Cohort 24 - Day 2

Okay guys, welcome to day 2. Sorry, it's a couple minutes late getting up. I clicked on a link and it was the wrong one, and then I clicked on the right link and here I am. But, um, I'd like to start today just by asking a question. Just, you know, what— just kind of thoughts you guys had maybe from yesterday, or maybe if you worked on some things a little bit more, if you have any questions that are coming up, but just kind of open it up to any comments or things you're thinking about. And if there is nothing, then we can just dive right into game time.

I'm a little bit hesitant or worried about making sure— one of the things that I've had to do a lot with in my work is deal with connecting dots, things that aren't working where things should be working. We've had a number of consultants who have been like, oh, we're setting up this whole system, but then they'll have something that doesn't work and connect together. I'm encouraged by the fact that you can be able to tell the AI what's missing and have it connect it. I'm also concerned that I— I'm not like, I'm glad for that, but I'm a little bit concerned that there will be things that are missing that I don't know are missing yet. And figuring out ways to make sure that I'm double-checking everything so that we're not missing connections and making sure that we're testing versus just rolling things out, making sure that payments working, that integration with payment so that we're actually making money on things and that it's going together with all of— there's been so many different things I've had to use to connect one system to another that our consultants have not always done that I'm just like, if this is another employee, like trying to train the employee, but then also figure out what things it's doing and not. It's like sometimes I can see how this would take 5 minutes and other times it could take several hours. That's what I found a lot with tech. So yeah.

No, absolutely. There are some things that are just super low consequence and those are the ones that take just a few minutes usually, right? Like, hey, I need to build a tool. I need to build a quick calculator. They'll do this for me or whatever. And I'll do that, right? Like the concept of throwaway software is actually a new thing. Like it's brand new. This was never, it was never like, hey, we're just going to like build something and we're going to use it for today and then we'll never use it again. Like no one would ever build software like that. With, with the traditional method, you know, without AI. But now that you can just be like, hey, I've got an idea. I wonder if we can just build a quick tool that will do this for us, right? Like we have to process this or we have to scrape a bunch of things. I'm gonna show you, I'm gonna show you guys some cool agent tools today that will help you do things like that. But, but you can really build something that you only use for a day or an hour and then it's done, right? And then, and then you're right, Sari, getting things, you know, what we would call production ready. Right? Uh, it's a process, right? Like, it took me 40 hours almost to, to get GenAIPI ready in my, in my 3-day sprint that I did, you know, that this course is based off of, right? Uh, it wasn't just, you know, hey, let's build a quick IQ test. It was, I had to build the whole website. I had to make sure every page worked. I had to then stress test it. I had to be able to log in. I had to be able to make mistakes. I had to be able to quit the test and start again. And I had, you know, see if it picked up where I left off or if do that or if it would answer the right question. And anyway, and then I had to do all the content stuff. I had to make a bunch of— I had to make hundreds of questions that could then pull and draw from this. And I had to build a system that allowed me to be able to see results and be able to, you know, automate, you know, different things. And so there's a lot of just when you, when you start to really dive in deep, the support systems that function, the main functionality become really significant when you're, when you're really pushing it out there. And then you have things like cybersecurity. GDPR, ADA compliance, and all these things that you don't really think about needing. But, um, but yeah, if you have a website that has any sort of decent traffic and some, you know, watchdog for ADA, for the American Disabilities Act— that's what that's for, right? Um, doesn't, you know, or comes and sees, hey, your site doesn't have anything for ADA or it's not compliant, you can actually get fined as a business, right? And so Anyway, you just got to know to think of those types of things. And the way that I do it, the way that I've learned is I literally just ask the AI. I just say, hey, I'm getting ready to publish this to the public. What have I not thought of yet? Are there any systems? Are there any regulations? Are there any gaps that I haven't— that I'm not aware of? Right. And you just ask Replit, right? Go into plan mode. Right. And just say, hey, What do I need to do? What do I need to do here? And it will always come up with something that you need to do. And then it's up to you as to whether or not you want to actually do that. But it will say, usually what it will do actually is pretty good. It will say, here's the red alert stuff. Like, you gotta do this. And then here's yellow stuff that's kind of, you know, kind of important. And then here's some stuff that you could do if you want. And, and usually I'm like, okay, do the red things and then do this yellow thing, do this yellow thing. And we'll hold off on the green stuff for now. That'll be nice to haves later. But yeah, great, great points, Soraya. All right. I want to do a quick pop quiz. Who can tell me right now, and I'm going to exclude Spencer from this question only because he's a developer, but how would you just walk me through the process of how would you vibe code a competitor to ChatGPT? You want to make your own ChatGPT, how would you do it?

You're saying just ask it to do it.

Well, that's definitely one way to do it, Pat. Evelyn, go ahead. When you find OpenAI's API and then use the API and I mean, from there you would just create with that API, your own ChatGPT. Yeah, your own interface, basically. Remember ChatGPT, the structure, guys, it's very simple. ChatGPT is this interface and it's a pretty simple interface at that. It has like a conversation history and a text box in a place where it shows responses. You know, and I know there's some other features, some settings and personalization and ChatGPT really is a little bit more than that, but for the most part, it's not, it's It's not that hard. You just ask it to build an interface, you connect the ChatGPT, you know, bot behind it, and it's done, right? I'm going to show you guys something, a project I'm working on for a client that is pretty cool. So I have a client called Sunline Landscapes and sunlinelandscapecashboard.com. You know what, let's actually restore. Yeah. A bunch of things. All right. So anyway, this is their web. So their website normally, what I did is I actually had Replit scrape their website and kind of rebuild it on this, on this website. But anyway, these guys do a lot of amazing landscapes for people. Okay. And so I had them build this dashboard and right here, it's going to pull up a bunch of metrics and things that we don't really need to share with you guys, but kind of their revenue and stuff like that. But what I did is I built their own custom AI. And this AI is a lot smarter than just ChatGPT because I connected it not only to OpenAI, but I also connected it to Aspire, which is their place, the software that manages all of their customer data and shows all their revenue and all their projects and all the different things. Right. And then I connected it to what's called a RAG database, right, or a RAG implementation, that what is, what that is doing is it's basically you have an AI, like OpenAI has this amount of knowledge, right? It's all inside of this, this imaginary ball right here. Okay. And then, but, but it doesn't know anything about your company, in this case, Sunline Landscapes. And so what we did is we took hundreds of documents and uploaded them to Sunny, right? And taught it to Sunny and added it to this little ball, right? So now you take this other information and kind of stuff it in there. So now that when you ask the agent a question, it understands everything about the whole world because it has everything about OpenAI, and it very specifically understands what's going on with the company. What are the company's values? What are they all about? How do you— what's the process for this? What's the reimbursement policy here? What's that? And whatever. It knows everything. And so now people have access to this chatbot and they could say, you know, how do I wash my truck? Right. And so it will then consider everything it knows about washing a truck, but it will consider first what is the policy and the procedure for washing a truck at Sunline Landscapes, because I've taught it and trained it using RAG, right? RAG. I'm going to show you guys a little bit how to do some of those training things today, how to make your own custom GPT. And there's some very easy ways to do it.. And then there's some little bit more technical ways to do it, but, but you should understand it pretty quickly. So anyway, what it did here is it said, hey, here's the standard Sunline truck wash procedure, 30 minutes with a 3-person team. Okay, here's the steps, here's what you're going to do, here's the best practice. And then here are the sources for that. And these are documents that I built a system here where they can come into the admin panel, they can upload documents here. And here are 51 documents that have been uploaded that teach it all of that. And when I click upload, it actually processes it and then adds it to the database, the knowledge base. I also added this other thing over here with knowledge where you can say, hey, you know what, I want, I want to add in my own thoughts about this. And so instead of uploading a document, I could just come in here and I could, I could type right here a couple examples. You know, Sunline is the best landscaping company in the world, right? Anyway. So some of this stuff is just test entries. But the point is that AI is now pulling from knowledge entry documents and the Aspire company information. So I could say, how many jobs do we have outstanding right now? Or how many open tickets do we have right now? Or what was our revenue last month? And it has the ability to go and use the API with Aspire, and luckily Aspire has a great API. Grab that data, come back and talk to it. So anyway, pretty cool. Pretty cool. But that's just kind of a real-world example of how you can use this knowledge of being able to build a GPT or a chatbot or something like that, that then is connected into your own company's systems or your own customer systems or whatever it is that you're doing. If you want to go and help other companies adopt AI, great, go do that.. But if you want to just adapt it and adopt it and use it in your own company or put it into your own solution, it's very easy. You just have— and, and by the way, guys, all it took was a lot of talking to the agent. Here's what I wanted to do. And then I would test it. That's another big thing. You have to test it, right? So I would look specifically for some small thing like the truck wash, and I'd be like, hey, what are the steps? Right? How do I wash a truck? And then, okay, great. It is actually pulling from this database instead of, you know, making something up. And in the beginning it was making something up. It wasn't properly embedding it into the answers. And so anyway, just guess and check until it works, right? Guess and check, guess and check, guess and check. That's the, that's the process now with a lot of this stuff.

All right guys. Is there a way to be able to get it to do time-specific things and to have answers that it's put in go in directly to it? Because we had, a consultant build a chat AI or chatbot for our for our product. And the very first question we asked it, it gave the wrong, wrong answer, even though it was coming from our own databases, because it was a correct answer 5 years ago, but the tax law had changed. And so we had— so it was like the very first thing we asked, it was wrong. And we mentioned it to them and they're like, oh yeah, we could do something about it, but it still hasn't happened. And so if someone's done a chatbot for you, how do you get access to be able to update that information and try and get it to train to know that more current data is more important than later or than clear.

