you know, actually spending token price on them. At that [27:00] Matt Slager: point, do we continue to use Claude and all that, or are we gonna go to these more efficient, possibly data mining sources like, like Deep- [27:07] Dave Pengelley: Or self-hosted. Are the, are the self-hosted models gonna be good enough even on the smaller, um, sort of 8 billion, 16 billion models without going to the full 24, 32? [27:18] Matt Slager: The, uh, the, the Qwen-3.6 27 billion dense I want to run right now, I just don't have the infra to do it. I... Yeah, [27:25] Dave Pengelley: yeah. Um, yeah, so, 'cause I, I think that's the other thing that, running your own models, um, and, uh, we're gonna, we're gonna see that. As we've been used to using this broad brushstroke thing and just throwing everything at it, I think it's gonna come back to being more tactical. [27:38] Dave Pengelley: Um, which, you know, that, that brings us to some of the, the, the... Ah, do I run the news stinger? It's not really news, but it's kind of what we're gonna talk about next, so, uh, let's, let's do it. I'll, I'll, I like the news stinger. Hit [27:50] AI VO: the button. Welcome to the AI Update. Let's look at what's happening in the news.[00:28:00] [28:01] AI VO: So, uh- I'll give you [28:02] Dave Pengelley: a second to plant This is, this isn't necessarily like news news, but this is, uh, conversations that have been happening on, on socials like LinkedIn and business leaders. I'm seeing a few common threads here talking sort of a continuation of what we've been talking about all show so far really. [28:17] Dave Pengelley: Uh, but if I bring this up, and it's funny you mentioned Canva at the start, Richard. We did not actually rehearse this. No. But I saw this post from Cameron Adams, the co-founder and chief product officer at Canva. [28:27] Richard Webbe: Yes. [28:28] Dave Pengelley: Uh, and I read, "Everyone I speak to feels more productive with AI, and yet many teams don't feel faster. [28:34] Dave Pengelley: Here's what actually happening. You're talking to your ChatGPT, your colleague is talking to her Claude, someone else is deep in their own assistant thread. Everyone in, everyone's individually humming along, generating ideas, drafting content, moving faster than ever. But none of those conversations are talking to each other, and neither really are the teams. [29:00] Dave Pengelley: AI has expanded top-of-funnel thinking. There are more ideas, more inputs, more possibilities than ever before, but the constraints at the bottom actually getting the work live haven't changed. So we end up with more output and the same [29:00] Dave Pengelley: decision-making pressure. What needs to change is how AI shows up inside shared work, not just individual work." [29:05] Dave Pengelley: That's what, blah, blah, blah. Um, I would actually challenge and add to that as well in my own things that I'm seeing, is that it is putting... It's showing up in individual work, and so I, I, you know, being out of tokens, decided to type my own Substack article yesterday with my own fingers, with no... I didn't even run Grammarly across the top of it. [29:26] Dave Pengelley: I was like, "Screw it. This is raw. This is... I'm, I'm raw dogging this thing." Man, you're working hard. But, but thinking about the fact that I have, you know, my dozen team of agents running hourly synchronizations and heartbeats and running through linear tasks and generating documents and generating content for me to review, it creates a massive amount of work just for me to actually review all the work the agents are doing, and I... [30:00] Dave Pengelley: And the human becomes the bottleneck in that instance as well. So, is there a benefit to generating 20 times the amount of content when you still have the same number of humans that have to verify, [30:00] Dave Pengelley: confirm? Or do you go the full autonomy, just sort of guy getting his car to drive him to Whole Foods for his magnesium thing that we talked about last week? [30:08] Dave Pengelley: Like, that's the two choices. You go full autonomy and just trust the robots to do whatever they like, or you still have to be the backstop and become the bottleneck, reviewing and chasing and confirming all the things. And I find, am I more productive? I, I can certainly build things I could not build before and generate some of the things, but in other ways it becomes this argument between busy versus productive. [30:36] Dave Pengelley: And are we really, really busy? Yes. Are we really, really productive? Maybe not always. I don't know. What do you guys think? [30:44] Richard Webbe: That... Yeah, I think that that's a really good point. Um, a while ago when I was talking about networking, like 30 years ago and people didn't really kind of understand what it was, and I used to talk about information islands. [31:00] Richard Webbe: So, the PC came along and allowed us to automate tasks. Everyone thought they're an accountant [31:00] Richard Webbe: when they got a PC with Excel. It accelerated their individual performance, but