Latest / Hardly Artifical Podcast / AI's Billion-Dollar Reality Check

Transcript

AI's Billion-Dollar Reality Check · Hardly Artifical Podcast

• Episode 7(1): Hospitality's been good. They give us free water. Is that what hospitality is? Thank you for the gift of water. Thank you for the gift of water. And yeah. As you suggest, sponsored by Desani. Sponsored by Desani. All right. We're just not going anywhere. Let's just get the show going. Yeah, let's do it. All right, everybody. Welcome to episode number seven of the Hardly Artificial Podcast. I'm Derek Major. And I'm Mike Bakowitz. And we are live in Calgary, Alberta. we're actually here for the I think it's like the nineteenth or eighteenth. I can't say it's the eighteenth summit, but it's definitely like the sixth or seventh summit that Eligio's had in a row. But it's certainly Eligio's nineteenth year in business. So Everyone's gathering here in Calgary. So but of course we're here to chat AI. of course the podcast is all about learning all the cool things that are going on in the AI world. And you know, we have lots to talk about today, as usual. It feels like every week goes by. main topics today. we'll touch on anthropic. the latest news. Of course, last week we talked a little bit about it. The whole IPO situation. new numbers came up for Q3, a little scary. again I'm not really into the idea of that stock there, Mike, but it's no we'll get into that one. we're gonna also do a little bit on the some of the use cases I've been working on. I I have this cool new setup on my Mac Mini that I've been playing with. so I'll go through a use case related to hiring. So I thought that was great. And you would do you have today on the show. You know, it's it's kind of a related story. Especially with the alphabet release and re earnings as well. So a lot of the big firms are releasing them right now. And I think that comparing the numbers, really seeing the reality of the situation is gonna help folks understand that, yes, it's scary to see a number in itself. Yeah. But un but once you compare the front the landscape, what's actually on, right? I don't think it's gonna be as as scary. I'm not saying it's not scary. I I'm I'm gonna clear enough. I'm gonna say that. and then I have a fun story. I also am gonna share a little bit of what I've been working on in. in the background here. Cool. But there's another one with Amazon hardware. No and it's an interesting one. It's more of a something people can relate to and it's their creepy alarm clock. Ooh. And I'm just gonna leave it at that because it's kind of an interesting I wanna see what the listeners think of this approach because I can see why it's called creepy alarm clock. And that's the that's the term that people have coined. That's not the name. So I I haven't heard about this one. So I'm interested to hear about that one. course segment today. We're doing prompt roulette again. we did that a few weeks ago. So we're gonna do today with Replit. so so I got it fired up. We've we've I don't know what prompt we're gonna throw in there. Mike's gonna throw one at me. We'll make one up on the fly. we're gonna try it out and see how what we can come up with today. We did Hugging Face last time. so that'll be kind of cool to see where that goes. So course for those in the United States this week, it's midterm week. So It's gonna be crazy. It was earnings week last week. Now it's midterm week in the US. So there's gonna be lots of stuff on the political side. Of course, AI is definitely front and center. do you hear like just going out and about, like for you? Are you d do people talk about it as an issue? Like, do you hear much about it? It's it's a joke, unfortunately. Like I and I actually I I'm a little disappointed that people are taking it lightly. Okay. I'm not a I'm not heavy on either side necessarily in my beliefs, but What I can't say. We won't get into politics, I promise. Absolutely not. But it's really heavy handed on the messaging, which has been very stark differences, which I I don't think either or is wrong or right. I think it's just this is a challenge in our world right now. And I don't know if there's a right way to do things anymore. And yeah. As we see in AI, right? There's so many different ways you can do the same thing. I don't know. Interesting. So I I I do feel for the folks that are feeling the pressure, and we're here with you. Don't worry about it. Things are gonna be fine. Yeah. I just want to say that out loud. We're gonna be fine. Things are gonna work out just like the A world. And I wanna think that's a great segue into what we want to talk about today. Yeah, exactly. So let's jump right into anthropic. Yeah. That's a very fascinating one for me. you know, some of the notes here that I have, of course this just came out in last couple of days. so their revenue surged big time, seventeen billion dollars in the quarter. So that's you know, forty seven percent increase. Which is massive. but the devil's in the details as usual. they're filing exactly. So the reason what's happening now, again, I'm I think I talked about this last week. I'm not really a you know, stock expert, investor expert and all that stuff, as you can tell with my very complicated terminology. but the the thing that I understand is they did their filing and their IPO filing, but now they also have to report on the Q three, which just finished up. as part of this process. So this month they will do the actual IPO. But what's happening here is that growth has slowed too. So that's a significant issue. Yeah. So there's concerns about sustainability. the other thing too on here is that the computing costs is are s are skyrocketing. And it's not even just the the cost for the compute, it's also the commitments they've made. I remember I said that it it sounds like they have a lot of debt they got to repay next year. Yeah. And they needed this, otherwise they're in a lot of trouble. Yep. So there's what is it? They have nearly nine hundred billion dollars in future infrastructure commitments. So isn't that crazy? That just gives me heartburn. and and just just being part of the the like the high end of like, you know, you start thinking about the bigger numbers, the biggest impact. Like just just that's a daunting number to have to It's numbers most people can't really wrap their head on, right? But it's like I just hear risk all over it. but you know, there's there's some though that are s suggesting that the underlying business can't be more profitable than what the headline is is saying. Because part of this, they're in a huge race. They're they're competing against open source models, Chinese models, they're competing, I guess obviously OpenAI, Gemini, Muse, and Grok. but really, you know, before we get into more conversation on like think th their their biggest challenge, I think, is not necessarily generating the revenue. It's gonna be con controlling the infrastructure costs and sustaining the growth. Because the reality here is that the the losses are headlines here. but the the main thing is they could probably dial things back and have a really great business. But it feels like they're almost tackling two two things here. They're trying to stay ahead of the curve, tons of R and D, tons of tons of infrastructure costs, but at the same time they're trying to create a business. So what do you think? Do you think I think I asked