The Chainsaw Paradox: Why Your New AI Tools Are Making You Slower

[00:00] Dave Pengelley: Good morning. Um, afternoon. It's afternoon gentlemen. Of course it's the midday show. It's always afternoon. How are we going guys? [00:50] Matt Slager: We're going good. It's afternoon for us. It's not afternoon for mano who couldn't bring it to bring us. Yeah. Very good. Good point. [00:56] Dave Pengelley: Yeah, no. So I can, I can actually legitimately say, um, good afternoon, gentlemen, today and it's true. [01:01] Matt Slager: Yeah. [01:02] Dave Pengelley: Um, yes, no. Um, for, for those joining us, our regular visitors, viewers, attendees, uh, mano had a calendar conflict, uh, and will not be joining us this week, but, so it's back to, um, three Horsemen this week, uh, with, uh, myself, Matt, and Richard. Gentlemen. Uh, how's that? It's hump day. Does it feel Humpy? Uh, I mean, does, uh, do you feel like you've, you've had good momentum going up the hill during the week and now you're ready for the downhill run of Friday? [01:31] Matt Slager: Yeah, definitely. Uh, I, I get my Mondays at like Wednesday, so yeah, today for me was just frantic trying to catch up. I gotta travel on Friday, so I've got like a three day week now. Yeah. Work, work [01:46] Dave Pengelley: or pleasure. What kind of, what kind of travel do you got? [01:49] Matt Slager: Uh, it's not, it's not pleasure. It's a family thing and it's a, it's a funeral related thing, so Yeah. [01:56] Dave Pengelley: Yeah. [01:56] Matt Slager: One of those commitments. [01:59] Dave Pengelley: Sorry to hear that mate. And of course, uh, as well as any emotional toll that those sorts of events take, the workload toll is also real of working out how to do more in less time, right? [02:09] Matt Slager: Yeah, a hundred percent. Yeah. Keep the plates spinning without falling over. [02:14] Dave Pengelley: Yep. Yep. Um, yeah, I, I, I try something new this week, which I'll talk about in a sec. [02:18] Dave Pengelley: Uh, Richard, how's, how's your week been? [02:21] Richard Webbe: I just keep spinning out the amount of conversation around AI in the media, on the web, social, like it's, and people still aren't getting their head around it quick enough. And then just as we do, you know, Matt and I were talking earlier about. Another takeover, another change, another company merging. [02:43] Richard Webbe: Yeah. It is just accelerating along and we can't get our heads around it quick enough. [02:50] Dave Pengelley: Definitely. Yeah. Yep. I, I think that that's very real. And, uh, as, as we've talked about previously on this show, one of the issues we'll say, our technology is you're not actually necessarily getting time back. You're just doing more in the same amount of time. [03:04] Dave Pengelley: And so you don't get that little break because you don't go, ah, cool, I finished all my work by midday, now I can take the afternoon. Off you go. Cool. Now I've got like another five hours to get more stuff done. And so then you just keep like, just cramming and cramming and just, just sort of doing more and more and more. [03:19] Dave Pengelley: And we're not taking the break. We're just, I mean, it used to be that whole thing. Work bled into real life, like real life being outside of work. Um, when people started getting laptops and they were taking their laptops home and then they were getting mobile phones, they were taking their phones home and they were doing emails on their phone and blackberries and so on. [03:38] Dave Pengelley: So for the last 25, 30 years, we've seen work go from that thing you do from nine to five of the office to slowly creeping out into all parts of your life. Generally speaking, depending on what job you do. Um, I know obviously if you are on site building a house and as, as a chippy, you're not doing that at home, you're doing that seven to three. [04:00] Dave Pengelley: So, um, it is trade specific, but um, for many people working sort of desk jobs, we've seen that, that scope creep of, of, of work and life. And I think ai now it's not giving the time back, it's just meaning that the ones that are using it are accelerating and doing more and being more productive. [04:17] Richard Webbe: And it's a really, it's a really good point. [04:19] Richard Webbe: I remember I did some work for Cisco Systems. Who proved what an intranet did. Matt, 27 years ago, and I think his name was Svic, was the CIO for Cisco at the time, and he got a 1400% ROI by putting in an intranet. No one ever knew what an intranet was. Of course, it was all filing and physical filing and digital sharing or not just then. [04:46] Richard Webbe: And he got, he did got such a big return internally at Cisco. The, the, uh, the story as it goes, I think it was Peter Salick was the guy's name. Apologies if I got that wrong. I think the board said to him, okay. You can do whatever you want, provided you get that same 1400% return on your investment sort of thing. [05:07] Richard Webbe: And I got empowered by Cisco to travel around the world and talk about this as a value proposition so people could put the technical commercial value together. And I was traveling around the world talking about VPNs and, you know, VPNs as they were sold to you, rather than build it yourself, but still using the Cisco technology, which was embedded in a lot of the partners, like the telcos and that. [05:28] Dave Pengelley: Mm-hmm. [05:29] Richard Webbe: The first thing I got back from customers and CIOs and partners and that, well, if I do that, I'm gonna have to sack all those people. So we came up with a story about repurposing those people because there was a, a tiered chart to say, you know, you won't need this. It was like the, I think I mentioned the typing pool last year, last week. [05:48] Richard Webbe: And so everyone was getting ready and we had to change the narrative completely. Of efficiency because CIOs didn't wanna sack their staff. They were building empires. Uh, we're talking 27 years ago. They weren't, they weren't wanting to downsize. And what you are saying is the same thing. We're doing more with less, more efficiently, but it doesn't mean we stop doing more. [06:09] Richard Webbe: And the whole story about am I competitive advantage? Am I getting well? If you are doing it and someone else isn't and you're getting more done, there's your competitive advantage. So one of the key things about AI was as it was for internet and intranets, if you're not doing it, you're screwed. You're in the back of the room. [06:27] Richard Webbe: If you are doing it, you're just keeping up. But if you do it cleverly, there's your competitive advantage and you need an intimate understanding of the technology that you guys often talk about and the business processes. And it's not just about getting home to your wife or family earlier. It's about what am I doing differently? [06:46] Richard Webbe: It using this technology to disrupt the world I live. A [06:50] Matt Slager: thousand percent. ROI is pretty intense. What, um, what, like for people that don't know what an intranet is versus the internet, what, what even is that? [07:02] Richard Webbe: Yes, so it's a very good point. I did say to someone at a talk I did in Europe in front of a lot of people that as soon as people stop using the term internet, we will know it has arrived. [07:12] Richard Webbe: You know, 'cause we don't say internet anymore or as, uh, as I think, uh, president George Bush once said the interwebs, uh, 'cause he wasn't sure if it was a worldwide web on the internet, which was very funny. So the internet is just the web technology, but used for information within your company and it became this accelerated way that people had never experienced to share information. [07:33] Richard Webbe: And we started off walking around and handing files and using the photocopier, and then we got a little smarter and we started loading and saving files on a server here. But you had to log in, had to be in your area, and you had to get permission. Or we used to use what I called sneaker net, where you had a floppy disc, sometimes a CD rom, and you burnt the files on there and you ran around the building to share it. [07:54] Richard Webbe: So we called that sneaker net. That was a pretty good one. So you shared all the same files. So an intranet is just an information, uh, sharing technology architecture that mainly uses web, hyper text, and off you go. [08:07] Dave Pengelley: It was private, private webpages just for your company to share announces announcements and, and, and knowledge share. [08:14] Dave Pengelley: SharePoint is the modern version of it. [08:16] Richard Webbe: Well, that is exactly, but the efficiency it delivered to business was through the roof. [08:20] Matt Slager: So I, um, when you were talking about that, I just had a bit of a light bulb, like I'm, I'm hoping a lot of the, uh, audience did as well, where the idea of an intranet is the