AI Hype, Broken Tools, and What Actually Works

**[00:00:00]** Hello everyone, my name is Armando Prescareno and welcome to the web talk show. Today we're joined by the legendary Mike Carlo. You might know him from PowerBI tips, Carlo Consulting or the explicit measures podcast. Love the name by the way. But in all seriousness, welcome Mike. It's a pleasure to have you. Well, thank you very much. And you you're way over speaking uh how great I am on things. I'm just a normal guy doing a lot of things around data. Uh so yeah, those are those are the social media places that I play on. And I I'm a I'm a big data nut. I love data. Uh, and I've been huge in the data warehousing. Microsoft PowerBI is the place that I play. A lot of the the output that we do is from that space. And now we have these new things. Microsoft fabric. We're looking at a lot more data warehousing and what does this look like for organizations. Now we've been exploring app building. And so Armando, you have a number of great topics around just things on the web that people need to pay attention to around web building experiences. And this is exciting. I've been following you for a while now since you started. um we've been working together on some projects together in the past. So this is wonderful to be on the on the podcast and talking to you. Oh thank you Mike it's pleasure and you're very humble but really people Mike is like I know his stuff. So thank you. Uh let's just get like brief preface of how you got into because I really want to know how you got into all the BI space. Yeah this is great question. So my background is mechanical **[00:02:00]** engineering. I came from the mechanical engineering space. That was uh my bread and butter. I built and designed products for different companies and I got I kind of landed in a call center job. So in a call center job, you have a lot of calls coming in. You have to funnel a lot of data together. You're basically grooming a lot of information to to dole that out to individuals. So again, being a manager of a call center, I got all the hard questions. Uh so I was the one at the top. The buck stops at me and when someone's really angry, I had to deal with a lot of angry customers uh to figure things out. But that kind of pushed me into a career around data and analytics. And so I I quickly attached myself to the analytics team for their sales and marketing department. Sales and marketing needs a lot of data. We were in a business industry that was doing a ton of comparison between different industries, category management, a lot of data pieces there. And I just like this is what I want to do. Uh and at that time I I made the decision I'd want to be moving from mechanical engineering to data science. I said data science is what I want to become a data scientist and I started taking paths to go down that route. So uh went out got my MBA uh then thought that okay that was good now I know the business side of things and then I continued on to get a masters in data science but I never finished. I got a half a masters in data science. I got through all the hard stuff and then I decided after going through **[00:04:00]** that experience I'm like okay this is cool. I really like it. I think I can make a business out of this. And so actually I left corporate America. uh I joined a small consulting firm to only help them build PowerBI practice uh just around PowerBI and report building and that's where I really started you when you go when you move from working in the corporate space to working as a consultant I really feel your knowledge level has to really increase quickly uh in in business world you experience one or two problems fairly frequently and you solve those problems and move on but when you're in the consulting space I feel like the amount of issues and problems that you experience is amplified. You're you're dealing with many different clients. You have to pivot very quickly. And that just fit my my brain, my culture, the way I like to work. I liked having a lot of variety of problems. And so then after 2 years working at a consulting firm, I said, "Look, I can do this on my own." I found a client and I went off and I started my own business. And so since 2019, I've been running my own um consulting firm. And now we're moving our consulting firm uh from 100% revenue in consulting to now a portion of our revenue is consulting. and our portion of our revenue is now app development. So now we're trying to figure out all the things we've learned in the last, you know, the last number of years I've been consulting, let's say seven or eight years with consulting, I'm now turning those into products that we can go sell to our customers to solve problems faster than it would be if I **[00:06:00]** was just a consultant showing up and building things. So that's kind of been my progression. We're now a software development company that also does consulting. Um, I'm heavily involved in the PowerBI community. You'll find me on LinkedIn. And I have a podcast, the explicit measures podcast, which is a it's a direct play on words around PowerBI. There's this thing called an explicit measure that you build inside PowerBI. And so, uh, it's it's interesting. You have to get the whole phrase in when you look for us on Spotify because if you don't, you get some really interesting podcasts that are out there. So, Explicit Measures podcast is out there and it's super fun. So that's kind of me in a nutshell like how I got to the data and analytics space and now I do pure consulting and apps development for companies who need data solutions. Nice. That's amazing. And the part of I think there's a very important point that you touched on early at the the onset which is when you move into the consulting space and now you're working with multiple customers. Yes. It's a whole different ball game like you were saying and a lot of people don't don't see this if you're not in that space and not everyone has either the skill set or or the joy or love to be working on multiple types of different yes problems not just different problems just multiple variations of them and so I think you have to sort of have a knack for that don't you think yes I would totally agree with that and I think that's something I've I'll give Another example here. When I was in the mechanical engineering space and I was working with my team, my managers **[00:08:00]** and my other employees always said, "Michael, you're such an energetic engineer." And I didn't really was that a slight was that a compliment? I couldn't quite tell. So I I think there's this idea of being very personable with people and being able to like very quickly I have this knack of being able to listen to someone have empathy for their situation and can rearticulate the problem in a way that I understand what the challenges that we're trying to solve is. And I have this ability to kind of like step back a level. Sometimes as as you especially when you walk in as a consultant companies focus on a very specific problem. This is the thing I'm trying to solve. But when you when you bring an outsider in, you can actually take a step back a little bit and look at that and say, "Oh, what we're actually trying to solve is something bigger. There's a data culture problem. There's there's a you know, well, why do we have all this bad data coming in?" Well, maybe there's a problem upstream that we should be addressing and focusing on instead of just trying to solve the problem after the bad data comes to us. Like, there's all these kind of other observations that come along with this. And I think that perspective, the ability to do that kind of pulls you up. And so you have to be I I think in consulting space you have to be really personable. You have to be one that someone can easily talk to you and trust you immediately uh and have build that rapport very quickly. But then once you're done there, you need to deliver real value. And I can't I've been in with firms. I've **[00:10:00]** hired firms that have done consulting things. They promise all this stuff and they don't really deliver anything. And I feel like I've just wasted my money. I don't really get anywhere. So, I'm trying to make sure that our company is not like that. And and I I want to just be um I'm going to use all the buzz words here, a force multiplier, right, of whatever you're doing on your company. We try to show up and not just be another uh person you pay, but actually come in and deliver real value. Let's let's reduce time to produce data. Let's make it simpler. Let's remove these other things that are noisy and don't add value to our actual process. How can we simplify what we're doing with our data or our apps for that matter? So true. How do you find now that you're in the app and software development space clients responding to your way of doing things which is a lot of communication and empathy and that's not typical from a software development firm? So do you get any feedback in that sense? Oh yes um definitely a lot there. So I think that's a that is a a learned skill to be able to do that but I think what that does is it provides a lot of trust from the customer side. So I think you know you hear companies and and when you're in the consulting space you hear this word term a lot trying to be your trusted adviser trying to be this partner with you these are the buzzwords but in my experience no one really is actually able to do it or maybe there's very granted there's probably are companies that are doing this but I found it's **[00:12:00]** been few and far between to where we can actually add value to these companies and organizations. So again, I think I think that empathy comes across as a look, I'm going to do the right thing regardless. Um I I do a lot of YouTube short watching and I I watch a lot of people like Alex Hermoszi and a lot of people that talk about business and things and one of the things I can continually hear over and over again is whenever you're running something, it's integrity. 100% it's integrity, right? Um you know, I' I've I've been able to work with customers and meet them where they're at. And you know, I've I've b I've bent on some things. I've given away a couple hours for free here and there because the customer needed it and it needed to be done right and and I they didn't ask for it but I just did it right. And I think that's that integrity part. If you can convey that integrity back to them through how you're speaking and how you're interacting with them um that goes a long way. Again, it's a progression. I'm still learning. I'm a I'm a young business owner here. Uh I'm not a prof, you know, I'm a professional, but I I'm still adapting, learning the industry, trying to do the best we can here as well. And you know with consulting it kind of es and flows. Um the money comes in quick and fast in consulting uh when you do that and then there's long periods of time where there's no consulting and it gets very light and thin. Uh and so that's where we're trying to strategically as a business move a bit more towards a mix of app **[00:14:00]** you know licensing and software in addition to um what comes out of the consulting space which I have found when you build apps they are quite uh time and in like capital expensive to get them going right it there's definitely some investment that has to be happening there to really get them off the ground but you you pay dividends on uh in the future. So the work you do today provides results for tomorrow. So it's like the years of work that I've been building these apps and softwares and developing now they're becoming products that people are purchasing and now the ramp up is occurring. Now people are finding the value. So it takes time to really like land a software piece uh and actually make real money from it. And this is where we may get there in some of this conversation. But again, my YouTube feed is just littered with AI everything. And I I don't know about you, Armando, but every other day there's like I made this app and it went like $2,000 a month and all I did was talk to an AI and I'm like, dude, I am so skeptical of these things. I It'll either last for 3 months, make some money, and then