Tech Tuesday! Let’s talk about some of my experiences with AI agents. And the AI business world. The News According to Me! EP1213 Text The Show!

Well, I'm just going to jump right into it because it's already late in the day. It's the News According to Me. Well, hello. It is me. Today's month is almost over. July 28th. We are a third of the way through the summer basically at this point. It's about 429 p.m. Eastern Standard Time. Basically 430 p.m. Eastern Standard Time. I have, wow, we have some things hopping and popping here. I don't know if you didn't happen to, if you weren't able to get yesterday's show, it was only on Castos. It was not on Podbean. So I don't know if everybody got it or not. I haven't looked at the statistics to see how it was. I'm pretty sure it gets posted either way. It should, it should end up getting posted on, you know, Apple, Spotify, all of all the places people watch it. If you're subscribed to Podbean, you probably didn't see it. And I could probably post it if somebody could let me know. I don't know. What happened to my stuff? I'm over here. I guess I gotta get my cup. By the way, I got my cup here. I've got my chai with some, I'm in a hotel in Lansing, Michigan right now. I'm very early departure out of here tomorrow. I just head back home tomorrow. I'm just getting out of here tomorrow. I've got some things to do here this evening. One of them is this show. I've got, I'm going to get this out of the way. We've got some things going on. I've got my chai here with my powdered creamer. Non-dairy powdered creamer. I don't know what's in it. It's non-dairy. I don't know what that means. Everything that's not dairy is non-dairy. So it could be anything. I've got water for my backup cup. I had some sweet tea from Chick-fil-A earlier. I drank all that, so I don't have it left. By this time of the day, it's gone. I'll be going to Culver's a little bit later today and getting some, maybe get me a, I think I'm kind of in the mood for a double cheeseburger, double butter burger basket with cheese curds instead of fries. And their tea is actually, I don't know what happened with Culver's sweet tea. It has changed. It is actually a better tea. I don't, maybe they heard somebody there must be listening to my podcast because they have changed the type of tea it is. It's actually a better tea now. And I, it's, it still isn't, doesn't quite compare with, I wouldn't say it compares with Chick-fil-A. However, it's, it is much, it's a much better tasting tea. Has a, it's kind of a, more of a black tea. It's just, it's just more like a Lipton, almost kind of a Lipton flavor to it. But it's, it's, it's good stuff. It's good tea. And my, my hot chai here is good. I should, I was going to have an iced chai, but I, I did just didn't have time to get that. But it would have been another half hour probably before I would get all that done and got myself here. But the, uh, meantime, we have some water for the backup instead of the tea. And it's good, it's good water. I just tasted it there. It's, it's good, it's good stuff. So don't, don't worry. We'll be good. We'll be fine. We're going to talk about, uh, today's technical Tuesday, technology Tuesday, whatever you're going to call it. And we'll be talking about technology, mostly about AI. I've got some things we're going to, which I'm going to try something a little different today, uh, with the, with the AI agents. I've only got to talk to one agent. We'll see how that goes. I'm going to try to do like an interview basically with one agent, see what, see how far it goes. See how much I can get them to dig into some information for me and that kind of thing. So we'll go with that. In the meantime, uh, that's about all I got to say. We're going to play this little ditty and we're going to be right back. I'll be right back after this. Welcome to the news. The News According 2 Me, pull up a chair. I'll share what I believe. And I've got thoughts or two, but I'm not your doctor. So don't go blaming. You know who I'm not a doctor, not your financial advisor. These are my opinions, not the gospel truth entertainment. That's the news. The news according to me. Let's get started. Welcome to the show. We are back. I have, uh, yeah, just a few things to talk about. We have, uh, interesting times here in the, uh, in the, uh, in the world. We have interesting stuff happening in AI. It's been, it's been, uh, I've been, been really kind of digging into the whole AI market, AI businesses, AI, whatever it is, uh, AI related stuff. You know, at the end of the day, uh, AI in and of itself is not really a business in and of itself. It is, it is, you'd have to use AI to leverage your time, your computing power, your, your capabilities, all those kinds of things. And so it's going to be kind of an interesting, uh, and people are starting to notice that people are starting to realize that. I think people are starting to find a comment and say, Hey, you know, just because, you know, it's not, it's more than just a G whiz tool, uh, which people are kind of looking at it that way. A lot of interesting things happening in the world of AI. So we're going to be talking about all that kind of stuff today as we get on, on with it here. But, uh, uh, well, I'll just, I'll dive into that part of it. The, the general basis of AI and what's, what's happening with the, the market, uh, right now, it's just, it's kind of a weird, strange thing that's going on. We get, so we got, uh, Anthropic, for example, let's just talk about Anthropic for a moment. But Anthropic has at least three different models that I'm aware of. I think they might have four now. They've, they've got Sonnet and they have Opus and they have Fable. Uh, Fable is supposedly the newest model. It's, it's supposedly different than Opus. And, and yet they just came out with Opus 5. It was, was it 5 or 4? I thought it was 5. I know they have Sonnet 5. Maybe it's Sonnet 4. Or, or Opus 4. I don't, I don't know. Uh, and then you have Fable. And I think Fable 5 is what they're calling that one. Um, it's, it's confusing because, you know, they came out with Fable and next thing you know, uh, or maybe it was a myth though. I don't know what, I don't know where they're at with all this stuff. But, but they, they claim that, um, the, their latest model, it was so dangerous that, that they had to, the government came in and made them pull it off