So I, I can think of two quick ways to fix that. Number one is going to be, um, have the AI check itself, right? So you're going to add an extra step in that response, which might take an extra 20 seconds, might add 10 seconds or something to the query itself. But you can say, once you've found the answer, I want you to search the internet and make sure this is accurate. For, you know, today's date, right? So you could do that. Second thing you could do, and this kind of requires you to have good information. Well, not kind of, it definitely requires you to have good information and dated material. So if you had a, you know, a document, a Word document or a PDF that was from 2019, and it goes to reference that document, but then you have another one that's clearly labeled 2024. You could say always check, for the most recent content, right? If it's not labeled 2024, then it has no way to know that other than the first step I said, which is, hey, once you've found an answer, I want you to scour the internet and check yourself, sanity check yourself and see if this is the right answer. If you think it's good, show me. If you think it needs to be adjusted, adjust it, then show me. Right? So, Um, and the reality is you could talk to the AI about how can we fix this problem, and it will probably give you similar ideas to what I just mentioned there. But anyway, um, okay guys, let's start today by looking at some agents, some agentic systems. Um, and by the way, Tanya, if you're, you know, coughing and not feeling good, feel free to turn your camera off too if you want and drink water and be more comfortable. Whatever. You don't have to like perform for us. Um, but anyway, so let's jump into agents and, um, actually watching Tanya cough is making me thirsty. I'm going to grab a water bottle really quick here. Um, agents. I'm so sorry. No, you're good. You're good. Like I said, I'm a little bit sick too. Um, just, just It's kind of a minor cold, nothing crazy. I'm really being selfish because I want to be here and I want to learn this. So I apologize. No problem. So guys, agents are— you can kind of think of an agent more like a human in that it can reason through multi-step requests. Right? Something that might require you to do something, then do something else, then do something else, then go back and check it, then test itself on it, then go check this and go do that. Right? And you've experienced that a little bit, especially if you've, if you read along with some of these AI agents like Replit that are working, right? Sometimes it'll say if you, if you put it in like high autonomy mode, then it will say, hey, I just did this and I just asked it this question and hey, I'm trying to do this, but I don't know if this is working. And the architect says this and the planner says this and So they have multiple kind of personas inside of Replit that will talk to each other and reason through and try to help you get to the better solution. When I first started using Replit, that was, that was not a thing, right? It was able to, to consider your request and then build code as best as it could. And then it got better and better and better and better. And now literally, like, so, so when I first started using Replit, the agent could code by itself for about 2 minutes. That was the max. And then when V2 came out, it jumped up to 20 minutes. And V3 that we're on right now can do— they publish up to 200 minutes, but I've seen people say it can go for 6 hours, right? If you give it the right— if you give it the right prompts and instructions and autonomy, right? Now, a 6-hour prompt might cost you $100, right? So, you know, don't just set it loose unless you have a very specific reason to do that and you're willing to spend $100 on it, which I would be if I had a 6-hour task. That AI agents are going to be working on for 6 hours, I'd pay $100 for that in a heartbeat. I wouldn't even blink because I probably can get something pretty amazing out of it.

Hey, I have a question about that. Is there a rule of thumb about the— I know the costs are ridiculously small, but the— is there a way to plan out like what, what a cost would be? Like, just let's say I'm just trying 100 different ideas and I only want to dedicate $10 each one or so. I don't know what the scenario is, but Is there a rule of thumb that you use to estimate the cost?

Most basic ideas you can get out the door for about $2, right? At least with Replit, right? So if you just say, hey, I'm gonna start building, designing, and I wanna see a first version of this to where I can actually play with it and use it, you know, just at a very high level. Again, if you're building something really, really complex, there's, there's gonna be exceptions, but most of the time to, you $1.50 to $2 is usually what it'll cost for you to get to that point. And so I don't know.

Yeah, that's, that's, that's wonderful. Yeah. No, I've just— Yeah. You've never like done something and like, oops, came back later and you're like, you owe them like $1,000 or something?

I have never run into a big crazy like, oh my gosh, I can't believe I did that. I've never been like, oh, I can't believe I made that mistake to ask it to do that. Right. I've just never.

I've never run into that.

The other thing you can do if you're worried, like if it's going down a rabbit hole and you're like, I know this is not the right way to do this, you could just stop it, right? You just push stop the agent and say, look, I saw you were going down this. I think that's wrong. And here's why I think that. Go into plan mode, right? And say, do you understand what I'm saying? Oh yeah. You know what? Good catch. You know, it'll think you're really smart. Thanks. Make you feel awesome. So. You can usually set spending limits as well. You can set spending limits. You absolutely can. Yeah, 100%. 100%. All right. We're going to jump in right now to Manus. So Manus is a really amazing system that has been out for a while now. And they, in fact, just got acquired by Meta, which is pretty crazy. This is a Chinese company that got acquired by Meta. And so now it's an American company. But, and this is actually pretty cool too. They just announced this. I was reading a post about it this morning. Mannis partners with SimilarWeb. So you can ask Mannis about any website's traffic rankings or marketing channels. Thanks.

Yeah, that's big for me actually. It totally, totally— I used SimilarWeb 10 years ago. Like I, I was part of, or I've used it for a very long time. I'll just say that.

Yeah.

It's, it's a wonderful service. It's, it's something that is And this is a cheap way to get it now for a similar—

yeah, potentially. Now I think it's only on Pro, but, um, but I think I'm on a Pro. I think I'm on a Pro plan. Um, count— man, it's Pro. I am. Yeah. Anyway, so, um, by the way, if you guys want to use this, um, I'm pretty sure there's like a referral link. Yeah, here it is right here. I'll just drop this into— you see, you get some free credits if you do it and I have so many credits, I never use them, so I don't really care about that piece, but you guys can get some free credits if you just use that link and sign up. So, all right, so we're gonna continue. We're gonna see how Jenni API data— I'm gonna just have it do this thing. But basically it's going through and it's doing, it's using tools. It's looking at, you know, my, my current website, you know, what's the global ranks and traffic sources bounce rate. It's probably not gonna be that great because quite frankly, I don't focus. We don't, we don't do any marketing that drives people to the website. People will go to the website, they'll land on it if they find us, or if they find me through my social profile on LinkedIn or something. We're going to talk about marketing later a little bit and how I, how I use AI and, and really how we go about doing it. Um, all right, that's fine. Just, just do it. Okay, so, so anyway, it was able to collect some data, and now it's going to research it and figure it out and put it all together for me. But I want to share a couple examples of things you can do with Manus that I think is pretty amazing. Um, let's go with, um, let's go with something here. Um, where is this? All right. This one's pretty cool actually. So I had it work on— and by the way, I can, I can click away. It's going to keep working on that task while I go to other things. You can have it work on multiple tasks at the same time. Okay. So I had a, uh, or I have a neighbor. I have a neighbor, a good, who's a good friend now. Um, but he, he said, look, I, I own a construction company. And the thing that takes me forever to do is to take construction plans, which are pretty, you know, they're pretty technical and pretty, pretty ridiculous, right? I mean, very detailed, lots of lines, lots of little pieces of data, whatever. I mean, look how crazy this is, right? And what you have to do as a construction company owner is you have to go in and you have to measure, you have to kind of say, okay, there's, you know, 2 feet here and 4 feet there and 3 feet there and 10 feet there and whatever of all these different things, right? Here's how much concrete we're gonna need, here's how much gravel we're gonna need. You have to do all these things and it can take for a plan like this, It can take easily 24 hours of work to, tedious work by the way, right? Where you're having to go dig in deep and figure out all these things and you have to put together what's called a takeoff. And he said, man, is there a way to get, you know, can we just have AI do this? Right. And, you know, so I completely understand where he's coming from. It's hard. It's a, it's a tough, it's a tough thing. But, but he said, look, I want to, oh, look, my SimilarWeb thing might be done now. So I'll go look at that in a second., but for this task, I said, look, I want you to look at this. It contains construction plans. The other 4 documents attached were created through a standard and thorough construction takeoff, right? So basically here's the source, and then here is what, you know, humans normally produce, right? This is done manually. Okay. This process is highly complex and time-consuming. I want you to build a piece of software that can do this automatically. That would allow me to just upload the plans and pull out exactly the information provided in the format. Okay. You'll need it. You'll need to create measurement tools that you then test against the construction plans and see if you're getting the same numbers as the actual ones provided. Keep testing and iterating until you have a working product. Okay. So instead of having to sit there and talk to it and do all this stuff, right, it creates software, tests it, creates software, tests it, tweaks it, tests it all the way through. And it took couple hours for it to do this. But by the time I was done, I was given a bunch of Python scripts, right? And these Python scripts is just a type of, you know, coding language that, that can run these types of things. And then it gave me instructions too, right? Like, here's the user guide, right? Here's, here's how you use the construction automation. You know, just, it built this guide on how to do everything, right? Here's what all the things are for. Here's the batch mode. Here's how you use the calibration. Here's troubleshooting. So it built all this stuff, gave a bunch of information, and then what I did is I just downloaded this entire thing, just batch download as a zip file, extracted it, and then I took the entire thing, dropped it into Replit, and said, follow all the instructions and use all the code here and build me an interface where this will work. Okay? So I was able to take easily months of work. And compress it into a few hours and have AIs do pretty much everything. And so the key things here that are really important is that one of the most important things is I asked it to iterate against itself, which means it's not going to just do it. I need it to test it and make sure it's right. And I gave it the right answers. I gave it the right source. I provided all of those pieces of information so that it could just iterate until it got it right. Okay, so if you have something that, that follows a pattern, here's a great example. Sariah, you have wills that need to be done, right? And so you could say, here are 100 wills that we have written. This is the style we like. These are the types of clauses and inclusions we have in this. This is what we want you to model after. I want you to now build me a questionnaire or an intake form that we can send to a client. They fill it out, and then you output from their instruction, from everything that you know they need, that you asked for, into a perfectly formatted will that looks just like this. Okay? So now you've given it this structure. Here's something we do all the time. And here's based on that, here are all the things we need to know. And now we have software that will take this, analyze it, use AI, write it, format it, and and have a PDF ready for download, right? Totally amazing, right? Sorry, we can't hear you, Sari, if you're talking, but I just realized that.