die. Like, I don't I don't understand. Like, how are these It feels like It feels like it's all being made up. Like, it doesn't feel real to me based on my actual experience of like building apps and software. And maybe I'm missing something. Maybe I'm not picking like the fun little goofy app that people need to build something with, but at this point I'm like I don't see how these apps can just be like landing really big revenue very quickly by **[00:16:00]** just using AI to help build them. Now I do have some opinions about where AI is going to help us build apps in the future, but for now I I think it's a little overstated in my opinion. Totally agree. We had that same conversation yesterday, my wife and I, of something basic like thumbnail. You know how hard it is to make a YouTube thumbnail, right? Because cuz there's if people you're listening, you have never tried to put a video and like you create a thumbnail. It's very important. Very important. And extremely important. It's like the it's like the thing. It's like the one thing that you got to get right. Like in shorts, it's the hook is the thing, right? But in in like regular videos, the thumbnail is just just everything. And so creating a good one is hard. And so everyone talks about AI. It's doing everything. So and so you use these tools and my wife was like I I mean I told the AI and I still had to edit a whole bunch to get it to look like how I wanted to. So what's the point? Like what are they why do they say it's so easy? And and so that's just a very basic example. But it's the same with the lovable and replet and all these like cursor etc. Yeah. We use them a ton and we might get into like the details in the nitty-gritty. They're great, but do they work for everything and for everyone as fast as people are claiming? No, not really. I think it's a lot like you said, the gimmicky little thing that everyone needed. Um, and and there's a space for that, right? There's this like that's why these kids are **[00:18:00]** building these sort of million-dollar businesses on on nothing because they don't care. They don't wait. They don't plan. It's like I'm just going to do it and they just do it and it's that's great like keep doing that right we lose that as we grow a lot and so you do that you don't care you try you fail you don't care you try you fail and then suddenly something pops right now to your point yes how software works we don't know how those are going to last right so they might do very well and then they might crash and burn because there's no stability nobody knows how it was built build, there's no QA, there's like nobody understands the code and so that it could never happen. It could just be well forever. But there's that part that you've experienced with building I think and again I'm I'm very much in the Microsoft stack. So I'll talk a little bit about the Microsoft stack from my perspective, right? I think Microsoft landed Power Apps. Power Apps is like their version of like, hey, we're going to build an application for you. We're going to be a user interface. You can kind of drag and drop fields and make this app very quickly. I think they had the right idea with Power Apps. Power Apps is a good way of building an application on top of a database and it just kind of works nice, but I think they missed the mark on how people were going to build it. Like they they Microsoft invested a ton of time and money on like building the UI and so it's it's low code. It's not really no code. But but now with the AIS and what they **[00:20:00]** can do, you can sit down and again I've done this like this is this is where it floores my mind and it's I'm not sure if I'm I'm it's probably a two-part thing. The part one is I'm still trying to figure out how to use the AI effectively. So I'm learning and adapting myself to like figure out what's the best way to use the AI. And then on the flip side, the AI is getting better all the time. And so I find I've I've done a really I've had really good success. Simple example. I was I had a website that I've already built. I was trying to add Google Analytics to it. So I was I went I'm oh I'll try the AI. We'll see if it can just build me a little component and we'll just whip it out. So we go to the the uh I was using cursor at the time. Use cursor brought it up said hey I need to add Google Analytics to this project. It said let me read your project. Rent through read my project. Did some things boop boop boop. Okay. I've added a component. And so then I was testing it and I'm like, eh, it doesn't feel like it's it's sending the Google Analytics, but it's like waiting a long time before the page actually sends the note. I'm like, doesn't feel like it's implemented, right? So, I didn't know how to debug it because again, I'm I'm not the developer. I'm I'm a one of the leaders of the company. So, I kind like just play around these things trying to figure stuff out. I could read code, but I'm not building every little detail of it. But then I I said, "Okay, well, **[00:22:00]** I've spent about four hours trying to debug and fix." I'm like, "This is not helpful." And then I decided to say, "Look, I'm just going to start over." And so I built a I literally prompted from a scratch project. Just said, "Build me a documentation site that has these features in it and has this left nav and this right nav. Oh, and by the way, make sure it has Google Analytics." And I I built the entire thing from scratch over again in the same amount of time it took me just to debug and add it the first time. And so it this is so my my aha moment here kind of was wow I think we're going to be able to get to a place where and I I liken this to like Star Trek a little bit when Star Trek for those who are super nerds and watch that as a kid growing up uh you'd have like this computer in the spaceship and you would say hey computer and you would just tell it to do something right I think the AI is going to get so good it's going to be like that kind of experience but instead of us just building like a question and answer kind of thing the computer will be able to develop a piece of software like literally write an application for a single question that you're asking like again I'm thinking like you know a question something like hey look I want to build this u hey I need to get a map or directions to this thing whatever the thing is right well now the AI can just build you like a 3D generated map it's using this data it's then building these things it **[00:24:00]** has like a whole UI and interface like you can get this really weird like here's the answer to your question but you've built an immersive app on top of the question and then because it's so cheap and easy for the AI to build hundreds if not thousands of lines of code, you just throw it away when you're done. Like so the app itself becomes a commodity. And so again, I go back to the Star Trek analogy. Like I'm thinking like the AI is going to physically change how we build things because the the AI is going to build entire applications for a prompt or a question and it's going to be useful in the moment, but it's going to be so easy to create a new one. You're just going to throw it out when you're done and go build something else for the next question. Like I think that's what's going to happen. I like that. I really like that. And it's it's not far off. remember Magnus all all the noise that it made sorry so so I got access to Magus out of the blue pretty quickly and I was just went in and it's a little weird now now it's a little better but it was capable and one of the things it had was like oh I need to analyze I don't know the population of this thing and instead of giving you data or a chart it actually built a small web app that you could browse around and it shows you all the graphics and everything and it was like oh Right. So cool. It's like this explore. It's like this exploration. It's like a little mini app and things. So, you know, when when do we get **[00:26:00]** to a point where my phone doesn't have a whole bunch of apps on it, right? I can like again I'm I'm thinking about all like these gimmicky games and people like what's going on here? Like these things like YouTube's got it too. Like if you scroll through YouTube, you can play these little mini games. Like these mini games are simple. like AI can now generate 3D worlds and have all the so why not talk to an AI and say like look I've got 10 minutes to burn build me a game that does something like Pac-Man but 3D version and just say and then it just spits out what you want and then now you're interacting with this thing for like 10 20 minutes okay I've entertained you move on and you just throw it away and then you know later on you can come back and build something new again like so it the the attention economy which is a very big I think thing right now everyone's trying to kind of gain your attention. What happens when creating an entire app experience becomes the commodity and I think that's where I'm I'm I'm getting my moment right now where I'm like this there's an epiphany. There's something here with this where the more we bring down the price, the faster we can produce these applications. Um they're going to be like water. They're going to they're going to be everywhere. You can just build one whenever you want. Uh and then it'll build middle apps. You'll use it for a bit and you'll just throw it when you're done and you'll move on. That's very interesting. What do you think would happen to the whole like app store economy in that sense? Would it **[00:28:00]** just be you just have like a home screen, nothing in styles? You just basically run it when you need it and then just throw it away maybe. Honestly, I mean this to me there's like some level of like there's like, you know, AAA games where there's a lot of people doing like styling. there's a lot of like personality put into these games and things, but like you know, as the AI gets better at at being able to read and develop these things, I mean, why why can't we just ascribe you these larger, more elaborate apps to the AI in a way? So, yeah. So, what does this look like? You know, can someone build So, here's one here. I guess I'm speculating here a lot a bit on this one, but what about that whole arcade space that Apple has developed and built on their phones, right? What does that really need to be there anymore? Um, in the future, you know, if you have people that are, you know, what will probably be happening at some point, it would be cheaper to to build the app that lets you produce these things efficiently, very fast, let you hang on to them for a period of time. But what about, you know, Amanda, you found, hey, I've built this really cool game. It's very intriguing. It's around a topic like it's like a chess kind of game. I really think it's intriguing. Check it out. Here's my prompt. and you pass me the prompt and I just hit the prompt and I just run the prompt and my phone produces an app that's very similar and again we're using similar language but then it produces the app and then I could even tweak it per **[00:30:00]** my style like the scoring is off I don't like how the scoring works or you know I want more levels can you hey AI can you make another three levels on top of this game that would be interesting to me like it should be it's going to be like this super highly stylized customized experience that now we can play the same game and maybe I pass back to you. Hey Armando, here's the prompt that I use to make it build this cool level. Check it out. And then you run it in your LLM and it builds it for you. Like this that could be I mean again this is I don't know how far away we are from this but this could be the new world we live in but we're just passing around prompts to build what we want to build. Um and another one I'm I'm really excited about is uh when AI can be localized on your phone. So when the AI doesn't have to go to a cloud service to run, we can then run it on our devices. Um that will be I think another really big game changer because me as a software developer I could then the cost for me right now as a software developer whenever I build AI I have to pay for some machine