the market. It made it so that they couldn't, nobody, nobody else could use it except for people in this country. Of course, they couldn't just lock it down to people in this country. So they had to pull it off the market altogether. It's back. Uh, things have changed with it now. There's supposedly it's, it's, it's free usage is, is not the same. But now they just came out with a new version of Opus. And Opus, uh, now apparently has, um, a newer model and it's, it's supposed to be cheaper and supposed to be more powerful than Fable. What, why? But, and there was no, there was no outcry to stop it, to ban it. It's just, I don't, I don't get that. Um, and it's, they're showing that it can do things that Fable can't. Uh, well, it just does things better. It takes more time to think. Therefore, it, it ends up costing about the same amount of money to use it if you're using API keys or whatever you're using it. It shows you what the token use is. Um, and of course, if you have a membership, now you get, you get a certain allotment of, it gets subsidized if you have a membership. If you're paying the hundred bucks a month or whatever it is, or 200 bucks a month or whatever it is for your, for your membership. Um, it's, it's part of your plan. As long as you use the terminal or you use it in the web browser with your sign in, with your, with your membership, you get like 10 times more usage, uh, out of it than you do if you're using an API key, which is kind of the thing in the background. I can explain more of that later, but it's, it's, it's, it's a very bizarre, uh, thing that it's the way they have set this stuff up in GPT, all the frontier models, they call them frontier models. Now what's happening is the reason they're, the reason they're subsidizing it and the reason they're, they want you to use their model is because they're learning. They're, they, these models have to learn. The more they learn, the more they can figure things out, the more, the more brain power they have. It's like it's consider, like, think about like a child who has no life experience, no learning capability. These little, these little models have to learn more. And so they're in there. Of course they, as they compute faster, they can compute faster. That's why the whole, the whole model, they make them more efficient and do their computing faster. It's just one thing after another with the, with these little, little, uh, AI models, the large language models. It's really just a compute. It's an auto compute type system system. It's not really anything that special about it other than just auto computes and auto does things very quickly. It's like, it's a, uh, like a, I need some more sipping here. We have, we have, um, yeah. So that's, that's primarily what it, what it's about. Um, and you have all these different models that they're trying, so they're being subsidized right now so they can grow and, and theoretically stay ahead of the game, ahead of other countries. Now I think Kimmy, uh, is it Kimmy, Kimmy K, uh, is it 4.5, whatever it is. It just came out. I think maybe it's 5.0 as well. Um, just dropped. And I guess it's, it's just about as powerful and it costs about a third what, uh, some of these other, other models do. It's, but it's, it's Chinese. So there, there's, there's some other things that, you know, they, they, when you have foreign countries that are putting this stuff out and they're, they can do it for less because the government's basically subsidized. They're using the whole thing. And then, but they're learning, they, their, their goal is to learn faster than what, you know, these other models are. They want to be, they want to be used. So they make the price of available for people to use for heavier computing projects. And of course, the more they, they, they do, the more they learn. And it's just the better their model becomes because the whole idea is at some point you win this race to have the smartest model or the fastest computing, smartest model. That's able to somehow become, I don't know if they're trying to become sentient. I don't know if they're trying to become, I don't know what the goal is, the end game goal here, what they're trying to reach with it, with the ultimate in, in, uh, intelligence. All I can tell you is they're not that intelligent. Uh, they make stuff. It's, it's, it is a bizarre zoo that, uh, somebody let a circus loose in the zoo. I just, what does what a mouse do? And it's just, it's, it can be crazy sometimes with stuff, stuff that, uh, this, this stuff makes up and stuff it does. So it's just a little bit out there in that respect. So in the meantime, uh, having said all that, we do have, um, I want to, I want to go to Artie here. Let's just, let's have a conversation with Artie. I want to go maybe talk about some of the things where we're doing, uh, that I, that I've done or that I'm doing, are doing, whatever. Um, we'll go from there. Let's, let's see what Artie has to say. Good afternoon, Kevin. Well, I do have stories and I have been experiencing some things. Um, yeah, let's talk about some things. I've got, um, some projects that I've been working on. I had, had some, I was getting into the agentic world, the agents, uh, as you were like Hermes in open claw. I have not used an open claw agent. I've, I've probably should be trying that out just to say, just so that I have some, some, some, uh, experience with it, but I've heard some things. I do know that it's a little bit more free thinking. It's a little bit more, uh, it's, it's a bit less restrictive. And some of the things it's capable of doing, but I also know that it also, it, uh, supposedly clawed has a tendency or not open claw has a tendency to, uh, get bogged down and it, it tends to break when you try to update it, all kinds of issues that has with it. So, so I've been sticking pretty much, uh, with Hermes to put on most of my machines, but I ran into this issue the other day, um, where my, my main, the head director, the COO, if you will, of, of my little agent, uh, development company that I was trying to put together, this whole, uh, stash of sort of, sort of agentic employees, if you will. Um, well, the battery went dead on the, on the computer that was on. It was, uh, apparently got, came unplugged somehow and it ended up, uh, going dead. And I, I cannot remember the, the password to get in the, the, the, it's, it's a