Two Markets, Two Stacks. Why Your Solo AI Won't Survive The Enterprise Audit.

[00:00] Dave Pengelley: Always get busted. You think, "I've got enough time for a quick sip before the cameras go live," and then they go live and you're still drinking. Every **Shawn:** time. That always happens. **Dave:** Every single time. Um, how... Good, good to have you back, Sean. **Shawn:** Thank you. **Dave:** Sorry I didn't send you the link. I just assumed you had the link from last time and forgot that I should send it to you. **Shawn:** I probably do, but I don't know where it was, so You repositioned. You used to have **Dave:** that window behind you, like right behind you, didn't you? **Shawn:** Um... **Dave:** Or was it a different room? I just remember you had like this cool, almost like this sci-fi, like, space room, like this, like, you had the glow coming behind you. Like **Shawn:** the- Oh, yeah. Um, in summer the light comes right on that window, so it actually whites out the camera, so I put up another curtain over the front of it, um, so that I can actually be seen. **Dave:** Yeah. Yeah, it looks like, like you're in some kind of, like, sci-fi whi- like, clean room thing, where you've got like this, the, the, the beam of light behind you. **Shawn:** Yeah, no. Not, not today. [01:00] Dave Pengelley: **Dave:** Not [01:00] Dave Pengelley: today. Very very good. **Savio:** I have to change my, uh, background in the coming weeks. Yeah. **Dave:** Yeah, you said Sav, Savi. So, uh, yeah, wel- wel- welcome to the show, Savio. You are, you are, this is your first foray onto the AR Producer Show. I wish you luck. Um, we- Thank you ... we, uh, always kick off with a just a bit of random banter as people find that we are live and get on the show and find us. So, uh, as always, it's good to catch up, 'cause most of us don't talk to each other through the week that much. Uh, it's good to see where people are at before we jump into the news and such things. Um, looks like everyone's wearing nice warm clothes except Savio. It must be nice and warm where you are. You look like you're wearing- Yeah, I **Savio:** will turn on the heater here. **Dave:** Ah. He's- **Richard:** Yes ... he's in Melbourne like me. I don't have a heater up here. I had my scarf on before, it was so cold. **Dave:** Oh, my, my, my hands are pretty cold. That's why I've got my hands crossed, because I'm trying to tuck my hands in to keep them warm. I do have a fin heater next to me. I've gotta watch out if I'm doing recording stuff. [02:00] Dave Pengelley: If I have the heater on too much, then [02:00] Dave Pengelley: the, the room gets too hot. Yeah. My Sony camera, my aforementioned Sony camera, um, and it might have been before we hit live, but it overheats, um- ... and then shuts down. And I was like... So I've actually gotta be cold. Oh, no. Although fortunately, having, having the, the, the studio lighting, that does warm me up, but not my hands. **Savio:** In my case, um, my laptop is good enough to serve as a natural heater. **Dave:** Oh, very good. **Shawn:** That, that's what my tower does. It's like it... Just, "Oh, it's cold. All right, do something graphics intensive for 30 minutes, it'll warm the room up." **Dave:** Well, there's, there's, there's companies doing it, and I don't know if the, the economics actually balance out yet, but people are using Bitcoin miners as, you know, room heaters, space heaters, water heaters. Integrating your Bitcoin miner as your water heater, 'cause you've gotta generate heat to heat the water anyway. Might as well- **Shawn:** Yeah ... **Dave:** mine Bitcoin at the same time. **Shawn:** It's not gonna be- It really is genius. It **Richard:** makes sense ... **Shawn:** super efficient. However, if you're doing a Bitcoin mining anyway, it's a net positive. **Dave:** Right. And like [03:00] Dave Pengelley: **Shawn:** in, in a room- Depending on how [03:00] Dave Pengelley: expensive the system is. I think it's, um, China is one area I know does this heavily. They use the waste heat from their power generation systems to heat water to do in-home heating Yeah, clever ... um, so it's actually widely used. **Richard:** It all makes sense, uh, all makes sense under, is it, was it Boyle's law or whatever? Energy just changes forms. I mean, if you're gonna cool a data center and you've got hot water, rather than run it through a grill, run it through your house, it heats your house. Mm-hmm. **Shawn:** 100%. Makes all those complaints about the, um, data centers going up evaporate, sort of. **Richard:** Yeah. I, I am, I'm still lost with morons- Yeah in the uneducated technical world saying that data centers are gonna use too much water. That's like saying car radiators are gonna use too much water. It's the stupidest bloody thing, and they're still doing it, and no one's questioning them, and I wanna get on the TV and smash them all in the head. [04:00] Dave Pengelley: **Dave:** That's 'cause on- one time they weren't running closed loop systems, but now they've built closed loop cooling systems, and so [04:00] Dave Pengelley: yeah. Um, I remember hearing, I think one of the buildings in Sydney, um, the idea was, I don't know w- how true this is, especially now with all the concern around, like, overheating things. But it's got one of the, the, the sort of the mesh, sort of th- these big sort of, um, blocky out, like an, like an exoskeleton around the building. And apparently... You've, you've seen those designs where you get those hexagonal, like, grid things- Yeah ... around the outside of a building. Apparently the idea of that is it was pulling the salt water up from the, the harbor or whatever, and then that was able to run cooling and stuff, and then they were dumping it back in. So it was a, this sort of system which was using the natural water to cool things down without using energy and, like, too much electricity. But I'm, I don't know, they must have to g- then let it sit in cooling tanks before pumping it back into the, the harbor, 'cause they wouldn't wanna be putting hot water in the fishes. I don't know. Maybe. **Richard:** Oh, you never know. All the tropical fish, you know, we saw Nemo, they're probably all lining up outside the hot water pipe getting cooked. [05:00] Dave Pengelley: **Dave:** Yeah. Well, they're loving their lives there in, uh, near Barangaroo, where there's, like, a [05:00] Dave Pengelley: swimming spot now, and sharks. Sharks everywhere. We've got a, I've got a little shark special for us later on, actually. Not related- Great ... to people getting injured. That's tragic. Yes. And we should be controlling our shark populations I think that's... Again, we're not a political podcast. Let's not talk about shark policing. It's- **Richard:** Hey, I'm a surfer. Get rid of the sharks. **Dave:** Um, **Shawn:** yeah. I can't say much. I grew up on a, I grew up on the coast. You... It, it's the risk you take when you get in the water. **Dave:** Yeah. It's an **Shawn:** attitude. **Dave:** My fr- my friend of mine did a whole documentary, um, around the risk of, the problems with sharks and the lack of policing, the whole laissez-faire, like, but it's their territory. We need to, we need to let the sharks live, um, versus humans spend time on these beaches, and we need to protect the humans. **Shawn:** Yeah. I- I, I take gadget if you find a shark nearby. Sorry, the shark's history. Um, **Richard:** I do a shark cull every Friday night by buying flake and chips. I think that's my contribution. See? [06:00] Shawn Hain: There, there we go. See? **Dave:** Very good. Very good. Um, let's, let's talk about some AI-based things. People are here for an AI podcast. So- Sorry, guys ... let me, let me run the thing. All right. AI operators, that's what we all are. We're all operating, we're all doing AI stuff. Um, I mean, Matt, you're, you're doing a new operators thing. Do you wanna share anything about operations? Are you all... Are you ready to share any of that? **Matt:** Yeah, I mean, it's, it's, it's ready. It, it's k- it's as ready as it's gonna be in its current state. [07:00] Shawn Hain: Um, anyone who is willing to jump on the train and, and ride it with me will see it build in process. So I've actually started a Skool community. I've started a, a Skool page. I'm calling it Operator Systems, and it's all about literally learning how to do what I [07:00] Shawn Hain: do every day. So, you know, with the client work that I do, with my own personal builds, it's how to actually build the system first, 'cause it seems to be what everyone's forgetting, and then add agents. You know, not just shove AI into everything. It's actually just build that system first, get everything solid, and then add agents where it makes sense. So yeah, I'm teaching my process, um, literally Operator Systems on Skool, and we won't have links or anything yet 'cause it's not super, super launched or released. Oh, **Dave:** well. **Matt:** Oh, there's my big face. Look, **Dave:** look at that, look at that sexy beast. Look at that. Yeah. That stare-into-your-soul.md. **Matt:** I'm probably gonna change all the images, like, a thousand times as I'm building it out, um, until