you this last week and you laughed at me, but would you invest in anthropic when this IPO hits? it's an adamant no. Yeah. And again, risk averse. We we made that clear last time. But there's a there's a bigger choice to that one and a bigger reason. And I think it's the the amount and level of investment Of the risk taking there they're actually taking on themselves to really just try to solidify themselves as the choice, right? And I'm actually curious, there there is a question that came up when you brought up one of those numbers where they made a huge commitment for years, right? I mean, it's it's a long-term commitment, yeah. But I think that was in the thralls of their massive growth. And I think they're thinking, well, how could this slow down? And the diversification and the non-monopoly that they've started to see unfold in the market. I'm almost wondering if this is just they know what's coming or maybe they're just making an educated guess at this point. Yeah. And maybe there's something going on in the background where they see something that what's the next level, right? Because it changes so fast. I I don't I am not educated enough in that world to know even what's going to happen. So I can't tell you guys what's actually going to unfold. But I would foresee something changing fast. I feel like something's gonna shift and these open source models are still gonna be in play. But maybe there's an advantage, just like they had the advantage with anthropic, where they were the ones talking the news, talking the town for how long, like six months. And all of a sudden it was like everybody else is starting to do the same thing. Right. So I there's there's something going on and it'll be interesting to watch. I I I'm really fascinated. Well, I I think the way I look at it is that the technology If you if you really boil it down how it works, right? the whole you train a model, now you have a model with all this data in it. And the way and then you're you're executing tokens to get responses back, whether you're sending an image, whether you're sending whatever you're sending to the system. And that that's we've talked about creativity, right? Like AIs today cannot be creative. It's just not possible the way they're designed. So you still need that creative thinking. But at one to your point, like is there something else that's being worked on? That that we don't even know about, or maybe so other companies figure it out on how to actually create, you know, a creative system. So it's not relying on all this trained data. So I mean it's almost like simplifying it too much, but I do simplify it to down. It is software at the end of the day. It's we've created this thing. It is just a giant again, I'm sure like anthropic people be freaking out hearing me say this, but it's like a giant database. It's a really big database. And it needs GPUs to execute on, to actually understand, to actually to to train it and actually to to actually do inference to actually get results, right? So it's but is there something that else that's being worked on that can actually address the creativity side, which would then present, you know, new challenges, I think, for these open weight models and all that. Because really the Chinese models, was it I can't remember if it was OpenAI or Claude, they went through this too. They they had they're reporting, hey, China's hacking us, China's hacking us. Now they're all the same and they all come from Deep Mind, right? So if you look read up on Deep Mind from Google, that whole world when Google acquired Deep Mind and brought that under the umbrella, all these folks are tied back to Google, right? That's why the concepts are all very similar between OpenAI and Anthropic. Of course, Anthropics history is Dario and his sister came from OpenAI. So it's, you know, that they they're all doing the same thing. They're just doing it, you know, in their own ways and they're training it differently. is it is it far fetched to say it's it's like getting a latte from Starbucks versus another another coffee shop down the road? they're different different flavors, but at the end of the day it's a latte, but one might have a different flavor versus the other one, the way they make it, different type of beans and all that stuff, and mixture. That's my analogy. Interesting analogy. I like that. And and I I wanna I wanna take that one next level. Because if you think about it, yes, the ingredients may all be similar, right? The type, but We know, and again, I don't know those of you that don't actually drink coffee, maybe, maybe you do, hopefully you do. the the amount of roasting, there's all these variables, right? You have to control. So I always think about it as, and when you brought that up, that's actually a really good analogy. It sometimes doesn't have to do with the material itself that's being put into the cup. It has something to do with how it's prepared too, right? Yeah. And I think that brings back to the point of deep mind is a lot of those models, if you look at it, compare it side by side with Opus 5.5, right? That's that seems to be the the popular anthropic model right now, especially with how much more efficient it is and running what it needs to do. It's almost like you think about it, the model itself is what produces it, but the model has to be trained and be able to be manipulated to that point. Yeah. So if we think about it, it's the same thing as like a coffee, right? You can go to any Starbucks, get a, decent. I'm not saying it's good. I'm not saying it's great. It's not bad either. It's good. But go to that specialty coffee shop down the road. There's something different, right? They do something different. So I think I think that's gonna be the new wave is yeah. No, you could go with the big frontier model or you keep tracking and keep monitoring, just like we are, honestly, day to day and go, we're getting pretty close here. What's gonna ch what's gonna shift it again, right? And it's daily. totally. Guys, I one thing I wanna reiterate is you're not gonna keep up with this. I can't even get up with half the time. I know Derek's constantly he and I talk about it all the time offline. We're just like, well, did you see this? did you see this? do you see this? And it's in hours. It's not days. It's not weeks. Yeah. It's hours. So it is it's it's it's nice. Yeah. But I I would I would be interested to hear your take on if they go IPO, what's that differentiator that's gonna allow people to believe, just like they did in SpaceX, right? If you were to bring up a very recent IPO that just happened. They're almost back down to the IPO price, SpaceX. And see, that's the thing, right? Is it was hyped. Yeah. And now it's back down to basically where they were at. Yeah. Is it on track for a similar? Or do you think there might be a different Man, I don't know. You know, I read how I can't remember that one of the investment firms, I think it Blackstone. And they they they basically said the second that that IPO hits, they're gonna they're basically gonna rate their their debt as junk. Like just right right out of the gate. Which is unheard of, right? So again, not an expert in this stuff, but That's how leverage these guys are. But again, people are looking at the potential and the and that's what they're looking at. But it'll be interesting, like I you know, that's the same conversation or sorry, same