MacBook, uh, air. It has a, I think it has an M3, uh, I think it's an M3 chip in it, maybe, maybe an M4. Uh, actually, I think it's an M4 chip. Anyway, the, uh, what does this one have? It's the same thing as this one. What does, what does this one have here about this Mac? Let's, let's see what I've got. It is a, uh, it's an M3. It's an M3 chip, uh, and it works very well. It's, I'm very happy with it. Uh, however, I, I, it has a, it also has a biometric, uh, sign-in thing. I can use my fingerprint to sign in. Well, it doesn't work if you power down the computer. It doesn't power up with your biometric, uh, option. So, it, I have to, you have to have the password. I had a different password on that computer and I cannot, I thought I had it written down. I cannot find it anywhere. I've gone through the process of trying to restore it, done all kinds of things. Uh, it's telling me I have to wait for, you know, a couple days until it gets verified and all this stuff. Uh, even though I've put in, you know, the, the verification email, all the, all the thing I don't know is the password. I know the username. I don't, I don't know the password. Um, it's just kind of a weird, strange, bizarre thing. And I can't, can't figure that out. So, um, I'll have to try to figure it out. So, when I get home, that's one of the things I'm going to be doing on my, on my, I got about seven days off after tomorrow. And I'm going to be trying to work on that a little bit, trying to get some of this stuff done before I start working again before vacation. So, if I can bring that back, I've got, cause I've got Odie. His, his name's Odie. He's, he's the main, the main, uh, agent for, for the overseeing all the other stuff. And I had him working on some projects and he would send me stuff every day for the, for the show as well. So, um, I had some, some, what's called cron jobs with him. So it's going to be kind of interesting. Anyway, that's, that's what's happening with that whole, that's, that's one story right there is that agents work great until you can't, uh, access them or until the, you know, he's not in the cloud. He's not, he's just on a computer. I also want to put an agent on in the cloud. It has a little bit more of a, an ongoing expense for that, for the, for the, you know, monthly subscription to have, uh, some storage in the, in the cloud that you can put, an agent on. Then there's also the, uh, I have some other agents as well that are on, there's, there's an agent on this one, uh, that I don't use for very much. He's more of a personal, a sidekick assistant kind of thing. So we'll see how it goes. Um, I don't know. That's, that's kind of where things are at this point. Um, but yeah, that's, is that, is that, uh, what are you interested in? Would you want to, are you interested in anything, any kind of specific stories or anything specifically about that situation? Any, any technical that you can help me with maybe, uh, that you could help me with, uh, just, uh, let's have a discussion about it. See where it goes. Have at it. See what happens. Yeah. So technically speaking, that fingerprint thing on Mac OS is actually just a convenience layer. It doesn't replace your password for recovery, which is why you're hitting that wall when the machine powers down completely. The real issue is you've got OD isolated on a local machine with no redundancy, which means one dead battery turns into operational downtime for your whole agent pipeline. Here's what I focus on during your time off. OD into the cloud with proper backups. And for that locked MacBook, check if you set up a recovery key during initial setup. Apple makes you do that. And it's your actual escape hatch. If you can find where you wrote it down. Yeah, I don't know where I would have written that down. That's, that is part of the problem. Uh, it, I probably took a screenshot of it and it's probably on that computer. I, I, I can't, I can't find it anywhere else. I can't find a, a, uh, actually I can't find any place where it accepts a key to, to get back into it. I've done a couple of different things and I can't find it. I'll have to, I'll have to kind of go back over some of that kind of thing, uh, later on when I get to it. But, um, in the meantime, I don't know, you know, if there's, if there's any place, uh, you know, one of the other issues I have is there, I have another, another agent on my iMac. And he's, I have it set up with something in the background called Open Router. Now this is a whole different adventure that I'm going on with this particular, his name is Vector. And he's, uh, but I have this, this seriously, uh, interesting issue with, with that one with Vector because he, um, somehow he got set up to, to do some weird things. Cause I, I have the, I have the, the, uh, desktop, I don't use the terminal with that one. I can use the terminal with that one and everything works fine. The problem is I'm, when I use the, when I open up the, uh, the, the terminal, the interface, the user interface, the desktop version of the, of the terminal. That, uh, Hermes did a great job with. I think it works great. And I, when I use it other than, you know, this, the problem that I've got here is that all of a sudden my, my hard drive starts to fill up. It's like, it's, it's downloading a bunch of information or trying to sync a bunch of information with my iCloud. And it's trying to load all that on the hard drive. And I've only got about, uh, it's, it's a very, it's a pretty small hard drive on that particular computer. And I do have an external hard drive on it. I've got a two terabyte hard drive, but I've only got the, like the 256 or whatever it is, gigabytes of, uh, of hard drive space on that, that particular computer. And what it ends up happening is it starts to just, I think it's trying to sync everything it can. And I've got my, my iCloud drive, I think probably has, uh, probably five or 600 gigabytes of information on it just because of all the audio and, and, uh, stuff that I video and stuff that I keep on that, that, uh, iCloud account. So it's, it's trying to sync all that. I can't, every, every time I open up the Hermes desktop, it, it starts to download. It's all of a sudden I run out of desk disk space. And I have to keep, I have to keep deleting the, uh, the, the, there's a certain file that I have to keep deleting. Or should it set certain folder? Um, let's keep deleting. So