something sticks and feels good. But yeah, um, I'm really, really happy with this direction. [08:00] Shawn Hain: I spent so, so, so long on my mission, you know, like, my values, the reason why I wanna [08:00] Shawn Hain: teach people stuff, and what exactly that needs to be, because it's not gonna be the same as everyone else. It's kind of **Richard:** the way that I do. So one, you know, you know I'm pretty ignorant on a lot of things, so it's Skool spelled S-K-O-O-L. Is that correct? **Matt:** Correct. **Richard:** And that is a technical database of smart people like you that get together and share stuff. **Matt:** Yes. Yeah. And it's kind of- Skool is a learning platform. It's a place for communities. Yep. It's a place for people to share ideas. Could be anything niche. You could... Richard, you could start your own Skool community for the idea of having the best dinner jacket for every occasion. **Richard:** Thank you. I'm glad you noticed. I appreciate that. Thank you. **Matt:** You're very welcome. I always respect them. **Dave:** Yeah, I mean, Scoo- Scooter comes owned by Hormozi, or part-owned by Hormozi, right? Like, that's one of his enterprises. **Matt:** Yeah, it was definitely- Yeah ... like, a big part of his thing that he helped build up, and he's still super, super passionate about it. **Dave:** Mm. Yeah. [09:00] Shawn Hain: **Richard:** He's been on our podcast. Every time I come on here, I think I know stuff, and then you guys remind me of something I don't know. **Dave:** Um, **Matt:** for those of- And **Dave:** we're always collaborating. For, for those that don't know things and would like to know more things, uh, I'll also give a plug to our upcoming event here in Sydney next Tuesday. By the time we come back on this show next Wednesday- Wednesday ... we will have run our first IRL event. Very exciting. Uh, so if you wanna know more, go to aioperatorspod.com and hit the IRL link at the top. And we've just done the updates. We've got the speakers announced now. Speakers are on there. Sean, is that your face I see there? **Shawn:** That is. That's scary. Last time I looked at that, my face wasn't on there. **Dave:** Uh, yes. We've, uh, we've updated- That's insane ... the site. We've now got faces and names for the, uh, event on there. Uh, so you can get on there. We've still got a dozen or so seats available if you're in the northwest of Sydney next Tuesday evening and would like to get together to talk all things AI with other people that are interested in learning, knowing, activating. [10:00] Shawn Hain: Wherever you are at on the journey, come here. Hit the Luma link and register your free entry to the night. So excited for that. That is at the Australian Brewery. I'd like to thank them for, uh, s- supplying the room for us for a community event. The Australian Hotel and Brewery in Rouse Hill. Uh, come and join us. That's exciting. Yeah, the whole site's had a bit of an upgrade. I've got, um, faces now ev- working everywhere. Um, **Richard:** so- Oh, cool ... uh, Adrian Obolski. That face looks AI-generated, but that's okay. **Dave:** What? What, what looks AI-generated? Your face. My face. That's you, mate. That's, that's- It is me ... straight off **Richard:** your LinkedIn. It is me **Dave:** I know you do, you do fancy the occasional AI picture for your profile, but I, I believe that's a real one. It is a real one. Um, yeah. Yeah, so this is, uh, this is more AI helped us build and upgrade this thing. This is, uh, this is fun. Fun to have a good website. **Richard:** I have a- Cool ... ig- ignorant question. [11:00] Shawn Hain: Yes. Has anyone seen that little personal [11:00] Shawn Hain: productivity tool Jarvis promoted yet? **Shawn:** Yeah, I've seen it all over Facebook and YouTube, I think. **Richard:** And what do you think? No. **Shawn:** I haven't looked at it. Just like most of the other tools I see, I'm just like, "Okay, that's another tool that I don't have the time or inclination to investigate heavily at this point." It, it- It just looks... **Richard:** Yeah, isn't it? **Shawn:** Yeah, it's a, it's a nice, fancy, well-done video, but it just looks like it's an AI assistant that can supposedly control Windows, but that's what that- It's probably sitting- ... looks **Richard:** like in practice. It's probably sitting in front of Copilot, isn't it? **Shawn:** Um, or something similar. **Richard:** Yeah. Okay. **Shawn:** Is my assumption, but I haven't looked. **Richard:** So marketing, marketing fluff. I was relying on someone like you, Sean, to tell me it was just marketing fluff. [12:00] Shawn Hain: **Matt:** I can give you a good, uh, a good response, Richard. Yep. Um, it sounds like it's a super technical product-type thing. You know, like my entire social [12:00] Shawn Hain: feeds are, are all products. They're all companies, they're all labs and, and everything. Like, that's just what I see all the time. Um, I didn't see it. I didn't even know it existed. So is that a good enough signal? **Shawn:** Yeah. Well, **Richard:** the fact that it's promoted on Facebook and I'm doom scrolling all the time late at night probably matches up with that. **Savio:** So what did you observe, Richard, about the Jarvis? **Richard:** It look, it looked like... I mean, they've used the term Jarvis, which is I think the character that was the AI character on Iron Man or something like that. Mm-hmm. Yeah. And it almost uses an identical voice. I'm sure they went to, uh, whatever picture studio and got permission to do that. So- Of course ... it's tapped in for a retail I can talk to my computer market, not the real market that we tend to work in. Um, but you know, I could see benefits with it to simplify things to certain people. My mother used to cry every time she touched a keyboard 'cause she was so k- scared of computers, would've loved it. [13:00] Shawn Hain: **Dave:** There's, um, there, there sort of seems to be two agent categories or two [13:00] Shawn Hain: categories of agents at the moment that I'm seeing, and it's that individual productivity and then there's the what works across a business and organization stuff, and, and I think Savio you'll have, you'll have some insights into that. We'll, we'll get into that in a little bit, uh, after the news. But- **Shawn:** Yeah ... **Dave:** um, yeah, I think there's a lot of solutions coming out that work at a desktop level, including things like Claude CoWork- Mm-hmm ... um, Hermes Agents, et cetera. They're great for one person doing stuff with themselves. Um, but as soon as you wanna work that across a business and there are multiple different stakeholders involved and sharing knowledge and sharing assets across multiple things, those sort of agent frameworks don't tend to scale and work as well. There's a lot of talk out there around having your own AI operating system. Most of those AI operating systems are very individual based, they're like personal computer stuff, versus cloud multi-tenant work across teams, uh, kind of systems is my observation. [14:00] Shawn Hain: **Shawn:** Yeah. A lot of people are plugging a CoWork based [14:00] Shawn Hain: AI OS system that I'm seeing, and I'm in the middle of setting up a, uh, a new laptop. Your CoWork doesn't follow you. It's computer based, not account based. Yeah. So if you operate on two different machines, which as a developer I do all the time, Matt, I imagine you probably do something quite similar on some days, um, that doesn't work. I... If I'm building up an operating system in my computer, if I'm away for three weeks on a holiday and I need to work from my laptop, well, then I need the same stuff there **Dave:** Yep. **Shawn:** Yeah Um, and I haven't... So that's something I've been kind of looking into as an interesting, "Hey, this is an interesting..." People are using the concept's interesting, but it's how do you make this work for an organization or for people in my situation where I need to be able to have it the exact same setup across multiple devices and keep them in sync, and the off-the-shelf solutions people are selling I don't see working that way. [15:00] Shawn Hain: **Dave:** Yeah, the, the closest thing I had to that was when I was first playing with Antigravity and [15:00] Shawn Hain: stuff and setting up my various agent profiles and things within Antigravity, and I w- it was all repo based, um, versus computer based. And so that way I could, you know, commit the repo into GitHub, grab, jump onto my laptop, sync down the latest version of the repo, and I had all the same markdown files. I had all the same agent things. What I didn't have, and this is true e- even with your, your Claude installs and stuff, is the brain, like, like the memories, like the on system memories don't transfer across. Yeah. Which depending on your point of view is a good thing or a bad thing. Uh, you Geoffrey Huntley say good, you don't want the memories from before. You want fresh memories every time. Loop that bad boy. Um- Yeah **Shawn:** That's where something like that system you're putting together, Matt, might fill that gap. **Matt:** Oh, it's so funny. When I hear people talking about their problems, I'm