idea of your business information, data files, all the stuff that you just said, being accessible from anywhere for those company team members, but also. [08:44] Matt Slager: For the company applications and therefore agents. [08:49] Richard Webbe: Exactly my thinking. That's exactly right. So if you've got an intranet, AI or sovereign AI that operates and you are making sure you've got access to the data that unique to your company, now you'll get competitive advantage. Mm-hmm. Is that your company's knowledge? [09:05] Richard Webbe: I read about a guy recently who was trying to create a narrative about what he did himself in his timeline. And he got all of his files, all of his emails, like hundreds of thousands of 'em, and threw 'em at uh uh, an agent ai. And it came back with a perfect description of him, timelines, where he worked, what he did, why he did it, what he was good at, what he wasn't good at. [09:28] Richard Webbe: It was, it was pretty clever. [09:29] Matt Slager: That's cool. [09:30] Richard Webbe: And if you did that for a company, wow, how powerful is that? [09:33] Matt Slager: Yeah. So that kind of thing, like this is sort of similar to what we talked about last week where, you know, external actors can. Effect and, you know, the prompt injections and all that kind of stuff. When you have your internal, like your sovereign agent running on your own data, that is like a next sort of force multiplier, you know, productivity increase, all of those kinds of things. [09:55] Matt Slager: Where do, where's that file stored? No idea. Ask the agent. [09:58] Dave Pengelley: It's, yeah. Um, the, the sovereign thing. There's some news come out around sovereign stuff. I, I'm gonna loop back on a little bit later, but just on the productivity thing with, with apps and, and tools and productivity and how AI is helping you do more in the same amount of time. [10:10] Dave Pengelley: It's easy to go, oh, I've gotta do so much more. And we do talk about it as knowledge work, as the cognitive load of, of having the next task and the next task. And there's certainly something real there. But I mean, if we think about technology throughout that ability to have things that do things faster, like a power tool, like all of a sudden now I've gotta drill, now I can drill a hundred holes in an hour instead of 10 rather with the old crank cranks thing. [10:33] Dave Pengelley: So you go, well that's actually a good thing. I can do more in the same amount of time. Did did you hear about the guy who, who ran like a, uh, lumberjack company? And he hired a new lumberjack. And this guy was meant to be the best. Everyone said, this guy is the best at taking trees down. Um, and so he hired him, gave him his tools, sent him out. [10:51] Dave Pengelley: Um, and the first day he, he fell like only a handful of trees and everyone else was, was sort of doing dozens of trees a day. And he, he was like, is everything right? He is like, oh yeah, I'm just, I'm just, I'm just getting the hang of like that. That's fine. Um, but day after day, this guy who was meant to be the best, he just was not delivering anywhere near the same results as everyone else. [11:11] Dave Pengelley: And so the boss goes, there must be something wrong with his chainsaw. Like, what, what's going on here? And so he says, can I just look at your tools? And then the guy goes, oh yeah, no worries. He comes over and the, the boss sort of grabs the chainsaw and he rips the cord and goes, r and Vick goes, whoa. What is that noise? [11:28] Richard Webbe: I feel a metaphor coming on. Go for it, David. [11:30] Dave Pengelley: But it, that, that, that, that idea, like he had this tool, but he wasn't using it. So he, this guy just hacking at trees with a chainsaw, like it was an ack. I mean, these AI tools, all these productivity tools, they exist there. But some people are gonna be hacking at the trees with 'em. [11:44] Dave Pengelley: And some people, people are gonna be firing them up and really seeing huge benefits from it. Whereas some people, it's actually gonna make them slower because they don't understand it. They don't know what it is. This, this lumberjack story, give the guy an ax, he would've done a lot better than working with a chainsaw, right? [11:59] Dave Pengelley: Um, and so there is gonna be a whole change management process and education journey that businesses and people have to go on in order to actually leverage these tools. So it's great to say, okay, let's now put, instead of an intranet, we've got our internal agent that has access to all our corporate data. [12:15] Dave Pengelley: But how do you make sure people actually know what they should be asking the agent for? How do they make sure they're actually getting the value and the return on the investment of installing that agent? Um, and so that, that, that, that, that you can have the best tools in the world, but if your staff are scared of them, don't understand them, dunno how to use them, don't recognize the value, that's a waste of money. [12:35] Matt Slager: It's true. That's a good point. [12:37] Richard Webbe: Can I, can you hear me all right? 'cause I think I've frozen. [12:40] Dave Pengelley: Yeah, your camera's frozen. [12:41] Matt Slager: We can you, yeah, you're looking very pensive. [12:43] Richard Webbe: It's, it's not the best frozen look, but leave it there for now. So we still communicate worse. [12:51] Matt Slager: It's you just judging anything that we say. [12:54] Richard Webbe: Yes. [12:55] Richard Webbe: I'm not impressed. [12:57] Dave Pengelley: Very disappointed. [12:58] Richard Webbe: I dunno how I fixed that, but, we'll, uh, we'll push on. So, David, you've been, you know, uh, you and Matt of the guys really at the cold face. You've both been working on some stuff this week. What's, what's the exciting thing you've been working on? [13:12] Matt Slager: I can speak to something hopefully that's relatable or understandable. [13:17] Matt Slager: Um, there's a, there's a tool called obsidian. That is like a, a note taking tool. You know, it's the idea of capturing your notes and owning the files, owning the data, as opposed to putting it in a tool like notion or something where it saves on their cloud. And if they die, then you don't have access to this stuff anymore. [13:36] Matt Slager: So the, the creator of obsidian, like the founder, he always talks about the idea of file over app, you know, like, own the files, don't worry about owning the app kind of thing. Mm-hmm. So I have been following obsidian for a while and ev everyone is pretty comfortable with the I of idea of markdown files now and that LLMs and agents love, you know, using markdown files. [13:57] Matt Slager: The whole obsidian thing is based on that. And they've recently come out with an early access CLI tool, so I had the idea of seeing CI. [14:07] Dave Pengelley: For the players at home [14:09] Matt Slager: command line interface. So like a, a terminal interface tool or, um, yeah, interacting with your obsidian vault. So yeah, I had this idea of basically rebuilding obsidian functionality into something completely different. [14:23] Matt Slager: But for agents, so for my agents, so for my local, you know, coding agents that I use, like Claude Code, open code, all those, as well as any kind of open cloth thing that I spin up on VPSs and they all can access this same knowledge base, which I'm calling my core, you know, like the knowledge core. And yeah, I've basically built this thing where it's like the obsidian CLI, so it's this agent specific tool harness that they can use with zero friction. [14:53] Matt Slager: And it's just like they're accessing a, like a hive mind, if you will. You know, so it's my notes, my knowledge, and all of these agents are able to interact with it, add to it, edit, you know, give me things to look at. Um, so yeah, like I built my own rust based, uh, CLI application for these agents to use and it's insane. [15:13] Matt Slager: It works so well. And then also my own TUI, so my own terminal user interface for accessing this as well. And you know, basically my own application layer on top of it. So yeah, it's been really, really interesting. Really fun. [15:27] Richard Webbe: You know, people always talk about the search engine. We're really building the intelligent information engine for instant recall, for everything we want to know and everyone wanna find. [15:36] Richard Webbe: And although the human brain is limited, I think by 11 main variables in I identifying a pattern with the support of something like that, it's kind of limitless what we can do, isn't it? [15:47] Matt Slager: Yeah. It's like when I open up my thing and run a search, it actually does three different search algorithms all at once. [15:54] Matt Slager: You've got the BM 25, which is like a keyword, and you've got a similarity search, which is through the vector stuff, embeddings, and then. Also just through