to run to do the AI or I'm paying for a service to run the AI that I'm going to use in my application or or whatever when I can just build the IP the technology and the software and I can give it to every single iPhone user who has enough AI processing power on their device to just do Mhm. Great. All the secret sauce I provide now is **[00:32:00]** I just download this mega large model for a period of time and then the phone just uses it, right? And then I don't have to pay for all that extra expense. So that's going to drive the price again of creating more software, creating more AI things. It's going to drive it to the floor. It's going to be a race to the bottom again. Yeah, that makes perfect sense. Especially with these chips, the the new chips are just being designed for this type of this. Exactly right. So it'll make perfect sense for that. That's why you see a lot of people don't understand this. If you go from hardware, like if you have an experience for with hardware, traditional hardware like computers, right? Yeah. And then you see the phone and then you see some apps that doing ridiculous things on the phone, you're like, but how how is it? It's just because the architecture on these devices so different and optimized for these graphics and things like that. And so people code with very constrained specs, but at the same time, they're extremely performant. And so it's equivalent in a sense to having that huge GPU card on on the other machine just because of how how you build these things. And so yeah, I think that that would make and I don't think we're far off by the way. I think people are already doing that with some like little games like hey do me a version of Flappy Birds that does this thing and it'll it'll actually do it and it works. But yeah, when it gets to a point like you were saying I think modders will love it. the game matters. It has existed forever and so everyone will have like **[00:34:00]** their own version. So maybe some of these huge multiplayer massive most multiplayer games will be more like prompt prompt based mods. Yeah, I mean think think about that. I mean that would be really a kind of a unique idea of like you know the the you know have a central something structure to like a world or a 3D space and then allow your you know use a large language model to develop your character. use a large language model to develop like the weapons and the things that you have and you know within certain bound like that that could be create like that could be like a lot of these games I remember playing it was just it was always limited by the creator of the developers like the developers had were the limits for how creative you could be in any any kind of game. Well, now with without that, now it could be the creative audience, a collective audience of all your people. That was what drives your games. And games are different now. Like when I was a kid, I had to go buy like a $60 multiplayer game to go play with other people. Now the games are free. Like you just go get the game, it's for free, and it's all these extra like hats and uh you know, skins and that they make the money now on all the like accutrama that goes along with the the game. Why? Why why limit the ability to create all the different uh artifacts and game elements by the computer? Just have people program whatever they want and now they can create really unique items whatever they want now. And so now that's that's the feature you pay for. You pay for the **[00:36:00]** customization of the app as opposed to just the core of the app. They give that away for free. Who knows? I don't know. This is potentially a thing. Yeah. Why? I definitely see it being a thing, especially with these games that that are like you said, like all loot boxes and things like that. And a lot of these people make these and then even even just regular people make these, not not game devs, then they upload them and then people buy them and they make money off of these. So, it's great that if if they add these this sort of and maybe they they are already working on it. I I wouldn't expect them not to. um such that every like you and me could just go in and without knowing game development create our own version of the little helmet thing with the screen that turns around or or whatever. Right. Exactly right. Who knows? That's that's very interesting. I I think the space for AI coding and all that is just getting started. Yeah, we are. We're on the we're on the we haven't even hit the there's like the curve, right? There's like the elbow and the curve when things really start ramping up. Um I think you're right. That's a really good observation. And let me give you just a quick analogy. I know you run a bunch of websites. Um have you seen Let me just ask a question to you now. Have you seen the chatgpt.com become a data source for your website traffic yet? Are you starting to see that uptick? Yes. for for many of my clients that have a large knowledge base or yeah knowledge repository. We'll get into data in a bit. Yeah, we have **[00:38:00]** some clients that are consistently showing up in perplexity and GPT search and things like that and it's great and you see that in the analytics as well and and people are now optimizing. It's like the new SEO is sort of the LLM SEO. That's what I'm getting into now. So like so, so the reason I'm saying this is like I think there's there's a monumental shift happening right now and and it's I mean I can literally see it in the data. So we're going to talk some data stuff here. I'm going to get super nerdy here for a bit, but I've been looking at my website traffic. If you look back in like February of this year, zero traffic from chat GPT. Almost nothing. Nothing's there. Okay, we look at it now. I just checked it like the other day looking at website traffic cuz I run PowerBI.tips which is like a knowledge based website and so that's a that's a place where I drop a lot of like insights things I've learned tutorials all kinds of those things as well. Well, I have found something has happened cuz chat GPT has gotten a hold of like the knowledge that I've been putting out there and now I want to say it's almost tied with Google's traffic. I mean within within this amount of months it's been what four five months now like we're almost tied like the amount of traffic coming from the growth on that is like I'm seeing like 32% growth month over month of like people coming from that space. So we are hitting a moment where somehow it's gotten a hold of it and I don't know how it does this but all I know right now is for me **[00:40:00]** personally I am highly motivated again as a knowledge provider to the community and things that I do. I've got over 430 podcasts out there. None of that's transcribed. None of that's searchable on the web. It's just kind of stuck in videos and either, you know, Spotify or YouTube. Like, it's out there. But how do I take that 430 hours of knowledge and conversation and chatting and summary? How do I spit that out in a way that the AI can go leverage more of that and I can drive a ton more traffic? I'm just doing this like for me, the content creator. This is going to be kind of fun here cuz we're content creators. We're creating knowledge and experiences. We're sharing these things through a medium that is either LinkedIn or through video. But one thing I think is incredibly important here is how do we automate our way into taking that one piece of effort into three, four, five pieces of additional content. And I've done some experimenting with this. you know, we we get a lot of our video views or podcast views on um Apple Podcasts and also on Spotify. That's where the majority of our stuff comes from. So, in doing that, what I've been looking at is, okay, well, how do I We don't get a ton of views on YouTube. That's fine. It's just one outlet of our medium. Um, most people run with us or bike or do something like that on the podcast side of things. So, how do I how do I reformulate this into getting more views or getting more engagement? Well, doing shorts is another way of doing that. So, yes, I'm speaking for an hour, but I'm I'm sending out lots of little **[00:42:00]** nuggets of information, key data points that are interesting. Well, those should be bundled up and made into shorts, and that is the attention economy I can go get from the other side. So, you know, while my episode may get a,000 views on the actual full podcast, I can rack up another 2, 3, 4,000 more by producing a number of shorts from that same episode. So now it's becomes for me I'm looking at okay I'm doing this content generation thing once how can I use AI or automation to develop 30 40 different outputs of the same piece of content so right now I'm really exploring like the reusability of content and how to best automate you know I've done the video I've recorded it how can I as fast as I possibly can get that into a place where chat GPT AIs perplexity like all the different AIs can search what I'm saying. And now that you know, okay, Mike's a kind of an expert in PowerBI, people have found value from the links and the time they spent on his pages, how do how do how do we route more traffic to that knowledge base that we're already producing? Does that make sense? It does. And I'll get exactly into how you can do that in a second. And but I do want to touch for the audience. It's it's not I mean yes Mike has a lot of traffic and so some people might say well it's Mike right or it's this other big account. Yeah. I'll give you an example. We had this is a relatively new podcast and that's very little traffic. However, especially at the beginning we had a guest came on the show. Nothing technical at all. It was actually **[00:44:00]** uh investigative interviewing. Yeah. like the police do. And we were talking about the Tik Tok and that sort of things. And so so as an experiment, as a thought experiment for him, I said, "Let's repurpose a little a little part." So something he said that was very genuine, coming back to to what you were saying at the beginning. Yeah. Something he said, I grabbed a little snippet of that and I put it on TikTok and it got 6,000 views in a couple of days. and and it was just a little from an episode that lasted an hour, right? You were saying. So if if if anyone if everyone does this, then yeah, maybe they won't listen to the whole show, but as you're speaking for a long time, you're likely to say a few good things at least. And so, as they say, a blind squirrel eventually finds a nut. Yes. So, so if you generate those as shorts, then then that's fantastic. And we were talking yesterday about there's dscript, there's opus clip, even riverside does this for you and a lot of different tools will allow you to make these sort of shorter versions, analyze them, give you some some that that are likely to perform, right? Yes. Yes. But but to your point about how do I automate this whole thing because it starts getting pricey, right? So if someone says, well, I'm pushing out like you guys, you're doing two episodes every week at least, right? And so that's a lot of content. The time commitment if you do, but I'm also doubling down on it. Like it's like I'm already spending the time commitment on it. Like I want to maximize that investment of time. Like I don't want to **[00:46:00]** just be like, "Oh, I got a couple hundred views and I'm done." Like you want to maximize that investment of your time. If that's becomes your mark, like I I'm firmly a believer of if you have something you're doing, you have to be regularly producing something to really get some attention to what you're doing. And that then that becomes part of your salesunnel. Mhm. Yeah. Yeah. And and unfortunately some of these tools are credit based. So when you start getting into like long form content like this then you start hitting some of those credit limits and you eat them up really quickly. Like it's it's not like five credits to do a little thing. you have to like, okay, it's 150 credits to go like read the whole video, download it, junk it up, build like Yeah. And and this is this is where I'm a little bit torn because this is why you pay for some of these services because it makes