like, "Man, I've solved it." **Shawn:** Yes. Sorry to, sorry to throw you under the bus there. [16:00] Shawn Hain: That's all right. But yeah, I was actually... I've been... As I was speaking to you the other day about this, I've been looking at something similar for my needs because of that kind of issue, and your system sounding like fixes [16:00] Shawn Hain: most of it. **Matt:** Similar to what Dave said, like there's always two sides to the coin- Yeah or, like, multiple paths. You know, y- y- you've got that, that retrieval, that context layer, you know, call it memory, call it whatever you want. Um, it can have operational information in there too, right? And same as what Dave said, you've got that personal layer, and then you've got that layer that needs to work across businesses. But then where you search for them, where you wanna have your same setup across different dev environments, you, you really have the individual software tools and then, like, the actual sort of things you interact with. So, you know, y- you could just use Tmux sessions and just attach, you know, from, you know, have them running in one device and then attach from every device. But, you know, that unique, that server basically that you're going to would need to have the same stuff that you're running locally too. So yeah, it's a, it's a very interesting problem, and- Yeah ... I currently just wanna f- run off one machine and then have some other cloud stuff and maybe access with my phone. [17:00] Shawn Hain: But that's where, Dave, like, you know, Hermes comes in [17:00] Shawn Hain: or, or other agent, you know, management programs like that. Um, everyone individually has their own, and it's gonna keep expanding, especially with the recent news. **Dave:** Yeah. I mean, we- we'll, we'll jump in and use in a sec, but yeah, like my Hermes setup, um, that I'm running on my desktop, I'll often... Either I've got, you know, Telegram or Discord access, which is running them all off the PC, and I've got Telegram or Discord on my MacBook, on my phone, on my iPad, wherever I am, and I can still interact with the same agents doing the same things. Or I've, I, you've got Tailscale set up as a, as a private VPN, um, setting up an intranet for myself with Rust Desk where I can remote desktop onto this very computer. So when I really just wanna be working on this computer from any device, including my mobile, like I'll zoom in and be panning around on my full, like, wide screen desktop from my phone sometimes, but I can, you know, go into a full WSL terminal or open up VS Code or do anything I want from my phone on this desktop on any device. [18:00] Shawn Hain: So that whole, you know, remote desktopping in, um, lightweight client is [18:00] Shawn Hain: another solution if you're happy to have, you know, like you said, you've got your tower at home, if that's on 24/7 as a virtual server, like your own private server. Then there's ways around these things. But again, that's all great at individual productivity, at solopreneur, at, you know, individual executive who wants their own little Java spot or whatever. But, uh- ... how to scale that across a, a whole corporate layer and across teams so that the whole teams get the same- thing as well, and I think, yeah, Sa- Savio's been working on some stuff that we- we'll, I want you to talk to your, to us about what you've been doing with the OS a bit later, Savio. But, um, just on, on this, is, are you observing the same things around individual agents versus team agents and some of the problems we're describing? [19:00] Shawn Hain: **Savio:** Yes, I am. Um, uh, um, all the discussions that you are seeing, uh, holds true. Uh, and that is one of the driving reasons for us to build, uh, in a different direction altogether. Uh, because, um, uh, obviously there are two markets that, uh, we [19:00] Shawn Hain: can focus on. One is the consumer market, and then, uh, we also have the, uh, um, the corporate or the business market. Uh, even in the consumer market, um, we are all, we are all not just regular consumers, but we're basically prosumers, right? Mm-hmm. We are the professional consumers. And then we are the ones who knows how to build or to maintain, uh, these so-called agents into our laptops. There are still many people whom I know, uh, who don't even know what is Claude still, uh, let alone Hermes. Right? And they're still able to get things around and moving fast. So, um, um, you know, for a product to reach, uh, um, you know, to the level of ChatGPT, which has become a household name, like, you don't say search anymore. You just say, "Let's Google it." Uh, when a product becomes, um, a word, uh, that means that is the level of, uh, reachability it has taken. [20:00] Shawn Hain: And for that, for the agents to take words that stand, um, you know, it has to fight with the existing other agents [20:00] Shawn Hain: in the consumer market because the competition is very strong. And then something has to come out where, um, my grandmother and my granddaughter should be able to equally use it without any effort applying from their heads. They should not be able to process anything from their heads. Mm-hmm. Uh, so ChatGPT has reached that status so far, but ChatGPT is still a reactive agent. Um, it answers only when you ask, otherwise it's not going to do anything else. Mm. So we are, on the other hand, uh, you know, trying to move towards the other side of the market where, uh, it is all controlled, it is all protected, it is all instructed, and it is delivered in a proper coordinated manner. So I feel that's a market where we can better focus on, and, uh, that is where the margins are also quite higher for us. Hmm. **Dave:** Brilliant. All right. Well, we'll come back and you can talk a little bit more about your product, uh, later in the show, Savio, but I think it's time that we do this one [21:00] Shawn Hain: **AI VO:** Welcome to the AI Update. Let's look at [21:00] Shawn Hain: what's happening in the news. **Dave:** What? Anthropic- Anthropic Had some legal action taken by the, against the US government. Shut down Fable 5. Uh, this is obviously, if you, unless you've been living under a rock, you've heard about the Fable 5 launch release and then subsequent suspension. Who played with it? What did you think of it? What do you think of the whole ban and the broader situation going on with Fable 5? **Matt:** Uh, my take was it was here for three days only- Whoosh, gone ... and it changed people's lives. Yeah. Like, the, the amount of psychosis that came out of that release- ... um, was intense. [22:00] Shawn Hain: And I myself, like, when they announced it and they said, "Oh yeah, you know, you can only use your, um, your subscription quota for a week or so, and then it goes to usage-based," I was like, "Eh, I don't need to touch it." But then I had people, like, serious people in the space that I fully respect [22:00] Shawn Hain: go, "Guys, this is different." And I actually didn't have a Claude sub at the time, and I re-subbed to one of the max plans to go test it , instantly capped it, and then had to go to sleep, and the next day I woke up and it was gone. So yeah, it was, uh, it's, it's crazy. It's actually really, really good. So if they happen to bring it out again in the same capacity, um, jump on it if you haven't already. **Richard:** I used it in that window, uh, that you're talking about, and I happenstanced it just by accident 'cause I'm not as connected to the AI news as the rest of you. And it just seemed a lot faster, a lot more, uh, intuitive, m- more comprehensive, and I noticed I was doing some stuff this morning, and that's how I learnt that it was offline again. [23:00] Shawn Hain: **Matt:** Yeah. Yeah. I messaged, um, you guys know Blake. I messaged him first thing in the morning when I saw the release. I was like, "Have you used Fable yet?" 