connections, through the actual knowledge graph itself. So yeah, it's pretty crazy. Um, [16:09] Richard Webbe: that is, that is very impressive, Matt. I like the way you talk. [16:11] Richard Webbe: Thank you. That, [16:12] Dave Pengelley: that, that's wild because I, I've, I've been saying to a few people recently, like, search on at the operating system level has, feels like it's got worse. So I remember back in the, the nineties, I was like, back in the dos days even, and, and through the early Windows days, I was very regimented with all of my file structures. [16:30] Dave Pengelley: Like when I was in uni and stuff. I'd have this year, this semester, this term, this subject, this, and like I just had like deep hierarchies of all my folder structures so I could quickly and easily navigate to things and, and find where my data was for any given project on any given subject. Then as search technology got better in the, in the operating system layer, I kind of became lazy and just went, ah, just have a documents folder. [16:52] Dave Pengelley: It's all in there and I can just search the name, bang, it's gonna find me. The thing. I don't need to worry as much around having my, my detailed file hierarchies, but I found in the last couple years. Maybe because there's just so many tiny files and things and all the little node and web files and all the other bits and pieces that are involved in being in an operating system now that get indexed. [17:12] Dave Pengelley: The search engines at the operating system layer on Windows, on Mac, doesn't matter, are just getting dumber. And I can't find files quickly anymore. They're just, they don't surface things. And I've gotta go back to having an old school F folder structure. And the LLMs, the LLMs are great with stuff they've been trained on, or if they've got a rag database they can sort of vector and search in. [17:33] Dave Pengelley: But have you tried searching for old chats inside chat GPT or any of these things? It's rubbish. Like just try to search through your old chats is so impossible. And I'm like, are you kidding me? Can you not give me a better way of searching through my old conversations? [17:49] Richard Webbe: So let me ask, let me ask the obvious question that some of the punters will want to know. [17:54] Richard Webbe: So I'm sitting here, I've got. Uh, Mac, I've got a laptop, I've got iCloud, I've got, um, Dropbox, uh, I've got all my emails here and inside Gmail and a few other places. Is there an agent I can buy or install? 'cause I'm not technical. I used to be, but not anymore. And will, it, will, will be my personal LLM to cut across all of that data. [18:21] Richard Webbe: How would I do that? [18:22] Dave Pengelley: What, what you've got is fragmentation. Um, yep. I mean, Matt, Matt's probably got a similar or different point of view. Step one is rationalization. How many of those places do you really need? And doing some consolidation would be the most probably impactful thing you could do. And then secondly, and, and, and I'll let Matt talk to this, there's plenty of like what we, what I call hands you can give to agents that would interrogate those different data sources. [18:48] Matt Slager: Yeah, you're right. Yeah. The, the hands thing is like such an easy analogy for people to think of. I guess like you, you mentioned the chainsaw before. You know, like the idea of cutting down a tree, you basically need something to, to hold with your hands in order to actually get it done. So the, the actual sort of tool dev language of that is just the simple API. [19:11] Matt Slager: So as long as those software layers give you APIs that you can access or that your agents can create tooling around, that's what you need. And that, that last sentence is the key. A lot of people don't realize that if you just run up your, your cord code instance or anything like that and you say, Hey, I have a Gmail inbox, I would love you to be able to search through it. [19:31] Matt Slager: How can I give you that access? It will pretty much say, well, you probably need to go to the, the, the Google Cloud platform and give me some credentials and you know, I'll walk you through it. Um, like that's how I learn. Pretty much everything that I've done is just saying, I have this thing that I want to do. [19:49] Matt Slager: So let's research and figure it out together. And you have to treat me like an idiot and do one step at a time, otherwise I'm not gonna keep up. Um, but yeah, like the, to, to get like a, an instant all in one thing for what you said, Richard, about how do you just buy that agent that does all those things, that's what all these companies are trying to create right now. [20:09] Matt Slager: Um, you know, you've, you've got so many, I, I could probably try and list a few, but the thing is that they're gonna keep coming out and there's gonna be these weird new names, like, so and so agent and you know, this company's bought this agent and you know, like we keep mentioning open core a lot over these last few episodes. [20:26] Matt Slager: That's just an example of somebody going, what if we just did it and didn't take long for that to get acquired? So [20:34] Dave Pengelley: e even tools that have been around for a bit longer for automation stuff, things like N eight N that allow you to connect into these platforms and systems and then connect that, like run agents across 'em, do things. [20:44] Dave Pengelley: Even we talked a lot linked last week around my little DC superhero team for my internal little, um, business operating system sort of things I'm doing with, um, antigravity, what I call anti claw. Oh, which I'll, um, I've now put a version on Git as an open source repo for people to go and pull down my little DC superhero skills. [21:03] Dave Pengelley: I sort of stripped all my IP out of it and made well outta my business data and made it sort of a generic starter pack that people can install and go through an onboarding process to explain who your company is. And then you've got my little team of agents to help you out. [21:15] Richard Webbe: Mate, your story has been doing the rounds. [21:17] Dave Pengelley: Um, so that's, that's actually something people can install into Claude Code or Anti-Gravity or Cursor or anywhere else and, and Codex. Um, it works across all these new LLM coding platforms and know what skills are 'cause that's the technology. Um, where was I going with that? Um, but I wanted to give it hands. [21:36] Dave Pengelley: I wanted to give it hands, uh, and so I wanted it to be able to check my calendar and see what events I've got coming up. I wanted to be able to scan through my inbox. And I didn't want to go to the Google API and work out how to write a full API connection through Google, but I knew in N eight, NN eight N had pre-configured Google Mail connections. [21:55] Dave Pengelley: It had Outlook connections, it had all these things. Plus I could then use manual APIs if I needed to. And so I said, here is a, here's a N eight in workflow that can read my inbox and can read my calendar. Here's another N eight in workflow where you can send emails. If so, you can send me updates to my inbox from your own email address. [22:14] Dave Pengelley: Um, I wanted to, to review all my web stats and things, so I did a little bit of homework on this and, and worked with it. And I built some manual connectors to Google Analytics, um, and to my, to the YouTube, API, to pull my YouTube views and so it could go and check all my, my content analytics. And so I used N eight N as sort of that middleware to use that, that old school tech term, um, to give my AI the hands and that, that that could apply to any ai. [22:40] Dave Pengelley: So, so those. Make Zapier, n any of these things that let you create webhook endpoints, you can use those and any of the things they can connect to in a drag and drop workflow manner you can use to build connections into any of those systems using any LLM [22:56] Richard Webbe: hundred. So this is RPA on steroids, right? [23:00] Matt Slager: Pretty much for those. [23:01] Matt Slager: Don't [23:01] Richard Webbe: remember that. [23:02] Dave Pengelley: It depends on your definition of RPA selling. [23:05] Richard Webbe: Yeah, I was involved in selling it a number of years ago. I'm assuming the world's moved on, but it's robotic process automation, reducing civil chair and just api. Everyone together. [23:14] Dave Pengelley: Yeah, I think, I think some of the RPA, I think Pega would argue RP a's still got a place in the world. [23:20] Dave Pengelley: Um. Yeah, I mean e exactly. And I think what's happened is we've seen the accessibility and consumerization of these tools. So rather than needing a massive a hundred, multi a hundred thousand dollars Pega investment to put in BPM and RPA, now there's platforms like N eight N that allow people to do it cheaper, easier, faster, at a smaller scale. [23:40] Dave Pengelley: So SMBs can now access this technology that once upon a time only enterprise has had. So it's a bit of that democratization of these technologies, uh, is what we're seeing. Yeah. [23:50] Matt Slager: So you feel that we're still at the point now where the, the humble no code code automations have a lot of place? [23:59] Dave Pengelley: I think. I think so. [24:00] Dave Pengelley: I mean, obviously like the more vibe coding picks up or sort of, sort of ag agentic engineering Yeah. I think is the, is the preferred term now. Um, as, as that gets more sophisticated, that's good. Um, but at the end of the day, you still then need to host those connections, how you're managing your security, where those workflows are. [24:20] Dave Pengelley: Sometimes, you know, it's easier to just use a visual builder depending who you are. And like, like, you know, for me that was much easier. Just go into N eight N and create a few little workflows to give my anti-gravity agent hands For you, Matt, it might be quicker to smash up some CloudFlare worker APIs, um, and, and push them up to, to CloudFlare. [24:41] Dave Pengelley: I don't know. They each to their own, but I think there's no one way to roam [24:46] Matt Slager: like the, all of these APIs are usually just HT TP requests of some variety. [24:51] Dave Pengelley: Yes. [24:52] Matt Slager: Usually a post or a G or something like that. Just super, super simple. And those, your agents can just do like, like a good example is like I, I, I still build N eight N backends for various things like voice agent companies that need tooling and stuff like that. [25:09] Matt Slager: And that's just like a good preferred way to do it. 'cause like you said, it's easy just to visualize it and see the flows and, you know, build the, the cases, but. I use my local agent to curl those endpoints with the expected payloads to then check that the workflow works, you know, run this through all the different cases. [25:28] Richard Webbe: That is very clever [25:30] Matt Slager: and [25:30] Richard Webbe: I used to get overwhelmed merging two databases between Microsoft Outlook and, uh, you know, pick which field goes with name, first name, I mean. [25:39] Matt Slager: Yep. [25:40] Richard Webbe: That is very cool. [25:40] Matt Slager: Yeah. Then like you have other applications like Linear, for example, which is good for just task and project management through teams, and it's kind of primarily built around software and dev teams, but mm-hmm. [25:52] Matt Slager: That they have a an API interface that doesn't really exist in many places. There's an official MCP, but it's kind of slow and clunky and it just clogs your context window. So I went in search of a CLI version of the same thing, so that my agent could just call commands that hits those endpoints accurately, directly with the right auth. [26:14] Matt Slager: And everything just works instantly. I like, I could have set up multiple different N eight end workflows and end points and different little, like entry points, you know, with the web WebBook notes, but I was like, this is what I want to do. Can you just go figure it out? And then it asked me a few bits and pieces. [26:31] Matt Slager: I had to give it a few little bits and pieces. And at the end of it I've got this perfect, uh, agentic just interface layer now between my coding agents locally and then the, the linear app itself, which I don't even open anymore. Like, I basically say, let's, let's organize our work and it'll just find the lo like the most relevant issues to go through, batch them and put together a plan and go and just complete them at the end of it. [26:57] Richard Webbe: Let me ask a question of you two. I'm just thinking people might be thinking about AI and copyright, not the copyright that you wrote a song. It sounded like Kenny Rogers and the Kenny Rogers family's not happy about that 'cause it's used a library and pinches the sound. Or I think there was some lawsuits going around a a few months or 12 months ago. [27:17] Richard Webbe: What I'm interested in is in the software world, we've always said, here's an idea. Wouldn't it be great if someone wrote a bit of software like that? And then the person who went and wrote it, which took effort and money and investment and testing did it, and then they went to market and became a company. [27:34] Richard Webbe: And like some of the companies you're talking about, it came something since vibe coding is so quick, right? Just like you said, gotta be smart. Like you guys, like people like me can't do it. We could suggest it, but we would know where we're going. But within context, if you vibe code a, a widget, an app, an interface, an agent that works really well, what if someone does it at the same time? [27:58] Richard Webbe: And where does copyright come into with that sort of stuff? What do you think? [28:01] Matt Slager: It's a really good time for you to say that. 'cause there just was this massive thing about Anthropic having a massive whinge about the Chinese companies stealing, uh, data from their models where they were the ones that stole all the data in the first place. [28:15] Matt Slager: So, um, yeah, like the, the whole copyright thing, it, it almost is gonna become a myth. It's my, my, my closest analogy that I think most people will be able to like, paint a picture of is music. You know, if I go and I take one of the main melodies from one of the old classic Beatles songs and I remix it and turn it into some club hit, you know, people aren't gonna say, Hey, you stole that from the Beatles song. [28:39] Matt Slager: That [28:40] Richard Webbe: I recognize it. Yeah. Because I think, uh, my brother's a Muso and he's been a recording artist and performer, Chris Webb, uh, on the road for, you know, 40 years. And he tells me I'm, I'll probably get it wrong, but he, he basically said to me, there's like a limited number of notes and courts. [28:56] Matt Slager: Yeah, [28:56] Richard Webbe: there's a lit, like, it's like, [28:58] Matt Slager: yeah, [28:58] Richard Webbe: only a bit above single digit. [29:00] Richard Webbe: Right? And then the arrangement of those, of course, therefore must be limited. Uh, and therefore, um, o dear, uh, how many new tunes can you actually come up with? There is a finite number. Uh, and then on top of that, we know that copyright under music and art, oh, I'm not sure, matter under music and songs and Right. [29:18] Richard Webbe: It's, it's not a forcible after 50 years, is it? That's why people get Mozart's music and reconverted into a rock song and do exactly what you just said. And I had this feeling the same way, Elon Musk and that was saying that money will no longer be a value. Um, i this feeling that information will, will be totally democratized. [29:38] Richard Webbe: So when you are in the information industry and you make money out of sharing information, if anyone can get access to information, you're not gonna make money out of that anymore. [29:48] Matt Slager: A hundred percent. That's what a SI is. [29:50] Richard Webbe: Sorry. [29:51] Matt Slager: That's what a SI is, you know, people talk about a GI, artificial general intelligence. [29:56] Richard Webbe: Yeah. [29:56] Matt Slager: Which we basically have now. It's very close. It's just fragmented. But a SI is that, that level where we hit that singularity and everyone has access to everything at any time. [30:06] Richard Webbe: Totally democratized. So, so all of a sudden we have to think like tradespeople, right? Uh, and I mean that literally as in the trades person doesn't make the brick. [30:17] Richard Webbe: He just lays it a certain way. [30:19] Matt Slager: Mm. [30:19] Richard Webbe: That's why we pay him. 'cause he's a, her, I should say, uh, slap me down. Um, you know, that's the value add. So our new world of working is gonna be, if you get that information, how are you adding value to it? And of course, information's being recreated all the time. So we're not saying human creativity will stop, we're just saying that. [30:38] Richard Webbe: Okay. I've got this. 