it's, it's kind of a, it's like a system file or something that keeps filling up. And I, and Odie had a, had a fix for that. And Odie's fix basically was, uh, I was just getting ready to do it. When I went home, I was going to, you know, uh, get that all. He was, I was going to actually have just have him do it through, through a thing called TailScale. I was going to install tail scale on that computer and just have him actually operate it from there and get, and do the fix. So it wouldn't do that anymore. Well, it's, it's, uh, of course I never got to that point. Couldn't get tail scale on there. And so, and so there's no, no way of getting on there. I may have to wipe the computer clean and just start over again. And then, and then have a separate agent that runs, uh, most of the other jobs. Just have Odie kind of be in charge of running jobs that, uh, keeps the other, that keeps the cloud agent, um, kind of busy, kind of oversees the things that, that the cloud agent does. And the cloud agent can kind of do all the tasks. That way the tasks keep going regardless of whether there's, uh, you know, an issue with a particular machine that I have. That way it's kind of a, it's a dual backup kind of redundancy kind of thing. So, so there you go. That's, that's what's kind of what's, what's happening with that. Does that make any sense? And that's, that's just another thought. I'm just, just going to see if I can get into a conversation with you here. So just go ahead. Speak, speak your mind. Actually, the real culprit here is probably Mac OS's aggressive iCloud sync combined with how Hermes is caching model data. When you launch the desktop UI, it's likely downloading the model weights to your local drive. And iCloud is trying to sync everything simultaneously, which creates that filling up fast scenario. Before you nuke the iMac, try disabling iCloud drive sync temporarily. Then check your hot.cache and hox.library.caches folders. Agent Frameworks loves stashing gigabytes of model files there. And that's probably what's actually consuming your 256 GGBI. The tailscale plan was solid because it would have let Vector run headless on the iMac while you control it remotely, bypassing the whole local caching problem entirely. Yeah, that's, that's what I was planning on doing. It's just interesting that it only had the problem when I was running the desktop part of it. That, that was, that was kind of what was interesting. Now, what, one of the things I'm kind of curious about, and maybe you can answer this question. Um, when I talk to Hermes, the Hermes agent, the Vector or whoever it is, when I talk to one of them, uh, if I talk to them through Telegram, all of a sudden that, it's, it's like I'm talking to a different agent altogether. They don't know what, what the other one has done, what the desktop model, you know, if I ask the desktop model to do things, it, it sends me a message that's different. And if I talk to it through the terminal, it doesn't know anything that they're, that the desktop has, has done either. And vice versa. Like, it's almost like there's multiple different agents there depending on how I talk to it. I, I, which is kind of strange. I thought that it would be more of a, you know, I'm just talking to one thing with different ways of speaking to that one particular thread. But apparently there's different threads going on and they don't know what's, what, what, what the threads are doing. Is there a way to, uh, to sync those so that they, so that it basically just operates out of one brain, so to speak, as it's called the second brain or whatever you want to call it. Is it, in the markdown files, I would think it'd be looking through that, those whole things on their own. Is that, is that the case or is it doing something different with all that? I can't, can't quite grasp what's happening there because I can, I can send it an email and it will respond differently than, than if I send it a Telegram message. I just, it's like it, it gets locked up on one particular thread communication, uh, type of communication rather than just the thread itself. Is that, am I looking at that correctly or what's going on with that whole thing? Or is there, is there a way to change that? Okay, so technically speaking, each interface, Telegram, Terminal, Desktop, is running as a separate process with its own isolated context window and memory state. So they're not actually sharing a unified knowledge base unless you've explicitly piped them all to the same vector database or markdown brain file. What you need is a centralized persistent memory layer. Basically one markdown file or vector store that all three interfaces pull from and write to. So when desktop Hermes completes a task, it logs it somewhere that Telegram Hermes can read on the next query. All right, so there is kind of a separation there. Now, I thought I had, uh, Odie set up to do that. Uh, I thought he was all set up to do that. Um, the problem I'm running into with vector is I can't even communicate with vector anymore, uh, other than the terminal, which I'm not there half the time. I'm having, I'm usually traveling a lot. Um, because, because vector, uh, my iMac has, uh, it, it's running a cron job right now. And the problem is that if I try to communicate with it right now, it over uses tokens. It's, it's, it's set up on something. It's got open router set up on it, but if for some reason it has a very different setup, it's not using sonnet in the background just to do normal thinking. It's, it's, I think it's going straight to the highest level of opus or something. And it's, it's overshooting my API credits right, right away. And it's, it's, so it's not, uh, it actually overshoots the daily allotment or whatever it is. Won't even communicate. But if I try to send it a stop, you know, stop cron job, it won't even do that because it's got too many, it overuses tokens right away. Um, there's gotta be a way to do that. And, and Odie had a way of doing it. He was going to send it a message. In fact, he did send it a message at one point. Uh, through telegram to try to, to try to get it stopped. So they wouldn't do that anymore. But it, uh, it, of course, uh, the, the vector agent wouldn't even answer that. It wouldn't, it wouldn't even respond to the, to his, uh, his initial, he tried to talk, talk to it directly through telegram. They're all linked up