maybe not the company's performance, 'cause we created information islands. Mm. And people didn't always trust the data, 'cause it came from their computer and not mine, and I haven't reviewed it, and all that sort of stuff. [31:15] Richard Webbe: Fast-forward from the old sneakernet, which is, we used to run around before a network with a disk and stuff like that, fast-forward to something like autonomous driving. Okay, so I'm driving my car, but I don't completely trust it, so I'm keeping an eye on the road and I'm checking- Mm ... I don't wanna be one of those four surgeons in the US that were drunk and turned their car, their, you know, their Tesla onto full autonomous and accelerated and crashed and killed all four of them. [31:41] Richard Webbe: Um, someone did find out they hadn't bought the full autonomous pack, so that was a, a bit of lack of trust. [31:46] Dave Pengelley: It's like Cru- the Anchorman 2 scene where he's got cruise control on and they're all in the back having a, having a laugh. [31:53] Richard Webbe: Can I just tell you? You can't believe four very senior surgeons and specialists did that with their Tesla in the US, and they all died. [32:00] Richard Webbe: So, now [32:00] Richard Webbe: we come forward to- When I get data from someone, do I trust it? Don't I? I mean, when I'm driving a car, I used to have friends when they were first driving an automatic, they'd look at the revs on the gear change on the automatic, and they were second-guessing would I have put it from first to second then? [32:14] Richard Webbe: It is a cultural evolution, and it's an evolution of trust, and where it comes from, we have to be sold on and get to. I spoke a while ago about, um, uh, my old, uh, uh, bosses, uh, Marc Andreessen and Ben Horowitz, when they first set up one of the first cloud companies called Loudcloud. It failed, and they took all the EDS's processing. [32:36] Richard Webbe: Why? 'Cause no one trusted someone else to take their data processing and do it offsite for them. Today, oh my God, everyone. So I think there's a cultural impact to what you're talking about, Dave, and it's really important. Do I trust that data, and can I get it to sing like a concert and get everyone working together? [33:00] Richard Webbe: Because you're not gonna be more productive if you're second-guessing where the data's coming from anymore. [33:00] AI VO: Hmm. [33:01] Matt Slager: I want to say my thousands of thoughts, but you know I can't get them out. First of all, I have to humor what Sean's saying in the comments, which is pretty much what I was gonna say as well, like, where his, his, his mention of internal benchmarking was to do with cost u- uh, cost- Yeah [33:18] Matt Slager: and token usage, but it's also relevant here. Benchmarking your productivity, you know, your metrics, like the whole point of KPIs existing in the first place. You guys can have KPIs for your systems that exists. You know, you can actually measure how productive your agents are. You just gotta define that in the first place. [33:34] Richard Webbe: Make a bill- That's a- ... out of this. But- That's a really good point. Yeah. And, you know, someone mentioned can we augment humans' ability to review the output, right? Yeah. 100%. That's what, that's what we're doing. [33:44] Dave Pengelley: Well, and, a- and, and an example is, is I, I wanted to create some social shorts and content for my, for my, for Chart Reporter, and I said, "I wanna lean into the meme culture the traders have on X and stuff," and it came back with this generic, horrific sort of stuff. [34:00] Dave Pengelley: And, like, as my, as good as my, as my agent team are, [34:00] Dave Pengelley: they don't get humor, they don't get meme culture. And so they're like, "You could do this comparison thing and do..." I'm like, "Dude, no. I'm thinking, like, using clips from sitcoms and, like, other bits and pieces." And it'd go, "Oh, okay, that's a hard pivot. We're gonna do that." [34:13] Dave Pengelley: And in the meantime, like yesterday, I had no AI, so I was like, my highest leverage move now is to just go and pull that clip from Arrested Development that came to mind and throw that into a short and push that up, and in like three hours, I had 1,000 views on this short. That was all 100% me, whereas if I rely on the AI to try and generate the meme ideas and then follow that, and I've gotta review 10 crap ideas, I'm better off in that instance just going, "No AI. [34:38] Dave Pengelley: I will just mine my own brain and just generate five good things versus reviewing 10 crap things." [34:44] Richard Webbe: And that takes us back to Wall Street in 1985. I made that number up, but somewhere about then- ... when everyone was using their automated stock buying and selling software, and of course- It's like with AI, it's the sum net average of all the Venn [00:35:00] diagrams, and as soon as the shares started to drop, everyone started to sell. [35:04] Richard Webbe: And as everyone started to sell, the shares dropped and it went down, and they had