sense, but you could be shilling out a couple hundred bucks a year, a couple hundred bucks a month like for these services. And again, it goes back to my investment, right? Am I if I spend that money doing these things, is it enough to push me forward to get the sales that I need for my business? So, it's really kind of like a a balancing act. You've got to really evaluate what's the tool doing? Is it adding enough value to what it the outcome should be producing for you, which is ultimately sales, knowledge, expertise in that industry, whatever you're working in. That way, people can trust what you're saying. And like, you know, I love I love being on on video calls and like, "Oh, wow. You **[00:48:00]** run a podcast?" Yes, they'll, you know, I've had one call where they're like, "Yes, we also run a podcast." I'm like, "Really? Tell me more about it." And they're like, "Yeah, we've got like around 75 episodes." I'm like, "Oh, wonderful." I didn't say how many I had, but I was like, "Dude, I've crushed I've crushed you on number of episodes." Like, uh, you know, if if we're going to be spurting metrics here around podcast, I'm pretty sure I've got you beat in that space. But it it it's it's now becoming part of like the the marketing, the salesunnel, the media uh space. you're you're starting to spend some time as a business becoming a little bit of a media company and again what we were talking about earlier with AI and building games like this is the same thing now like the AI is now allowing us to be able to take uh video content which is now becoming predominantly I think the main source of consumption of content now that that piece of content now is um able to be made easier cheaper faster from more people. So there's no excuse now. I mean with a with a a decent investment on a camera and a microphone and some time anyone can be producing media. And this is where I think YouTube for me and now a little bit LinkedIn as well is really interesting. I as the content creator the the person who's coming up with the ideas the the line of sight between me and a consumer is like nothing. There's no media company. I don't have to get a contract. I don't have to go through and ABC, NBC, Fox, like none of this. I can I can literally produce the **[00:50:00]** content myself, spit it out there, and now I'm directly in the hands of people who may want to talk or engage with my content or buy from my business. So that's that that cycle, that sales cycle has gotten extremely short from my opinion. It has. And there are ways to make it more economical if anyone's trying to do that same thing because I I think it makes perfect sense to repurpose the content. It's so important. And now remember there's no like, oh, I pushed it, that's it, like you were saying. No, no, no. I mean, if you push it now, those people saw it. You can upload it again and more people will see it. The same thing. Especially if you repurpose it and you put it on different platforms, then it's always more people. And we always get stuck on the, "Oh, but it only got 200 views." Okay, put 200 people in a room that want to listen to you, even if it's for one minute. Yes. 200 people for free. Yeah, for free. And all you had to do is turn on your camera and get going. Like, so yeah, we we um on our podcast and and when we do a lot of things like educational things, we're we're continually getting like hundreds of people now tuning in like so other perspective, right? I've also had to evaluate like where do I spend my time, right? So we would run a PowerBI local community user group in our area in Milwaukee and we'd get you know anywhere between like you know 10 to 15 people to show up and we would talk about PowerBI and data things and just generally build community and it was fun. People would travel you know **[00:52:00]** 20 30 minutes away we'd have a presentation we'd talk about some topics some interesting uh content and then we'd kind of open up for Q&A and you know discuss what challenges we're facing and how we how we're using this cool tool to help us build our businesses. Excellent. Well, what I found was during um the pandemic, we were able to like go online and do everything there. And that immediately pulled us out of this room with 15 people. Well, now we could reach hundreds of people in any time zone. And now the content's always out there. And then once I so I've quickly found that by moving my stuff more online. So we were do an inerson and an online event together that really started stepping up my game and my my impact my my space my influence grew greatly when we had that experience. Now I could put these content out on YouTube. I could put it out on LinkedIn. And so now across multiple platforms I'm getting hundreds if not thousands of views per episode. So that was like a huge moment like, "Oh my gosh, I'm spending so much time and effort doing this one little thing where I could 10x this just by moving it online." That's amazing. That's a very good point. And of course, the in-person events are fantastic and and there's different type of engagement, but it makes perfect sense, especially if you're trying to see how you're going to maximize your agree your reach, your whatever you're trying to do. if you're selling or if you're helping whatever you're doing it'll it'll and then you can have some inerson events that are even bigger perhaps because now you have a crowd or a community behind it which **[00:54:00]** is very interesting. Yes. Can I ask you a question around this a little bit? So I've had a couple experiences now because I've done a podcast I've had some episodes. I've been in conversations now especially and this is maybe kind of like the around the podcast and sharing candidly and very openly, right? So, I share what I know. I'm very happy to share what I know to to people in other organizations. I've had a number of conversations where people come to me now looking for work or or a project. Hey, I heard what you said on episode D. I heard what you said on this this thing. And I really resonated with that point. And what I'm finding is again as a business owner in this space, I'm looking at this going, "Wow, this is interesting." Because now before I had to go hunt for leads and go find people, but now people have been kind of following me on LinkedIn and maybe I've popped up a couple times or maybe a couple shorts have run through and they they thought, "Oh, that's a good point." Or, "I really disagree with that." If nothing else, it's just becoming part of their mind as that I'm around this space. And so now I'm getting calls with people where they're one paring back to me what I've said on podcasts and and it's been hard because I'm like okay what episode were you talking about? I need a search index on this cuz I don't know what I don't know what I said. So one that's that's part of my you know oh no I get nervous when when people do that. But on the other hand people have been forming a one-sided relationship with me potentially **[00:56:00]** through hundreds of episodes. I've got some people who followed who have followed us since episode one. And so we I have met people who who've been following this. And so I've never met them at all, yet they know a lot about me, my business, what I do, my perspectives on how to do business. Like there's a lot of um conveying of information that they already understand a lot about what's going on here. So I'm no longer trying to figure out and prove who I am. I'm not giving out my resume anymore. Like no one's asking me for what are your qualifications. just send them, hey, do you have 400 episodes of podcast? Have you talked about this topic for something? So that becomes now a part of your resume now. And I think that's this is changing the dynamics for me personally as I'm looking at around um this this mediabased experience that we're producing. Definitely. And that's something that Gary Vee talks about as well a lot. Oh yeah, as you said, I like Gary V stuff. He's good. They they I like them both because they are both very genuine and and like no nonsense and that that's it. It's if if you're putting out the content then people know who you are and exactly it's not it's a whole different dynamic. It's a completely different dynamic. And people even like if you're doing a local thing, if you have a local business that tailor to local crowds even more sometimes because people will then recognize you on the street. Like I remember this stuck with us for years because when I used to do all the Facebook ad stuff and I had the book and we used to do seminars and things **[00:58:00]** like that. I remember one time we went to one of like a little convenience store. We stop like quickly grabbing something and then someone stops me and like hey you're the Facebook ads guy right like what because they were seeing the ads on the phone constantly right and so you saw those videos and so it works if you're doing it in a local environment or even if you have a large enough audience then anywhere you go. So to your point, yeah, definitely helps out because now either they've heard of you because of your show and then they might just reach out or tell their boss or whatever like, "Oh, oh, PowerBI, talk to Mike. I mean, who else? That's the guy, right?" And then Exactly. Or if you do get to meet them by coincidence or chance, it's no longer, "Oh, so what do they do?" You can you can just say, "Oh, yeah. I run PowerBI tips or product consulting or the the explicit measures podcast." And so if they just look it up for a second, they'll be like, "Oh, yes. Oh, okay. So, and now you have that background." And what are you doing? You're just sharing what you learned and what you know. Yeah. And how valuable is that? I mean, how I mean, if if you're just doing if you're doing it for free, like again, this is what I think this is like an Alex Hermosi point, right? Um he said one someone was in again I love his stuff. He's he's got a ton of really good shorts. He has a lot of like thoughtful words and things and a lot of things that he does I subscribe to. Right. My basement here is the same environment. There's **[01:00:00]** no windows. Like I'm that's what he says he does. Like when he has to go do work and has to focus on things he's like in his zone. It's this thing. It's super set up. It's super comfortable. I know when I get here I got to go do work. And you know it's hard getting into the seat but once you get into the seat and I'm in the mode like we just go. It's go time, man. I know what I got to do. I build what I got to build. Like, we're just clipping along. So, from that regard, like I I subscribe a lot. But he was he was mentioning this one thing uh one time on on one of his episodes, and someone asked him like, "Look, Alex and Rosie, you you have you're you're about ready to leave this earth. You're going to you're going to pass away. So, what's your one kind of nugget here that you would give to other people?" And he basically said, "Don't listen to anyone else." It was kind of along this vibe. Don't listen to anyone else. do um do so work harder than everyone. He's a a huge work harder kind of guy. Work harder than everyone else and build so much volume that no one can argue with what you've produced, right? So just basically outproduce everyone else. Make your time as effective and efficient as you possibly can. And I feel like, you know, in some of these spaces here, just sitting down and cranking out it's it's routine. I mean, I've been doing the podcast for like 4 years. Doesn't it takes a bit of time to get to that space where you've been able to like keep it going along and **[01:02:00]** there were times where we look at it going, we're not really growing too much as a as a podcast, as a as a a media thing. Should we keep going? I'm like, I just enjoy it. Like, I'm You have to be able to sit into these things that you just love to do. And honestly, if we had no one listening to it, I'd probably still do it cuz I just enjoy talking about these aspects of data and and getting into the data side of things. So, I'd still do it. I I would still um people will approach you and they'll like, "Let's buy your business." But why? Like, yes, I could sell