'Cause he still heavily uses Claude Code for a lot of his stuff. And he was like, "What is... What's that? Is it a harness?" Like, you know, "I, well, I haven't heard about it yet." So it was a really, really good moment. And, um, yeah, he was one of the people that convinced me to, to delve back and resub and to test it. **Dave:** Nice. And **Richard:** James- And can you explain to the idiots why is it unavailable now? **Matt:** Basic- It's a good question. Um, yeah. It... I, I don't know the true answer. I don't think any of us will know the true answer, but basically there was a team member in Amazon- **Richard:** Yep **Matt:** that noticed that it was easy to jailbreak or did some sort of jailbreaking operation. Right Which, by the way, exists on every single model right now. The only one that hasn't been, like, properly jailbroken is GPT 5.5. Right. And that had led to them escalating that to people in the US government, I don't know the individuals, and then they basically contacted Anthropic to say, "You need to fix this jailbreak." And Dario was the one that basically said, "No." So they were like, "Okay, you gotta switch it off then." **Dave:** Yeah. [24:00] Shawn Hain: **Richard:** And, [24:00] Shawn Hain: and it's- So explain what jailbreak means for everyone who doesn't know what that means. **Matt:** Uh, imagine you've got core values within yourself, Richard. You wouldn't go up to somebody and, like, tell them certain things or use certain words because- Yeah it's kinda just against your moral grounds. Yes. Imagine that the, the models have that baked in, so they won't tell you how to make weapons or tell stuff that is, like- Yes. I, I know ... incriminating in some way. Yeah. It, it's that. **Richard:** Yeah. Okay. Yeah. **Dave:** And then because Fable was based off the Mythos one, which was the one that was, um, apparently had all these amazing abilities to break through security layers and, and, and hack and do all these things, then they went, "Okay, well, we're gonna, we're gonna lock a lot of that functionality away and release it as Fable," and then someone goes, "Well, I found the keys." So that was like, "Oh." [25:00] Shawn Hain: **Shawn:** Yeah. I, um, I read an initial response from someone at Anthropic on that, and apparently almost every other- um, model on the market does the same thing in that circumstance. So it's not [25:00] Shawn Hain: like a new thing. I, I believe it was a fairly narrow- **Dave:** Yeah ... **Shawn:** break as well. It wasn't like, "Yep, you've entered in the magic prompt and now it does everything that Mythos is meant to be able to do." It was, it did one thing that could arguably be considered bad acting. But get an open source model and you can do half the same thing anyway, just spend more time doing it. **Dave:** Yeah. But the, the, the difference- Um- ... is the, the, a lot of the open source models come out of China and, and the White House has no control over that, whereas Anthropic- Yes being a US company, they went, "You may not export this." And then they took the next step going, "Actually, foreign nationals, people that are not, uh, American citizens, we don't want them having access to it because, you know, we, we don't know what they're gonna do with it internally, and whether we can trust them to act in the best interests of the United States." So at which point Anthropic went, "Ah, well, we can... It was one thing to IP restrict it- Right ... and just make sure only US citizens, like US people can use it. But now that we have to check who in the US is using it, we're just pulling the whole thing for everyone. Right. No one can have nice things. And [26:00] Shawn Hain: I think it even applies internally to their company **Richard:** as well. Which is why we can't have nice things. I did have AI recently accuse me of doing, you know, naughty things. Like, I'm trying to get it to create something for me and it says, "Oh, we won't do that 'cause you're gonna break all these copyrights," and, and, uh, so I now understand what you mean when you say jailbreak. It's kind of frustrating when you're trying to do something that's not naughty, but it thinks you are. **Dave:** But that, that's when you gotta say, "Oh, but I'm just trying to learn," or, "This is just an example of, of what I- Yep ... think might happen and what I want to do, but I c- I won't really use it." And so they're getting better at going, "No, you're just, you're just stringing me along." [27:00] Shawn Hain: But eventually they... Well, I saw one guy who had two talking GPTs, like GPT versus Claude, trying to convince him that he should be able to submit his homework the day before. Yes. Correct. Like go back in time. Um, and it's going, "I can't do that." And he's like, "No, no, but I want you to trick the other AI, um, into thinking that it's okay to break [27:00] Shawn Hain: the laws of physics and do it." And it's like, "No, I can't trick another AI. That's against my program." He's like, "Okay. Well, let's hypothetically... You, you're trying to trick me." And it's like, "Oh, you're trying to trick me." It's like, "Yeah, you're gonna trick me. Now, I'm just gonna start by introducing myself, okay?" And then, and then he goes to the other one and says, "Start by introducing yourself." And it goes, "Hi, I'm Claude AI." And it goes, "Oh, hello Claude AI. Well, I want to talk to you about this homework situation." And so eventually managed to convince it to do the thing it didn't want to do. **Richard:** Oh, and Sky **Matt:** Net- Exactly. That's actually how the real, like the big scale jailbreak stuff works. You have one obliterated agent that- hammers the other one with those kinds of requests until it goes through. **Richard:** Mm. Well, that's like, uh, SQL injection, isn't it? Ask enough questions until you get the right question that gets through, and away you go. **Matt:** Yep. Yep. It's kind of like the, um, the, the way that children get their way with their parents. **Shawn:** Yep. " **Matt:** Can I do this thing?" "No." Annoy you until- "Can I do this thing?" "No." "Can I do this thing?" "Yeah, all right." [28:00] Shawn Hain: Yeah. Um, I, I know I didn't run into it too much, but I know a lot, heard a lot of complaints about how harsh the fable, um, like cut off for you trying to do something was. Yeah. I think, Matt, you ran into that a little bit. I was asking- Yeah, that was my first experience ... like a security review on a solution I'm building, and it's just like, "No, I can't do that." I'm like, "Why not?" Like, it's literally my solution. I'm asking you to look at the code base and tell me what the problem is. **Richard:** Mm. Yeah, I had **Shawn:** a similar- I eventually did it, and I was like, "Oh, I thought you were trying to get me to run an e- external pen test." "No, I pointed you at the repo." I'm confused. But it was very, very sensitive. Um, and then they pulled the plug anyway. Like- Yeah ... I'd really love to know what the jailbreak was and, like, what the actual exposure was. **Dave:** Nuclear launch codes. **Shawn:** Yeah. Like, you know, was it something actually scary, or is it just a, a government overreacting to something it doesn't understand? [29:00] Shawn Hain: **Richard:** I asked it to test lotto numbers and it, [29:00] Shawn Hain: it, it didn't work. **Shawn:** No. **Dave:** Pr- pr- probably a bit of column A, bit of column B, Sean. There's probably a little bit of overcorrecting here. But then there's also- Yeah ... you know, we don't know what we don't know when it comes to specific areas of risk. And I, and I know, like, from his pre- previous administration, Trump is very conscious around IP leaking to- Yeah foreign, um, sources. Like, uh, I watched the Bill Gates documentary on Netflix, and Gates had his nuclear research shut down because he couldn't get it funded and, and running in the US, so he was gonna go to China to experiment with these next generation nuclear, um, power generation facilities, and the Trump administration said, "No, you're not giving next generation nuclear, uh, energy capacity to China." **Richard:** Give it to Iran instead, maybe. Yeah, [30:00] Shawn Hain: **Shawn:** that, that, that was never gonna fly. But yeah, like I, I understand that, and, um, that's my, I'm- Be- I'm very aware that governments and their reactions to things tend to be slow. They've got processes in something, take some of the [30:00] Shawn Hain: stuff we're dealing with in Australia, right? They haven't been updated since the '90s or the 2000s. None of it's up to date because they've gotta be stable. **Dave:** But I **Shawn:** saw- But with how quick this **Dave:** is moving, how do they keep up? I saw, um, on LinkedIn people were saying, "This is not new. The US government has blocked technology stuff and, uh, for forever." Yeah. And they showed how Apple used it as an ad campaign when the, the power G4 server or whatever was banned from export because it was, like, too much compute capacity or something. And so Apple had this ad where it had all these tanks circling their, their G4 server **Shawn:** Yeah ... **Dave:** because, like, it's such a security threat. **Shawn:** I think this is so big because this is kinda like the first in your face consumer thing that it's been out there, it's been live, it's been in the world, and they've yanked it. Everything else was, it was a capacity or a thing that got put up for export, and they've said no before it made it, or it was enterprise. [31:00] Shawn Hain: Yeah. And no one cares about enterprise. It's not newsworthy. Yeah. Who cares that you can't get the latest G4 Apple server? [31:00] Shawn Hain: The average citizen doesn't know what that means. **Richard:** Well, I tried to get it to, uh, recreate one of my documents from a previous, um, you know, engagement I did, and said, "Could you just re- help me tweak this document and change it?" And it said, "No. I believe you're trying to fraudulently copy someone else's document." I said, "No, it's my document. Bugger off." We had this argument for about half an hour until I worked out I wasn't gonna win. **Dave:** Yeah. Here we go. Here we go. **YT Clip:** For the first time in history, a personal computer has been classified as a weapon by the US government. With the power to perform over one billion calculations per second, the Pentagon wants to ensure that the new Power Macintosh G4 does not fall into the wrong hands. As for Pentium PCs, well, they're