'cause I, I, I keep thinking, I mean, I got up this morning and, um, I've been writing a book for about four years, and with the help of ai, I, I've got a lot of it done in one morning. Now I already had the content there, I had the stories, the ideas, but to assemble it and, and add a little bit of that information that I didn't have, that's now democratized. [31:00] Richard Webbe: It just appears. It's just amazing. [31:03] Matt Slager: Yeah, it is nuts. [31:04] Richard Webbe: It is nuts. I like that too. Very nerdy term. It's nuts. [31:09] Matt Slager: Well, you should say hi to Joe as well. She's commented in and Oh hey Joe pieces. Um, [31:17] Dave Pengelley: yeah, she's, uh, here we go. Each, each of our vibe code projects have our own nuances. Similar applications, but minor variants. [31:23] Dave Pengelley: It's interesting, I, I don't think Richard is a copywriter is gonna be trademark or patent, like as patents. You can patent like a design, the design pattern. So it's not the exact code, but it's, you know, that, that A, plugs into B, which generates C and that pattern of function is something that people patent. [31:43] Dave Pengelley: And I know there's, there's, there's varying arguments on this. And I say this as someone who is a patent holder. Um, but, uh, I don't own the pattern that my ex company paid for it, but it's got my name on it. Um, but that idea of, of encouraging innovation by allowing the inventor and the innovator to profit from it for a fixed period of time encourages more innovation. [32:01] Dave Pengelley: So there is an importance to things like patent law and copyright and trademark to encourage innovation and invention. But at the same point in time, if five people invent the same thing at the same time, because the technology is a, is allowing that to happen. Like if five people create an open core, at the same time, if one company like Open ai say they bought it and trademarked it and, and whatever, 'cause they bought Peter Steinberger, they, they, they're putting it in foundation. [32:27] Dave Pengelley: They're not doing that. But say they did, then anyone else who creates an age agentic interface to have an age agentic personal assistant that has a heartbeat function and can be wired up to tools, can they then. Go and litigate anyone who wants to create something similar. Um, that, that's where I [32:46] Richard Webbe: I it's a very good point in the patent world and the copyright world, there's a few basic rules. [32:51] Richard Webbe: And like yourself, David, I've been down this path, a friend of mine who created one of the very first online auction websites and he wanted to patent the technology. And, uh, Chris land, if you're listening, how are you doing? And what's interesting is he and I went and saw his patent lawyer with him and he looked at it and he said, oh, you've got a patent. [33:10] Richard Webbe: This, this is really important. So we did that. And then someone came up with a much worse clunkier, you know, straw man type, pulled together bits and pieces. And the minute we went back to the lawyers, oh, it's very hard to defend this, right? They took the money for the patent. They didn't wanna defend it, they didn't want to get clots. [33:25] Richard Webbe: You know, you can't copyright or patent the fact that the sun's coming up. 'cause the rule of copyright and patent it. If it's a known fact you can't patent and facts. So if serendipity. Creates a situation with I AI where at the same time the fact exists that you are creating something that confirms that fact you end up in a world of pain. [33:46] Dave Pengelley: Mm. Yeah. [33:47] Richard Webbe: I think it's hard to understand. I think timing used to be the one who, who loads the patent first? [33:51] Dave Pengelley: There, there, there, there is something to be said. I mean, some people, um, like really wanna hoard patents for that very purpose to to, to litigate and make money off the, the fact they own the idea. [34:01] Dave Pengelley: And I know a lot of businesses, um, they have defensive patents, so if there's an idea, they would've patented it, patent it. So that way if someone tries to say, Hey, that was my idea, they can go, well, hang on, we've got the paperwork that says, no, you can't, you can't take us to court for that because we own it. [34:16] Dave Pengelley: You can keep doing it. We're not, we're not gonna stop you from doing it. But you can't, you can't come at us for [34:20] Richard Webbe: that. Some copyright and patent. And I'm just confirming for I'm not a lawyer and don't take my advice. Um, copyright and patent and all that stuff can be trumped if you can produce a device that was built. [34:32] Richard Webbe: Well before they, you can't, so it's not like going out and registering website names, you know, harrison ford.com. So I make money on the back of Harrison Ford. It's very, I dunno if you remember the Rotary Engine created? [34:44] Dave Pengelley: Yeah, yeah, yeah. [34:45] Richard Webbe: Uh, Mr. Mr. Winkel was his name. Mm-hmm. Well, what was funny is my dad designed one and built one that was the same basic functional rotary style engine. [34:56] Richard Webbe: Well before Mr. Wankel did it for the, uh, the Reno organization or whoever it was. And I kept saying to dad, dad, why don't you just go and sue the ass off these guys? And I think he actually did speak to someone, but it's, it's a very long, expensive, difficult road and maybe that way on purpose. Because if you invent something that's great, if you invent it, work on it and take it to market. [35:19] Richard Webbe: And it helps people. That's a different story. So, hey, we don't know the answers, but they're important ones. Matt, you're gonna say something? Sorry, I think I cut you off. [35:27] Matt Slager: I said Winkle. Yes, [35:31] Richard Webbe: I said it too. And we're not swearing. It's the name of the man that the luxury engine. [35:37] Matt Slager: No, that's true. It's like the thing that I built, which is like, you know, quote obsidian for agents. [35:41] Matt Slager: It's probably not very nice legally for me to use that terminology 'cause it's, you know, it's playing on somebody else's IP and that kind of thing. But it's the analogy that I'm sort of using to do with how can I give ev everything access to its own note system, but also share that note system, you know? [35:59] Matt Slager: And then, you know, like your intranet, I, I'm gonna be installing agents for my clients. You know, the people that I speak to in Slack that are relying on me for work. You know, those companies are gonna have their. Like mat agent that has access to its own version of that obsidian for agents where it's its own little knowledge core for that company. [36:22] Matt Slager: You know, the sovereign data like you mentioned before. And yeah, for them to be able to just instantly look stuff up, it's, I think we've mentioned in the past, like when people could all of a sudden chat to A PDF, it was really exciting and really fun, but now we can spin up multiple PDFs with, with specific data, um, variables, you know, to help, uh, search through things and explore things and wonder and also be able to create and, uh, recall anything you need. [36:49] Matt Slager: So, I don't know. It, it's exciting. And does that become something that I'm now like able to be liable for because I've not copied, but it's like a similar approach. You know, like, and then [37:02] Richard Webbe: if, if we follow Elon Musk and, uh, a few other futurists view of the world with ai, the automation and the technology, and they say money won't become a part of it. [37:14] Richard Webbe: So then if money's not a part of it, patent copyrights don't matter. [37:17] Dave Pengelley: It's all bitcoin in the future. That's why money doesn't, fiat currency is dead. [37:22] Matt Slager: It's a good point though, because it makes you go like, you know, if I, if yeah. If I was to try and commercialize that, how would I monetize it? What would the, what would the, the business model look like or the money model as Alex Mosey likes to say. [37:36] Matt Slager: Um, and then is there any point trying to put a payment behind it or do I just put it out there for free and those that want to install it can install it and run it on their own hardware? Like where, where is the economics at scale when everyone can just. Make things and provide it for others. [37:54] Richard Webbe: Well, he, here's an interesting perspective. [37:55] Richard Webbe: I'll show up in a sec, Dave. Sorry. No worries. It just jumped into my head. [37:58] Matt Slager: I don't wanna interrupt. [38:00] Richard Webbe: Perspective. So when musicians used to make money out of the music, right, and they don't anymore, they make money out of the performances. Mm-hmm. Because we, you know, the amount of money they make outta music now and everyone was scared the creativity of music was gonna be killed. [38:16] Richard Webbe: Because if, you know, you know, capitalism saves everything because money was put behind medical research and it just accelerated, all went forward. And, uh, not saying capitalism's perfect, but it was a way of financing and motivating people to do new things and create new things. And so when I think about what we're talking about with AI and the music industry, so they went from selling records to performing. [38:39] Richard Webbe: So maybe the currency of the future is getting recognition that you are the person that created that and you become like a rock star where people want you to come around. Pass, you know, judgment on what they're doing. I, I, I, I'm thinking, I [38:53] Dave Pengelley: think [38:53] Richard Webbe: intellectual currency. Sorry, David, you go [38:55] Dave Pengelley: for th th this is the open source thing. [38:57] Dave