the biggest crash in Wall Street's history. If you just rely... See, I, I, I, I see AI as this. Your experience, particularly you two, and the people who are listening to us, and the people we work with, it is unique. You're a unicorn. [35:21] Richard Webbe: You'll always be a unicorn. You're a human. You've got all this weird shit going on in your... Sorry, weird stuff going on in your head. And Matt and I always allude to our brains are running very fast. So, don't discard that. Push that to the front and just get the automation happening. Imagine if we were handwriting all our emails. [35:42] Dave Pengelley: Yeah. And, and, and I think, I think this, this, this comes down to what we would say with our, with any of our customers, uh, and unfortunately, you know, doctors are normally the worst patients, um, is you're running so fast and you, you set up these systems and then you expect a lot from them, and you don't always go back and reorder them quickly enough to go, "Is [00:36:00] that still adding value?" [36:00] Dave Pengelley: Like, if we're gonna go and implement workflows and automation and AI into someone's business, we find the things where there's high leverage, where there's high ROI. Internally, part of it is the R&D process of experimenting, and that's what we have to do, is living on the, on the bleeding edge, and we sort of go, "Okay, can I automate that? [36:15] Dave Pengelley: What if I have an agent that does that? What if... Can they, can they autonomously be, like, pursuing that path for me on my behalf?" And then you come back and I find out, you know, after two, three, four weeks go, "No, that's rubbish. I cannot rely on it to, you know, help me generate humorous meme ideas. That's on me." [36:31] Dave Pengelley: And go, "Cool. Let's, let's kill that and not have to worry about it coming back to me with like, 'Check this one, check this one, check this one.'" 'Cause then that's creating more busy work for me, not actually productive outputs. And so I think it is, part of it is internally we are sort of dogfooding and, and testing these things at the bleeding edge, but also then occasionally going back and auditing and going, "If, if I was my own customer, would I be automating this process? [36:54] Dave Pengelley: No, probably not. So why do I think it's okay to automate it in my own business?" [36:59] Richard Webbe: And look, [00:37:00] so, so what AI for us is, and for a lot of our customers as we know, is it's business process consulting. Push AI to one side. What is the problem I'm trying to solve? As you're saying, Dave, where is the biggest delay in me getting my services efficiently to the marketplace and getting paid for it? [37:18] Richard Webbe: That's, that's, that's all it is. And you have to look at a business holistically, break it down into its elements, and then go, "Right, I'll plug this in here, I'll plug this, and I'll get that, I'll get that." And as we said, if you can't trust the data, you got a problem [37:34] Matt Slager: Hmm. There's a, another, like, vibe that's happening in the really techy developer community. [38:00] Matt Slager: I'm sure Sean, if he's s- still here listening to this, will, um, would appreciate and, and, and understand hopefully where this is going. This is a quote that I've captured from a recent video I watched, um, from a d- a developer guy, an influencer that I follow and I respect a lot. And he talked about... [38:00] Matt Slager: He, he brought it into the concept a lot of people are feeling that their, their edge, their, you know, shine that they bring to the table, that unicorn, um, Richard, where that's diminishing the more that they use AI. [38:16] Matt Slager: And like Dave, if you're a, a master meme creator and all of a sudden you're starting to use AI for all of it, you know, where does your spark come from anymore and do you feel that start to atrophy over time? So this quote, I'm gonna read out just two paragraphs if you guys will humor this, and then I'm curious to hear your opinion. [38:31] Matt Slager: Um, let me just try and read it. So first of all, it says, "AI disincentivizes you from learning about the pieces," and I think that's where the biggest problem is. "Humans trust pain. Feeling dumb hurts. When you're trying a thing and it doesn't make sense, you feel the pain. And when you try the thing and it goes as expected, you feel good. [39:00] Matt Slager: AI has made it easier to avoid that pain [39:00] Matt Slager: and feel that reward." [39:01] AI VO: Mm. [39:02] Matt Slager: Wow. So when... There's, the second part of this is, is pretty intense. Yeah. So when what used to be an upfront cost where you would pay to learn the pieces and then you could get that res- that reward of solving the puzzle, it, it's now a slot machine, and your choices go from learning the pieces so that you can actually solve the puzzle correctly or just keep pulling the slot machine until hopefully the correct answer comes out because each pull hurts a lot less than reading the documentation for a language you don't