it. Maybe that would be an interesting opportunity and cash out a little bit, but what would I do after that? I would turn around. I would build the exact same things someplace else. So, like it I wouldn't stop. I'm I'm going to keep building the website. I'm going to keep doing podcasts. I'm going to keep learning about these data things because this is something that's interesting to me. Um, and I feel like as as I've grown older here, I I have younger people that come onto the business and like your 20s is just figure out what you want to do. Like just try as many different things as you can. Your 30s are get really good at the thing that you kind of think you're you're you're aligning to. You're getting really good at um, you know, working with others and, you know, your technology stack or your area of expertise. Get really good in the 30s. And in your 40s, I feel like it's like, okay, you've kind of done the experimentation thing. you're now in a **[01:04:00]** place where you're getting good at that one thing. Specialize now. So now I feel like I'm in this space where I'm just trying to super specialize. And I read my wife will yell at me and she she well not yell at me. She she laughs every time. Um we'll we'll be going to sleep and and she'll say, "What are you doing?" I'm like, "I'm watching a YouTube video on learning how to database." And she's like, "What are you doing?" It's like it's after work. Why are you watching more data videos like late at night? I'm like just what is it's just what interests me. Like I just I consume it all the time. Uh and now with YouTube and other things out there, you can do that a lot more than I probably should be healthfully. But again, I'm trying to pull the Alex from Rosie mode here is uh out volume everyone else. Just do it to the nth degree and become that expert. In order to do that, you have to be able to like fully dedicate yourself to the specialization that you're getting into. Yes. Yes. and love it that that you're doing it because you love it. That's definitely like just don't not don't just work for working something that you love. Of course. Yes. But let me bring it back. Yeah. We were talking early on about totally. No, this is perfect. This is perfect because if we bring it back to what we were discussing on chat search, right? and everything related to how the knowledge graph is growing, right? So, you've not stopped creating what 430 episodes or more, right? And so, maybe they're not all transcribed. Maybe by now the LLM have transcribed it already anyway, but **[01:06:00]** that is there and it exists. So by you not stopping, you have four years of content or whatever the number is. And that content is full of important helpful information, many of which is persistent. It's just and it'll it'll work regardless of the platform changes. And exactly all these people that didn't do anything in that time, well, they're just consuming. You were producing. And so when the time comes as it is coming right now for all the chat stuff to be how people are prompting and getting their answers. So either now or in a few months whenever I ask Chad PT something or Perplexity or Gemini who I'll be the source. You're going to ask the source. Yeah. Exactly. It's going to it's going to add to this like whole attention information economy of things that's happening here. And and so this is where I'm at right now. And we're I think Armando, you had some thoughts around how to like reduce the price and cost of using AIS to automate things, which is I'd love to get into some of this because this is an area that I'm exploring actively now. I mean, I've got a huge beefy computer here at home. I'd love to start leveraging the graphics processor unit or maybe I upgrade it to like a better um you know, GeForce graphics card so I can do the things that I want to be able to start producing. Um, and again, right, my current experience right now is I know I need to build a whole dedicated website to just the podcast. Like that's something that needs to exist. And every episode that comes out, I should have a full transcript in text written form. I should have a summary at **[01:08:00]** the top. I should have key insights that are coming out of this as well. And what I want, I want all of this to be written in an automated way. I want to be able to record the video, pop it down in this folder, and then say run. And then it just goes through be it does all the information that it needs to, builds the files that it needs, produces the output, and basically makes a blog post. So what I'm trying to do now is I want to have a whole like think of a factory. It's like a data factory. I want to be able to record the episode. I want to have multiple shorts and clips edited, ready to go, just approve them, maybe tweak them here or there. I want a full blog post written. I want summaries. I want, you know, groupings of information. And then I want all this information to get sent back to a website that can be like searchable, right? I should be able to search for any key phrase. You know, whenever Armando says data warehouse, I should be able to find through all your podcasts where it is and I'll be I'll be dog on like I was getting I was impressed. Like I was just sending some very basic things to AI to do these kind of things. It's really good at taking like media and distilling it down into other information that's reasonably good and close to what it actually is saying enough that like I can't discern that a human I mean I can maybe discern that an AI wrote it but it's coming from a human that has produced the content originally. So like I said you know what I wanted was **[01:10:00]** I literally made a problem. I said, "Go use this video from YouTube. Transcribe the whole thing. Add these sections, a summary, key phrases, key topics." That's what I want. And then I want you to add timestamps like to the YouTube video at what point in time the YouTube video occurred when that time stamp occurred. So that way every sentence you could go basically dynamically search this whole document and say, "Look again, every time let's let me see a list of all the things that Armando said that was data warehouse. I should be able to see every every sentence that was said using that and click a link and go to the episode with the link like boop I'm right there. So like this is the kind of stuff that I want to start building and I I'm I'm starting to experiment with how can we make this a reality. Oh I have some nice things for you. So okay good. Let me share something. This is not the LLM related part but just to you see do you know syntax by West Boss and Scott Stalinski? I have not heard about that one yet. Okay, so they're like a JavaScript kind very very popular podcast. It's amazing and fun. They're fun. Okay, but they had this whole thing where they were redoing the site and they they talk about it on podcast as well and how they wanted to have it all search and this was before the LLM stuff. So we'll talk about all the other but they wanted to have precisely like that cuz it makes sense to have all that index. It totally makes sense. I'm doing again I'm doing all the work. I'm doing so much work to make this all **[01:12:00]** happen. Like again, I'm spending like hours of time every week to produce this content. Why can't I just maximize? Yeah, this is it. This is exactly what I'm talking about. They have for anyone who's just in audio and if you go to the syntax.fm website, regardless of if you like the podcast, if you do a command K, which is the developer thing, but if you click the little search icon, it does the same thing. You can type anything you want like spoon, right? and it'll it'll show you like they're t they talked about the children's spoon set of whatever, right? Because they they sometimes after the show they suggest like something related to camping or whatever. And so you'll you'll find based on anything that was said within the show, any episode you can go to it and it'll link and it's just so practical because because the content is there. That was something Wes wanted to do, right? So I I mean I've been probably they they worked a ton on on getting that but then nowadays the generating all this can be automated in such an easy way once you build the tool. So there's this little sort of uh toolkit approach and I I'll share with you offline in more detail but basically we can create a little air table for for those who don't know this air table is like a Google sheets on steroids. Uh basically allows you to create your own little tablebased database. No coding involved and it's official. You can have but the great thing is these fields are not necessarily just text. They can have images. They can have dropdowns. They can have whatever. It's really really nice. And so you can build those sorts of **[01:14:00]** interfaces if you want as well. Basic we're just thinking it of like as a data dump and so basically using something like NHN or make.com you can and going back to cost by the way N8N can be hosted locally on your machine completely free no tokens no credits no nothing you can use as many cycles as you want and so if there's there's like a license thing just look at that but you if you put it on your computer then you're saving on that part and then if you build this little air table so basically you would build an air table that has a uh NA10 will be looking at let's say your YouTube or your Spotify or whatever you're uploading things, right? And whenever you upload something new, it will see, oh, there's something new. Let me grab it. It will download it. It will put it in the spreadsheet with the name, the length, the uh encoding, the resolution, everything, right? And and then once it has that, it'll start making your list of of all the things, right? Then you can have it run every few days or every few hours or whatever. And so it will automatically process each time a new video comes in. It will then send it to another process again running on your local machine that will read the whole thing, get a transcription and then it will automatically put the transcription back into that database and then it will go in and get all the with an LLM all the important parts that make sense and get all the key points and it'll put all that in the spreadsheet. So you'll end up with a thing that's automatically generating everything that you just said all the **[01:16:00]** time every day by you just posting content basically. And then it can go further and it can grab the video the original video file and then cut it in those parts, make it vertical, add captions to it so repurpose into shorts which then you can just see a list of all your ready shorts and publish the ones you want. Or it can do it for you. But I mean that's might be taking it a little too far. But I mean this is this is the point of like so what you're what so when I was talking earlier about like you know hey I'm going to make this you know large language model that makes these little mini games and all this kind of like automation of things. I'm finding that the there's so many patterns of this is happening over and over again. This is another good example of like where AI can be applied to do like simple automation tasks and like this is the same thing that we're finding when we when our development team is using and writing code like I don't want to write a function anymore. I can just describe what the function should do and have it you know and now I'm even learning better practices where I don't just tell a large language model like you know I'm using like uh claude or I'm using uh GPT mini or 03 something like that you know I'm using different I'm trying out different models but what I'm finding is um not every model is good for every every single thing right so sometimes I'm now actually transitioning from okay well I need to talk about what's the plan for this project what what you know so now I'm I'm **[01:18:00]** starting starting to change my prompting and saying I don't want you actually to build the code immediately. I want you to first here's what I want you to produce develop a plan and how we would build the plan to go build this app or build this function or build this