articles I've reading. Is they they've been talking how that the impact of this this recent news too might even affect the valuation. So they might have to come in lower on their IPO. Interesting. but they might not have a choice because they're gonna have to raise the money to, you know, meet these obligations they have. So It'll be interesting. So but yeah. So anyways, well hopeful I don't I don't I don't know when the IPO date is. I I but they just filed last week. Yeah and it was early last week. So I honestly I would say there's probably gonna be some time with the review and whatnot. I know it's this month, but it'll be interesting to see. I guess where I was gonna go. Hopefully I'm not talking about enthropy again next week, but we probably will be. We'll or something related to it, but yeah. So Anyhow, let's let's turn around. What do you got that what do you going on? it's a similar trend. And the quarterly earnings came out with a lot of the bigger firms. Yeah. And I think one of the things that really was interesting was Alphabet took a hit last week too. And it was related to the earnings. Now, I want to caveat it. Alphabet is the owner or the pr the parent company of Google. Yeah. Alphabet itself obviously houses the Google search, which has been a topic of interest, especially with how AI has changed the way people interact. And, Google basically dominated that space for years, right? So it's changing, right? It's evolving. They know that. but on October twenty ninth, one of the biggest companies basically on Earth at this point reportedly bet had the worst single day on track for market history. So this, you know, the reason it's spending too much in AI, but is that I mean, we just talked about it anthropic. If we think about it, they're also spending a lot. So what is what's underneath? Like what what's really housing this? It's, you know, they did basically double their s expenditure from like two twenty-four billion to like fifty-one point four billion, which is insane if you think about it in numbers purely. Yeah. But one of the things that's interesting is Alphabet's not shrinking necessarily either. So if we think about it scaling, right? Like you're you have to spend more to make more. I mean that's a pretty typical risk that you have to take on. But their quarterly revenue was 127.6 billion up overall 25%. So We have to look at the numbers underneath, and I think that gives you a good feel for news likes to extract what they want you to see to be able to get that reaction. Yeah. You gotta take a step back and realize these companies have been doing this for decades and they're not they're not gonna go under because they had a bad earnings quarter. They're gonna say, here's an adjustment we're gonna make. Totally. So the biggest thing that I wanted to bring up is since nineteen ninety-nine in their peak, right? You know, they they closed October thirtieth. definitely way below where they should have been. But if we think about it, they lost a bunch of cash, but they still have cash on hand. So if they're almost it's that whole term and again, some of you may know this others don't almost too big to fail to a certain degree. Totally. They can buy their way out of this massive thing. Exactly. But I think investors love that, but I and honestly be in that position than anthropic. Right. Yeah. But like that's the thing though is I I wanted to kind of put this in perspective. And I think if you want to know more about this situation, I would honestly look at your retirement account. How many of the accounts are leveling into your retirement that are AI right now? Yeah. Because there's a lot out there. Even ones that are not talked about are out there right now. And you just go, what is this? Yeah. I went through my portfolio. I don't know if I told you about this. I went through and I took my portfolio and I and I asked AI, can you take my portfolio and assess what this the impact on my portfolio would have been as though this were the dot com 2000. and it ran through the analysis on obviously a lot of a lot of stuff in my portfolio did not exist back then. Yeah. I get that. So it's not perfect. But it kind of gave some ideas. It's like, well, if you stay on the trajectory you're on, you're probably gonna drop like forty, sixty percent somewhere in there. It's like, ooh, okay. It's which is in line and it probably did based on the industry exposures and all that, right? Right. So I started moving things around. I went back in after I moved things around. So now what it's exposure gonna be like it's about 20%. I'm like, all right, now we're feeling better. I'm feeling a little bit better. So but I'm probably missing out on the upside on a lot of these things too. I did NVIDIA at the beginning of the year and I it popped when I was like I was at 90 bucks. I I I bought bought in on the stock and then I think I sold it. I can't remember, it was maybe double that. So yeah, but yeah, it's definitely good to take a look at that because I think it's worth being exposed, but Yeah. for you and I at our age, it's like we can ride it out if you're sitting if you're relying on retirement. I don't know if that's a risky bet, but I I think it speaking of something that is consistent, and again, I'm talking the US terms in terms of the market and whatnot, the feds didn't increase nor did they decrease the rates inside the US. And that's a good indicator of the market is strong still. it's not trending either way at the moment. It's uncertain. I will give it that. Yeah. But we don't have anything that's indicating it's a it's a bubble pop, right? It's not usually feds are pretty like aware and they're pretty conservative when it comes to that stuff. So in my eyes, I'm seeing it as an opportunity to take that risk before it happens to say, Hey, what could be, right? But again, I'm not I'm not here to give you guys No. Well, hey, you know you know what's interesting is and on the Google thing, like part of what I was reading, there's a meme going around See if I get this right. It's the Google traffic went up for Reddit. So Reddit's like, screw that. I'm gonna we're gonna now force people to log in, so it won't be publicly available. But the problem is the Google traffic went up, but the amount of visitors actually coming from Google dropped because they're s they're crawling more. But I think it's kind of a double edged thing because and I think this is I think it's just a rumor. I was reading this. It could be true. I I I'm pretty sure it said rumor, but They're talking about that, when you search on Google gives you the AI motor off the get-go. Yeah. They're talking about removing that as the default. Because their ad revenues are down. Right? So and again, traffic to your to people's websites are down. man. So it's like this combination of things going on with Google where which everyone, it's obvious why. Like that's a conversation people always have is that why would I go to Google and ask it a question like I used to when I just use Chat GPT and it gives me the answer. And I don't need to actually go to anything. it just gives me the data I need to know to actually understand. I don't need to go to the webpage. No, I I do because I'm like I'm paranoid about sources, so I'm always like going after that. It's it's our debate all over again, right? But most people don't No, they they trust it. Don't they they they trust it face value