understand, or learning a library that doesn't map to your mental model, or debugging something that feels hopeless. [39:48] Richard Webbe: Gee, that sounds like my brain. [39:51] Matt Slager: It's crazy though. Hey, like, this, this is the thing [39:53] Richard Webbe: with smart people. I, I, I think like that, to be honest, and I, you know, the three of us have discussed that we have brains [00:40:00] that we think run fast, maybe not smart, but they run fast and we jump between things. And I think you're right. [40:06] Richard Webbe: You try and explain to someone who's never driven a car before- what the gears do, right? It helps them be a better driver. But when they've got an automatic, do they need to know that? Or maybe they need to know when the engine's revving, they've kicked down to a... So a lot of that is you have to have context to what you're doing and why you're doing it. [40:26] Richard Webbe: One thing I, I think is okay, and I'll go with Einstein on this one, 'cause he, he was a bit smarter than me. He famously said, of course, when someone asked his phone number, he couldn't remember, and they said, "Why don't you remember your own phone number?" He said, "Because it's in that phone book over there. [41:00] Richard Webbe: Why would I bother? I've got more room in my brain for other things." I personally believe that your pain-pleasure point is true, but I also think it frees up a large part of your creativity when you're focused on what the, the, the problem is trying to solve, is freed up. You, [41:00] Richard Webbe: you've got a little... You're less overwhelmed, you've got a little bit more room to play and try, and you have more resources to try things out because it's automated. [41:08] Richard Webbe: That's my view. Dave? [41:10] Dave Pengelley: It's, it's horses for courses, right? I, I think there's times and places, and I think if, uh, too much of anything can be a bad thing. Like even a good thing, too much of a good thing can still ultimately be a bad thing. Um, it's, so it's a, it's a mixed bag. I mean, for me, the fact that AIs can generate and create things that I couldn't, like I don't have to go and learn the pain of writing syntax and writing thousands of lines of TypeScript in order to build a great web app. [41:35] Dave Pengelley: That's, that's leverage for me. Um, but in the same point in time, knowing some of the underlying, as Matt would say, prim- primitives of where, of database design and data flows and thinking about state management, all these kinds of things really, really help make sure that the thing I'm engineering, it comes back to that engineering versus, um, coding, really, really helps me. [42:00] Dave Pengelley: But then on the ideation point of view, it becomes really easy to become [42:00] Dave Pengelley: lazy and just trust that the magic AI knows all the answers. And so, yep, go research this problem and, and tell me, like, all the specific things that my app should have, and it comes back and goes bam, bam, bam, bam. And you go, "Okay, cool. [42:10] Dave Pengelley: Now go and code those things. Thank you for researching. Yep, that all looks good." But I have joined some forums and things, and having some conversations around different data points for analysis, uh, and things. Someone threw up this, like what they're doing with their own tool that they're building just for themselves. [42:24] Dave Pengelley: And I was like, ooh, ooh. I hadn't even- A new angle ... that hadn't come up in any of my research, in any of my AI-assisted research that, that here's all these ideas. And so I was like, cool. So that interacting with people and exploring the thing and that getting back to humans and other humans and the way they solve problems, uh, can be really handy, and we, we lose that. [43:00] Dave Pengelley: And I think even I mentioned yesterday for Substack, I wrote my article by hand, um, and this was, this was the last couple paragraphs that I wrote. Uh, "I don't trust the AI to fully unleash it publishing content on my behalf, but it generates so much I don't always make the time to review it. I'm sure that in the six weeks [43:00] Dave Pengelley: since I last posted here, it's generated a number of different drafts that I've ignored, glanced at, and ultimately I find myself here today." [43:06] Dave Pengelley: typing my own one out instead of leveraging one of those. And I think personally, part of that is just the catharsis and the creative ability to type and just put your own thoughts out rather than just tweaking someone else's. Yes, my AI's trained on me. Yes, it's read all my stuff. Yes, it's listened to all these transcripts and it knows my voice and can write a reasonable facsimile of me, but it's not my words, it's not my actual feelings about a particular topic. [43:32] Dave Pengelley: So I think, again, reconnecting to who you are with these creative aspects and going, "I actually wanted to write something yesterday. I wanted to share how I feel about all this AI stuff. I wanna share my thoughts, not, uh, just interpreted thoughts