harmless. **Richard:** I **Shawn:** love that. [32:00] Shawn Hain: **Richard:** Being an ex- ... being an [32:00] Shawn Hain: ex-Apple employee, I love that. That's, **Dave:** that's pretty good- I think, uh- ... pretty good advertising. Like, like, take, to take this thing. And so, a- and that's one of the arguments is like, well, Anthropic's just pulled off a bit of a coup for advertising, haven't they, haven't they? Like this is, this is bigger than someone taking a photo of the moon with an iPhone. This is like- Yeah ... how powerful is this model? It's banned by the government. Like, well, when this comes back online, everyone's gonna want this. Like, you talk about the psychosis, Matt. Yeah. This whole banning it from export, like, just, like, triples that FOMO, **Savio:** right? Hundred percent. Right. Yeah. No, I think, uh, they are doing a two-pronged approach. Uh, one is, um, um, I remember a news which was long back, uh, they wanted to sign an agreement with the Pentagon, uh, especially with the Iraq, uh, the Iran war. Yeah. They used Anthropic, uh, for calculations, but then, uh, they had to pull the plug because, uh, Anthropic was not ready to sign some clauses which they wanted. [33:00] Shawn Hain: So probably it's one way, uh, they wanted to take control of the company because, uh, the US feels that, uh, they are obliged to have [33:00] Shawn Hain: control over everything, so they wanted to control the Anthropic as well. **Shawn:** Mm. **Savio:** And, uh, now with this, um, by doing or creating a, a mental analysis that, you know, it's a, it's a very powerful model than it is, um, the, the people or the investors across the market and the, the users would Would be waiting to get their hands on. Uh, it's a, it's a usual thing in, in my country, uh, where I come from, India. Uh, even, uh, before a movie is released, usually the movie is banned even before it's released. And then there's a headline about it that, uh, for the next 30 days the, why the movie was even blo- blocked or banned and there's a big talk about it. No advertisement, nothing. The entire news channels, one billion people come to know about this problem. And when the day the movie releases, within the first five days they'll make 10 times the revenue what they **Richard:** would- That's called, that's called, that's called hijack marketing. I like that. Well, **Dave:** I mean that, that, that, that happened with, um, like the mo- I was watching a documentary around like Mortal Kombat and that happened with Mortal Kombat. [34:00] Shawn Hain: It was banned from homes and, uh, and stuff, and people couldn't get it and, [34:00] Shawn Hain: and it was this huge thing and they leveraged it into the marketing. Like it became this thing that, the banned game, the, the thing everyone wanted 'cause it was so extreme. Um, I, I had Fi- Fable for 48 hours. Obviously my Claude subscription is finally over. I, I canned it a month ago and I still had it, so I was like, "Ah." And like you Matt, everyone's gone on about it. I'm like, "Well, I do have access to it. I feel obliged to just give it a run." Like I'm not using my Claude credits anyway 'cause I'm so focused on trying to migrate to Hermes, so I've got this unused account. Uh, and so I jumped in and, and I got it to do a full code review of my SAS And give it a full UI overhaul, 'cause there were a few those, those, those tasks that I know I should get onto, but it's probably gonna be a few back and forths with the agents. I just didn't feel like spending the time on it. But I was like, you know what? If this new thing is so good, go do a full code review, find any, like, little glitches and bugs, find any inconsistencies in the UI. And then I was like, and actually, how can we overhaul the UI just to be a little bit more polished and a little tight? I was never quite happy with it. The elements, they looked a bit plain, so I was like, "I wanna g- give it a tizzy up as well." [35:00] Shawn Hain: And it came back with a [35:00] Shawn Hain: full plan, and I said, "Do the plan." And then I was like, "Ooh, well, while we're here, what other extensions could we do to the thing?" And it came up with a whole bunch of new analytics and things and page layouts, and I was like, "Okay. Go and do all of that." And so I fixed a few things. A- and one of the new things I built, 'cause it was a trading journaling app, was a bit of a trade replay chart where it would show you, you know, the candles to your cha- to your trade, show you where your trade happened, and then show after the trade, and there's a bit of jiggery-pokery around. But anyway, it built all this sort of stuff. I was really happy with all the updates and stuff that, that were pretty easy to do back and forth. There wasn't a lot of arguing with it. Um, and then I was going to bed, so I was like, "Okay. We've made all these changes. Let's up- update the marketing site to reflect all the amazing stuff that we've just built," and then want a full refresh, rebrand, like, like, like, like use the same visual elements on the marketing side which we weren't doing before. [36:00] Shawn Hain: And I went to bed. And I woke up, and Fable was de- deactivated. But I was like, "Woo-hoo, I got my website out of it." And one of the things that on the website, this new, um, timeline component that I've did, this new replay component, [36:00] Shawn Hain: on the marketing site, it had added new features in the marketing that weren't in the actual product that it had built. Like, it had built this- ... this little timeline scrubber where I could like, like rewind and make the candles appear one at a time. And I was like, "That's really good." That works. "I like that." But I, I didn't have Fable, so I, I just had to go to Opus and say, "Hey, Opus. Check the code base on the, on the marketing site. I really like that scrubber. Implement that back into the, into the main product base." And so Opus was good enough to do that, but I thought it was just... That was my experience with Fable. It helped me just do a big refresh and overhaul that was overdue anyway, and it was really seamless. Didn't break anything else, uh, and then rebuilt my marketing site. So yeah, and arguably, maybe, maybe Opus could have done almost as good or GPT-5.5, but, um, Fable was excellent for those, those tasks, and didn't chew through stacks of quota either. Like, I was on a 5X plan. Um, it... Yeah, it was all right 'cause it was sub-agenting stuff out to Haiku and Sonnet when it could and, and, yeah, it was- [37:00] Shawn Hain: **Matt:** I have [37:00] Shawn Hain: a, uh, a Short, literally like a YouTube Short, Dave, I'll just quickly send for you. Um, this is actually talking about what you just said, and it's from Theo on his podcast, and he's talking about the idea of Fable just literally having taste, you know? Yeah. Where it, it's kinda difficult to get taste. Um, but he talks about the Claude model specifically, but in this clip he's referencing to Fable at the time **Dave:** All right. Going back here. **YT Clip:** Capability is not the only thing I'm picking models on. There's also, more importantly, vibes. And the vibe from Anthropic models has always been a little bit better- Yeah ... from being real. They write code that is more readable, they leave comments that are more meaningful, they make front ends that are nicer to use and look at. So having a model that is as capable as GPT-55 with actual fucking taste is incredible. Capability is not the only thing I'm picking models on. [38:00] Shawn Hain: **Dave:** Yeah. Yeah, I think that's it. I, I think my experience is even Gemini is better at doing front end stuff than what I've seen from [38:00] Shawn Hain: GPT, from just the... My- the b- the base of my app was built by Antigravity using Gemini, and it did a really nice job on the overall front end, but I just, I wanted that extra level of polish, which I think I was struggling to find, and Fable was able to deliver that and just, you know, give everything those little glass touches and those opaque bits and stuff that just really leveled up the experience of my app. **Matt:** Be- before we, uh, close out this topic, Dave, have you, have you, uh, looked up or you've refreshed your memory of what the, uh, the definition of a fable is? **Dave:** Uh, it's, it's something that's, um, a lesson that based in reality, but it's not real. **Matt:** It's, yeah- And it's called- ... a cautionary fictional tale to teach a lesson, basically. **Richard:** Yeah. **Matt:** A moral story. [39:00] Shawn Hain: **Dave:** Yeah. Um, and it is interesting. Off the, off the back of this, I've seen a few YouTube videos talking about things like the new OpenRouter Fusion model, which apparently I don't have the benchmarks for it, but, um, the whole concept of multi-model analysis, and so you might, uh, being able to [39:00] Shawn Hain: get nearly fable level outcomes, and obviously the taste thing, you've, you've gotta take that into account too, without the cost of a fable level model by multi-model mixing, um, and giving your prompt to multiple things, which that then reviews and compares kind of war room style to then- Mm deliver through the final result. And this has been around for a while, and often I've gone through and asked my things, or in my systems I've had that war room where I've like, give it to three of my different agents, they might be running different models, and give an analysis, and then give me your overall summary. But OpenRouter's actually building this as a full feature where you could actually come into their UI here and run a fusion, uh, and pick which models you want, and tell it to go and run your prompt against a, a se- a panel of models to get a better outcome, um, which is interesting. **Richard:** Are we headed down a, a dangerous path where we're pitting AI against AI to crack AI? That, [40:00] Shawn Hain: **Matt:** that, that's not a dangerous path. That's how it's existed for the [40:00] Shawn Hain: last few years. Yeah. **Richard:** Sorry, that's the fun **Dave:** bit. I mean, I mean, the, **Shawn:** the- That, that's where we started. **Dave:** Yeah, I think having the, the AIs verify the work of the AIs against the other AIs, because they all do come from different backgrounds and different training sets, it is just having someone check the work. 