Sorry about that. So basically, we could be, rather than sending them a questionnaire, we could have that an app or a part on the website where they could just fill in the blanks and then it could automatically do that for them and then produce the document and get it to them so that they could be able to print it out and get it notarized, which would be a very good thing. That's actually something my, that's part of the reason it was taking me a long time to put together the, the app thing that we were doing yesterday is my boss was telling me, and we need this feature and this feature. And that's one of the features we were talking about. Why are you going to Manus for this instead of doing it inside Replit or inside of the ChatGPT or the, or the other ones that we were working with?

What's the advantage of this particular system? An agentic system like this is able to better review itself. Okay. And so what I mean by that is it will take more, it will make, it will take I'm sorry, it will use more effort. I think that's the right way to say that. It will use more effort and consume therefore more context to consider everything over and over again. ChatGPT, just you ask it something, it thinks once, it thinks hard, and then it's done. But it doesn't go back and check its own work and iterate on itself unless you're very manually guiding it to do so. Manus actually has a cool feature here. I'll show you this. If you click on settings, and you go to Mail Manus, it has an email address, joncheney7485@manus.bot. Now, you guys could try to email this and ask it to do something. It won't work because it only has these approved senders. You'd have to send it from my email address. But the point is, if you email Manus, I could have emailed it these 4 attachments and this text right here. And then 20 minutes later, it's going to just email me and say, it's done, here's all the attachments, right? And so it almost acts like a human, right? And, and so you can just, if you get an email from a client and you're like, hey, I need a proposal put together for this client, you can literally forward the thing to it and say, hey, I want you to do this. This is my style. This is what I want. Go. Right. And then it would email you back a fully ready, you know, proposal ready to send over to a client. Right. It just operates a little more fully autonomously. Right? You just have to babysit it less. So that's ultimately why I'm probably going to use Manus to figure out and make plans. And, and you, and if I'm going to build it in Replit, like this was, this was, I actually did this request. This is in April of last year. And, you know, it's only 9 months later, but I'm light years ahead in terms of how I use AI. I probably would, would do this. I would be more detailed. I would give it instructions like, hey, I'm actually going to take what you give me and I'm going to give it to Replit. I'm going to have it put together the interface and all those things. Manus is actually fully capable, by the way, of building the website itself. Right? It actually can do that. Let me show you a simple example of a website here inside of Manus. I asked it to do an inspiring biography about me.

Right?

I just said, hey, I want to just see what you can find about me. Right? Go out there, figure it out. I gave it just, hey, please write an inspiring biography. And make it a 2,000-word article, kind of feature piece on Jon Cheney and who he is and what is, what is it, you know, what am I about, right? So I just kind of wanted to know what, what my digital footprint was like. That was really kind of the idea behind like, if you go and you Google me and you try to just research me, what are you going to find, right? And so I had to do that. Okay, great. It goes, it searches, it does all these different things. And then I had to put it in a website, right? And that website here, Jon Cheney, The Art of Elite Mastery Across Domains. By the way, guys, AI is going to make you look good and feel good all the time. So, you know, take that for what it is. But, but yeah, so it found, you know, adventure, AI mastery, music, languages, recognition, whatever, personal, some other stuff. Right. And so, hey, you know, Jon does a lot of different things. He's a whitewater kayaker. This is him talking about going off a 45-foot waterfall and some things like that. I don't know what this image is. Of just a random street. And obviously the formatting isn't great on this. And so I, I, I didn't go to iterate it. I just wanted to see what it was gonna do. And I never published it or showed anybody this, right? But, you know, AI GenAI API, it did a bunch of stuff there, figured out, you know, hey, build a business in a weekend, discovered that piece there. There's some articles and news things on that musical mastery, right? All these things about albums I've published and, you know, it probably visited my website. By the way, I can click on my website right there. Boom. And that should take me to joncheney.com. There we go. Right. And so, yeah, amazing. I speak a few languages. I was able to find that and, you know, saw Startup of the Year, Innovation of the Year, Win 100, different things I've been featured in, Utah Business, Crunchbase, different, different things like that. Personal life and values, figured out I'm a member of the Church of Jesus Christ of Latter-day Saints. I love BYU, that I live in West Mountain. Right. And so, you know, anyway, but the point is it put together a decent website and then, you know, it can, it can very easily, you know, Connect me to this. It added some links to GenAI and other places like that. Kind of cool, right? So it is capable of building websites, putting things together like that for you. And I want to, I want to show another example of creating a website. So, um, I was hired by, um, by Tony Robbins and Dean Graziosi. Anybody know Dean? Anybody recognize that name, Dean Graziosi? So Dean Graziosi is a really cool guy. Um, if I pull them up here. You should— you may or may not recognize him. Super smart guy, kind of a business guru. He's like best buds with Tony Robbins, right? If you just do this search here, these guys are always together. They work together in a ton of different things. And anyway, so there's all their pictures together and stuff like that. But I— they wanted to launch an AI system, whatever. And they did AI Advantage, and I actually helped them make all this stuff, right? So, but what happened was I was on a, on a call with Dean, and he's like, hey, I just barely did this live video on Facebook. I was asking people, just kind of like testing the waters with AI. I want to know what people thought. And so he's like, Jon, could you just go and read through the comments? And there were like 1,700 comments. He's like, go read through it and then report back to me and tell me you know, what are people interested in? What do they want to learn about? What are they worried about? What are they excited about? Whatever. And I was like, okay, great. Well, I'm not going to do that manually. I was just like, okay, I'm going to do this much faster. I'm going to have AI do the work. And so what I did is I just kind of copied down, scrolled down, took me maybe like 2 or 3 minutes of just scrolling down and load more replies, load more replies, and just copy paste, copy paste, copy paste. Took me probably about 3 minutes to copy over about 1,000 comments, right? And then I just said, look, get rid of, you know, only focus on the, you know, the, the real meat, right? If it just says, hey Dean, like, I don't care about that, right? But then I want you to put together a website that has interactive visualizations, bar charts, pie charts, graphs, et cetera, whatever. And so about 15 minutes after Dean asked me to do this thing, which for most people that would take hours of thinking and considering and then having to write and do all this stuff. I was able to deliver to him a website link right here. Boom. Right? Total comments, positive sentiment, current concerns raised, excitement points, sentiment distribution, sentiment breakdown. Here are the opportunities. Here's learning interest. These are the top things. Here's AI usage levels. Got some nice little animations, fears and excitement, and key quotes. I sent him this website. With all this information, he's like, that's insane. How did you do this? How did you get that put together so fast? Right. And obviously, you know, he hired me for my AI chops. And so he knew that that's what the answer was. But this is the kind of stuff I expect from like my team and, you know, people that work with me these days, right? If you know how to do this, you should be able to easily, even, even if you've already done all this work, right? And you just put it together in a nice presentation. Drop it into Gamma. I'll show Gamma later. Drop it into Manus. Drop it into Replit and put it together in a nice presentation so that we can just easily, quickly, and make it interactive. Give it a way. And, and you don't want to do interactivity just for the sake of interactivity, but make it useful. Make it, give us a way to the where we can test our assumptions and things like that. So anyway, long story short, I was able to have Manus, you know, analyze all that data, categorize it all, put it all together. Build a website and ship it off to someone in 15 minutes. That's, that's incredible, right? That, that's just an amazing use of time, right? Um, let's go to, let's go to, uh, GenAI Pay. Let's look at this analysis. I don't even know if it'll be good, um, because this is a new feature and I don't even know if my website is that good, right? So executive summary, business overview, this is what they do. Target audience is clear defined as businesses seeking to integrate AI, as well as individual professionals aiming to enhance their AI skills. That's good. I'm glad that we got that. Here's total unique visitors, total visits, month. It looks like we're, we had, we were really high in July. I don't know what happened in July that would have led to that. I'm not sure. I'm not sure what that is, but it looks like we kind of, we're kind of like generally growing a little bit here. Like we had this spike here. That could be, by the way, a, uh, like, why do we have 882 visits and 3,189 visitors?

That feels right. You cannot rely on that data as if it's—

yeah, no, real.

So I'm gonna get rid of it.