that the agent has created on my behalf," it becomes a different conversation. [44:00] Dave Pengelley: Uh, so then, then I wrote, "So I'm always busy feeding the machine, giving it more things to do, but each of those tasks becomes a rod for my own back, just kicking the can down the road and multiplying the things I [44:00] Dave Pengelley: need to finalize and complete. I have no doubt that having robots yell at me keeps me more productive than ambling along on my own, but is it just satisfying a dopamine loop or is it actually driving success?" [44:10] Richard Webbe: Good question. Yeah, it's a really... It's, I, you know, I'm glad you raised this, Matt. It's a good question. Where does our creativity stop? Where does our value stop? I still stand by the fact that if we're all using the same tools, then the same garbage will come out, and you will deliver some baseline value, but nothing creative, nothing unique, nothing competitive advantage, whether it's a business or as an individual. [44:36] Richard Webbe: And, and I like your approach, Dave. It's, it's, you gotta be really discerning about what you accept and what you put out. And so the concentric ring story, we're automating the complicated or the volume part of creating content, but our brains are still adjusting that so it's unique and we've got the right perspective. [44:56] Richard Webbe: Otherwise, you're just regurgitating the average of what everyone else says. [44:59] Dave Pengelley: And, [00:45:00] and some things you wanna be personally creative, like I wanted to write a Substack article. I wanted to share how I feel about the way I'm using AI. That's different to, I'm doing three to five streams a week now showcasing my chart reporter app, where I'm doing live trading sessions for 30 minutes to an hour. [45:14] Dave Pengelley: That I just wanted to literally pull down the transcript, write up a summary, subject, and description, and repost that to YouTube. That is like a low-value, high- I just say something ... I volume think ... program. Yeah Where I just, I just want the algorithm, I want YouTube when people search things to find the right keywords, not, "Dave did a live stream tonight." [45:32] Dave Pengelley: Like, I want it to go, "Oh, he w- looked at this currency pair, and he did this, and he traded that, and he was looking at this indicator." Bring some of those keywords up auto-magically for me. I don't wanna sit there afterwards and do another half hour of going, "What did I talk about? What did I trade?" And it puts all the chapter timestamps, it just does it all auto-magic. [45:48] Dave Pengelley: That's an instance where, you know, that's a less creatively intense thing where I just want it done. [45:54] Matt Slager: Yeah. I just, um, while you guys are just speaking about those things, I just had the [00:46:00] best analogy come to mind. Oh, please. I love good analogies. Now, if Kate Fabian's still here, she's gonna love this. So what Richard just said about, you know, the same rubbish, you know, everyone's using the same tools, the same rubbish comes in, the same rubbish comes out. [46:14] Dave Pengelley: Yeah. [46:15] Matt Slager: I have the perfect rebuttal in the shape- Good ... of an analogy. [46:19] Richard Webbe: Good. [46:20] Matt Slager: So, cooking. It's [46:24] Richard Webbe: a good analogy. [46:26] Matt Slager: I like it. Culinary skills, food, ingredients. Everyone's using the same stainless cookware, the same gas stove. Same cook. Yep. And then you have some people able to create objectively, subjectively, doesn't matter, you know, better cooking, you know, better food. [47:00] Matt Slager: It tastes better. It gives you more of an e- enjoyable experience when you... The aromatics, all the stuff. Mm-hmm. Now, so that analogy so far makes sense. If you guys haven't been aware, um, today, yesterday, there's, there was a, a, a [47:00] Matt Slager: huge security breach through various supply chain stuff, dangerous worm, spooky things, you know. [47:06] Matt Slager: Um, I won't go too deep into it, but if you think of that, that- Mm ... that kind of developer land supply chain risk is based on the ingredients. It's the- Yeah ... ingredients that you're using. [47:16] Dave Pengelley: Is, is it bigger than the Axios thing from a few weeks ago? [47:20] Matt Slager: Uh, probably not bigger, but scarier. Okay. So i- i- it's a, it's a worm thing where it actually will embed itself and, and go deeper and continue to spread. [47:29] Matt Slager: So, you know, this isn't a security lesson in, in, in the world of technology, but i- what this analogy is trying to steer towards is you can get food that is contaminated, that's gonna affect your downstream product- Yep ... and potentially infect your staff- Yep ... and then your whole kitchen's screwed. So- Don't go on a cruise. [48:00] Matt Slager: Yeah. So with, oh, with this, this whole thing of creativity and all that- I think it is still possible [48:00] Matt Slager: that the individual can use the same tools and use lesser tools, use a subpar, um, set of