Pengelley: The, the, the, that comes up a lot. And I mean, you mentioned, I think last week, Richard Red Hat, red Hat, red Hat Open sourced, um, Linux or Ubuntu or whatever it was back in the day. Um, but then they sold professional services on top of it. They sold support packages. So here's the thing for free, but we know you can't do it yourself. [39:14] Dave Pengelley: Um, and so pay us for some guarantees to help get it working and make it run effectively. Uh, and so, uh, more recently we saw Open Claw Peter Steinberger, and I was talking to chat go like, how did he make any money off OpenCL? And he didn't, he got money by being Aqua hired by Open ai. Yeah. But he developed this thing and put it out there. [39:37] Dave Pengelley: Now in the future, people like Matt, um, who create agent systems and, and you know, I'm, I'm working on some stuff as well around the, the. Anti claw idea I've been talking about and, and seeing what I can do. Productize with that again, I mean, unless you wanna have a fully hosted SaaS where you wanna create a platform where people don't have to do any hosting, they can just log in and pay you money and you manage it at all and they just run it on your systems, that's one monetization path. [40:02] Dave Pengelley: The other one is you do give it away and you can also sell professional services on helping people set it up and implement it. So here's this thing. Sure, you can have it free, but if you want our best practice knowledge and guidance to implement it and set it up and do the education and the onboarding so they know how to pull the crank on the chainsaw, then give us a call. [40:22] Dave Pengelley: Um, [40:23] Richard Webbe: yeah, so it's, it, it is that depth of understanding. We see those dummies books and stuff like that. Excel didn't make us all accountants. It just made us more accurately count things. [40:35] Dave Pengelley: Mm. [40:35] Richard Webbe: Uh, and so you, you, you guys are saying that, okay, yeah, you'll be able to create this yourself, but then what do you do with it? [40:40] Richard Webbe: How do you manage it? How do you service it, and where do you go from there? And we do need people with experience, perspective, and creative minds to help us navigate where we go in the future. Of course, this is only gonna roll out slowly over the next a hundred years, so now I should panic now that we're getting rid of money. [40:55] Richard Webbe: But it's an interesting perspective to have. [40:58] Dave Pengelley: Yeah, I mean, ev even my anti clause system, um, that I, I've built up with my little skill files with my superhero agents. I didn't share that straight away because I know it's a, a limited audience, people that can actually comprehend what that even is and how to use it in a tool like antigravity. [41:14] Dave Pengelley: I mean, like, it's one of those things that that's set for a specific audience that are comfortable in a code editor. Um, I mean, you are, you are very competent, Richard, but I, I think you don't probably wanna be living in anti-gravity code editor to talk to business files and things. That's, that's not where you're gonna be your happy place. [41:31] Dave Pengelley: And so as much as I could deploy this and give you, give you a version of this that's not actually. Suitable. Um, I, and so like, if, if you, even the concept of the, this sort of business in a box hierarchy of agents. I was talking to some people around that concept and then they were saying, do you think people would even know how to use that? [41:50] Dave Pengelley: Do you reckon like that would be even a thing? And it's just because it's so beyond what people understand. The way people like us in our little AI bubble are talking about these tools and how to use them. The regular Joes out there are just like, that's too much. It's like the chainsaw. Like, what do you mean? [42:09] Dave Pengelley: That this thing could actually fire up and spin fast and cut things out fast. It's just incomprehensible to them. Which is, which is, I'm gonna, I'm gonna try and do something that, um, I think is relevant here. Can I find it and bring it up on that we're seeing this? [42:28] Matt Slager: Yep. I actually saved something on x uh, earlier today too, to see if I can share the link. [42:35] Richard Webbe: People giving open claw root access to their entire life, giving them a gun and they're gonna kill everyone. [42:41] Dave Pengelley: Like giving a chimp, it's not like giving them a gun. It's like giving a chimp a gun who has no idea what it's doing, has no, no guardrails or controls, is just, um, gonna be unfettered chaos. Yeah. [42:55] Dave Pengelley: And so I, I think we're seeing there's people at that end of the spectrum going, oh, I'll jump onto this thing. And, you know, they're holding the chain while they crank the thing and they're gonna cut their own hand off. [43:05] Richard Webbe: Yeah. [43:05] Dave Pengelley: Um, so yeah, we, there's all gonna be all different technical levels as we go through here. [43:11] Dave Pengelley: So that, that whole, how do you productize some of this stuff is gonna be interesting. 'cause there's some people just build it themselves. Some will want to grab something off the shelf and use it, and they will have the technical competence to use it. Some will be like. The chimp thing there where they're just gonna cause themselves a whole world of pain because they don't dunno what they're doing with it. [43:30] Dave Pengelley: And many will just go too hard, too hard, um, [43:33] Richard Webbe: time to market, which is a really, Matt, you know, you, you specialize in that area. Helping companies, you know, time to market is, is a critical factor, isn't it? [43:42] Matt Slager: A hundred percent. Yeah. When, when you have a founder or a company lead go, I have this amazing idea, let's find somebody to do it. [43:51] Matt Slager: They want it done that week. And a lot of the time, even with all of the AI speed up, a lot of the time, these, these large projects are delivered over, you know, a month at least. And like, yes, that's fast now compared to what it used to be, but you still gotta wait that period. Um, and on the same token, uh, Dave, with the idea of handing. [44:13] Matt Slager: Over the, the gun there. I just shared a quick link that you might be able to share with people as well. It's, Peter actually responded to this, the creator of Open Claw and said, this is good. People need to know that this thing exists. You know, this is somebody that had a scenario where their open claw kind of went a bit rogue and they were actively trying to stop it and you can even see, so there's a little bit of tool call notes. [44:40] Matt Slager: Basically the thing started saying, yeah, no worries. I'll clean up your email inbox for you. Let me go find everything and clean it. [44:47] Richard Webbe: Do no good, then [44:48] Dave Pengelley: stop. Don't do anything. Stop ilaw [44:56] Richard Webbe: the processes. I'm the host. You know, I always think about you and, and this the policing, the security. Uh, all of that are extra things that we need the experience on. You may all remember the when, um, the UU US Stock Exchange, everyone was running an algorithm on, I think it was on a sun platform from memory many decades ago. [45:20] Richard Webbe: And it was a buy, sell algorithm. Mm-hmm. All the agents had it, you know, if it drops below this percentage sell, if it goes above buy or vice versa. And I think the US Stock Exchange had to shut it down 'cause everyone was running the same algorithm. With the same parameters at the same time. So if a share price slightly moved, it went bang because it started selling, it went bang and it was like, man, Poseidon, it was mad. [45:46] Dave Pengelley: Well, now, now they still do a lot of that, I think, but it's, um, high frequency trait, like, like the nanosecond. So they, the ones that win that sort of arbitrage trading are the ones with the fastest connection to the stock data where they can make the transactions before anyone else. Wow. And they're in and out. [46:04] Dave Pengelley: So they, they talk about this high frequency trading, uh, and there there's people that are running super computers close to the exchange data centers so that they get those sub-second times and can execute these buy and sell trades very quickly. And just, you know, it's the old Richard Price sort of shaving cents off every dollar kind of strategy that they're doing. [46:23] Matt Slager: I have a, a, um, [46:24] Richard Webbe: coin mining Go on. [46:25] Matt Slager: There's a another question from from Joe here, which I feel like I wouldn't mind taking the reins off, but Yeah, I I we need to [46:33] Richard Webbe: get Joe on the program, by the way. She's [46:34] Matt Slager: great. That would be good. Yeah. She could join us. Um, so yeah, the difference between all of these different agents really comes down to how you are accessing