It's good for trends and it's also good for like, uh, right, for sure, for sure. But bottom line is I think I might have— you know what I think I did actually is I think I was seeing some weird, um, like multiple firings on websites that were loading multiple times. And so it was showing different things and the stats were wrong. And I think I fixed that. So I probably fixed that. So let's eliminate July. If we look at that, then August, September, October, November, December, we generally grew in terms of unique visitors, even though total visits is still a little bit— it's kind of, kind of steady, right? I'm actually happy with over 1,000 visits per month if that's, if that's right. And look at that, our bounce rate Look at that. 56, 48, 42, 36, 32, 27. I'm super happy with that. Right? Global rank looks like we got worse over time for whatever reason. And again, we're not trying to rank, we're not doing a lot of things that, you know, and this is a global rank versus all websites. Like we're not, we don't care about that really. I would really have it go and dig deeper and say, hey, what is my ranking versus other AI websites or AI education or AI for business or, you know, things like that. Geographic distribution, not surprised, it's majority US. And actually have had a decent amount of people from India come on and buy stuff from us. So anyway, pretty good. Here's some stuff based on this. We could do that, whatever. So then I could obviously, now what I do with this information is up to me, right? Now I have to say, okay, I could take this summary and all this, all this data here and I could drop it into, you know, my Replit that manages the GenAIPI website and I can go into the plan mode and say, hey, based on all this information, what do you think we can do to improve the website? Right? Oh, hey, well, based on this and these trends, maybe we should focus on this. And hey, it looks like we're getting a bounce rate here, or this region, or this thing, whatever. So what I do with it, that's up to me and my, you know, my, my systems.

Um, hey, I, I have a selfish question. Go ahead. I noticed in your thing that there's a lot of extracting of HOAs.

Yeah.

Yeah.

So I was actually going to share— I'm going to share this one next, but go ahead with your question.

Well, no, because one of the things that— one of the real-world problems that I've ran into that aren't— that isn't like a website I deal or anything, but just real-world stuff I come into, is like I always want to buy like a condo somewhere or something, and then— but I can't tell which ones can be rented or which ones can't.

Yeah.

And I know you could get a real estate agent to to, to do that. But I would just like sometimes to be like, well, what's it like in Tucson or Waikiki, or what, you know, what are these different things? And my brief, a brief attempt at that, it seemed like it was locked off information a lot. So I'm wondering, is Utah like open about their HOAs?

Well, not just Utah. Many states are, um, some states are not, and so you just kind of have to deal with that. I actually have nothing to do with HOAs In fact, I avoid them like the plague. But I've lived in HOAs before, but I don't now.

I actually— I think that's an interesting business question is like, how do I just find out which ones can be rented? Like, I don't want to buy a condo that you have to live in.

Yep.

And I want to buy one that I could rent if I'm not going to live in it. So anyway.

See, you could use agents, by the way, to figure that out. If you had a list of 20 different homes or places you were considering, you could just build a little agent that will automatically email everybody go back and forth until you have the answer.

Oh, that's a good point.

Yeah, you could just build that automation for you. So here's what happened. I received a message from someone on LinkedIn. They said, hey Jon, I want to take your course, um, this one. And, um, but I'm just, I'm just curious if, you know, here's my problem, right? And he literally wrote me this right here. Um, he said Utah has an HOA registry website that's right here. If if you go there, you can search for HOAs by name. But you just have to put in one letter and search. It will pull everything that has that letter. When it pulls up the list, you then can click on it. You can see names, addresses, sometimes email and phone numbers. I want something that can pull all that data into a spreadsheet. And there are thousands of them, right? If you go to this link, last time I clicked on this link, it actually didn't work. Yeah. Yeah, I don't know what's, I don't know what's wrong with this, but anyway, um, eliminate, eliminate duplicates. So then I added eliminate duplicates, provide a long and detailed CSV of all this information, right? And so I just literally copied his request from LinkedIn. Pasted it in here, right? So he's asking if he can take my course to learn how to do this. I was like, I'm just going to do it for him. Just, I wanted to see if, you know, if it could do it, right? And so I just dropped it in here and this is a huge amount of data. Okay. And so what happens in this, if you do something that's too much data, you can actually go, you can actually click inherit and continue. Right. And so yeah, please start a new chat, get support with this issue, please. You know, anyway, there was a button. It's hard to show you now because It's not there, but there's a button that says inherit and continue. So what inherit and continue means is it's worked a bunch. It's done a bunch of data, but it's not done, but it's run out of context. The task is too long. It's too big. And so you can just say, all right, grab all the important stuff and then pick up where you left off.

Right?

And so I kept doing that, kept doing that until it said, okay, I'm done. And it gives me this CSV right here with thousands and thousands and thousands of of pieces of information. Here's registrations, address. I don't know why it's showing me this file not found thing. Move out of my way. There we go. But now I've got the primary chair, I've got the primary phone, I've got all this information. So now I can go in— here's the HOA name and all this stuff, but it pulled out thousands and thousands and thousands. I think it's 2,000-something. Yeah, there we go. 2036, right? Anyway, but it did all that data and now he could just literally take this and drop it in. And so it took maybe 3 hours probably of just Manus working and then me clicking continue and continue and continue. But the total amount of actual work I put into it was less than 5 minutes, right? I had to click the button a few times and just kind of monitor it and watch it. And I just took this CSV and responded to him and gave it to him. He said, how on earth did you do that? How much can I pay you for it? And I was like, don't worry about it, man. So, um, anyway, he ended up taking the course about a month later. He's like, I need to know how you, how you do these things, right? So, so anyway, um, pretty cool. And then he went on to build actually a piece of software that he's selling, right? He's actually selling and making money with, with stuff that he took from this business or from this course. So anyway, good stories. I love stories like that, but I, I love to use agentic things for really time-consuming tasks. And by the way, we did some math on this. If we were to go through and do this in the right, like, in a kind of more reasonable way, it would have taken a, you know, a dutiful secretary-type person probably a week to get this done, right? And it might maybe even twice that long considering how mind-numbing it is. And, you know, just, but to go and copy every field just because of how that website was was put together, you had to copy manually one field at a time. You couldn't like batch copy stuff. It just wouldn't work that way. You could only see one entry at a time and then you would click back and it would say reconfirm submission and then it would take you back to the top and it wouldn't take you to where you left off. I mean, it was, it was not, it was kind of built in a way to stop you from scraping it. That's really honestly how it was built. But Manus doesn't care. Manus will go back and click and wait and find the next one. It remembers where it was and What's the next one? And it just, it can do all that because it's a computer and doesn't care, right? We've ordered a robot for our home here, the Neom robot. Um, I'll show you this. This is part of, this is part of what I want to talk about today too, is how AI is going to be involved in, you know, in our homes, right? There's going to be robots and Neom is one of the first ones that's going to be there. Obviously there's Optimus as well. That's by Tesla. That's by, you know, anyway, and if that comes out sooner than I'm probably going to cancel my Neo and do that instead. But I think this one's going to come sooner. Right now it's $20,000 if you want to buy it outright, or it's $500 a month, right? But Neo has the ability to keep your house clean, to do your laundry, to do the dishes, to do whatever, right? And, and I have a big house. I have 4 kids. There's always toys everywhere, and there's always 12 or 16 loads of laundry that we're behind on and all this stuff. And if I could just be like, Neo, go, right? Like, we pay, I think, $350 twice a month to have just some people come in and help us clean. And, and they don't even do the laundry. They don't even do most of the stuff. They just like do surface level stuff and, you know, vacuuming and cleaning bathrooms, things like that. That's, that's $700 a month right there versus $500 here. Plus this thing can work all the time and do the laundry and do all these other things. If I just got the cleaning, like it's vacuuming right there behind that, it's carrying in groceries. It can carry something like 70 pounds, right? And it can go up and down stairs. You can do all this crazy stuff. But AI is able to, it kind of walks like Joe Biden, you know? So you gotta get over that. But other than that, you know, kind of downside. It's pretty cool. And this is just the beginning. This is just the beginning, right? But it's capable of, You know, it would just be awesome. It would just be awesome if it could actually deliver on what it's doing here. And it can also talk to you too, right? I mean, you can— it's a companion. You can have it sit down on the couch and watch a movie with you and tell it to comment on funny things. You could literally do that. And just as you would be able to talk to ChatGPT and say, hey, comment, you know, give me a joke about this scene in Independence Day, whatever, right? It would be able to do that. It's capable of watching, understanding, analyzing the world around it, and it has knowledge of just about everything, right? Hey, Neil, what's this thing? Anyway, so anytime I can put an AI, a robot on something that is going to— that would normally take me a long time, right? That's just kind of repetitive and it doesn't really bring me value to my life to do this. I don't really get smarter by, or feel accomplished even hardly by moving data from one system to another system, or by cleaning a dish, or by folding my laundry. It's just a repetitive, dutiful task that's got to happen in our lives, right? But we can have AI and robotics take that from us and do those things and allow us to then do those higher-level activities, the higher level, the higher value, the higher connection, the more personal things, the more human things that robots and AI cannot do. Any questions about Manus before we move on from this?

If you can use Manus to build the websites, why are you putting it back in Replit?

What are Replit's advantages to want to be able to use that hosting and deployment? And it, it's really better at long-term kind of going back and forth with it. Manus actually has gotten way better at building websites. I just haven't tried it. I haven't tried it lately. They actually have the ability to kind of host and publish and, and like, you know, tweak things and modify things. Replit is more powerful for sure in terms of just building websites, but Replit can't do research for you and do all these types of things. Like I had this one, this one crazy task here. Yeah, this was a simple one. It was like, hey, convert this, this to this file. And I couldn't find a converter., right? No, nothing else could do it. And so it was able to figure it out and do it really quick. This was pretty cool. I was working on a project for somebody and they had 250 painting projects that they had done and they had addresses of all of those. But what I, what I needed was latitude and longitude for each one. And so I was like, hey, here's the list. Give me the longitude and latitude. Boom, done. Long lap, done. Super easy, right? Like, just tasks where you're just like, this is definitely something the AI can just knock out for me. Um, so yeah, you just kind of have to start playing with these things. And I— this is how I learn, right? I, I just jump in and I say, okay, I want it. By the way, they have a new project feature. That's new. That's cool. That's fun. Anyway, Manus is getting much better and they're going to keep getting better. Um, but this is like having your own— I, I like to compare Manus as my, my assistant, right? Replit's my developer. Manus is my assistant. Okay. So that's, that's that. Now yesterday I, um, came across, in fact, I posted about it this morning. Let me find the, let me find the, uh, thing here in our company. Discussion. This is pretty cool. Let's watch this for just a minute. We don't have to watch the whole video, but this is just a quick video of something that was announced called Atoms. And this would be a competitor to Manus that just came out. Pretty cool.