tools and get a better result than somebody using the best tools, which has been the case with any kind of craftsman or- Abso- [48:15] Richard Webbe: absolutely. [48:16] Richard Webbe: I think, I think that's natural evolution. I think you're right on the money there, Matt. Uh, and it's very true. And I always said when I, you know, I ran an AI company about five years ago, I always said to the clients, uh, as a, as it was a BPO outsourcing AI company with, you know, automate... I said, "It's all about the question, not the answer. [48:36] Richard Webbe: And if you ask the right questions, then you'll get the right answer." And like, what is the problem I'm trying to solve? So when people are cooking, to use your analogy, everyone's got a different objective in their mind, even though using the same ingredients, same cookbook, same cooker. They've got a different perspective. [49:00] Richard Webbe: And I always hang my mom out to dry for this one, 'cause she clearly made the best spaghetti [49:00] Richard Webbe: bolognese in the world. Um, and I knew that, and she knew that. We had to convince other people that. But when I went to her place, it tasted great. But when I cooked mine, her recipe at my place, it didn't taste as good. [49:11] Richard Webbe: And of course, it took me 40 years to work out that she actually didn't give me all the ingredients, so I always went home for spaghetti bolognese. And that's the same thing, right? If we look at it, individuals hold those secret ingredients just by what they do. It's going to be 100 years before, in my opinion, with the evolution of AI and technology and access to data and formatting of data and opinions and perspective for it to even come close to being as good as a human would when they managed it. [49:42] Richard Webbe: That's my opinion. It'll be over 100 years at least. [49:45] Matt Slager: There's also the, the, um, like you just said, what's the... Like, what is the question? The question is the most important thing. If you think of, like, we're, we're discussing what these things are outputting and whether or not it's the same spark as what the human can do, [00:50:00] that's judging the result as the thing that we're looking at. [50:05] Matt Slager: Yeah. What, what happens when the result doesn't matter anymore? Similar to what Dave said, you know, like his, his throwaway streams. He just wants to get some AdWords, you know, just some SEO, whatever. Um, there's another guy that I follow a lot, and he had a tweet, um, today that he, he basically... It's the same thing that Dave just said before. [50:24] Matt Slager: His literal quote is, "It's weird how much I spend trying to get agents to write code like me." Maybe I should just [50:32] Richard Webbe: write [50:33] Matt Slager: the [50:34] Richard Webbe: code. Well, it's, it's the icing on the cake story, isn't it? Getting cake out, 'cause cake is pretty genuine, just do that, right? Yeah. 'Cause you're the icing. Your stuff is the icing, and you'll differentiate yourself usually on the icing. [51:00] Richard Webbe: Or with a pizza, everyone knows how to make a pizza, but there'll be a sauce you put in there that will differentiate you. And I don't think that's ever gonna change, right? Yeah. I don't think it's ever gonna change. And yes- That might- ... there are some things we should automate, and there are some [51:00] Richard Webbe: things we shouldn't automate. [51:01] Richard Webbe: My, my old, [51:01] Dave Pengelley: my old boss at the pizza shop, he'd, um, occasionally go and consult and help other pizza shops out a little bit. Um, and the first thing he'd do is, uh, "Well go make your dough." And they'd be, "You, you make it." He's like, "No, I'm not giving you my dough recipe." Exactly. Exactly. [51:16] Matt Slager: There's a, um- Bite the pie [51:19] Matt Slager: there's a... For, for those people that listen to our show that are trying to figure out their place in the world, um, looking at the, the listings on SEEK and, and all the places and nothing makes sense to them, there's a, a new position, a new role that I've heard a company that I follow talk about recently that it interests me a lot, if I ever went down the route of being fully employed in a big company again. [51:42] Matt Slager: It's called a forward deployed engineer. [51:45] Dave Pengelley: I knew you were gonna say that. [51:47] Matt Slager: Yeah? [51:48] Dave Pengelley: I knew. That sounds pretty great. I, I was actually talking to an FD last night at the, uh, AI meetup I went to. Please explain. [51:53] Matt Slager: Do you think I would fit that, Dave, from your definition? [51:55] Dave Pengelley: Yeah, yeah. It, it, it's basically, it's kind of almost a pre-sales, post-sales. [52:00] Dave Pengelley: Not fully [52:00] Dave Pengelley: post-sales, not implementation, but it's being on site, helping customers make sense of the technology that they're buying and getting it implemented, but not really implementation. It's a bit more innovation, um, and experimentation. Uh, yeah, so it's, it's kind of like a