part of my code. And so what it does is you can now first have a conversation of okay let's just bash the idea a little bit. Okay, now I see the plan. I don't like this plan. We need to tweak it. Like I don't want you writing the data to air table. I wanted to do it a different format or like okay when you do produce the data it should be in this format not that format. So it's making guesses on things on what you think you may want. But if you refine that planning statement up front more you get much better results on the back side of it because now I've been able to describe the steps or the sequences of things and then you can go back to the AI and say okay now I want you to um implement this plan that we produced and then use it. So you may use, you know, Grock Deepthink to build the plan. You may use Chat GPT Mini to do something else on the plan. And then, you know, once you've gone through a couple of these different experiences, now you have like a couple people all thinking through these AI agents. And I think um one of the the other apps that I've been looking at is web UI, I think. I don't know. Do you know about that one? Yes. So web UI is the ability for you I think **[01:20:00]** again I haven't stood a ton of time on It's an open web UI is is the um the URL. I think it's openwebui.com. But the idea is you can then host large language models on your machine. You can run a server of them locally and then now that you're running a server and you're already paid for the machine and the compute to run them. You can just run like the same prompt through six different models and see like which one gives you the better answer. And now you can start. So that's one thing I've been like missing in this whole paid subscription space of things. like I just want to be able to run multiple models on the same prompt or know that there's I'm going to build this plan for what I'm going to build, but I need it to kind of talk to a couple different models. Maybe I want to start the prompt with one model and I want to pass that to other models and say, "Here's the plan I'm thinking. Would you make any changes? Would you make any suggestions to this one?" and actually have multiple models thinking through like this and now put out an answer that I've gone through three different models and now I have a different plan that uh this is just weird to me like it's it's amazing what's happening is because it's it's almost like these are really sharp experts and I can I can pick their brain for very specific things and the amount of knowledge they have is ridiculous like I don't think in my human lifetime, I could even learn as much information as these things have been trained on, which is incredible to me. Yeah. And the the whole **[01:22:00]** agentic thing where where you have this thing that can talk to multiple Oh, yeah. is is is so good because like you were just saying and for anyone who's listening who's sort of interested in this and you're like, I don't know how to code. What's this data? So NA10 again or make they now have these sort of agent steps and so you can have this step that you can tell it your job is to orchestrate and you're going to do this like Mike was saying and you're going to send it to DeepSync and then you're going to get the result and you're going to send it to Llama and then you're going to get the result and you're going to send it to this and then you're going to ask each one what they think about like you were just saying but you can literally tell it that and it will orchestrate. even see it doing its thing. It's so much and they can talk to each other and then they generate this polished output like you were saying. And that's a great tip for anyone what Mike was just saying about the coding. It's it gets confused. You mentioned it earlier like sometimes you give it a code and it just messes it up sometimes and but if you give it a reference document, yeah, it'll work with it a ton without messing it up. And then once you like it now, you tell it to build it. And like your example where you told it from scratch, it worked better. Yeah. It'll do it. So that's just a great way to handle that where where I mean we're all new to this and it's evolving so much that there's no way someone will **[01:24:00]** know all of this. So no, these are great for people to to see. It's been constant like a feed. Like again, I'm watching like Twitter a lot for this. I listen to a lot of and my feed is just because I watch, you know, the algorithms know what I like, man. They they just do. Like every algorithm is like Mike likes AI. Any new AI video, we're going to give it to him, right? If it's something AI related, Mike will watch the whole thing of it. So the a the algorithms know what I like now. And because of that, because of what the algorithms have been adjusting because of what of content I consume and I feel like I'm able to now, you really do need like especially in this space where things are evolving so fast. You really need some media real time newsy type thing. Again, it's going back to our earlier conversation. It's connecting experts that are playing with this stuff to me. How do I find them? Where do I discover these types of things? And it's on social media platforms where they're doing shorts. They're doing small videos and small pieces of content and now, you know, I'm people laugh at me because like, wow, you're so much you spend so much time on YouTube shorts. Well, that's because I'm learning. Like, I'm I'm actually my feed is filled with Yeah, there's some dumb stuff in there. Like, there's people driving cars and things exploding. Like, okay, fine. There's some entertainment that's going on there. But I would say the majority of most of my feeds because of the way I consume content, it's literally around I'm feeding myself all these brand new things. And again, as someone's learning this **[01:26:00]** going back to like knowledge and things, right? Before I had to learn how to Google, Google then had to go index the site and you know that would maybe take a couple days or a couple weeks and then Google had to then eventually figure out which content experts were producing content that Michael would want to see, right? Right? And they're doing this at scale for millions and billions of users all the time. What is happening now is we have social media platforms that are able to do a turn on that information much faster. Like it's now Armono makes a podcast or YouTube video on LinkedIn. Great. Armono is going to take this episode, chop it up into like 10 shorts, right? You're going to we're going to use this. Now you can turn that around and now the algorithm knows, okay, new content came out. Who do we serve it to? who's going to watch it the longest. And so now this this whole speed cycle of like learning and information is getting much faster. And I think I I'm I'm really close here. The next big industry that I think is going to be disrupted with AI is going to be education. I think education is woefully behind in in the educational experience area. And and how do we you know as as technology people as as family members as kids we have kids are doing this. My kids are going to interact with technology way different than I did at at their age. And um already they're using chat GPT to write letters to each other and and communicate with. I mean this is going to it's going to even now even now my kids don't have phones but they have a watch **[01:28:00]** they can talk to like to Siri and stuff. They don't even type anymore on their phones or their watches essentially. So now they'll just press the little button and it's almost 100% accurate and then maybe they have to adjust a couple words here or there. But we're going to get to the era where you're you're not going to type things. It's going to be all communication. Uh I'm even now experimenting a little bit with like I'm using VS Code. VS Code has the ability for you to talk to it like talk to the agent and just say what you want it to do. I'm even like I'm going to the point of like, wow, when when do when does this flip, right? When do we switch from I used to type everything and now I just communicate to my microphone and talk to the computer and it transcribes what I want. It puts it in the large language prompt. It sends it to the the large language model and starts building all these things for me. We're really close to I think another leap in in technology here. Yes. And the education space is ripe for it. Like you were saying, I have a lot of clients in the sort of adult learning space. So, all that sort of stuff. We've even worked on a project together uh in that space. And there's there's a lot of overlap here for creation of content because a lot of these companies just have a PowerPoint and someone doing a voice over. Well, now there could be B-roll, there could be AI generated graphics that support you could that and then uh examinations, quizzes, that sort of stuff. You can now generate them with a AI in real **[01:30:00]** time, right? Automatically automatically. So that whole part where the the learner traditionally would have a hard time understanding a concept and now even in voice using something like voice flow trained on the knowledge base using rag of course it could then actually oh I I I don't I didn't really get that uh that part about the mass times gravity. Oh okay and it'll explain it to them with voice in another way. They'll be like, "No, no, I don't I don't get it." And they're like, "Okay, you've got a ball." So, it'll it'll do the whole exercise with them and then it'll be, "Okay, I'll get it." And and maybe they'll pass the test, right? So, it it turns into a learning experience more than a test taking experience or or just reading experience. That's interesting that you bring that cuz so that's the way I like to learn. I don't like to learn by tests. I'm I'm horrible at taking tests, but you give me like concepts, you know, thinking things, thinking like tasks and like I I need to do stuff. That's why I think I'm I love this computer space because I write code and I can hit run and then I get an immediate output. Did you do it right? Did you do it wrong? I get like this really like I love this feedback loop of like that instant gratification of like getting something built. But you're hitting on a really really interesting point here in especially the education space. Like I'm again I'm just I'm totally spitballing now here, right? So I'm just again I love idea and and being creative with things but like this makes the total sense. You should be like there should be if there's not **[01:32:00]** a platform maybe I'm we need to build one here but you know in these educational things the video the training the transcript there we're getting so good. I've seen little YouTube shorts where the I've seen I don't know famous actors like teaching yesterday. Have you seen that? So good. I was I was like I just kept watching it. I was like, "Okay, I want to learn like this." And all the comments are like, "I want to learn like this." Yes. It's like it's like like um uh who's the other guy? Who's the guy from the famous like I don't know famous? Yeah. Like Drake. Drake just shows up and he's like, "Okay, the you know, we all know that a squ= c²." And then some other actor like other famous person shows up a cut scene and go yeah. And then so they're like doing this really interactive famous people educating you on it. And again, I'm looking at myself going like, "Okay, I'm almost 100% certain this whole thing is being generated with AI. Can't quite be for certain." But it's like a whole bunch of cut scenes of famous actors teaching you a concept around how you find the diagonal of a cube. And they're just it it was uh or even I guess it's like the direction of a vector, the length of a vector. They're breaking it down step by step. There's little animations. And I'm thinking to myself, this this is going to change things. Like when you can have one person creating a piece of content and then the AI can like you can talk to uh the large language model ask it questions around what was just presented. The AI can then again to your point what I **[01:34:00]** was talking about