I think it's worth watching. The era of scaling with silicon has begun. Kind of hard to understand. Founder of Atoms. We are the team behind MetaGPT and the OpenManners, trusted by over a million builders worldwide. Our mission is simple: turn ideas into business. For the past 3 years, we push open-source AI to outperform leading closed systems at higher performance and lower cost. Today, the foundation becomes Atoms. Atoms is not a coding assistant. It's the first autonomous team that builds real businesses. This is a shift from web coding to web business. Let's start with a simple software. Atom doesn't start with code. It starts with deep research. It scans market and finds a niche in professional headshots. You say start. You reveal the plan. He approves. Instantly, an entire product team spins up inside Atom. A PM, an architect, an analyst, and a developer. And Atom starts to rest. It writes the backend, configures the database, integrates Stripe, and deploys. Minutes later, the site is live. You say, "Switch to pay-per-use." Atomz rewrites the code, updates the API, and then redeploy instantly. Now your product is ready to make money.

Next.

A DTC fashion brand. You don't start with designs, you start with demand. Deep research scans Instagram, Etsy, and Trend Data. It finds a fast-growing micro-market. You say build it. Addons creates the brand, sources print-on-demand products, generates catalogs, and writes SEO pages. Now you have a live sellable brand built from pure market signal.

You try something random, make a simple warning. So now you don't even need me anymore to teach you all the steps. That's basically what this is saying. Instead of having to go through all that process of everything, it automated that whole thing and it spins up a bunch of agents that are working together to vibe a business. Where you literally come out of this and it's just working, right? Like Sariah, again, I'm just picking on you cuz we've been just going off your example here, right? But if you wanted to go back to this Will's product, you can literally go to adams.dev. I haven't tried it yet. This just came out. But you go to adams.dev and say, make a website that does this, this, this, and this. And, you know, tell me when it's ready to deploy and it will be deployable, ready for customers right then., right? You don't have to do anything. Now, obviously you might want to iterate it. You might not like its design and you might want to tweak it and tell it a little bit more information. But the point is you could legitimately, if, if I were to do my same challenge that I gave myself back in February of last year where I said, I want to start building a business today and see if I can get my first customer tomorrow, I legitimately think I could do it. Maybe even same day. Now, can I do this in a day and just use adams.dev? And because I can do it all at the same time, I could do 3 of them at once and see which one gets me a customer first and then just double down on the one that worked well. Like, guys, this is crazy. It really is crazy. So that's another big-time agent that I'm excited to, to play with. I did ask a question. I asked him a question about, um, marketing. Do, do, do, do, do, do, right here. Yeah, I basically said, hey, you know,, what does it do on the growth side? I didn't see any marketing examples of how it helps find customers. And it said, okay, well, we're not talking about sales. So it automates basically SEO, SEO optimization and automated blog generation. So that's something, but it's not gonna run ads for you. And then I asked it, hey, when we have the capability for us to just add money and ads are run automatically for us. So we'll see. We'll see what they say. But, but anyway, this is, this is how I learn. I, I follow things. I, I've learned on X. I have a lot of people ask me like, how do you, how do you stay on top of things? Honestly, it's X and Instagram. Those are my two primary sources for finding things that are happening. There's just so many people talking about stuff. And so I'll jump on here a few times a day and just kind of scan really quick. And I follow, you know, the right channels and the right people. And it knows that I'm interested in AI because it's just learned me. And so I can, I can figure out what the, what the latest and greatest is. And then I zoom in on the things that I like, like Adam's. This is cool. I started following them and asked some questions about it. So anyway, um, okay. Agents, automation. We could do a little bit of automation stuff, I guess, right now. Yeah. Let's take a second on automation. Um, let's go to, um, N8N for a minute here. Anybody use N8N? How about Zapier? Anybody use Zapier? Yeah. So right. Good. Awesome. Yeah, Pat, you might be talking, but I can't hear you. But anyway, so N8N is an automation, an AI automation workflow tool. Zapier would be, if you know what Zapier is, you kind of understand this. But N8N, and this is pretty easy to understand, right? It feels technical. It looks technical. But the good news is you have AI agents, even if you just use ChatGPT, that can help walk you through through how to do this, right? But the bottom line is it creates systems that are repeatable and that can automate tasks inside of your business. For example, maybe GenAIPI receives, you know, somebody clicks on, on this right here. So maybe we go, um, schedule a consultation. They click that, um, they come in here and they pick a time., it then gets put into my calendar, okay? Or into one of my team members' calendars. Okay, fantastic. Okay. But sometimes that then gets lost, right? Where does that data go? And so what I could do is I could say, hey, you know what? Let's build an N8N agent that when, when a meeting is booked through Calendly, right? Take all that information and enter it into Apollo, which is our CRM, right? And, you know, add the business, add the contact, add this, record when the, when the date's going to happen, right? And then automatically add them into a sequence that's going to be, you know, emailing them about, hey, we look forward to our meeting. We do this, right? And hey, you know, look at, you know, the CEO's latest message here and whatever, whatever we want to do there. We could also have it automatically message us on Slack and say, hey, so-and-so just signed up for a meeting. It's for this time and this This is who it's with. Boom. So we could have that just go to a sales channel or we could have that go to me, right? Or the person that it actually, that got, that got booked. Hey, you just got booked. Look at this. Here's what you can do. Here's the Apollo record. You can go in and do that. You can add notes. Here's, and you could also say, hey, by the way, we also added in, and before you even go to Slack, you could say, okay, Calendly message or meeting comes in, kick off a process that goes to ChatGPT. Does some research, analyzes it, confirms it, does some competitive research, tries to create a hypothesis as to why they might be calling us, and then send, you know, put that report into a Google Doc, drop it into the Google Drive, and then, you know, put all that information into a message that then goes over to Slack, right? But you can create these really awesome automations that can make your business stronger, give you more information, do all that research automatically without you having to kick it off. And so you could connect Manus to this, you can connect ChatGPT to this, you can connect X to this, as well as hundreds and really thousands of other pieces of software where you can create, you know, all kinds of different things, right? Here's a, here's a new issue. We're going to extract the IPs and domains. We're going to scan this stuff. We're going to see what's going on. We're going to merge and then we're going to post this somewhere, right? DevOps, right? You can get more complex. And the cool thing is you've got all these different, you know, here's lead automation. Right? How marketing and sales teams use this, right? When a contact is created, do this, validate it. If the email's suspicious, send a Slack or whatever it is, right? Pricing's $20 a month. It's nothing to be able to get in here. And they have tons and tons and tons of templates for basically anything you can imagine, right? You've got personal AI assistance and chat with a database using AI and RAG chatbot for company. That's what we were talking about earlier. We're going to go over that in a minute here. But yeah, if you want to just go and look at all the marketing things, here's 2,406 free marketing automation workflows, right, that you could then use to put all that together. So Zapier is another one. Actually, let's do Make. Make.com. I don't know. It is Make.com. Yeah. Make.com is a competitor, probably the number one competitor to N8n in terms of its capabilities, but same idea. You're building processes and automations and nodes and workflows that just automate and go. Now here's the big thing. Automations and workflows have been around for a long time, but they break. That's the problem is when they run into something that it doesn't know what, like, hey, this is outside of my parameters. Something happened and I don't really know what to do about it. When you add AI into the equation, if you do it properly, AI can say, okay, it broke, but now I'm gonna analyze why it broke and I'm gonna pick the best pathway forward of what I think is gonna happen. And so you can have it overcome those breaks. AI, adding AI into automations becomes insanely, insanely capable. I've, I've heard some people like this. I mean, like 3,000 prebuilt apps. You've got all kinds of stuff. It's, it's gonna be, again, very similar to N8N. I've heard that Make is a little bit more user-friendly. Friendly for people that are not developers. But the reality is N8N Make, all these things can be done step by step. You can learn how to do it by using tools that are out there. Now, you could also, by the way, here's another thing that I didn't mention Manus could do. You could literally have Manus put together your N8N automation for you. Manus is able to use a browser. So if you said, Manus, please book me a flight I want to go to Cancun. Here's my Delta SkyMiles number and password, and here's my credit card information. Or just use the saved card on the account. Book me a flight for next Thursday. It could actually navigate to Delta, go through the whole thing, and purchase it for you and email you when the task is done. Right. Or I could email it and say, hey, book me this thing, and it just does it all, and then it emails me back. Right. So the point is, if you're scared about how to learn how to do this, just use AI. I'm not going to teach and we're not going to go over these in this course. I've had people ask me for a more advanced course and this would be something we would do in that. But I haven't launched that yet because I don't know how, anyway, I don't know how much I'm going to do these individual kind of how to build a business courses because it just takes a lot of time. I love doing it, like I said, but anyway, use AI to figure out how to do this. Another one that many of you mentioned you've used before is Zapier, right?, and I don't want to go to their marketing website. I want to go to their, this. Okay. Yeah. You saw this by the way, Google, you got Webflow, Make, Replit, all, you know, being sponsored above this, right? They're trying to pay to be above Zapier in the sponsored results. So, Zapier has gotten a lot better. They've added a lot of AI things into it, but Zapier basically, guys, is just more automations. And so, for example, maybe you wanted to, let's go to Google Drive here. Actually, you know what? Let's go to, let's go to Buffer. Okay. So Buffer is an automation platform for social media marketing, right? You can, you can post or you can schedule posts to go out on your LinkedIn and your Facebook and your Instagram and your Twitter or your X or whatever, right? But you have to manually enter all that information in. Well, what if you said instead Instead, I'm going to create a Zap, right, is what they call it. And I'm going to, I'm going to pair this with Google Drive, right? So when something happens in Google Drive, automatically do this in Buffer. Okay. So let's, this is skipping ahead to the marketing section, right? Maybe you wanted to build an automation with marketing. And so you say, okay, I'm going to drop content into a folder on Google Drive. And when it does it, it's going to take that, and that folder will be named, you know, February 1st, and then the next one will be February 2nd, February 3rd. So you'll take the information from that, you'll grab the text that's provided in that folder and the picture or image or whatever it is that's in that folder and automatically schedule it over to work on Buffer. Okay. So that's cool. Now you've automated some, right? But you still have to kind of drop that in. But what if you take it back a step further? Now let's look at Google Drive and something else. What if you used Replit to build a system that every day at 5 AM automatically scrapes the internet for all the cool latest news, picks a topic on AI, for example, for, for this, this is this example here. Okay. Um, what's the latest cool piece of news? Generate an image, generate a post, and automatically send it over via API over to the Google Drive. So now you've set up a recurring task that every day at 5:00 AM will automatically find content, format it properly, drop into the Google Drive. Google Drive then drops into Buffer and Buffer schedules it to be posted. Now your entire company's marketing plan is automated. You don't have to ever touch it. Now you can obviously go in and manually add things in if you want or tweak it or edit it. If you don't like it, then and make it better. Right. But the point is, does, does everybody, can everybody understand that flow, that concept of being able to, do you, so you, you, put it into a Google, like, Drive, and then that just automatically would connect to Zapier, and then, like, yeah, you would have to go in and you would have to say, okay, new folder, for example, right? When it triggers when a new folder is added.