salesperson that's embedded at the company temporarily- Yeah [52:19] Dave Pengelley: in some ways. Yeah. Yeah. It's [52:20] Matt Slager: curious. I, I find it very interesting, 'cause I don't think I could ever just be a cog in the machine anymore. I think I have to be, like a, like a, like a touch everything person. Yeah. Um, and I always have been that way. Even when I worked at this shed, I, um, I literally was that person that knew where everything was and could fix anything for you. [52:36] Matt Slager: So- Yeah ... um, yeah. The forward deployed engineer is such a curious concept. For me, it, it kind of moves in the same space as someone that is the, the actual, like the product person. Yeah. But the product person is less about the engineering, it's more just the user experience, I guess- Yeah ... and the feature stack and, and making sure that the reality comes true. [52:59] Matt Slager: Whereas the, that forward deployed engineer, the FDE, is super customer-facing engineering stuff. It's about that user experience and- Yeah ... the agent experience. [53:10] Dave Pengelley: Yeah. It's, um, uh, it's kind of the evolution of pre-sales blending into post-sales. The technical architect, but not necessarily super architect-y. [53:21] Dave Pengelley: It's a bit more experimentation. It's, it's being rolled out differently at different organizations. But yeah, I think it's, it's interesting to see what's happening With this FDE role versus the traditional SE role, um, yeah And that, [53:34] Richard Webbe: that plays up into the organizations at a higher level, because they demand that support and that explanation and that context. [53:42] Richard Webbe: And I was doing, you know, you announce the news, I always do a bit of research before we get on for our session. Um, and IBM's just announced that 76% of companies have now established a chief AI officer role. [53:55] AI VO: Mm. Kind of like [53:55] Richard Webbe: what we went from IT managers who were playing with their PC to [00:54:00] CIOs. And I think a chief AI officer role will probably sit well above the, uh, chief information officer role from a strategy point of view, in my opinion. [54:09] Richard Webbe: Because they're looking holistically at what is the problem I have to solve? What are the tools I have to do it with? And they're gonna need one of these forward deployment engineers you're talking about. Yeah. So that they get the impact of, okay, if we go down this path architecturally, and we grab this data and bring it there, how's that gonna change the way the company operates? [54:28] Richard Webbe: Well, and- Are we gonna be flexible? [54:30] Dave Pengelley: And you think about, you know, I'm just gonna pick a handful of modern software stacks that companies implement, you know, whether it's Notion or Miro or Asana or any of these types of things. The, the one I was talking to last night was Loreakeet. But imagine you're putting that in, and there's, there's different ways and different users are gonna use that in different ways. [55:00] Dave Pengelley: And so it's not that you've gotta go in and engineer the deployment, it's helping the teams there understand how they can leverage this flexible platform solution internally. Um, it's different from here's your deployment and here's how you use it. It's like, okay, you've got this great tool now, let's experiment [55:00] Dave Pengelley: and help your teams work out different ways you can execute and leverage it most effectively. [55:05] Dave Pengelley: Um, because I mean, and, and just, you know, bring up another link. This is a local Australian software company, um, Elmo HR. And Justin, whose name is obscured by our logo, um, he's just written, uh, on LinkedIn, "Most organizations are no longer asking if AI matters. They're stuck on something harder. Why isn't it actually changing how work gets done? [55:24] Dave Pengelley: Here's the uncomfortable truth. High adoption doesn't equal high impact." So they've got 75% of their workforce actively using it, um, but it's not the same as enablement. And so he sort of, he'd written an article talking about some of the ways they're helping make sure their 75%-plus team are actually getting some value from it. [56:00] Dave Pengelley: Um, but, uh, yeah, there's a, there's a lot going on in the space. Um, we are, we're, we're running up to time, guys, so I did wanna just wrap up and mention that, uh, and you teased this at the start of the show, Richard, we are having our first AI Operators [56:00] Dave Pengelley: IRL in real life event coming up in June, uh, in the northwest of Sydney, because I'm sick of going into the CBD for all the events. [56:08] Dave Pengelley: Uh, and there's a lot of people out in the, the suburbs that won't be going in. Business owners and people that are in this space that don't have the time or energy to make it all the way into the city for all these meetups and events. So- Uh, we're gonna trial one up here in Northwest Sydney, and then we'll move them around. [56:23] Dave Pengelley: We'll, we'll come down to