and that's the problem that we run into these days with big time big time misinformation, right? So Yeah. No, I it's it's gonna be interesting to see how this evolves and I think To be honest, these Q3 earnings I don't take very seriously whenever I look at them because typically it's a point in time. I think the evolution is gonna come as soon as the hardware catches up too, which is one of the biggest root one of the and a not to bring in another story, but an interesting point is the at the Apple new Apple CEO basically just said, We're not getting into the hype, we're focusing on the hardware. That's a that's a huge statement in this era right now. And I I would I'm I'm interested to watch that. They'll be interested to see what happens over the next eight, ten months of you doing next, right? How are you how are you supporting that? Well people get scared about that. And so yeah, it's it's true. The hype, the the FOMO, fear of missing out. I talk to business owners all the time and they're they're just they're they're scared, but they're they they're overcomplicating it. And I was just I was reading it, I think it was up here I think it's National Post in Canada. I don't think it's in the States, but they they said that they quoted Actually it's funny, Canadian newspaper if I quoted someone in the US, around ninety-nine percent of AI initiatives by enterprise companies actually bear no fruit. Yep. Like there's no they're not getting any impact out of it. Yeah. So which is crazy. Like it's it reminds me of CRM. In twenty ten, that the thing was sixty six, seventy percent of CRM implementations failed. It was poor execution, poor change management, whatever it might be. There you go. So it's I I feel like it's very similar. It's 'cause you hear how many people are like, Hey, we just bought a thousand cloud licenses and rolled them out. We said go use it and without any direction. Mm-hmm. another example is another guy came up to me the other day and he's like, Yeah, I hired this intern to come work in my accounting firm and I asked him to go and do stuff. And then he's like, I don't know if I'm gonna be able to keep him around and like well, why you why do you think that? Because you I don't really know what to tell them to do. I'm like, yeah, you you need you need a strategic layer on what are you actually trying to do? Just because AI is there doesn't actually change the fact that you have to have direction and leadership and and and strategy, right? So I mean, just to close that thought, I mean, that's why all the big firms are starting to become more aware of the need for that enablement. Yes. Right. And I think that's, you know, again, still developing story. I have no idea where that's going, but it's interesting to see the focus is starting to shift to the human side of the adoption. And you know, it it always has been there. It's just now I think it's more prevalent with where we're at right now. Totally. Totally. Yeah. Cool. Should we do some more fun stuff now? Let's do fun stuff. I want to do fun stuff. I just dragged everyone down like a doomsday again. no, we got some funny stories now, right? Yeah. Yeah. Let's do it. So yeah, my next my next story is on Just solving a problem that I I had to deal with. So we're we're looking to hire someone new for the Ligio team. And I'm helping the Ligio team out on the corporate services side. so I said, Yeah, you know what? Let me set something up. I'm helping our CEO out. And we did Indeed, posted the job post on Indeed, all that. And all these resumes are coming in. I'm like, I I don't really want to go through this. I don't really want to have someone go through it. but I'm like, you know what? I have an idea. So I have this Mac Mini. I might have taught that on previous episodes. My Mac Mini sitting in my office in Vancouver and hooked up to my phone. I have Hermes on that. That's where I run Hermes. I also have ChatGPT installed on there too. And I have the mobile connection set up. Mm-hmm. So I kind of alternate depending on what I'm doing. in this case I just use the chat GP chat GPT remote because the computer use function is fantastic on it. last I checked with Claude, it wasn't as good. But the chat GPT one does it does it perfectly. It literally mimics what a user does. So it gets over a lot of these different issues where the where where re web browsers can actually cause it's not using the chat GPT web browser. Like I literally log into the Mac mini through screen share, lock into Indeed. That's nice. Go to town. There we go. So I didn't share a password with it. I just logged into Indeed and let it do its thing. So what I asked it to do is to go through all these resumes. And actually classify them and store them in Monday.com. Okay. So typically what we do is we go into Monday, we go and download stuff, and then we take that and we throw that into Monday.com all manual. You just go back and forth, upload the document. So I went through and I created a prompt. I said, I would like you to go to Indeed, that's open on the browser. And then I want you the MCP connection, which is like an API for those. It's just the computers talking to computers. It's con it's connected securely. That's exactly the key. Everything's secure. And we're on enterprise accounts, so everything's secure. And we and then I asked it to on the Indeed webpage, go through that specific job post, collect those resumes, and then I want to take that information and put it inside of Monday, and then I want you to actually score the resumes. So go through the resumes and look at different characteristics. So I'm looking for things like how many how many years of experience do they have? Relevant to the job post. Not in general. Relevant to the job post. So going through getting it to go through all the jobs. Like, okay, this is like if they have a dishwasher on there, it's not really relevant to corporate services. So not gonna apply. if it's if they're not in Metro Vancouver, which is where the role's been posted, it doesn't even bring them into Monday. Okay, just skips it. and then basically it just loads it all in there, loads the resume in there. So now I have a nice board to kind of clean through. So what I also have it doing on the Indeed side is I have it marked maybe on all of them, the ones that I like. interesting. Yeah. And then ones that, like I said, are not Metro Vancouver, disqualify. Right. Just get rid of get them out there. so I'm starting to iterate on that. I was literally iterating it all the way to the airport today. That's where I added those criteria pieces. Nice. And I'll probably iterate on it more just because I'm like, yeah, there's instead of me, because I'll go through resume and be like, Yeah, you know what? I gotta do I should do this on all of them. Okay. Chat GPT, I'll go to Codex. It's through Codex, which hooks in the computer, go do your thing. And then it just messages me a little bit later. Okay, I'm all done. And then I go and check the board and everything's ready to go. So that's a good one. So it's great. Like it's you know, the flexibility of you know dealing with those mundane tasks. I was thinking about it like What did I used to do in the past for this kind of thing? Because I haven't had to do hiring like this in a long time. Yeah. And what we would do is one of our admin people, they're going through indeed one by one. Or if they use LinkedIn job