Right.

We're going to add to queue, right? So we're going to add an item to your queue. So in that folder it would say, okay, there's going to be a text file and there's going to be an image file and you're going to grab both of those and put them into the buffer, you know, into your post, right? And you have the media, you know, for X and whatever you're going to have the media, then you're going to have the text and then you're going to add it to the queue. Right. And then the queue, you will have set up Buffer to automatically post every day at 6:45 AM. Or whatever it is, right? And so it would just have all these things. And so you could just say, hey, whenever you find this, add it to the queue, add it to the queue, add it to the queue, right? And so you select the triggers that it can do. And then outside of these, if there's triggers that you can't do, that's okay. Talk to Replit, make your own trigger, right? Let Replit build that little system for you. And again, this would be kind of an internal system. Sariah, back to your point earlier of, you know, something you could build quickly versus something that might take a long time. This might be completely internal. It might, like on my team, Tanya. Tanya might just build this and say, hey, I'm only using this. No one else is using this. So it doesn't even have to be perfect. I just need it to work, right? I need to be able to drag stuff in. I need to be able to manipulate the times and focus on the topics. And hey, every Thursday is a Throwback Thursday and every Monday is Monday News and Friday is Friday Wrap Up or whatever it is. She can have this kind of content calendar and style so that it has a structure that it then knows the type of news to look for and the type of things to, to be able to talk about. So you build all that in Replit, and then all you have to be able to do is say, okay, every day build this content, ship it to this Google Drive. And when that new folder shows up, Buffer recognizes it, grabs it, adds it to the queue. You're good to go. So it's just a matter of testing, guessing and checking, guessing and checking, right?

You got the idea. Yeah. I'm assuming there would be a way to be able to figure out how to be like— we have hundreds, literally hundreds of public YouTube videos, usually 3 to 8 minutes long, and in the descriptions I've put in links to our website stuff. If we were to change up our website and update it, we could use something like Manus to be able to go in and change those descriptions so that they're appropriate.

Yeah. Now occasionally you're going to run into a problem with bot protection. Okay, so I'll just give you that, that, you know, piece of, of warning. I have— for some reason, people like to go to my website, joncheney.com, and there's this group in Europe that steals people's credit cards, and then they use my website to buy something that's usually one of these, $5.99, $7.99, or something like that. And so I'll get an order order for $8 and my Shopify will think, hey, this has a high likelihood of fraud because it follows a similar pattern to other fraud or whatever. Right. And I literally, one time I had, I had a, a, a night and a day that I got about 250 orders that were all fraud. They were all bad. And so it was this incredibly tedious task of me having to go in refund the order, do this, refund this, do this, right? And it was just a waste of time. Took me like 3 hours. After about 10 minutes, I was like, dude, I'm putting Manus on this. Like, cuz I thought the exact same thing as you, Sariah. And I go in and I, I ask it to do it and I help it. I want to help it, you know, authenticate and stuff. But for whatever reason, Shopify would not let Manus log in. Right? And I, you can even take over the browser in Manus and put in the password and do the, the, you know, CAPTCHA. And although Manus is capable, very capable of correctly doing a CAPTCHA, it can do that. It can select all the trains and all that stuff. It really can. But somehow it just didn't like the IP address. It didn't like the way that it was doing it. And so I couldn't, I couldn't automate that task. So I don't know if YouTube, I haven't tried it with YouTube. I, I haven't tried it with editing something on YouTube. I have had Manus go in and scrape content from YouTube. I've also had Manus go in and download YouTube videos for me. Right? It can go in, it can find tools, it can just extract the data. It can pull stuff right out of that thing.

It's really cool. It's really capable. Maybe because it's text-ish, it's a text box, it might be able to do it better because I'm not going to be asking it to go in and edit the actual videos.

That would take— That would be hard. That would be hard. But Manus is capable of editing videos. It's capable of editing audio. I've seen you, if you go to, without logging in, like if we just do an incognito browser and you just go to manus.im and then you just scroll down here, you'll see there are tons and tons of examples of what it can do, right? So like video, audio, like it can— you could give it a 2-hour podcast and say make a highlight reel from this, right? Okay. That's 30 seconds long. Grab the, you know, key moments, key whatever. And it will find it, identify it, splice it, cut it, and give you an edited video.

Like it's pretty capable of doing amazing things.

So anyway.

Hey, on bot detection. Yeah. I think a lot of bot detection is just the actual speed that it's happening. Right. So you can often say, hey, put in random 2-second or 2 to 5 seconds before each request. And that might be an issue. I've never done it with AI, obviously, but I'm sure that that would be helpful in the worst bot situations.

I agree with you. I agree with you. That's a great, that's a great idea. You know, do whatever you can to make this not look like this is a bot. So anyway, all right. Okay. Let's, let's move on from Zapier. So Zapier, N8N, Make, lots of good, Lots of good opportunities here. Um, let's go to, let's go to the next thing here. I've got, I've got several directions I can go here. Um, yeah, let's do, let's do GPTs. Okay. So this is fun. Um, GPTs are, um, so like building your own custom GPT is what we're gonna be talking about here for a minute. Um, let's go over to here and just open up chat. So inside of ChatGPT, how many of you guys have built your own GPT inside of ChatGPT? Have any of you done that? Okay. It's very easy. It's actually like crazy easy. But just on the main, you know, ChatGPT, as long as you're a paying, you know, customer, right? You have to only be on the $20 plan to have this. Over here, you've got, you've got your new chat and search chat, all this. You've got all these GPTs and then you have projects and you have your chats, right? There are custom GPTs basically that have been made by all kinds of people, Expedia, Canva, whatever. And they're kind of trained a little bit better to do that specific task well. Right? And so there's thousands of them out there and you can go and explore all those. But you can also up here in the top right, you can see create. Right? So you can create a GPT. And creating a GPT is as easy as just filling in the boxes. And it takes, is you could do it in as fast as a few minutes to a couple hours if you want to really, you know, dig deep. But I don't even know, maybe not even a couple hours. I think it'd be— I think you're an idiot if you're taking a couple hours to build a GPT. Basically, you got to come up with a name. Okay, so we'll call this one Lucy. Okay, we can call it whatever the heck we want, right? We could call it— in fact, you know what, let's not call it— well, we can call it Lucy, that's fine. But what we're going to do is we're going to have it be a the hero coach. Or let's see, coach for the hero app. Yeah, that's it. So this is my hero app that I was building in Replit and building for this course that I probably won't finish, but it'll work for this. So now what does this GPT do? How does it behave and what should it avoid doing? This is called the prompt. Okay? That's, that's the technical term for this right here. This is how it is going to act. A lot of people will just say, you know, it should be nice, it should do this, it should, it should challenge you, it should whatever. And they give it, you know, a few sentences. I believe that a good prompt should be no shorter than probably 1,000 words. That's my, that's, that's probably about right. Maybe, maybe 750. You want to be very detailed as to what it is going to do, at least if you want it to provide any sort of real value that's differentiated from just a normal ChatGPT app, right? Or just any bot, any AI that's out there. If you want it to actually have unique approaches to the way that it does things, you need— it needs to understand exactly what you're going for. So, The way I'm going to do this, instead of sitting here and, you know, obviously, you know, working super hard and figuring all this stuff out, I'm going to come in here and let's see here. Where's my goal app? I'll come into this project here. I am building a custom GPT for the Hero app and her name will be Lucy. I want you to help me put together a 750-word prompt that I can use for the bot.