you. Richard had one somewhere around Melbourne. Matt, we can, we can run one in Albury. We'll get, get Mano running one out in Kalgoorlie somewhere. Um, we'll, we'll shuffle these things around. But, uh, yeah, yeah- Karl Stefanovic, [56:35] Richard Webbe: uh, did his, uh, show in Albury the other night. Who did? [56:39] Richard Webbe: Karl Stefanovic. [56:41] Dave Pengelley: Did he? Karl. Yeah. Okay. He did [56:43] Richard Webbe: a podcast in Albury. We should have jumped on the back of that. We could [56:46] Dave Pengelley: have- ... [56:46] Richard Webbe: bragged the quality of this show up a bit. [56:48] Dave Pengelley: Uh, but, uh, yeah, we're, we're running the first AR producers IRL night in Sydney's northwest in a pub. I'm glad to hear that. Um, so, uh, we're still finalizing, uh, [00:57:00] some of the details, the exact date and the speaker list, but the idea is we'll have a few people giving some short little lightning talks, doing a few little demos of things they're doing and how that could apply. [57:10] Dave Pengelley: Just some, uh, ideation, inspiration, followed by maybe a short panel with some Q&A where you can ask people that, and then networking over drinks, beers, whatever you wanna buy. Um, it is a, a... Yeah, I think it'll be good to just do some in-person networking. Uh, there's a huge demand. I've been going a bunch of meetups in the city, and they're always packing out 40, 50, 100 people at these various different meetups and, and builder nights and things. [57:35] Dave Pengelley: So, uh, I think it should be good. And if you're interested- How do people come? Is it just, like, keep follow [57:41] Matt Slager: your stuff and look for an announcement? [57:44] Dave Pengelley: Yeah. So we'll eventually put up a Luma link for registrations 'cause the room's got limited size, so it'll be sort of first and best dressed with a wait list. [58:00] Dave Pengelley: And, um, but as we finalize details, we- we'll, we'll put more stuff out. But for now, follow me on LinkedIn. That's the main place we're [58:00] Dave Pengelley: sharing this- Cool ... um, as, as the first event. Uh, obviously- I feel like- ... you guys, back channel- Like- ... I'll, I'll keep you guys in the loop. [58:07] Matt Slager: I, I feel like I'm gonna sneeze and it's gonna be June. [58:09] Matt Slager: Like, the, the speed at which everything is going at the moment- Right ... um, as a- Yeah ... as a quick, like, I don't know if this is a, a, a, well, something worth mentioning at the end of the show, but, um, yeah, I had a really weird end to last month. I only just realized now, and it's been two weeks, like, I didn't, I didn't invoice one of my main clients, which is, has been my biggest source of income so far. [58:35] Matt Slager: So, so far for April, I've got $0 'cause I took on a couple extra things, which meant that it took me away from that. And I was like- Yeah ... " [58:42] AI VO: Well, [58:44] Matt Slager: I didn't do anything in April apparently." That's- But, like- That's- ... time just keeps going so fast. I- [58:49] Dave Pengelley: Yeah. When I, when I started mapping out and going, I wanted, you know, five or six weeks of lead time up to the event, I'm like, "Oh, that's like, I need to do it in, in one of the, like, earlier weeks of June 'cause I'm not doing it right on the, um, the [00:59:00] edge of the financial year end." [59:01] Dave Pengelley: So- Yeah ... time is moving fast. But we'll, we'll get one of these running here. Um, and that'll be a good night, uh, for whoever turns up. And I think we're gonna roll these out, um, to have I- more IRL events, which it's not a podcast. We, we might see if we can record, maybe stream. Probably not stream, but we'll record the talk and put it up. [59:19] Dave Pengelley: At least just, yeah, random record I think would be good. I think I'll try and get there for sure. Yeah, I mean, I've got... Where is it? I've got, I've got cameras. Ooh. So yeah, we'll um, we'll, we'll sort that out but, uh, look forward to seeing people there. And, uh, for those on the chat who interacted with us today or didn't interact, great to have you. [59:39] Dave Pengelley: Um, I mean, Manjula- It's great ... and Sean, who were on the show with us last week on camera, great to have you in the chat today. Kate, uh, we will get you on here at some point soon. Uh- We do indeed ... well, and anyone else who has watched, joined after the fact, the replays, remember if you join us live you can interact in the chat too, and we love to see that. [01:00:00] Dave Pengelley: So make sure you have liked [01:00:00] Dave Pengelley: this video. Tell the algorithm to get this out there. Subscribe to it. Hit that bell icon so you get notified whenever we go live, and, uh, we'll see you all on our next show. Thanks, boys. Appreciate it. [01:00:13] Richard Webbe: Good [01:00:13] Dave Pengelley: evening, [01:00:14] Richard Webbe: everyone. [01:00:14] Dave Pengelley: Morning. Afternoon. Afternoon. Depending on when you're watching. [01:00:18] Richard Webbe: You sound like, uh, The Truman Show