like these little mini apps that you just throw away. The AI could one moment let me generate something for you could actually produce a whole new script, new people, new interactions and so you can ask a question and then the AI can generate a a basically a a video persona of someone me, right? They could spoof me and deep fake me and say, "Oh, I see your question here. Let me reexlain this to you in a different way." And then it provides this and then it provides additional infographics and all that information can be then again like a little mini app that you build for that user during that learned experience. And then boom, now you're done. That that piece of knowledge is now used and we you know not throw it away but like it's done. It's over. We don't need to save it. And now the user understands like there's that's the kind of stuff that we should be doing um uh with learning experiences. And the reason I'm so interested in this is because I'm generating a lot of the content, the expertise and the data. How do I take that information and dovetail that into AI that doesn't lie to you, but in fact does some really interesting like use my content to produce new new content that educates about things. Yes. Yes. And for those who who are not familiar with this, if you've heard that LLM hallucinate, they do. So if you just talk to a gener generic LLM, it it has a very large knowledge base, but when it doesn't know something, it will make it up. However, if you are using a customized type of voice agent or chat agent using something like voice **[01:36:00]** flow or any other tool like that, you can tell it to use retrieval augmented generation which basically means first it will go into what's called a vector store database with all the knowledge and it will find what's closest there's a closest match to whatever you're asking and and it does it really well and then an LLM can take that and and use that to generate an answer and so it if you do it properly in the in the business space or in the enterprise space you you cannot have it hallucinating so it has to be able to say I don't know or I did not understand or here is the response and that response must be based on real data and we can do that now so it's I if if you were sort of slowing down on your implementation of AI in your business because you thought it was just going to make things up that's not the truth if you if you use the right tools like retrievable augmented generation then you could do what Mike is saying and then if you're teaching someone if there it's customer support or whatever it is it could be based on real data to your point Mike that would be fantastic like just have I think right now you could get a like get away with something pretty nice with just like like V3 have you seen what they're doing now oh dude that's amazing I I could I mean I'm going to try it later today like just within the course content yes you can just say, "Oh, can you explain this visually?" Right? And then the prompt would then go to the its knowledge and grab all the knowledge and then it could feed **[01:38:00]** that into VO via the API and then give you back a rendering of Yes. like a take the take the question, right? take the question out of the the person's question, go go into the uh the vector database, search for the things that are most relevant, and then say, okay, based on this information, build me three cutscenes that does one scene that does this, one scene that explains this, one scene explains this. And then you can describe, okay, then you can go again deeper and say, for this cut scene, you know, again for explaining like geometry or something like that, provide an infographic that shows a triangle. in that triangle showed this this and like so you can literally break down the training steps um potentially there and have the prompt you know the prompt would be something along the lines of like okay in this scene I have a student is asking this kind of question you are a teacher you're trying to explain this information teach it to them like they're at this grade level with these things and you know assume the stuff make a graphical representation of what's actually happening do not lie and all these kind of like no be concise be per you know But there's a this is where stacking and and this is where I think this this concept of like stacking lots of models together to kind of play well together um is going to be very important. But I also look at this from the perspective of the user. The user is not going to sit there for 5 minutes while it just thinks right. So one of the big hinging points for me a lot of this is is how fast can a lot **[01:40:00]** of this stuff be produced. Right? So to me, you know, a one or two second thinking or even just, you know, it may take a little bit longer. Um, but what may happen is I actually am much more tolerant of things taking longer if I can watch the large language model type things out to me like, okay, I've taken your question, here's what I think you said, and it writes a little blurb and like here's a little thing. And again, it's it's this idea of like these little mini data applications like the response is now becoming interactive. Again, when I'm doing code in like VS Code, I'll ask the question and the large the again cloud whatever it is. I'm using cursor a lot right now. So that's my main development platform. I feel it's slightly better than VS Code at this point. Yeah, it's it's just it's really good in like showing you like here's the changes I'm going to make, but then it starts making changes to files and it actually breaks the app in the middle of making those changes. But I can start reviewing the changes while it's still thinking ahead. So I'm like I'm clicking on this file. I'm seeing what the what lines it's adding, what lines it's deleting. Okay, that makes sense. Yeah, approve that one. Move on. So, there there's really creative minds thinking about what this experience looks like um and how to do this. Well, I will say this, I've gotten myself burned multiple times using large language models cuz it just broke things atrocious atrociously. And so one thing I will just definitely highly comment here if you're doing app development things is make sure you have attached git repos or a repository attached **[01:42:00]** to it and um constantly when you when something's working right if you get to a point in the app where it may not be exactly right dude commit like crazy. Uh you want to commit every little thing because the amount of times I've gotten like oh great it works and then I like three four more prompts later I'm like crap I just broke everything. I don't know how to go back. So, so as a as a word from the wise here, make sure you commit and commit often. Small changes. Make a little tiny change, make a commit. Make a little tiny change, make a commit. And that way, if you ever get stuck or the language model does something stupid and you don't know how to debug it, you can at least roll back to where you were at a good state and then try to reprompt again to something else. That's excellent advice. I I do that all the time. I not from experience of course but it's it's a I think it's a different way of coding perhaps if you're coding manually you have a different structure of what your commits include and how you build them but if you are vibe coding like some people call it v coding yeah totally totally doing that commit like crazy because if if you don't know if you have source control within your cursor for example you're you're writing the code in a in a in a development environment and so it has syntax highlighting and a lot of nice little things that are relevant for coding. So if you do commit and to a repository and you make any change even if you delete something then when you save the file you can see **[01:44:00]** those changes and you can click a little thing and it will show you the difference between the lines before and after and you can see the whole thousand line file and just see a little green thing or red thing where where things change. You could just go to that and see. And so to Mike's point, yeah, the thing will might make a few changes. You're like, "Oh, what did it change? I clicked something." You just go in and read that difference and be like, "Oh, cool. Yeah, it just did that. It makes sense." Or just revert it completely. Now, cursor has a thing to it could roll back steps, which nice, but it can make mistakes. And so, definitely commit often. Well, and I've I've gotten to the point where I've made so much chat history that it starts losing those commit changes. Like, so there's like this kind of concept of like revert. Like, so you make a prompt, you have it do some things in the agent mode, it'll make a it'll kind of make a temporary commit like locally that you can roll back to so you can get back to the file states. What I have found sometimes is when I was reviewing the code or accepting files or doing things in the space, you know, sometimes it would change something in a different file that it thought it needed to address to solve the problem I was asking for or add the new feature. In reality, it actually broke something. And so I've had to like figure out, okay, what did you change in this other file that I really wasn't paying attention to that actually impacted the loading of of something? And so, you know, cursor is pretty good **[01:46:00]** about when there's a problem, you can go ask it and say, "Look, it, you know, you made a change. This code you just made, the prompt just broke and now it won't render the page or the the blog article, whatever." And it's like, "Oh, yes, I see it. I'll go fix it." And then it says, "Oh, I'll try a different approach. Oh, I'll try a different approach." And sometimes you get into this loop of like, "I've tried three different approaches and none of them are working." And I'm like, "Dude, you're not fixing it. You've gotten yourself into like a stupid loop and now the thing's even more broke than it was before. I just need to start over. Like, so this is where I'm like, crap. And this is why the experience piece of like commit often, right? Make lots of little mini commits because then, you know, um, this is a state in time when it was actually working decently. And and I I can then just go back to that step and say, okay, look, my prompt was totally wrong. And then I take the approach of like, okay, build a plan on how to do this. Let's review your plan first and then we'll try to implement. Yes. Yes. And and if one does not have the time to do that, then the next best thing is what you originally hinted at, which is now considering you're committing, if your chat history is very long and it's just saying silly things, clear the chat because you have the commit, right? So go back to your commit, clear the chat, just make a new one and just ask it again, and it'll be like, "Oh, here's the answer." And it'll be perfect. Why? **[01:48:00]** Because it doesn't have all the trash of the rest of the conversation. So that's a really good point, too. Yes. Exactly. Exactly right. Cuz it's it's it's trying to use that additional context to like help you build the next thing. So like yeah, you need to clean it out and like Yeah. And I've also found switching. So my developer that I work with um on my team, he's actually said switching between large language models. So like you if you're sitting in, you know, chat GPT 03, right, and you're getting stuck and you're getting into a loop like ah just it's not quite getting it. Just switching to a different lang language model again. It reads your project. it figures out your context and then you ask it sometimes it will it'll do a better job and so sometimes even just transitioning to a different language model helps out there as well. Yes, that is that is definitely true and and even for some reason it's they're they're the same thing. Well, I'll explain the reason. So there's Yeah. So there's like multiple thoughts happening at the same time. This happens to me too. How many places I can go? Arondo, plan the plan. It'll prompt the prompt the brain to plan. Yeah. Before before take talking about it. So I have to do that part. Okay. Thank