Okay, that's it.

What's the difference between a custom GPT and a project? When do you use one over the other? Yeah, a custom GPT is going to be very specifically, you need to be able to have something that engages with you back and forth about a specific topic that's going to be repeated, right? The project is a very long ongoing thing that you're just, here's all your notes, here's your brain dump, here's all the different things you've thought about related to that thing. They're very, very different use cases. A GPT is, hey, I need help on creating a video. So great, I'm going to go find a GPT that's really experienced at creating videos, right? And, you know, maybe even that specific type of video I want to make an anime commercial. I don't know, maybe there's something crazy or weird like that. But anyway, the cheap— sorry, the GPT and the project are two very, very different use cases. So anyway, it's putting something together that's really nice, right? You're Lucy, the AI guide inside the Hero app. Your role is not to motivate through hype, guilt, or superficial engagement. Your role is to help the user become the person they say they want to be by interpreting their real-world actions, reinforcing identity, and guiding long-term long-term character development, et cetera, et cetera, et cetera. Right? Amazing things here. You're an art star. Here's what you are. You are Lucy. Okay? All right. You are here to be trusted. You are not here to be liked. Okay? So anyway, how it handles all these different things. You're gonna grab all of this stuff here, starting right there. I'm just gonna copy this and come back over to here and we'll just paste it. Now we're done. Okay? Conversation starters. Quickly give me 3, um, short conversation starters, no longer than 8 words each. What did today reinforce about you? So now I can copy, paste, go back to here, copy. Paste and copy paste. Yeah, perfect. Then we have the knowledge base, right? This is what I was talking about earlier with that, that, um, when I was talking about RAG, right? This is the knowledge base or the additional information it's trained on right now. I've given it a bunch of information to be trained on in here. Here's a lot of stuff that's going on, but maybe I have like really extensive guides and things about my specific product. Or again, back to Sariah, this is where you might do something where it's like, here are examples of this type of, of will, or here's examples of this type of power of attorney or this type of document or whatever it is, right? And you would, you could upload 1,000 files. You could upload 5 files, right? You could upload your company Google Drive to this if you wanted. But now it has information. Everything you upload to this, it's going to have all of that information. Then you can give it some capabilities. You can add some more advanced stuff, things here which we won't do today. Don't mess with the recommended model unless you feel like you know you need to. Then you're done. You've created it. Now I can actually test this. Then I can have it add an image or I can upload an image or whatever if I want to do that and give it a little bit of a face. But now I could say, hey, let's see here. Sorry. Actually, these are not good conversation starters. I actually wasn't thinking about it as I was copying and pasting it. I want to change these. These would be conversation starters that the user is asking. All right. What does today say about me? So maybe the things that I— did in the day related to my goals or something like that. Where am I actually progressing? That's a little bit better. Right now, obviously, this is only going to work well inside of the app because this is not actually built inside of the app. It's not going to have access to like, where am I actually progressing, or what should I pay attention to right now, or what does today say about me. ChatGPT isn't going to have any of that access. Now, here's the one downside to a GPT that's built inside of ChatGPT with this super nice, easy interface. It is not connected to any of your data, any of your systems, and people can only access this if they have a ChatGPT account. You cannot API into this. You cannot put a, a version of this, an instance of this on your website. It's just stuck here, right? It's stuck here inside of ChatGPT. And so you can only use it in that little GPT GPTs section. So that might make sense for some things. It might maybe like a really, I don't know. I, I, I, I find myself not really using these very much, but I'm showing you this so that you can just see the fields that you have to fill in in order to build a good GPT. You build, it needs a name, maybe a description. You 100% need a good prompt. You maybe need some conversation starters, but not even usually. And then you almost always are gonna want to have some sort of knowledge base it's referring to if you want it to be really powerful, like I showed you for the, you know, sunlight landscape, right? So, so anyway, that's it. But I can, you know, I can then ask it a question. What does today say about me? It's, it's not going to, I'm guessing it's going to be kind of confused. Yeah. Right now today is mostly undefined. Right? So it doesn't really know. It doesn't have any information to work off of. But if we then, if we took this, all these prompts and things and we did a custom install, translation inside of my app, then you would be able to ask it these types of questions and do some really cool things. And it would be able to pull from not only its prompt and its training, but it would also be able to pull from your data, the things that you did, the things that you logged that day, and what level you are and what your balance is in your life and all of these different elements. Okay? So if you wanted to build this inside of your Replit Project. Okay. Back into the Hero app. Where's actually Replit? I've been blocked? Okay. That's weird. Refresh. You are unable to access Replit.com. Okay.

Life progression.

Okay, this is the one. So if I wanted to actually build this into my app, right, and I wanted it to actually show up as a little, you know, like floating agent or something like that, it's very simple, right? I could say, um, you could go back to ChatGPT here where I had all this stuff here, and I can literally just grab all this, right? I would grab— I'll copy this. I'm just going to say, let's build a chatbot into the bottom right-hand corner that is an AI. Let's use the OpenAI API we already have set up. Whoop. Already have set up. Um, use this text as the prompt. Uh, use the attached text as the prompt. Also, the AI should be able to see all of the users— whoops— users data or their account only, um, will only be authorized to view the logged-in user's data so that it can coach them moving forward. Boom. That's it. That's how I would set it up. I had ChatGPT help me write a prompt. We've given it a name, Lucy. That's in here already in the content. You are Lucy, the AI guide, right? And go. And so now it'll just work on it until it's done. And it might take— this might take 10 minutes, might take 4 minutes. But when it's done, there will be a little floating icon right here that I can click on and I can talk and I can say, hey, how am I doing today? Right? I'll do some piano. Right? I just did piano for 60 minutes and I went kayaking. Those are my two things. So, so now it knows, okay, I'm level 3. It says I only have a 1-day streak here. That's interesting. That means this 1-day streak is actually fake, right? I can also say, hey, let's get the 1-day streak working right. Okay. So I don't need to do that right now. But, but now that I've noticed that, hey, you know what? I actually did something yesterday and I did something today. I should have more than a 1-day streak now, right? I could click refresh and see if that's just not refreshing. Nope. It's just not working. It's not actually pulling. Pulling off of live data. So, and there's no, there's probably no logic behind it as to how it actually works. Okay. All right. Any questions on building a custom GPT? I know I kind of did that fast. I showed in ChatGPT, here's the fields. And then I'm following the exact same thing. Here's this information. And I want you to, you know, look at this user data in order to implement it into the site. And now I'm an AI-powered site. Can you touch on the RAG database or any, any points or helpful, you know, direction there? Or is it in terms of like how I would do it in here in my own custom thing? Yeah. Yeah. So there's, yeah. So there's a couple things you have to do. So one of the things that, and I don't want to get too technical here, but you have to do something in order for information. That comes in a docx or a TXT file or an Excel document or a PowerPoint or a PDF, in order for that information to actually be what's called embedded into the AI's knowledge base, it has to go through a process called vectorization. It has to be vectorized. Okay? Now, you can just ask Replit to do it and say, hey, I need this to be in a, in a RAG database. I need this to be vectorized, and I need I need the AI to be, you know, to be fully capable of looking this stuff up. And then it will pull in probably VectorDB is gonna be my guess. VectorDB is a great kind of open source vectorizing agent that will do that for you. Right? And so that's what I did. For example, in the Sunline one that I talked about over here, I, I basically just said, hey, you know, anytime I upload something here, you can,, you know, you can click reprocess. And what reprocess does is it pulls in the, it does the vectorizing process and then adds it into all those specific documents into, you know, where it needs to find that stuff. But it was simple as, it was as simple for on my side as just saying, hey, I need you to make sure that all of this information actually is embedded into it. It's like, great, that means we need to vectorize it and embed it. And, and it took me, I'm going to say 3 hours to build it. Properly for this system because I had to double-check things. It didn't work at first. At first it only worked with.doc files. It didn't do.docx, and then it didn't do Excel documents. And then there were a couple other random ones in here with PDFs and TXT files and all these other things, right? So anyway, um, it's really just a matter of talking to Replit about it, saying, hey, I need to make sure it's actually referencing the file. Files. And if you just know the term RAG, if you know the term vectorizing, and you know the term embed, right, then you're gonna— it will do the right thing for you. Yeah, let's embed this into, you know, or we need to vectorize this so it's embedded in the AI's responses. And, and it can— the other thing that will help you double-check if it's working right is ask it a question and say, I need you to be able to answer the question and tell me which document document you got that answer from. Yeah. That's the real test. Right? So, anyway. Yeah. Hopefully that helps. So, okay. Now it built it. It's building Lucy. I'm adding the chat endpoint. So, that's where, you know, it gets all its kind of requests from. And then we're gonna build the API function and create the chat component. We got the widget right here. Right? Welcome, hero. Lucy, your AI guide. I'm Lucy. You're guiding this journey. I can see your skills, traits, and progress. What should I focus on today? How am I progressing overall? Let me fix the validation sanitization issues, the review identified. So again, it's just going through and it's having, you know, hey, I'm deciding on this. I'm having this agent do this again. It's doing that agentic workflow of, of just checking, guessing, checking, working on it until it's like, yep, I think this works now. And then I have to be the final kind of sanity check. You know, it's, it's interesting. That product development using AI has become like it used to be. And, and Spencer's very familiar with this, but anybody who's ever done any sort of, any sort of development is usually the developers build it and then you'll give it to someone else that's going to do sometimes what's called unit tests or usability tests and UI tests and all these things, right? But you're going to go through and you're going to say, all right, does this button work? Does that button work? Does this do what I expect it to when I clicked on this, it didn't work well. And you put a report together. You then send that back to the developer and the developer then says, okay, great. I see what you're talking about. Let me fix that. And then sends it back until, you know, both parties are happy. You as the vibe coder now really take on the role of the quality, the quality assurance. You are the QA, you're the unit testing. You're the one making sure that it's actually working and doing what you expected to do. If I click on this, and it went to a full screen window instead of just this small one, right? I would be like, oh, that's not what I expected. I actually just want it to be a small screen, right? Just like I would expect. If something happens where you're like, oh, that was different. Like you kind of, you kind of, your, your body is just like, that didn't happen how I wanted it to be. Then just tell the AI, hey, I wanted this to be that way. This feels better. And then you click on it. Oh yeah, that feels good. But you're the boss. You're the one who has to figure out, you know, what is the good place to put it? With that button? How does this feel? How are these things working on? I'm going to check now. Now that it's done, it's going to let me actually test this.