posts, they're going through it's because tied to their profile unless they're using recruiter version. They're going through all of these resumes. And what do you typically do? You gotta go download it, take a look at it, read it. I don't find the LinkedIn. One is very reliable from a from an analysis perspective because they have their own AI tools. I don't find that works really well. Whereas this just lets me custom tailor it within our environment, secure environment, talent to you. Don't do this on a free you're dealing with personal information. Yeah, no, don't do that. Don't do that. I'll do that as a reminder. But it it saved me a ton of time 'cause it and I have that job running at noon every day, just picking up all the resumes. Nice. I could probably have a check at it all the time as they come in, but I'm Whatever, once a day, then I then I know in the afternoon I can take a look at it, right? So so that's my new HR solution. Nice. I have to go package it up and sell it now. Although I'm sure everybody else is probably doing it. I'm sure it's out there. But I I do want to ask a question. And it's interesting you bring up, you know, selections, right? And if we were to think back in the day when there were those filtering softwares out there, right? The HR software that would basically do what you just built, right? You didn't take too long to build it. Obviously a lot of code related. but there was that that lawsuit a long time ago around bias, right? Of that of that software of like they were filtering out people they should have never filtered out in the first place. That's a good point. Yeah. I'm curious, how did you approach and there's a term that comes to mind, and this was years ago I learned about this, and this was in the project management world, unconscious bias, right? You can't call it unconscious bias. That's a human trait, right? You you you have a you lean a certain way around certain circumstances. Yeah. It's almost a conscious bias with AI or any other tool that has been, instilled by humans. How how did you approach that with that with that skill? Well, I never really put anything like I guess you could easily say I know. Just give give me men or women on a or something. something violating human rights at that point, but or what or whatever the whatever the employment codes are, right? So human rights tribunals be all over that. Yeah. Not not that I I mean that's not me, but you could in theory do that. Yeah. Right. You could I guess you could instill bias if you wanted to. But I think the way I handled it is like the categories that I I'm primarily using are the I mentioned one of them years of experience, but another one is just How accurate is this job post against the resume that's been submitted? okay. So it's like on a percentage basis. So is it like 90% close? Is it 40% close? Yeah. now keep in mind my prompt was pretty basic. So the AI is assuming a lot. Sure. Right? It's assuming that like I didn't tell it well, what other than saying 100% means that they match the job post 100%. Like if I said anything. five years of experience in this or that, maybe that's probably what the logic is doing. The only way to really know that is to test it. And I mean I didn't test. I just wanted something quick and dirty to get going. Sure. because it filters it down. Cause we we've had like two hundred, two hundred and fifty resumes on this thing. So it's a lot easier to actually pare it down that way. And then I can man do my old manual way. I definitely biases kick in for sure. Like if I see a sloppy resume, immediately my brain goes, This person's disorganized. They might not be disorganized. So if you slant in a sloppy resume and you're actually a really highly organized person, I'm very disappointed in you. really, right? I mean, it the way in which the world's turning, right, with technology. Like there's no excuse not to have a good looking resume. In fact, I have a criteria on one of them is is actually on on how professional and creative the resume is. See it. But again, I gave it I didn't give it for really specific. It's more of a it's more of a little measurement, a balance. Yeah. A little balance of that point. I said if you give a basic standard resume template that you find in Microsoft Word is how I described it. That's zero percent. Yeah. If a hundred percent is you've added a photo, you have color, you have sidebars, you have highlights. I just kinda gave some basic examples. 'Cause to me that's you want to put your best foot forward, so why wouldn't you put together a great resume? Right. By the way, no one's hit 100%. Well, that and that should be the case, right? Which is okay. But now it lets me and it doesn't even mean I'm not gonna I hate to say it, it doesn't mean I'm not gonna look at the 0%. Right. But it does help me categorize and I will sort by these categories and then I'll start work that that's why I start chipping away. Right. And so it's a nice way to kind of manage data in a way. It's a different way. I've ever like in a database, I write SQL query and filter it all out. Right. Right. Whereas now I can just ask AI to Put a new ki query in there, put a new calculation in there. Nice. I'd literally have to write SQL to actually do that. Whereas just doing this thing on the fly. Exactly. So common language, right? Right. I I'll add to something to that. And I I appreciate that example because that's approachable. That's something people do. one of the other things I was thinking about, especially with unconscious bias, was, one judgment, right? Just just straight up judgment. What does that look like? Right. So we have actually inside this event we're doing here in Calgary, is a workshop. That we're gonna be building agents together. And I was thinking about it in terms of if it's gonna be a competition and mind you, not a heavy competition. We're not talking people going nuts with it and you know, it's almost like a little sticks in the middle and see who dukes it out. I mean, you know, maybe with AI, right? AI agents duke it. AI sticks. but like I wanted to level the playing field to those that didn't necessarily have the time, nor does their job aim towards experimenting and make sure that those that were experimenting are being guided well, right? So I was thinking about I was going, to level that playing field one, a standard has to be set. How do you set that standard? And that standard is in fact going to be a prompt that everybody's using underneath the organizational hub. And in anthropic, Claude is skill. In I think they're calling them skills in Gemini now too. I think I just saw that. And then in OpenAI, I believe it's like a project or skill or something, something along those lines. Anyway, long story short, when I was going down that rabbit hole, I realized, wait a second. If we are to choose a winner, why in the world are we the ones that are going to have the unconscious bias to what we want to see? Why don't we measure it towards what the market's looking for? So instead of it being a manager's getting in a room talking about the presentations, maybe we're skewed a little way to somebody because we've worked with them, or maybe it's they're talking the the language we want to hear or whatever, why not have an agent, judge an agent to judge an agent? Like carrying it down the line with Real numbers, right? Just like you did, where the scoring mechanism's five