you, Mike. Now point one point.1 heading is heading size three introduction. It's total total CSS action. I love it. So, a lot of people don't understand that when you're chatting with an LLM, it's actually sending the whole conversation along for the ride each turn. So, if you're building a rapper, you'll learn this. If you build a rapper for an LLM, you'll see that when you **[01:50:00]** ask Chad GPT something or any model, typically it's just one turn. So, you're just asking it this thing and immediately it forgets it. There's no history, there's no memory, any of that. the context window exists and that means it what it can keep track of in that same turn in that same conversation. So, you say it hi, it'll say hi. You say I'm Armando, it'll say I'm John. And whatever you start that conversation. Well, by the time you get to it is 54 degrees here. Then next time you say, "How's the weather going to be tomorrow?" it will receive, "Hi, I'm Armando. Hi, I'm John. Hi, what's where are you from? This is that." And then and it'll every turn it's sending the whole thing. And so that's why it seems to remember things but really it's just reading everything again and just rereading it all of it. Yeah. And then the next thing is that's why it seems to be very good when you start talking to it and then it becomes sort of dumb as you talk to it for a long time. And for some people it's the opposite depending on if it's just information and you're just giving it. Yeah. It'll it'll get trained and it'll it's not really training but it'll it'll learn more about that concept and and you how you interact and and it'll give better things but it comes to a point where con context window will be full and so it has to start discarding or compressing part of that and so that's especially with code that is a lot of information. It happens a lot and it'll start just doing this the most silly dumb things. Yes. that you're like why don't do that and **[01:52:00]** I've I've had some very weird syntax issues a lot of times I like writing like documentation sites and we use a language called MDX like it's markdown uh it's an advanced language of MDX you can like write code with markdown attached and for whatever reason I've tried like three sites with this and the dog gone thing like like cursor or uh clawude could not figure out the syntax and it kept putting like the import statements in the wrong section And then and then the app was just like I can't run. Like I don't know what you're doing. I'm like dude like like go go go please go research what an MDX file is. Please figure out your syntax. You're literally doing everything right. You're just putting the import statement in the wrong line. Figure it out. Put it in the right step. And I even have example files in here. Just match the examples. It's like just can't it just can't figure it out. It's like not it's not trained on that information yet. So so weird. It's so weird. But they're helpful at the end of the day. that really you were saying at the beginning like it's a I don't want to write a function and I find myself doing that a lot. I obviously you're writing code. You're preparing. You're you're engineering what the code should do. But there are things that you're like, I mean, I don't I don't want to write a loop that does this thing. Just do it for me. You can read it. It did it correct. I can read if it did it correct, right? So, I mean, why spend five minutes doing it when it could do it in five milliseconds, right? So, there there's **[01:54:00]** a difference between context and and syntax, right? So what maybe what I mean by this is there's the context of what I want you to do conceptually I can put the concepts together. I want you to build a loop. I want you to do this thing. I want you to grab this information and then iterate it over it multiple times. It's like conceptually I know what I'm trying to get you to to build. The I think for me as a developer or writing code, the gap for me is I don't necessarily remember or understand all the syntax exactly how to write it in like .NET, exactly how to write it in C, how exactly how to write it in Python, how exactly how to write it in like JavaScript. Like th those are all, you know, Node has its own way of like doing they all technically do the same thing. I just don't want to remember the syntax anymore. So, this is where I feel like right now, again, at the stage we're at, and this is going to get this is going to change over time, but right now, again, thinking about my children and like what they're going to learn how to write and build and use with code, they're not going to need to learn the syntax. They're not going to need to learn um, you know, the the nuances of the different languages. Maybe to some degree, but they'll be able to sit back a little bit more and at the concept level say this is I need to understand what a function is. I need to understand how to build a website. I need to understand there are common repeatable pieces across the website like the header. We have **[01:56:00]** to know what that is. So there's there's concepts around these things that have to be um understood and then from that foundation we can build all the other things and let the let the AI build all the you know uh squirly bracket this brace that put a semicolon here like I don't want to deal with that. That's beneath me at this point. I want I want someone else to own that. And so this is where I feel like right now I'm in a stage where I want the large language models to take care of that for me and they do a really good job on the syntax side of things and I can focus more of my attention on the concept of things. And so to me this is like my aha moment. I want my team to focus more on the concepts and less on the syntax. That's that's a short right there. Yes, that's that's perfect. That is so valuable. The We built the language, right? You're like, "Oh, but yeah, the nuances of the code and like make it more performant." Humans invented the language so that they could talk to the machine in a way that humans semi understand and that the machine understands. But that's not the language of the machine anyway. The langu the machine is speaking in bits, right? And so if it's if it's doing ones and zeros, from there you go to assembly and then from there you go to whatever the operating system is doing. And then from there you go to whatever the engine that you're using to run the code is doing or runtime whatever. So there's so many layers. So you could say no but for performance I need to know like **[01:58:00]** I want to code it. But you could if for the love of it. There's still people I remember there was this guy who built a like a DAW a uh desktop audio workstation software to to do music production completely in assembly and it worked amazing. Gosh. Amazing. On like the oldest Yeah. lowest computer you have. It just ran fast. No delay. All the plugins because he run an assembly. Who's running an assembly? No one. But but I mean that guy if he wanted to and it was performant. Yes. Does it make a difference? Maybe. But that's it's the same s sort of thing. I think you're completely right. This is another So your analogy is awesome. Love it. Um did you ever play Roller Coaster Tycoon? Oo. a long time ago. Very long. Okay, good. Okay, so Roller Coaster Tycoon, the original ones, was I believe it was written in assembly. So, the gentleman who who originally created the Roller Coaster Tycoon, I watched a little mini documentary on it. Same concept. Same concept. The dude was like putting data into different registers and this is how he could make this really advanced application very early on and it ran amazingly well. It was like this super tuned, highly efficient app. But to your point, right, it was the fact that he was able to get all the way down and and communicate down to the lowest levels of the computer to to make this app highly efficient and performant. And it would run on the smallest, simplest machine with great graphics, lots of animation. There's like thousands of little people rocking around this little theme park, but all of it's being handled in assembly at this very low level of code. But to your **[02:00:00]** point, a lot of this stuff gets compiled down. You know, we keep adding more and more abstraction layers on top of like machine code, right? When I worked as a mechanical engineer, we had a developer on our team. He was like the guy like he would write on the on the processing chips that we used inside the devices. He would write the assembly code or compile things down to like assembly code level things and program the individual chips to make them very efficient. They're very cheap. They're very, you know, commodity level chips to keep the product running. So, like we had one guy that was like that was like and we're all looking at this going like when he retires like how's anyone going to write code for these devices cuz he's the he's writing like assembly level code and switching things around and fixing things and putting loops where they should be and and not and like really cool stuff. But I'm looking at it going like wow we have been so far removed from that layer of computing anymore. It's now not assembly code. It's now these abstraction languages JavaScript, Node, Python. And Python's even like another abstraction layer above that cuz Python compiles down to other languages that can then pile it down. And now we're talking about adding large language models and other like so we're just adding one more layer to the the the compilation of all this coding piece. And I think it's I think it's good. I think it's the right progression. I think it's where we're going to go. Um but every time you add another layer of abstraction, you lose a little bit of like what was really being done at those lower levels. And I **[02:02:00]** think at some point in time we're going to relinquish a lot of the um the the harder end development things and and again back to our kind of earlier statement here. All of this is doing is just making it easier for us to build more applications more customized to your need or your situation. So it's it's going to be I see a very bright future for creating apps and developing apps quickly. Um, make sure your teams are the teams that are learning how to build these apps this new way with this new language so you can start focusing more on the, you know, the concepts of the app. What should it be doing as opposed to just focusing on the syntax and making it run? Amazing. That's amazing. Mike, I don't want to take much more of your time because I know you have another engagement, but we need to get on the topic of BI stuff in another conversation because we we ran we ran pretty hard on AI for this one. So, um, we should definitely come back and revisit, uh, BI and development on that as well. That's again, it it ties so closely to everything app development. I mean, all this AI stuff and app stuff, it only runs the only reason all this AI stuff exists is because companies were able to program and code off of like all this code that's being built on GitHub and stuff. So, the the data side of things and where AI is coming from, like these are so closely tied together. Um, it's it's been it's a really exciting space to be in right now and I think it's been a lot of fun. I really appreciate the opportunity. This has been a **[02:04:00]** wonderful conversation. I hope you get many, many shorts out of this that are going to be great link juice for you later on. Uh we really appreciate it. Make sure you tag me in those so I can thumb them up and like them as well. So, thank you very much for the opportunity, Armando. Let's not make this be a long event again. We'll we'll have to do it in a couple months again and revisit and talk more things. All right. Thank you so much for coming on, Mike. It was great. And where can people find you? You can find me at uh powerbi.tips the website. So powerbi.tips is the website name. If you want to check me out on YouTube, we also have a PowerBI tips YouTube channel as well. You can also hit me up on LinkedIn. Uh Michael Carlo or Mike Carlo on LinkedIn. You'll find me there as well. Appreciate the time.