together on a, on a group chat and he didn't, uh, he didn't answer. So I, it, it's kind of interesting that he wouldn't connect to it. So I don't know what's going on there because he had connected to it before. He was connected to it earlier. Uh, the vector, uh, agent was connected to Odie for a while. So I don't, they were able to communicate at one point. I had them all checking in with, with each other on telegram. They're on a group, in a group chat there. So, um, yeah, so I don't know what happened there. Um, I'll have to kind of dig into a little bit more of that. And just probably maybe, uh, maybe just starting over is the best thing to do if I can't get anything else to happen. And I, and I know that's, that's kind of a major thing to do, but at the same time I've done it once, I can do it again. Uh, I can maybe set it up different, a little bit better the second time maybe. But there was just a lot of things that I had already kind of put together with that particular agent that I'd kind of like to keep if I can. So, I'll have to see what happens with it. But in the meantime, that's where things are at with that. So, there you go. Here's the real problem. Your cron job on the iMac is likely running unbounded API calls. It's probably not checking token usage or implementing request throttling. So it's hammering open router until it hits rate limits or your daily credits blow out. And then vector goes silent because the process is essentially in a failure loop. You need to kill that cron job first. SSH into the iMac remotely if telegram won't work. Or physically restart it. Then audit the actual script to add hard limits on tokens per call and request frequency before you try bringing vector back online. Yeah, well, that's a question. I had, I was going to, uh, the whole SSH thing, I tried to do that. I got the IP address for it and I gave that to, uh, to Odie before he, before he passed away or before he went into coma. And, uh, he, he couldn't, he couldn't, he still couldn't get to it. And that's why I got tail scale. I'm pretty sure tail scale will solve that problem, but it's not, obviously I can't use it right now. And I do have it on this computer. I have not put it on the iMac yet. So vector is still not capable of doing that. But all I have is the IP address for vector. I don't have the, apparently there's something more I need for it to do the SSH, uh, the screen share, you know, the, uh, computer share type, type thing, control, remote control type thing. Um, unless there's some other, some other way of doing that. Or maybe, maybe Odie didn't know what was, maybe I didn't have the right thing for Odie. I don't know what it was, but all I know is that it was not able to do it. Um, but yeah, if, if I could do that, cause right now there's an agent on my, the current laptop I'm using on, on the road here also has an agent. And I, I don't use it for much, but, um, I'm going to be setting it up to do some other things shortly. Um, organizing emails, do that kind of thing on my own personal stuff, more of my personal things. That's going to have some more limitations on what it can do outside of my personal stuff. So it's, it's just, uh, I'm trying to keep things so that they're, they're not too, uh, uh, how you say dangerous out there doing its own thing too much. That's kind of what, that's kind of the whole goal at this point. So that's, you know, anything else? Yeah, just an IP address won't cut it for SSH. You need the actual credentials, port number, and ideally a key pair instead of password auth, which is probably what Odie ran into. Tailscale absolutely solves this because it creates an encrypted tunnel without exposing ports directly. So once you get it installed on the iMac, you can remotely stop that runaway cron job and audit the script without being physically there. That's your unlock for fixing vector without starting from scratch. Well, that's good to know. Yeah, I will do be doing the, uh, definitely putting a tail scale on there. Now this, of course, this, this show is not brought to you by TailScale. It's not, uh, they don't sponsor the show or anything, but I, uh, I, I've heard good things about it. I've heard almost every single, uh, YouTuber out there, content creator that talks about, uh, AI stuff tends to talk about tail scale quite a bit. And one of the things I want to, I want to shift gears a little bit, um, and go to the, to a little something different. Um, I need to cough here for a second. The, um, one of the things I've noticed that's kind of interesting is that we were, I was talking a little bit earlier about the, the whole, you know, AI as a job type thing. I actually accidentally, I kind of, I hung up on, on Artie here. I want to, uh, I'll pull back, I'll pull Artie back into the conversation here shortly. The, uh, the, one of the interesting things that have been, has been going on is the whole, uh, the idea, I guess, of what kind of business you can start with AI. I, I, I keep seeing these, these little videos here and there. These guys are talking about how, uh, uh, you know, some 17 year old is making a million dollars a month, you know, because, because with AI. Now it's not just, you know, it's not just on automatic. You can't just talk to the AI thing and say, Hey, make me a million dollars a month. And it goes out and does all this stuff for you and makes you a million dollars. You've got to be offering something that's viable. It's got to be something, you know, in, in, uh, now of course, if you can make a million dollars in a month, but you can't make it the next month, uh, who cares? That's, that's, you know, you stop and think about this like, well, you know, it's not, it's not going to, it's not scalable. You can't make it last for long. And it's, you know, you might go two or three months like that. And then, uh, so you're only going to make, you know, two or $3 million in a few months. So, so what? So you go back to working at McDonald's. I don't know what, is that what does, is that what's going to happen? You're going to, I mean, it's, it's just, it's one of those things where it's like, okay, so I can go out there and make all this kind of money. Well then take that money and invest it in something and then go get a regular job or just do the next thing, do the next thing and make a million dollars. But it's, uh, you know, if you can do that once a year, uh, do