pillars. And I think that's something you gotta realize is there is a way to use AI for good. There isn't there isn't a best way necessarily yet. It's very experimental. But what you can do is it's that golden section of examples. So what I ended up doing was using a extract from what has impacted the market, not personal agents or personal assistants, deviated from that. I wanted something more focused. And what I found was there's a trend of like top five. scoring mechanisms. So I create a scoring rubric with that. And then now it's going to score it based upon how they deliver it, which is already standardized. So it doesn't matter. Do they have all the information we're asking for? And then excuse me. And then it goes into how are we going to score it? Cool. And so it's going to be fun because it's going to basically do the steps, but then it's going to explain it and audit itself. So what I want to get to is what you did is basically a version of what I did. Yeah. It's a standard we're setting now, which I think is going to Fabulous because we're getting away from the world of bias for bias' sake. It's not gonna be this, you know. And again, humans make good judgment, they make bad judgment. I've made a lot of bad choices, make good choices too. Yeah. It's just taking out in terms of keeping the playing field fair, maybe it's level setting it a bit too. And I I I think we do need to give it that. Well, it's in a way, it's like your superpower in a way, because it helps you make better decisions. Exactly. that's different than we are no different than the reason you know with me with the resumes. I mean, I feel more confident that we'll make a better hire because of that process. Yeah. Whereas if you I don't know if any of you had to go through 250 resumes before their listeners, but it's it's annoying and daunting and you're just you're exhausted. Like you can just sit there all day having to do it, right? Exhaustion plays into judgment. Right off the bat. Exactly. So you do that. That way you can spend valuable time on the actual things. And it's no different than anything. We might we'd much rather have time spent on value add than actually doing all this make work busy stuff that I just drives me bonkers, as you know. So Yeah. Cool. All right. Well you wanna the last fun one on the stories? Yeah. So this is this is a this will be a quick one, but it's a fun one. So going back to what I said before at the very beginning here, the the creepy alarm clock. And it it was it caught my eye because it one, it had a big name on it, Amazon. Not saying that's true. I have no idea what this product is, never seen it from the description. It's an alarm clock that sits next to your bed and listens to you breathe while you sleep. Yeah. So in the in this in the spirit of the how the Halloween just happened, right? on Friday, it was interesting to hear that. And I'm going, well, hold on a second. What are they actually doing? Right. And so evidently on December on December October twenty eighth, sorry, Amazon announced Echo Time. So that's what it's called. That's the product. And it cost two hundred dollars or hundred and ninety-nine ninety nine. It's creepy clock. Creepy clock. Why do I think frog? Does it look like a frog? Uh-oh. I'm picturing a frog now. Maybe because I have kids. Yeah, who knows? But anyway, anyway, it it's detecting different types of s sleep. So you the wearables, right? They're like a sleep tracker. Basically. Right. The sleep sensing is is optional. So you can turn it off. And that is something to be aware of. Okay. There is no camera. It's just listening, right? But the whole indication is that it's useful side of technology showing up in a very personal space. Right. Yeah. The sensor is not the scary part. The scary part is not knowing where the data goes. Right. So that's the that's the question people are asking. And I think it's supposed to help with the upside of real waking up in lighter sleep the rather than the deeper sleep. And I know in my watch I have the same kind of principle where it if it notices I the the internal clock starts getting me up, it's like, hey, it's time to get up instead of it being right at the same time because sometimes you're different places. But I think the the the biggest thing is, you know, putting a new vice in a place that is very private is most likely going to jar people a little bit. And I think it's a personal preference. I don't think it has anything to do with, you know, Is this gonna work? Is this something that people are gonna latch on to? It's a personal preference and it's always going to be. So just to put it put it to bed here, yes, you might be a little bit taken aback by it and going, I don't know. It's kinda like all those cameras out there capturing your face and identity and all that stuff. It's kinda like when Alexa first came out, to be honest. Exactly. It's always listening, right? That's the that's the funny term everybody says, even though it's kinda weren't they making jokes like on T V all the time? it's to say the key words and then they just S N L had a field day with That was SNL, that's right. Yeah, field day. But I don't know. Thoughts on that in terms of yeah, we talked a little bit around hardware last time a little bit, but well that reminds me of actually it's a funny thing that popped in my head. I'll I'll finish off with that but in a second. But I mean there's lots of wearable technologies for sleep. Like I I've tried aura ring, I've tried My wife uses Fitbit Air. She loves that. there's an eight sleep, the you know, the I did say that. Yeah. The beds. so I mean, people do care about that stuff. I mean, I know I like to track that information. I'm not sure. I I probably I pr I I mean you've piqued my curiosity. I think I'd like to learn more about it. Maybe I'll look it up afterwards, but 'cause how do they actually understand that what if you're like a really quiet sleeper? You know? or maybe you are I don't know, sleep apnea. I was gonna say sleep apnea is the one that came to mind. I was like, how does that track that? So it's like anyways, I'm fascinated to hear more about that. But I was gonna say th just to finish up the thought there before we get into segment was it reminded me of I don't maybe I frog stuck in my head, but there's this thing going on. I don't even remember where it came up on my computer or on one of my feeds. It's the the mini duck. I haven't heard of this one. It's a Dancing duck with AI built into it. Okay. Yes. Yes. You know what? It's so important, listeners, that I'm gonna pull up right now. Sidebar. Sidebar. It's a sidebar we're gonna do right now. I'm gonna share my screen. So if you wanna see this. Our producer's gonna kill us, by the way. Yes, she is. where is it? Okay, mini duck, let's go share. I have to see this. Okay, so Sorry, not mini. It's microduct. Micro duck. Okay. So I encourage you to look it up. But Micro Duck is literally this AI duck that learns and has an open source weight model in it. Okay. it's only 400 bucks. You can have you can have I mean it looks like a robot, but you can have your own micro duc. So if you want to check this out, it's pollen dash robotics, P-O-L-L-E-N-Robotics dot com slash microduct. Or just go Google Micro Duck and skip the AI mode, of course. but yeah, it looks pretty neat. It's Okay. So we we have a a sidebar story to throw on there for everybody. Yeah. But I