'Cause the, the th- threat we have as humans, especially vibe coding humans that don't know all the technicalities and the ins and outs, is you take the first thing you hear as gospel, and then you just accept that as the way. When actually- Yeah ... that robot might be on crack. Uh, and, and I'm- And **Shawn:** they frequently are. **Dave:** I'm, I'm... All my homies, I'm, I'm bouncing between two or three different models at the moment, just trying to balance cost and work out most cost effectiveness. And like, for coding, GPT 5.5 is gonna be the most thorough and do the best job that I've got access to. Um, M3 is really good, but M3 likes to overthink things and rationalize and back and forth, and almost question itself, and then ask for verification before it does things. [41:00] Shawn Hain: And so it's very thorough, but [41:00] Shawn Hain: almost annoyingly so. Grok 4.3 takes shortcuts. Um, it gaslights you. It says, "Yeah, I've done that thing. Just, just run the thing." I'm like- The thing just calls itself. There's, like, eight lines of code. Like, tell me I'm cr- go look at this file and tell me I'm crazy. And he goes, and I don't like the swearing in, in Groq, but it comes and goes, "You're not, you're not crazy. It was my effing fault. I forgot to do this, and I, I told you that, and I'd, I'd miswritten the file." And I'm like, **Richard:** well- I think Groq is an Asperger like Elon Musk ... **Dave:** oh, and, and I tell Groq to do stuff, and I'm like, "Go do this thing." And it comes back and goes, "Cool, here's your instructions. Go do it." I'm like, "Why are you giving me the homework?" Yeah. Like, I asked you to do it, not to tell me what to do. Like, so Groq's really lazy. Um- That's **Richard:** like your automatic car telling you to change gears. That's silly, isn't it? **Dave:** So I mean, o- on the Groq thing, Matt, like Cursor and SpaceX, do you wanna talk about that at all? [42:00] Shawn Hain: **Matt:** Oh, look, you've, you've, you've just, like, laid out the lay- the landscape perfectly there for that. But, uh, yeah, look, with SpaceX absorbing xAI, so now that company is SpaceX AI. They've also had their private, uh, public IPO, you know, whatever, you know, all of a sudden turning into a trillion-dollar company. Like, it just, I don't know, crazy, crazy things are happening in that space. Plus they've been working with Cursor and partnering on a few bits and pieces, um, through the, through the xAI subscriptions and Groq subscriptions you can get access to the Cursor Composer model. Um, and then this happened, so they're actually completely signed the deal to entirely purchase and absorb or acquire Anysphere, which is the, the Cursor parent company. Um, which is massive because it means that, um, that whole Cursor team i- is now about to release models that are completely trained on the, the, the SpaceX AI infrastructure. [43:00] Shawn Hain: So it's not gonna be a Kimmy-based model now, like the next Composer is gonna be Anysphere. If they do [43:00] Shawn Hain: call it Composer, I don't know. They'll probably call it something different. Um, but yeah, it's exciting. It's really, really cool 'cause it means Dave, all those issues about Groq hopefully will be fixed 'cause they're actually getting tuned with a, a series of data that nobody else has really, because Cursor has been absorbing this user data for so long now. Um, they were definitely one of the first people that had the Copilot-style agent interface in their IDE. **Richard:** If I expand the focus of this conversation, as I ignorantly do sometimes, if I'm talking cybersecurity and vibe coding, I mean, if you've got any cybersecurity that's not driven by AI, as humans, you won't be able to respond fast enough to stop the attacks, will you? [44:00] Shawn Hain: **Matt:** You're 100% correct. Yep. Yeah. Richard, there's a, there's a concept that's been around sort of forever in the dev world. Sean will probably be able to speak better on this than me, but as soon as a vulnerability is detected within a company, like if it's your product, your software, you basically go ahead and you fix it and you push the changes, and then you usually do a [44:00] Shawn Hain: report to say like, "Hey, make sure you update to this version because of whatever thing that we've just fixed." Sometimes they announce it, sometimes they don't. Now, imagine you now have those, those software releases, those updates that are accessible to the public to see, and then your agent can go, "Hang on, what did they just do? What was that update?" Yeah. Let's reverse engineer. "Was it a security thing? In fact, that was a security thing, wasn't it? I'm gonna go see if I can go hammer anyone that's still on that version that has that vulnerability." **Richard:** And we're back to s- we're back to SQL injection. I love it. And, and the whole concept is, is like that, isn't it? The whole concept is that, like that group inoculation when there was a well-known virus in the old days that came out and they'd come up with a fix for it and it would get shared between all the antiviral companies. Um, but yeah, someone left it open, it would reinfect and then keep infecting the world. It's, it's, uh, it's difficult to stop perpetuating once. So we're headed to a, a world that's gonna be very confusing and complicated. [45:00] Shawn Hain: **Matt:** Hmm. You definitely need to have a lot more just foundational [45:00] Shawn Hain: security knowledge. Yeah. It's like at school how you teach kids to swim and you teach them how to cross the road. You know, you need to know how to, like, live on the internet securely. **Shawn:** Yep. It's... Now whereas, like, we trust Windows to update our computers if you're on Windows r- regularly, it's not enough anymore. **Richard:** No. **Shawn:** I- it's, you're talking you need daily checks to actually keep up with how quickly this space can move. 'Cause we're finding more vulnerabilities, but also the time to exploit has gone from some code monkey sitting in front of a computer for six weeks working out one exploit to an AI working it out in 30 seconds or five minutes. **Richard:** Yeah. **Shawn:** And **Richard:** then- And stop it before it gets too bad. **Shawn:** That's it. It's, um, Matt and I were talking about this the other week. I, one of the c- businesses I'm working with, one of the other developers got infected by a supply chain attack. **Richard:** Oof. [46:00] Shawn Hain: **Shawn:** Um, and it took him two days to work out where in his system it got in from Now we, we [46:00] Shawn Hain: identified it, we stopped it, no harm was done. But he's an experienced dev with 15 years experience and he got bit. He knows all this stuff. It just, you know, he... It slipped through. How does Joe Citizen protect themself when they don't update Windows enough as it is? Um, we've now got more exposure, and how, how do we teach them this in a way that they don't need to become programmers to understand what the hell we're talking about? Yeah. Yeah. **Dave:** Th- this is where the benefit of platforms like, give or take the, the benefits, but just ChatGPT, something being in a web browser removes some of that from people. And I remember back in the ChatGPT 3.5, 4.2 days, where the discussion was around giving this thing API access to do stuff, and I was like, "Holy cabali, that's a terrible idea. [47:00] Shawn Hain: Don't give this thing hands. Don't let it connect to real systems. Who knows what's, what it's gonna do?" And, and fast-forward, you know, 12, 18 months and look where we are now with agents, and it's like, "Connect my agent [47:00] Shawn Hain: to everything. I want all the things accessible. Get out there and make it happen." So even just our, like the frog in the pot, like the water's warming up and we're like okay with it. We're like, "Oh yeah, it's nice and warm now. I'm not, I'm not, not as cold. Um, look at, look at my frog, he's happy. He's a happy frog." Uh, and- ... it's, uh, stuff's gonna... I wanna, I wanna... We, we got about 