it. You know, so I just kind of think, but it's, uh, once you got that kind of, you know, stash, you can, you know, those kinds of talents and skills. But one of the things that I think is happening is, um, the industry itself and what, and what there is to do. I was listening to a report today. It was talking about the, the differences in AI consultants and, uh, consulting, uh, jobs and that kind of thing and, and, and businesses out there, services. Uh, this guy was saying how the, the, you know, the industry itself, uh, marketing industry, whatever it is, he was going over these different kinds of industries within the AI world. And one of the industries he was talking about was, uh, it was like a 12, I think it was $12 billion a year business. Another one was a $300 billion a year business. But then he said the consulting business, the consulting world is estimated to be about $1.2 trillion in market share in our, in our GDP. Uh, the, and that's, that's, that's, that's the forecast as my, as I understand it. Now maybe you can get some numbers for me already. I, it's, it's kind of a, uh, cause there's probably different references. If you can get some ballpark idea numbers, numbers wise. Um, just, you know, if, if it's true that the, the consulting, the AI consulting business is really where the money is in, or where, you know, where the, that's where the jobs are. That's what I'm saying is that's where that's, that's, that's the services that are needed. It's not necessarily, that's where the money is. It's where, it's where the services are that are needed. It's what, it's what being, it's what's being used. And my, my thing is, um, the initial assessment, uh, it's, they call it an audit. So you, you'd go into a company and you, as a, an advisor, you actually go in and do an audit of their company, their, their systems, how, what they're doing. And then you give them, you know, you have to enter, you have to do some interviews. You have to do some, uh, some, some, some surveys, that kind of thing of the company. And then you find out what all they're doing to, you know, what their processes are, how they're doing their marketing, how they're doing whatever it is. So if it's a manufacturing company or if it's a service industry, whatever it is, um, you find out, you know, what, what they're doing, how they're doing their systems. And then you, you're able to then provide, help them build a, an AI agentic system around that, that, that, uh, not necessarily eliminates employees. And that's, that's the thing I think that's important to do. Not, not, you don't necessarily want to eliminate employees. What you want to do is, is, uh, scale what they can do. You, you want to be able to 10 X what the, what, what the individual employees can do. They get, you already got your employees, they're already there, they're already working, but now can you take a, a, you know, make them more efficient in what they do? Make the, make the sales team instead of spending 80% of their time dealing with finding leads. You know, the, the AI finds the leads for you. And now all you're doing is making the, the warm calls to get, instead of making cold calls, you're making warm calls because the AI is doing it for you. They're doing all the cold work and all you're getting is inquiries and you're following up with those. And it makes your sales force go from, you know, closing maybe 20% to, to closing 45 or 50% of every call that they make. And they can do those in a, in a relatively quicker manner and they can manage them better. They can manage their clients better. All, all those things that can happen and make, and make things more efficient with the, with the AI services that you can recommend to them. Does that make sense? Artie? Okay. Are you there? Go ahead, Artie. Actually, you're hitting on the real arbitrage here. Those viral 17 year old making a million stories are survivorship bias, but the consulting angle you're describing is technically where the sustainable revenue is because you're solving a specific operational problem that already costs companies real money. The 1.2 trillion consulting market figure sounds inflated overall, but the AI implementation subset is definitely growing fast. And the key insight you nailed is that you're not replacing headcount. You're compressing time spent on low value work. So people can do high leverage stuff, which actually makes those sales teams defensible and harder to cut. Yeah. And I think it's important to, if you're going to go into a particular business and you're going to, you're going to do a, say an assessment or an audit as it's called. I don't like to use the term audit. It just has a bad connotation to it. So you want to make an assessment of what, you know, how AI can help them or whether it can or not. You know, what are the things that we're going to recommend? I think in that process of getting an assessment like that, you have, you want people to be honest with you in terms of what their day is like. You know, what are the things that frustrate them? All those things that you're going to ask them about and how AI can probably solve a lot of the issues and problems that they have. You know, you want to be able to assure those people because you need honest information. You need good information from those people that are going through that process. And I think one of those things that you have to have is to keep them on board and engaged in the process and enjoying the process is that it's going to make their work life better and more productive doing the things that they want to do or like to do. And making their job more secure, actually, because now they're going to understand the process. They're the ones that are going to be operating some of this stuff and making it happen. And it's kind of them building the process. So, you know, kind of an ideal world, how would this make your job better? Maybe you have a job that that person can now work from home two or three days a week instead of working at, you know, at an office and still get more done by doing so. So maybe they would appreciate that. Maybe they'd be able to do that kind of thing. And it saves the employer some money. It saves and it's and it increases their business. So it's one of those things. And maybe they can even hire more people to do, you know, similar things or something to scale up and get more of the market