thought that was pretty cool. That is interesting. Yeah. So anyhow. All right. So anyway, should we jump jump in the side? I think I think it's time. I think it's time we jump in. I think we should do that. So let's pull up my screen here. Let's jump into replets. So listeners, today what we're gonna do is walk through an AI tool that I I've I've I've known about for a while. but I started using it after I went to Saster back in May. they were one of the the the the groups that were sponsoring there. And it's amazing. It's it's it's hooked into your most popular models and you can actually bring your own keys, your own API keys and all that. so if you want to use Claude or open AI on your own subscription you can. But it's like an orchestrator and it's really like a DYI a DIY system. You can go to Cloud and build artifacts and all this. You can generate code and whatnot. What I love about Replet is you can describe an application and it'll build it out for you. And then you can use it. Like I've read about some people that have gone in here and they've actually created full SaaS applications and they've a whole with payment systems and everything and and released it out to the world. Okay. So it's pretty cool. So so Mike, what I need from you is a prompt. we should try and be careful because if it's a really complex prompt, it could take a while for this thing to run. Okay. but we're gonna do is just like you'd be inside of Chat GPT or Cloud, it's asking us what are we working on? We're inside of Eligio's workspace, we're using the Ligio's count here, try this out. but let's give it a go. Let's let's start chat. Let's describe something we wanted to do for us. All right. So I have an idea for you. Okay. It's related to the sidebar you just had. What if we were an e commerce store? Yeah. That wants to create a customer facing, let's call it a website, just for the lack of better terms. That's a third party seller of MicroDuck. Let's create a third let's create a website as a reseller for the product. Micro. There you We're just gonna we're just gonna literally let's keep it simple. I I'm actually curious how much I assume how much it assumes. Yeah. Because that that's always interesting with these models. So I'm just kinda looking at I haven't been inside of Replit for a while, but as I'm looking on here, so on the screen it says Derek, what are we working on today? I've written, let's create a website as a reseller for the product microduct. Love it. And then along with the chat box I can add attachments. So I I would assume if I have like yeah, you can add a Figma design, which is I I know. We I I know some web designers have done that. You can use skills that you might have. any other integrations you have. I don't have any listed in there. then it has beside it output. Like what do I want to output? so I can actually Look at that. So I can leave it auto select, we're sh we're gonna leave this as bare bones as well. Let's do it for the experiment. But that's interesting website, mobile. You can do a design, a slide deck, animation, data visualization. We'll do dancing ducks later. Dancing ducks and frogs, because I have frogs in my head. So you can do that, you can choose what output you can have, what you want it to create. That's awesome. And then on the it looks like agent modes you have, so you can do the free one. There's no you know, lightweight models. Yep. auto power, so balance of speed and quality, best for most builds. again, we'll we'll use the free one. we're on a plan, so I don't know why it says free, but whatever, we'll see what happens. So and then it just, you know, I can even speak to it just like I would inside of Chat GPT and all that. Yep. But we're just gonna hit the start button and see what happens. So let's click on this. So what's happening now? It's saying starting. It's thinking probably. Probably talking to all the tools. Now what what it's done right off the get-go here is ask it's telling me you can find tasks on a project. So it's starting to showing features that it's pulling from. Exactly. Yeah, that's interesting. So first here, here we go. I'll build a reseller storefront for MicroDuck. I need a little more detail to make the product page accurate happens. So it didn't assume. No, that's that's interesting. Share a product link or short description. You know what? I'm gonna go grab that link of microdeck. We're gonna go toss that in there. Okay, continue. how should customers order? So my options here, I can contact order, I can link to store online checkout. let's just keep a link to store for the sake of we could probably looks like we could build a whole online checkout. I feel like we might make this the longest episode ever. So stick with Let's just let's keep that. So let's see what happens. Okay, now it's thinking, yeah, now we've given us answers. So that that reminded me of how a lot of the models are handling the questions now where they come up with the clarifying questions or anything. Which is great. I love it. I I I think it's important too. Yeah. Okay. so it's it's the screen's actually shifted over now. So it's it's I have the chat window on the left, and then on the right hand side, it's actually starting to generate the the application. So we're just gonna deny. so now it's because I'm using a free mode though you can see it's already it's already trying to ask me to add credits. Add credits here or some or earn credits. It's trying to get me to sell something here, but let's quickly look at the chat here. So I'll build so it's just confirming I'll build a standalone product a standalone microduct reseller storefront using the product information from PullMe Robotics with purcham butt purchase buttons linking to the product's official store page. So let's check in the source details and then create and preview the site. So starting the preview on the right. So it's giving us an analysis here too. So the product page describes a twenty five centimeter open source robot that can be retrained and links to a three ninety nine pre order. I'm shaping the site as an independent storefront that sends the order to the official store rather than claiming to process the purchase. I like that. Okay. All right. So it's you know there's some assumptions. I mean, we didn't give it much to go off of, so that makes it complete. You know what reminds me of is thinking of common sense is not so common. It sounds like there's common sense in this AI. Yeah. Yeah. it's interesting, going back to when we did the hugging face model of that PTO request form, right? Yeah. You know, that that's much more common practice and we we ask to be external facing and that's interesting. It's asking a question. I'd be curious if it does the same thing because it's external facing. That'd be interesting to see. So it's still building right now, listeners, as we're going through this. it's it's pretty cool. You know, I think I might joked with you at one point that the mic I'm gonna end up with a micro duck this week. and we'll bring it on one of these episodes. There we go. There we go. But Micro Duck, if you're listening, we can use a sponsor. but you can sponsor the Micro Duck segment. there we go. But I thought about like you go on Alibaba, like the on Chinese sites there, and they have tons of robots. Like I was thinking, wouldn't it be cool to have like an AI dog? And then I saw this thing pop up. duck wasn't