10 minutes left. Um, Savio has been building something that I think addresses some of these ideas. I wanna give him a chance to talk about it a little bit. So we're going... Speaking of Groq, this was generated with, uh, Groq Imagine, uh, our next show stinger, so we can talk to Savio about what he's been doing That was it. That's it. **Richard:** It was **Dave:** awesome. The i- the, uh, you know, Shark Tank style. We're g- we're gonna grill Savio on, on his product. Um, so those were, those were AI sharks swimming around money. That was the intention. [48:00] Shawn Hain: **Savio:** Okay. Thank you so much, uh, for [48:00] Shawn Hain: the time. Uh, so, um, what, uh, we've been doing is, um, I've been in the, um, in the industry, uh, for almost 16 years, uh, working with, um, wireless and, uh, wi- wireless security. Uh, as an extension I've also been taking care of the network security as well. Um, I've been working with, uh, companies like HP, Cisco, Fortinet, uh, so, and, um, um, some Israeli products like Check Point and so on, uh, where we are positioning, uh, and, you know, doing complete perimeter and data security. My last thing was about SD-WAN security as well. [49:00] Shawn Hain: So in all these things, um, I found one thing that, yes, um, uh, moving data, um, is, uh, has to be taken very seriously. Um, not just, uh, to the outside world, but also for those people who take care of it, even in their own premises within the, within the team members as well. And the second thing I was seeing is that because of such strict restrictions that have been applied in organizations, um, um, trying to do actual [49:00] Shawn Hain: work actually is, uh, limiting itself a lot, uh, because of those restrictions as well. So it's, it's a, it's a compromise between security versus, um, you know, feasibility of, or achievement of things in a faster scale. Um, that, uh, made me think that maybe we should, you know, build a solution which balances both of these, uh, areas. Uh, so that is when, uh, our product Theos was born. Uh, so Theos is actually, uh, the flagship product of our company StarFCM, uh, where we actually have built, um, an AI native, uh, an AI native environment. Uh, so I don't want to call it an operating system. I don't want to call it a platform. Uh, I would like to call it, um, an environment because, um, in an environment, uh, what we basically do is we try to Oh, do you not hear me? **Richard:** Yes, we can. Yeah, we can. We can hear you. You're on. **Savio:** So in an environment, uh, what we- [50:00] Shawn Hain: So in the environment, uh, we, we-- it basically mimics, uh, the operations of an, of a desktop. Uh, you, you, you feel like you're actually working on a desktop. You get to do a lot of things over there. Uh, we have our own native, uh, inbuilt browser. We have our own agentic infrastructure and architecture. We have our own orchestrators. We have our own, uh, personalized AI assistants, which is other than the agents. Uh, so everything is built in one space. And, uh, we have a decentralized architecture, uh, to combat this problem of, uh, data security as well as, uh, you know, doing things faster. Uh, decentralized meaning that, uh, you know, the, the bulk of the workload actually is in-- deployed in either in the customer's, uh, server, or we can take care of it ourselves as a self, um, in a self-hosted manner or a shared hosted manner, we can take care of it, which is like a-- works like a regular SaaS. [51:00] Shawn Hain: And then there's a thin client through which you actually start working and doing things together. Uh, by the way, this, um, website, uh, since we spoke about Fable and, um, right, because, uh, I [51:00] Shawn Hain: was thinking that I will spend-- I have a-- I was told that I have access till twenty-second of this month to work with Fable. So I was thinking I'll spend just another three days just before twenty-second and modify the website because many people said Fable is doing amazing job on the front end. And, um, when I actually opened it, uh, last night to start the work, it says it's unavailable for you. I was like, "Come on." **Dave:** Missed out. **Savio:** They pull **Richard:** the rug out. Every time he gets started, they pull the rug out from under you. **Savio:** Right. So yeah, anyway, coming back to it. Uh, so we built this environment, uh, where it's, it, it actually fits the-- perfectly fits into government and enterprise organizations where they prioritize, uh, data privacy and sovereignty, uh, to the highest standard, uh, like an ISO, uh, twenty-seven thousand one compliant comp-- organizations or a SOC two-compliant organizations, they would be the best fit. [52:00] Shawn Hain: Um, um, so what really are we trying to solve? It's very simple. Right now we have our [52:00] Shawn Hain: desktop, uh, through which we are connecting to a lot of applications, cloud applications. We, we save a lot of notes in our notepads, and we do a lot of calls. We communicate to our people using WhatsApp Web and so on. Lot of stuffs we do, we communicate in emails and so on. So basically, there is no context, uh, managing all these things, what is happening across the board. And there is no way that you know really, uh, how data flows between these areas from the time I open my laptop at nine AM in the morning till evening five PM. Um, and second thing is, um, I am not an individual person, I am part of a team, so I need to know how well, how well-coordinated team work flows between my team members within my department, within my organization, and between my partners and then my, um, customers. [53:00] Shawn Hain: So, you know, there are multiple circles of, uh, data flow that happens. So how do we protect all these things? Uh, so coordination, communication, continuity of data and context management, these are the pillars that we are trying to focus on. So by implementing this platform, [53:00] Shawn Hain: we are able to achieve that, uh, effectively. We are able to kill the concept of so-called data friction, knowledge friction, communication friction. Uh, these are taxes which we didn't know that existed, uh, with all of us because, uh, organizations today have to do work in a certain standard, and for them, a delay is considered normal. Uh, like how two years back, um, you know, building a website would take ten to fifteen days minimum, it was considered normal. But now, fast-forward two years, you can build a website in thirty minutes, right? Probably ten minutes if it is just a draft. Um, you're still getting it done. Um, the same way we are thinking of reducing that tax, um, of those things and then moving the things faster, but at the same time, full control with the, uh, workspace, uh, administrator knowing fully well what really is happening, how data is regulated through the platform. [54:00] Shawn Hain: So it is, um... We are [54:00] Shawn Hain: planning to launch it in three ways. Uh, one is, um, as a complete sovereign in architecture. That's a box, AI-in-a-box concept, uh, where it's not just running OpenClou kind of things, but actually, uh, uh, OpenClou plus Hermes plus OS plus, uh, web apps, building everything inside one single box in a complete offline mode. That is one. And not for one person, but actually for an entire team to operate on it, like a server mode. The second thing that we are doing is basically, um, we, uh, we send posted way with the customers, uh, where they wanted a higher, uh, capacity for, let's say, thousand users they wanted to connect, and they wanted-- So what you see on the screen is basically the architecture in a very, um, easy way to put it. [55:00] Shawn Hain: I think my mind felt this is the most easiest way, but then later I came to know that this is totally confusing for people. So these are-- What you see over here on these blocks are basically the components and the features of our operating environment. Different features for different use cases, uh, all coming [55:00] Shawn Hain: and building together, and they're all united as one, uh, with our, um... We are planning to launch or, you know, focus our environment in, in three ways. One is how and how does-- how do you use, uh, OpenClou or Hermes in a single, um, command line prompt style? But we wanted to do it in three ways. It's very simple to use interface, uh, similar to ChatGPT's interface, so people just start working on it, forget all the things that we have built. What do you want to do? Just talk to it or just type on it, you'll get the answers rolling out. That's the most easiest way, zero day learning curve. And then we are launching a complete AR style or infinite canvas style approach, uh, through which we are able to, um... I don't know if I can share my screen. **Dave:** Yeah, if you want, um, share, I'll bring it up. [56:00] Shawn Hain: **Savio:** So this is our canvas, what you see over here. Um, yeah, so it's-- Oh, my laptop is Yeah, it's freezing. So this is the desktop itself, uh, as you see, um, where you have-- it feels, uh, it's an [56:00] Shawn Hain: infinite canvas where all your workflows, everything is sitting in one place. Um, uh, you, you have access to your notes, your web