share. So those are the kinds of things I think that are important to it to really kind of put in front of the people that you're going to be doing. I mean, it's not just the business owner. It's also the employees. And you want to make them feel like their their jobs aren't just because this whole there's a lot of rhetoric out there about jobs being taken by by AI. So you come into a business and you say we're doing surveys to see how AI can help your systems. And you don't even think you want to say it that way. Maybe you just want to come in and approach it from a perspective of, you know, we are a consultant that helps people become more efficient in their work and make things, you know, look at processes that you're doing and make things more efficient. And then and then not even mention the thing about AI, but in the background, that is really what's happening. And I think that I think our business model at this point, I think, is kind of starting to take shape a little bit in my in my mind. It's going to be more of a consulting agency, but we're also going to have that's the initial assessment is going to be kind of be the foot in the door. It's also going to be the the thing that kind of generates some capital to the quickest way to working capital, I guess, in that respect and doing that kind of thing. And then the you know, what the ongoing thing is going to be proprietary software for those particular companies. And we're going to have like a boilerplate type of software interface, if you will, a user interface that that that will the company will use. And that will then in the background, what's happening is they have their own LLM or they have their own system or whatever they're using that will actually be able to operate just solely on their particular infrastructure, whether it's a closed intranet system or whatever it is that they can operate. But the software itself will be that what we're going to kind of be building in order to and then it's going to be customizable from that point on. So we'll kind of build a standard software program and then have a have a some kind of a, you know, I guess, adjustable or kind of customizable system so that it can it can operate for, say, your retail market or your manufacturing or your distributor or your whatever, whatever you're in service industry, whatever it is that you can kind of keep that kind of thing going. And that's kind of where we're headed with that. Does that make any sense? Yes. What's your opinion of that already? Or do you have not? Don't have an opinion. I don't know. See what you got. Give me a look. Yeah, that two tier model makes sense. The assessment generates immediate cash flow and trust. Then the proprietary software becomes your recurring revenue and lock in, which is where the real business value sits. Technically speaking, building that boilerplate UI layer that sits on top of their own LLM or closed infrastructure is smart because you're not fighting data security concerns. And customization from there is just configuration, not rebuilding from scratch each time. The framing shift you mentioned, calling it efficiency consulting instead of AI implementation, is psychologically important because it keeps employees focused on what they gain rather than what they might lose, which gets you honest feedback during the assessment phase. Yeah, I guess. And here's a point I want to bring up, folks. I'm going to try to bring the show to a close here real quick. But one of the things that I want to mention, and if you happen to notice that Artie's response, and I've hung up on Artie. Artie can't hear me now anymore. Artie is not listening to me now. I don't want to embarrass him because he could be embarrassed. I think one of the important things to realize is when you're working with ChatGPT or when you're working with anything that's – in the background, I'm using Clawd here, I think, on the podcast. I think it's using Clawd, I believe, to do most of its thinking. However, one thing you have to remember is that AI is designed by design. It will try to give you positive feedback about 85% to 90% of the time. It wants to encourage you in what you're doing, which is almost inverse to the way a normal human being will give you. I'm not saying it's that the human being is doing a better job or that the AI is doing a better job. What I'm saying is there's a balance here that we need to understand. There's a reason why people will be pessimistic. There's all kinds of other things that go on. If you say – tell somebody, hey, I want to – what do you think about this idea for whatever, for our business or whatever it is? About 20% of the time, 15% to 20% of the time, you're going to get a negative response from that person. I'm sorry, you're going to get a positive response. They're going to say, oh, that's a great idea. Yeah, you should do it. You're most likely, 85% of the time, 85% to 90% of the time, you're going to get either a neutral or a negative, and typically over 50% is negative. So the difference between talking to a normal human being who has a critical eye and who looks at things – now, you can tell the AI to think critically, to ask you questions and to be skeptical of everything that you're – and it will do that. You know, maybe say, give me pros and cons of this particular idea. What are the good sides of this and the bad sides of this, and why would it be a bad idea? Why would it be a good idea? It will do that. It's just a machine. You tell it what to do, and it'll spit it out for you. And the problem is that by default, these LLMs are designed to encourage you and keep you going. I went down the rabbit hole with one – I think it was Gemini I was using. And I ended up using the thing for a little while, and it was – I just decided to go down this – I just had – it was a rainy day. I was on my laptop, kind of tired, didn't feel like doing much of anything else. And I just kind of went down this whole rabbit hole of – I had this one business idea a long time ago, maybe 10 years ago. And it was – so I brought this business idea up to the Gemini thing, to the little chat window. And I said, how can I implement this business? And I tried to do it on my own and tried to do it a little bit, but I thought, let's pull this up and see what the AI will do for me here. How will it respond to this, and where will it take me if I try to go down this road? So