the first thing that came to mind, but it looks fun and cute. I saw one thing, and again, you can only take what you take from the highlights, but it's interesting that the camera itself is f it's like a following, the drones, a while back where they follow them around with the unit stuff. But curious to see how they how they process like what are they trying to produce there? that'd be that'd be fascinating. Yep. So as we're waiting here, I'm starting to load up the the site now. it's already done it. Look at this. Holy cow. That's clean. So it's definitely taken an influence from I was gonna say I just noticed that the colors, the you wanna bet it's been using Claude? a hundred percent. Right there. Yeah. So listeners, as we're looking at this, if you haven't figured this out, Claude has a distinctive design that it produces. Usually it looks like a subheader. It's the generic design. Yeah, exactly. And certainly it's on here. so it's come up with a website, meet your starter kit. I'm sure we can give it more prompts to iterate on this. So it's just created a whole web page. And I think that's great. I'm just gonna stop there. Actually, let's see what it's doing here. So so it's giving me some suggested next step step tasks now that it's kind of done the initial version here. But so I can reseller checkout, I can add more stuff. But you know what? Let's just publish this website. See what it does. So we're gonna hit publish. I have no idea what's going to publish. Doesn't mean we have a live site. If we start getting orders, hopefully we better get commissioned. So So I hit the I hit the publish button. And now what it's done on the right hand side, it looks like it's gone into a section called tools, and it's actually showing us the progress of the building. So it's calling the tool that it needs in order to accomplish what it's setting out to do. Now, what's interesting though is we're publishing what is the URL they're going to be latching on to. That's interesting. we're getting a subdomain, I bet you anything. okay. you what? It's even right there. So let's have a look here. And you know what, listeners? Just for the sake of experimentation. it's not live yet. But what we'll do I can't guarantee down the road when you if you're listening to this later on, if this website's still around, but I'll we'll get Katherine to to publish the URL. Why not? on the in the podcast notes. And you can own another micro duck. Yes, don't forget the you know, Mike and Derek sent you type of thing, right? So Hardly artificial slash microduck. Right? So the so as I pull the URL up, which is microduck dash marketplace dot replit dot app. Okay. It says the app isn't live yet, but it is it's certainly going through it, it's promoting it. There you go. there's lots of cool things in here as I'm looking at you can get technical. It shows you all of the the the the the log files. So you can kind of see if you're more technical, you want to see what's going on, you can get in there. Yeah. I wonder how they handle a lot of the Update stuff too, that would be interesting to know. Yeah. Is just like with PHP we talked about last week. That right there caught my eye. It's deep security scan. Just what's been happening with PHP these last couple weeks. So listeners, Mike just pointed out that run a deep security scan. So you can actually run a scan on your tool with an agent. So that's great. that's yeah. I often think now if I was gonna go build a SAS, I'd probably build on something like versal or replit. Or level. it's it's seems to why why wouldn't you? Even if even though I have the capability and knowledge how to code, like the thing that crosses my mind these days is the whole security side. Right. Having to a great quote I saw from someone on one of my feeds today was that I don't really want to maintain my own servers because hacking happens within hours. I I he does this person didn't want to be in bed, wake up the next morning to find out everything got destroyed. Right. So Yep, exactly. So this thing's still publishing. I think again we'll put in the show notes the URL. You can go check it out, just see what Micro Duck is and hopefully get our commission check. but you know there and actually before I actually I'm gonna stop before I close out because I just wanna point out inside of the tools tab inside of Replit, like there's things like publishing domains. I can go on looks like yeah, I can configure my own domain. So again, if I want to do my own SAS. I'll look at that. They even let you buy your own. They even suggest quackly.net. They even suggest the domain names that we can buy. So that's sold. They got monitoring. So growth, database management. Like that's that's pretty cool. But the database piece, I can go in there. So I can looks like I can access my database. And it says you don't have any tables in here, but that's probably because we only have a database. But I'm sure if we had a customer database in here, awesome. Yep. Users authentication, let people log into your app. So this is pretty awesome. I can see a lot of value in something like this to set up your own app in your own little app business if you have an idea. So I think that's cool. I think it's great. and you know what? As I kept talking, as just in time. So listen, we're just gonna go and pull up the website now. We're now on mic so we're on microduct dash marketplace dot repli.app. And there you have it. even animated. I like that. Did it it did. Look at that. There's some animation on there. With a a caricature of the robot or the micro duck. so I think that's pretty wild. I'm sure we could play with a little bit with this, but maybe we should get into e commerce next. That could that could be our next life. I bet you we can go in there and ask and hook it into Spotify. It'd be it'll be a piece of cake for us. So that's cool. So there you go. so that that's it for this week's segment of prompt roulette. sidebar, sidebar. Sidebar, sidebar, as usual. But I think I think that's cool. I think it's great. Some of the stuff you can build. my brain's already erasing through here 'cause I haven't looked at Replet in a while and it's so much in there. my gosh. So much. It's lots to learn, but we should probably bring this to a close just looking at the time. yeah, lots lots of interesting stuff. So I think next week, we'll see what's gonna happen with midterms and all that. I'm sure that's gonna affect markets, so that'll it'll affect everything. And then I I wonder too how many people or how many businesses are holding off till after the midterms to kinda make announcements and stuff too. Typically you never know what the the governments will do too. not just US based, like even I mean US affects everybody, right? So everybody affects everybody. Yeah. It's it's all connected. So it'll be interesting what's happening. So yeah. There you go. But anyways, we're in Calgary, of course, as we mentioned. love to get some good meetings here, get to see the team. Yeah. until next time. Until next time. Appreciate you guys listening and you know, just keep an eye out of the market, you know, just Always look at the data underneath. I just that's the one thing if I can leave you guys with. Look at the data underneath. And try out some AI tools. Yeah. That was fun. We had fun. We had fun today. We did. All right, everyone. Have a good one and we'll talk to you next week. Thanks again.