them and how they are accessing your files. [46:48] Matt Slager: So anti-gravity lives inside an integrated development environment. So an IDE, the place where you go write code, um, you can see files, you can do things. It, it's a big multiple function suite type thing. Mm. Whereas Claude Code, just using your example there is the same thing with all of that abstracted. [47:11] Matt Slager: So Claude Code literally just is a terminal agent that has the exact same functionality that anti-gravity has. But you don't see any of the file stuff. All you're doing is talking to it. Now, obviously there's layers to that, you know, those that know know that you can have various file watching applications and you can have your GI tracking applications through the terminal window as well. [47:34] Matt Slager: But the premise of it is you have an agent that can interact with your files in a specific directory. How do you actually interact with that and what do you want to do to interact with that? You could just pipe it into Telegram, for example, and you know, use that as your interface and it's running now on your computer and you can just talk to it wherever you want. [47:54] Matt Slager: Like that kind of selling point is pretty much why. I reckon Peter was like, let's go and make this open call thing. So it's this always running thing that has access to your computer in its own space. You know, the, the choice between tools is kind of like going to a bookshop and going, well, which book do I get? [48:12] Matt Slager: They all have pages and they all have words. It just comes down to which one you enjoy interacting with the most on which user interface level. [48:21] Richard Webbe: Matt, your view of solving a problem with abstraction and creativity is quite astounding. I enjoy listening to it when you say, oh, well, let's try this. I, I know why you do what you do. [48:32] Richard Webbe: It's cool. [48:33] Dave Pengelley: Thank you. [48:34] Richard Webbe: Actually, I should say it. Dave does that with his agentic AI family and his superheroes. He spins me out all the time. [48:40] Dave Pengelley: The, um, I mean the, the, the models are all getting better and closer and they can all kind of do the similar jobs anyway. They, they might execute some aspect better than others. [48:50] Dave Pengelley: So I'm building a complicated thing with a frontend and backend, and I'm living in anti-gravity as my primary IDE, uh, in development environment. And I'll actually say to it as I sort of plan out. Functions, I'll say which model I have access to within Antigravity, Gemini, pro Gemini Flash or um, sauna Opus, or I actually still have open AI codex. [49:12] Dave Pengelley: I've got that as an option as well. He goes, oh, for that particular backend thing, that structure and function, give that codex, oh, that's very front end react web interface thing. That's sauna. Okay, that's somewhere in between and requires deep thinking. Use Gemini Pro. And so I'm actually getting the models to recommend which model I should apply different tasks to. [49:30] Dave Pengelley: Um, if you are just in one, then maybe you've just got one and that's fine. But you've just gotta, there, there's no right or wrong answer I think. But it's finding where you feel productive and I feel really productive in anti-gravity and a little bit in Codex. 'cause I've gotta actually switch between the two. [49:45] Dave Pengelley: 'cause I'm hitting quota limits now. I'm coding so hard now that my, my, my paid plans on both. I'm just in the, the mid-tier plan. I'm just running out of. Actual token credit and asking questions, um, which is actually starting to slow me down. I'm like, ah, such a big jump from like the 20, $30 to like the 200. [50:04] Dave Pengelley: And I think with Gemini, two hundred's just there on special price. It's normally 400 a month for the Gemini Ultra. Like it's just crazy expensive. In which case you go, oh, well I did like working in anti-gravity, but if Claude Code Ultra is half the price, maybe I just need to move everything over to working out how to use Claude. [50:23] Dave Pengelley: Um, [50:24] Matt Slager: well, I think your best choice would be to move to something like Cursor. You know, like Cursor was one of the best first agent ideas that came out and they had a really good run. And there's still people that use Curse and they never stopped. I used it for a while and I'm like, eh, I don't need it anymore. [50:41] Matt Slager: The same with Antigravity. I was there and then didn't need it. [50:44] Dave Pengelley: Mm. [50:44] Matt Slager: So. I think we should probably talk to people more about tools that exist between the layers that they're aware of, because Cursor as a, as a tool is completely model agnostic. [50:59] Dave Pengelley: Yeah. [50:59] Matt Slager: You can run any model through it as long as they got it in there. [51:02] Matt Slager: So all of the Codex models, all of the Anthropic models, all the ge, all the Gemini models, you know, even they have their own, which they call Composer. Um, [51:11] Dave Pengelley: I, I did jump into Cursor for a little bit before Antigravity because I had Codex and, and Open AI didn't have their own desktop app at that point in time. [51:19] Dave Pengelley: I, but they had a Codex plugin for Cursor. So I started using Codex through their, their cursor plugin. Uh, and it was okay and I was getting reasonable results, but I wasn't really happy with it. And then Antigravity dropped and I jumped over there and the Gemini just blew me away. And I was like, [51:32] Matt Slager: so, [51:33] Dave Pengelley: oh, I live here now. [51:34] Matt Slager: I know a hundred percent because I would followed this pretty closely. You were using the rubbish version of Codex. Like you should have just been going straight through direct usage in Cursor because [51:45] Dave Pengelley: Yeah. Why, why, why would they put a rubbish version in the plugin? I would've thought the plugin should be using the C-L-I-A-P-I-I [51:51] Matt Slager: ask the open AI devs. [51:52] Matt Slager: They, they, why can't they name anything? I [51:55] Richard Webbe: just wanna, any of the companies that we've mentioned, we are looking for sponsorships, so please send your money. [52:02] Dave Pengelley: Of course. Yes. Always. Um, there's, it's interesting, although this plan thing and hitting quotas, I saw this post today. I'll hide his name because, um, I don't wanna embarrass someone unnecessarily, but I, this is the weirdest post that I've seen around all this sort of stuff. [52:15] Dave Pengelley: This guy feels like it's got, he's got cash to burn. AI power users need a hand. I've been hedging across a smattering of the $20 plans, kinda like me, uh, to stay on top of the best models for different tasks, but feeling a bit of fomo. It may be time to double down and jump on a single heavy duty plan and make sure I'm not missing out on anything. [52:31] Dave Pengelley: Thoughts worth going on? The heavy plans, even if you aren't hitting usage limits, are the conversation output's noticeably different? Is my fomo legit? So, I mean, he is not just doing it, but he is, he thinks that maybe to your point, Matt, is you using a crap version of the models because he's only paying 20 bucks? [52:46] Dave Pengelley: Will he get a better version at 200 bucks [52:48] Matt Slager: that exists? I can speak to that. Like in some cases, that absolutely exists, where the higher tiers will give you priority inference, which is faster and on better chips with better ram, which gives you a better result. Really inherently. Yes. Not every company though, he's not crazy. [53:06] Matt Slager: Yeah, no, not every company would disclose that. But yeah, 100%. And you even get early access features and stuff too, which I don't know if that's of use with regards to the FOMO idea, but, um, [53:16] Dave Pengelley: I just thought it was weird that, that he's not hitting usage limits, but he wants to spend five, like 10 times as much money. [53:23] Matt Slager: Yeah. My progression personally, like I was on Gemini CLI building, using just the Google $20 US plan, and all of a sudden my usage limits were pinned and I couldn't do anything about it. So I'm like, okay, let's, let's get into some, some Claude stuff. So I re, I restarted my $20 Claude and then just pinned it. [53:42] Matt Slager: So I'm like, okay, it's the a hundred dollars Claude. And then I pinned it and then I'm like, okay, I've gotta go to the $200 US quad. And then I haven't hit the cap on that yet. So yeah, for somebody that's not hitting the caps, I don't know, you would have to look which each, what each company discloses what they actually do and don't do at the different tiers. [54:01] Matt Slager: Um. You can ask your existing agents to go research that for you and then yeah, make that decision based on feature. But I always tell people, use the tools that you are currently paying for and if you