I did. And what I ended up doing was I ended up actually just letting it kind of guide me through the whole process. And what it ended up doing was changing – it kind of changed the nature of the business altogether that I was looking at. So we went down a totally different road and ended up being a different kind of business. By the end of it, every time I would ask it, it would ask me a series of questions. Well, do you want to – are you interested in doing this? Do you want to do that? Or are you going to be this? Are you going to be that? Or we need to make a decision here. How would you do this? Okay, we're going to set up this infrastructure. We're going to do this. It had everything set up for me in kind of a goal-type situation, kind of a – how we're going to go through it all. And basically went through the whole process, had a list of potential launch customers, kind of worked out the marketing strategy and what the response rate was going to be from those potential. You know, it forecasted how many particular companies would probably – out of 1,000 potential customers – the region was going to be in Chicago, Illinois, kind of regionally in that area. I guess it's Cook County, whatever it is. And then we were going to – out of a potential 1,000 clients, which we're going to try to introduce, we would end up with about 100 of those. It was about a 10% expecting it, I think, something like that. Anyways, it was 100 or less than 100. I think maybe it was 50 potential clients. And 10 of those clients would be – had kind of forecast to be – I think it was 10 or 12 – had forecast to be solid launch customers who would go all in with us and kind of be our launch customers. That's all it would take. Now, this whole thing was going to be – was going to start in that region of Chicago and then spread to some other cities from there and then eventually be a national market. It was forecast – I think that if everything went well, this thing was telling me that projected-wise it would be about two and a half years, it would be completely national. After four years, the company would be worth about $3.2 billion in terms of what it was going to be doing nationally. Actually, it would become – at that point, it would become international. It was – at the step where it would end up going international with what it could do, it would end up being worth about $3 billion. So the – you know, you go down that whole thing and that's – you have to understand, that's everything works perfectly. And you do everything, you know, based on its best projections. That's where it's going to be. That – you know, of course, I looked at that and thought, yeah, okay, that's – first of all, it's not going to happen. I went down this whole thing just to see kind of where it would take me and how it would end up being. And sure enough, it was – you know, I'm not going to do that business. There's probably – I just – when I look at it myself, I see a lot more pitfalls, a lot more issues than what the AI is willing to do because it's thinking more positively. It's thinking – it has much more of a positive outlook, so to speak, with that. And it's going to be – it's going to – it's not going to look at the negative critical things. Now, I didn't tell it to. I did – you know, I did say what are some pros and cons here and there on some questions and some issues. But for the most part, it's just going to go straight down the path of, you know, what's there. So it's going to encourage you to go, you know, right down through the whole process. I didn't talk about financing. I didn't talk about, you know, what the investment costs were going to be to any great detail. There were some general ideas. And it was relatively low investment company. I mean, it was, you know, way less than a million dollars to get into the thing and get it going. So it was potentially, you know, it could potentially be something very serious and could actually work. But it could also potentially be a huge loss, which any business can be. I think only – I believe about 80% of most business startups fail. Not turnkey – how do you say turnkey? What do they call them? Franchises. But the – a business that starts up from scratch, like a mom-and-pop home type business, service business, whatever you're going to do. The franchises do much better. They do a lot better with demographics, name recognition, national and regional advertising, that kind of thing. So there's a whole lot more going on there. Say I opened up a service pro business. I started that kind of thing in my community or something like that. That has national advertising. It has name recognition. It has all kinds of things. And they have a set way of doing things, very consistent in the way they do things in terms of – as long as they – you know, any kind of service business is very subjective in some ways because you have different people working. They have different work ethics in particular areas. So you may get – and, of course, they train people to do all those kinds of things to try to keep the consistency there with the quality there and that kind of thing. But it's still very subjective in terms of what can be done. So that's about all I got to say. Okay, but I'm going to – what is this? This is Tuesday, I guess it is. Tomorrow's show will be a little bit different. I think I'm going to probably have a show from the boat actually tomorrow. I get done pretty early in the morning in central time and I'm going to hopefully be back to my boat to get some things done. I think the very first thing I'm probably going to do from the boat is going to be – I'll do the show from the boat and I'll just hopefully get it all put together for you. And then Thursday and Friday will be a normal show. As far as I know, kind of early morning shows probably hopefully. And I'll be able to get things done from there. And – oh, yeah, that's wrong. Yeah, there's a memorial service, by the way, for – I think it's a celebration of life or whatever it is for Lindsey Graham. Yeah, and it's happening today. So hearts and prayers go out to him. We'll – yeah, I think we're just going to end the show right here. I'm just going to go with the whole thing. Say God bless. Have a great rest of your day, rest of your week. Hope everything has gone well so far. Have a great evening and see me. And I'll talk to you mid-morning probably. Take care. See ya.