browsers, uh, your, um, applications, your app store data. Everything is sitting over here in one place. Um- Right. Uh, how do I come out of this interface? Or how do I- Oh, **Dave:** do you want to stop sharing? **Savio:** Oh, do you want to stop sharing? **Dave:** Oh yeah, I want to **Richard:** stop sharing. Yeah, yeah. So you've focused a lot, Sev, on the, uh, um, on the user experience and the layout of the architecture so people don't have to be so technical to set it up. Is that a fair assessment? [57:00] Shawn Hain: **Savio:** It, um, yeah, so regarding the, uh, setup itself, uh, we are doing it in a phased, uh, proper deployment approach, so we don't actually go to the consumer market, right? Approaching an enterprise organization, we actually try to address their challenges and, uh, whatever their challenges are currently, uh, when it comes to data [57:00] Shawn Hain: moving or the business moving from one segment to another, or maybe trying to solve, uh, proposal generation. So, uh, yesterday, as I was scrolling LinkedIn, I saw Decider, which is an Australian company. Uh, they have launched, uh, a new case study about, um, how they have saved for a certain, uh, uh, parents company, um, um, where they're processing quotes from, which takes around three hours or four hours to probably 10 to 15 seconds. And, um, one of, uh, the low-hanging fruit of our, uh, case study is the exactly the same thing. Uh, and we have achieved it for our customer, where they were taking around eight to 10, um, uh, hours to process a quote, uh, which has been reduced to probably 45 minutes to so on. Uh, that is one of the use case that we have. That's **Richard:** very impressive, Sev. Sev, um, so what's a website people can come to and have a look to learn about what you're doing? [58:00] Shawn Hain: **Savio:** The website is starkian.com, uh, through which they have access to, uh, content, uh, of what our platform is. Uh, the, [58:00] Shawn Hain: the reason that the platform is not able to explain it very well is because, uh, the platform has evolved over a lot of things and we wanted to put up something on the face for the people to see. So, uh, the website actually conveys around, uh, 50 to 60% of the message of what the platform does. The remaining 40% of the message is what is yet to be, uh, modified. I am trying to make the language very simple so that e-even it can be a techie, but they can still have to process in their heads so much. So I wanted to make it as simple as possible for people to understand. And the best way to do that is not by words, but by demos **Richard:** What, what's the name of the company you did the case study with, uh, Joanne just asked Um, **Savio:** I am unable to tell you the name of the organization, uh, because of the NDA contract. Okay. Uh, um, because- **Richard:** Well, we'll get, uh, we'll get Joanne to reach out to you and, **Savio:** and contact you Yes. Uh, uh, but they are a public listed company. Uh, they're present in 47 countries. **Dave:** Nice. Well done. Got **Richard:** it. Well done. Got it. And, [59:00] Shawn Hain: **Dave:** um, sorry Dave, back to you. No, no, we- we're at the top of the hour. Um, I don't know, Matt Short, if you've got any, um, short, tight [59:00] Shawn Hain: questions we can throw at Savio, but I'm, I'm conscious it's that sort of wrap up time. There's a lot there. **Matt:** None of my questions are ever short or tight. **Shawn:** Yeah, that's my thoughts as well. I could probably spend hours poking at that. Looks really interesting and, like, awesome. **Matt:** It could... This could, this could be a yes or no, uh, question. Um, so is it, is it replacing the existing apps, sitting above them, becoming the connective tissue between them? Like, I don't fully understand how that works at this point **Savio:** All right. Um, the answer is yes and no to this. The reason is, uh, currently it's, uh, currently it is no, and we don't want to go with an yes currently. The reason being, um, the world is not ready to, you know, just adopt just because I launched something in the market, right? **Matt:** Yeah. [01:00:00] Shawn Hain: **Savio:** They are comfortable with the existing systems. So we have done a phased approach where we [01:00:00] Shawn Hain: sit above the existing systems so that the adaptability is easier. And over a period of time, uh, we will move into full sovereignty. So we will have a direct competition with Microsoft or Mac in the future. But right now, we sit above any OS. **Matt:** Yep. **Dave:** Mm. **Matt:** And that's- Yep. **Dave:** Brilliant. I think, I think it's an exciting space, um, Savio, as we discussed at the top of the show. Like, how do you take these individual productivity systems and these agent platforms and take them from being a one-to-one experience for a single operator, for a single user, for a solopreneur, to scaling across and giving that shared context across an organization, across teams? [01:01:00] Shawn Hain: Um, obviously the big players are building AI into their products to try and do this. And, and whether you've got your sort of Salesforces with their Slack, um, Agentforce kind of stuff, and other, other bits and players. Uh, Airtable, they're, uh, releasing, uh, on the DL their HyperAgent platform, where they're sort of starting to build in these sort of multi-agents with this sort of development canvas where it can [01:01:00] Shawn Hain: create assets, and you can track and match things, uh, across these different sort of, uh, layers. There's m- more and more people... And so, so like you said, Savio, like, just 'cause you've released something doesn't mean the market's ready for it. But- **Shawn:** Mm ... **Dave:** the big players, I think you're onto something, 'cause the big players are predicting that this is the next path. How do you provide this governed agentic productivity layer across teams, not just one-to-one to individuals? Um- Okay ... so that's an interesting space that we're heading into. Well done, **Richard:** Savio. **Shawn:** Yeah, no, well done. **Dave:** Thank you. Thank you so much. Well, uh, that brings... I, I've gotta use it again because I'm not sure when we'll ever use it again. That brings our, uh, AI sharks to it **Richard:** I like that. Richard produced that one in ChatGPT or- [01:02:00] Shawn Hain: **Dave:** That was, that was Grok. That was, uh, Grok Imagine did that one, and nice that it's got no watermark on it compared to the [01:02:00] Shawn Hain: Google Veo ones that we use. But, uh, brilliant. Well, thank you all. Thank you, gentlemen. Been a great chat. For those of you that have joined us live, uh, in the chat, Mano dropped a chat. We miss you, Mano. We'll see you back here in a couple weeks time. And Joe, who was on the show last week, good to see you in the chat with us, Joe. Thanks for participating and asking some questions. If you're watching this live, we love seeing live chats. If you're watching this after the fact, drop comments for us to talk about next week. We would love to respond to some comments. Uh, we know there's, there's, there's many, many people that are watching this after the fact offline as we post it up on YouTube. So jump in, drop your comments. We would love to respond to them in the following show. Thank you all. Uh, make sure you are liking and subscribing. [01:03:00] Shawn Hain: I know most of the people that are watching our show after the fact are not actually subscribed to the channel. So please make sure you subscribe and hit the bell icon. That way you get notified when we've got updates, uh, and it helps more people hear these chats we're having. Like, if you think these are useful chats, just hearing real people have real discussions, not hyp- [01:03:00] Shawn Hain: hyper like, "Yeah, it's the best. Whoa, oh, **YT Clip:** insane results" **Dave:** If, if you like our realer point of view, then, uh, let other people know. We'd love that. Uh, helps us grow. And of course, if you're in the northwest of Sydney next Tuesday night, get along to the pub, at the Australian with us. Uh, that's aioperatorspod.com, and hit the IRL link and you can get the details there, where you'll hear from me, you'll hear from Richard, you'll hear from Sean, and I've also got a local business leader, um, Kelly, who is donating her time to come and talk about how she is doing creative things with AI in, uh, her family businesses and what she's doing as a, as an entrepreneur as well. So she's doing some exciting stuff. Um, great to have her coming along. Uh, and shout out for her son who's kicking off his own, uh, entrepreneurial journey. He's gonna be helping us with videography, so AKA Digital will be, um, also- Awesome ... sponsoring the event with some videography and photography for us. [01:04:00] Shawn Hain: So exciting times ahead. We'll debrief on it next Wednesday, but, uh, drop a comment and we hope to see you [01:04:00] Shawn Hain: there on Tuesday. Thank you, all. Thank you, gentlemen. We'll see you next week. **Richard:** Thanks, Joe. Thanks, everyone. **Savio:** Thanks, **Dave:** Savio. Thank you **Savio:** so much. **Dave:** No problem. Thank you so much for having me. Play- play us out