haven't like capped to the limit and you wanna try a different one, that's fine. You can always cancel a tool and go to another tool. [54:19] Matt Slager: Like [54:20] Dave Pengelley: the other night, I, I hit my code, I, I ran out of Codex 'cause of the, the amount of deep backend work that I, I was doing and I was getting do some sort of Ralph style loops. And I'm fortunate, like I, I mean I talked about my business subscription and some of the pros and cons of that, but one of the pros turned out to be that you've gotta have two seats. [54:36] Dave Pengelley: And so I'm paying for two chat GPT licenses on the business plan. So I got my wife to gimme her password 'cause I gave her one of the seats. Uh, and. On that particular plan on some open AI plans, it's a global quota, but on the business plan specifically, it's per seat. So I logged in as my wife and the Codex environment didn't care who I was logged in as because it was still operating off the same native file system off my computer and I was back to a hundred percent quota and back the building. [55:01] Dave Pengelley: Nice. Um, so that, that was a nice hack With Gemini, I'm like anti-gravity. I'm, I'm pretty much running out of their, uh, Claude credits. I'm pretty much out of Gemini Pro credit credits for like four or five days and I think the solution's gonna be similar and I'm better off spinning up another $20 plan, like with Gemini to get a second full batch of quota than going to the $400 plan. [55:23] Matt Slager: I would like to, I would like you to get into the Gemini CLI and, you know, do the exact same thing 'cause they actually separate their quota. You've got anti-gravity and then you've got Gemini CLI and it's separate quota. [55:35] Richard Webbe: Joe [55:35] Matt Slager: just [55:35] Richard Webbe: asked a, an important question which I was gonna ask myself and that are the caps based on. [55:42] Richard Webbe: On, you know, time usage, [55:44] Matt Slager: it varies. [55:45] Dave Pengelley: Tokens, number of requests, [55:47] Matt Slager: it varies. So you, you have a, like a monthly quota, let's say, and that's usually not very disclosed, but then you'll also have a weekly and then usually a five hourly. So like, I know like on the lower, um, when I move to Claude, for example, on the lower anthropic, Claude Code plans you, if you cap that five hour, you then have to wait five hours before you can start going again. [56:10] Matt Slager: And then if you just keep pinning that, you'll hit the weekly and you can't do anything until you've like waited another three or four days. And [56:17] Dave Pengelley: that, that's where I'm at. Like, like it says five, five hour reset, but that rolls into a bigger number. And so if you hit the five hours too hard, too many times, all of a sudden it's like, okay, you are now in jail for five days. [56:29] Matt Slager: Yeah. Like, [56:29] Dave Pengelley: ah, come on. Um, but yeah, I like, again, that's the, the, some of the antigravity isn't just the models. It is the way that the system prompted it in to work with the file system to use its browser subagent and the like. With any of these models and the way they're implemented, it's like, unless you're going direct to API and maybe direct right via CLI where it's very much, this is the context you have. [56:53] Dave Pengelley: I'm giving you these specific files on this specific folder. Go do this exact thing. Um, with some of the IDs, like your anti-gravity a little bit like chat GBT with its magical context, it will. Proactively search and use a bit of its own inference or cross the top of your specific request and things like that, I think, which can be really helpful and useful. [57:14] Dave Pengelley: I, I dunno, Matt, tell me if I'm, I'm making stuff up there, but that's the way it seems to be to me. [57:18] Matt Slager: You, you're doing what I do, which is just letting your words change to one or the other. Um, I, you gotta try it, you gotta feel it. You have to literally do it and experiment. Like, uh, I would like to suggest that you can have the exact same experience as what you've been having inside into Gravity in the Gemini CLI. [57:37] Matt Slager: It just probably requires a tiny piece, more work to set up the same, like you said, the um. What was your words? The, the, the system prompting and like the layer control stuff. Like for example, [57:50] Richard Webbe: you sounds like a plan. [57:53] Matt Slager: You what? [57:53] Dave Pengelley: We, we, we we're at the hour. I do wanna wrap it up, but we, we talked at the start around, you know, sovereign data, having sovereignty, where is your data kept, intranets, all that kind of stuff. [58:03] Dave Pengelley: Now, when I was at Telstra, we were, there was like, we were still, people were still running small business servers in their own hardware. They had little hard, like every small business had a small little server rack in the corner that was running their own local exchange server for their email domain and everything. [58:17] Dave Pengelley: Or they were outsourcing it to a, to like a, an outsource provider that hosted it for them. There was no office 3, 6 5, Microsoft Cloud was very, wasn't really existing yet. It was just in an early infancy. SaaS as a whole 15 years ago was brand new and everyone was still running locally hosted servers and stuff. [58:34] Dave Pengelley: It was all on-prem as we call it Now. We are seeing, I think with this AI thing and with the amount of data and, and the, the risk and the leaks and all. We're actually coming full circle. We've gone from putting everything to the cloud to everything wanting to come back in-house, especially for your bigger enterprises and Microsoft have actually seen the wind changing. [58:56] Dave Pengelley: I saw this one come out 13 hours ago from Satia. We are building the most comprehensive digital sovereignty, so the most comprehensive digital sovereignty platform to help countries and organizations meet their regulatory security to operational needs, uh, across Microsoft Sovereign Cloud. So now they're basically making the whole Azure Office 3, 6 5 stack something that you can deploy in your own data centers, on your own metal. [59:20] Dave Pengelley: Um, and so they went from like, let's take all of the exchange and all of the, the Microsoft server stuff and put it in the cloud and you pretty much had to go to cloud. No. Like who runs their own exchange server anymore? It's all in, in there to now going, oh, now that people wanna run their own local models and their own local data and they don't want everything out there with AI training on everything, we need to give people the option to host it back on their own medal. [59:42] Dave Pengelley: Internally behind the firewall. So it's interesting to see Microsoft taking that approach as well. [59:46] Richard Webbe: Yeah, the concentric circle model, what I used to do myself today, I'll give to someone else to do so they can scale it across more people. [59:52] Dave Pengelley: Well, I mean, [59:53] Richard Webbe: think it's overused, I need to bring it back in house. [59:56] Dave Pengelley: Well, I mean, think about that with our power grid, right? We went from people had little steam engine things outside their house. Maybe if they, if they wanted to, to run something or they were doing a bit of fire stuff and then, you know, power plants came in and it all got industrialized and, and who's, who's gonna generate their own power? [01:00:13] Dave Pengelley: Who wants to be their own power station? And now, oh look, solar panels, batteries, we're gonna run all our own power back, back at the edge. Um, and so we, we go back and forth, back and forth on where does the infrastructure live? Do we want it on the edge or do we wanna centralize it? And we're seeing that now with ai. [01:00:28] Dave Pengelley: It, it's going back to the edge. [01:00:32] Richard Webbe: Amazing. [01:00:34] Dave Pengelley: Brilliant. Well, thank you gentlemen. Uh, thank you viewers. Thank you one and all, uh, wherever you've watched, whenever you've watched, if you joined us live. Excellent. Thank you, uh, for our, uh, our, our chat contributor today. Joe, thank you very much for, for participating in the show there. [01:00:48] Dave Pengelley: Um, anyone else who ever wants to join us live and contribute to the chat, you can do that. We are here at the same bat time, same bat channel, midday Sydney, Eastern Time Wednesday, and we [01:00:59] Richard Webbe: chat via LinkedIn if they've got a question. [01:01:01] Dave Pengelley: Yes, yes. All our links to our websites and things and our links are all in the description box below. [01:01:07] Dave Pengelley: So hit that up. And um, gentlemen, thank you for engaging and insightful conversation and we'll see you back here next week. [01:01:14] Richard Webbe: Thanks, mate, David, have a great week. [01:01:16] Matt Slager: Thanks [01:01:16] Richard Webbe: all. Bye.