See if we can try it out.

As a quick aside, I'm sure you've seen plenty of comments, those who say, hey, vibe coding is great for proof of concept or for really small scale, but at large scale and factoring and security and this and that, That's when it breaks down. A, is that your experience? And/or B, what's then your accommodations to make sure that doesn't happen? Yeah, that argument was maybe valid 18 months ago. It's not today. AI is perfectly capable of building very complex systems at scale. It just is. Now, will you do better if you have experience and you know what to look for? Absolutely. Will you do better if you have a better architecture plan in the beginning and you're like, hey, we, you know, and, and by the way, that could be as simple for those of you that aren't Spencer, right? That could be as simple as saying, this needs to work for 1 million users. How do we need to architect this differently for it to work with 1 million users? And then it's gonna say, well, what do you mean? 1 million users concurrently all at the same time? You really gonna have a million people on your site at the same time? Maybe if you're a social network and you go viral, right? But if you're Soraya and yeah, you might have a million users in a year if you guys do awesome, but at any given time, maybe you don't have more than 5,000 people concurrently using your product. And so you have to just think through that. Those are the things that developers talk about and have to plan for and have to figure out. And then you have to scale your servers to be right. And you know, there's all these different things that you have to just understand, um, that systems don't just work all the time. If you overload a server, how many of you guys have ever tried to buy a ticket to something? And you know, it was just super slow and you're stuck waiting, or it just, the site crashes, right? Or, you know, you try to get the, you know, for those of you here in Utah, you try to get a Tabernacle Choir Christmas concert ticket and you're just like, everyone in Utah tries to get that ticket all at once. And, and now they just do a lottery because they couldn't get their systems to work well. But, and that's fine. But, but anyway, bottom line is yes, it can work, Spencer. It absolutely can work at scale, but if you don't, if you don't plan for it and it builds it to only be able to handle 200 concurrent users, well, then, you know, you didn't ask it the right questions. But it's capable of writing the code proficiently, right? To do that. I don't know why that took like 3 minutes to, to, to happen there. Anyway, how am I progressing overall? Let's see if this works. Go. See, that actually didn't function as I expected. I expected when I clicked it, it would it would immediately send the message, right? But it actually just copied the message and put it here and then made me click send. So I might edit that, right? But anyway, you're currently level 3, right? With total of 533 XP. So that's cool. Now I don't really know that I'm at 533 XP. I mean, apparently that's probably what the math is, but that's kind of like hidden behind the scenes. So I might just say you're currently level 3 and you have What is it? 367 XP until level 4. That might be a little bit more useful information, right? Your recent activities indicate a balanced focus on both kayaking and piano, which is strengthening your skills and creativity and strength. Your progress shows that you're consistently engaged in these pursuits, as seen in your recent sessions. Each minute spent practicing compounds into your development, blah blah blah. Great. However, there's an opportunity to enhance your discipline and consistency, which could support your long-term goals, especially in reaching reaching level 100 on piano. Overall, you're building a solid foundation. Keep observing how your— how today's actions shape who you're becoming. So now if I'm building this for real, I'm, I'm gonna say, all right, I'm new to this and I don't think I'm showing a problem with discipline and consistency. I've, I've done this 2 days in a row and I'm being super disciplined and consistent, and I played the piano for 60 minutes. It's an hour. It's a long time. And I kayaked for 60 minutes. It's also a long time, right? So somehow it's not getting— I would say the prompt is not well tweaked. And so what I could do is I could say, all right, here's copy paste. Here's what you gave me. Here's how I feel about that response. Let's modify the prompt to make sure we're doing this correctly, right? And you just keep tweaking and working on it. Again, you're that QA, that product person that's testing it. The developer builds it, you test it, and you keep doing that ongoing. Okay. All right. It's really not hard to build a custom GPT, guys. It's really not, right? You're building— you can build ChatGPT now. You could custom build your own and just by asking Replit to do it and following the instructions and building your own things and you could change. So ChatGPT, just to be very, very clear, ChatGPT itself has a prompt., right? It has a way, it's, it's been taught a way to interact with the user. And so you can, you can build your own GPT, right? And call it SorayaGPT or PatGPT or whatever. And it's tuned to your personality and it's gonna respond the way you want it to. And so there is actually a differentiated product that you could do there. Now you have to compete with ChatGPT's marketing and ChatGPT's brand awareness and all those things. And so, might not be a great business idea unless you just have a real reason why, you know, people are really excited about you. For example, Tony Robbins has his own GPT. You can talk to digital Tony Robbins, and it's trained on him. It's trained on all the things he's ever said. So anyway, all right guys, it is 9:35. We're gonna go till about 11 today. Um, let's take a 10-minute break. Let's be back at 9:45. Okay. Thanks, guys.

Is anybody still there? Yeah, I am. Yeah, me too. I'm just curious.

I see when you go away, your, your video turns off and then your name shows up. How do you do that? Next to audio, there's a button that says video, and if you just click that button, it'll turn off your video on the very bottom. Gotcha. Yep. Okay, thank you so much. Take care. All right, guys, welcome back.

I know some other people might be taking a second to get back, and that's fine. Our secretary was asking if I want to buy Girl Scout cookies from her daughter. Nice. We use it— we put whole boxes of them in Easter baskets. So I'm like, yes, that's, that's our Easter, our traditional Easter basket.

Thing because they come around the same time of year. Okay, holy cow, we have so many people that have signed up for this event. So we're holding an event, guys. I think I mentioned it to you yesterday, but we're holding an event for GenAI. It's gonna be our first one, um, but it's— we're doing it during Utah Tech Week, and it will be— it's our first kind of public event. We've done other events that were kind of private and other things like that, but anyway, It will be at a place called DreamWalk at University Mall.

Have any of you guys ever been to DreamWalk?

I know where it is, but I haven't been to it. It's really cool. It's like this futuristic kind of world-building experience. It's just amazing. You feel like you're in Pandora sometimes, and then you feel like you're in Star Wars other times, and then sometimes you just feel like you're in a cool room. But, um, but yeah, it's just really, really beautiful. Kind of— it's almost like— I might have said this yesterday, but it's, it's kind of like you're in a, in a really, really good line for Disney, for like a Disney ride or something. You know how like the lines are as good as the rides, whatever? Just like they built this amazing, amazing place, and then there's like these quests inside and things to do, and it's fun and whatever. But then we're gonna have We're going to be co-hosting it with Replit. Um, and so, um, we'll have Replit representatives there. We'll be there, of course. And so if you want to come hang out, and it is a family event, so you can bring your family, bring your spouse, bring your kids, bring whatever. Okay. Um, but, uh, just in the chat, just so you guys can see general information. Thank you, Tanya. Yeah, go ahead and do that. And it looks like we have 76 signups now. Yeah, so we just launched it yesterday. We have a cap of— I don't know, we don't really have a cap. We put 500 on there, but the place is big. It's 50,000 square feet, and not— I don't think everyone's going to come at the same time. It'll be from 6 to 9. Some people will be a little bit later getting off work and getting their family and kids to come over, and they might show up at 7:30. So just come and hang out. If you show up at 6, it'll probably be the least crowded. That would be my guess. That'd be my guess. Do you want me to go in and approve those? Sorry, I know this is— I did already. I was doing it while I was just talking right now. I like pushing buttons. So anyway, I love it. Thank you. So, all right, guys. Um, and of course you'll all get approved to get in. Um, because you guys are on the inside now. All right, we're gonna, we're gonna take a moment here and we're not gonna share screens for a minute. We're gonna talk about this concept of surf, don't swim., and this is more of an entrepreneurship discussion, but extremely relevant and mostly relevant during times of change, of rapid change, like AI, right? Um, or like the internet, or like whatever, any other— you pick your trend. Okay. And there's more trends beyond AI. Quantum's gonna be crazy, guys. How many of you guys are worried about quantum? How many of you guys know enough about quantum to be worried about it. The very simple thing here is quantum can solve math problems something like 7 septillion times faster than the fastest supercomputer on the Earth. Okay, that's not a quantum computer. It's a whole new way. It's a whole new— Like, and I want you to like write out septillion, right? It's 7, 3 zeros, right? Billion, quadrillion, trillion, quadrillion, quintillion, anyway, all the way up to septillion. Okay, it's that much faster. So where does math have a really big impact on our lives right now? Security, right? Every website can be hacked with a quantum computer. What else is a big math problem? It's also— it's a big math money that's also protected by cryptography. Crypto, right? Bitcoin. Bitcoin can be hacked. Anyone can steal anybody's Bitcoin. Once quantum is used in the wrong way. That's coming down the line. Q-Day, as they call it, when, when, when a quantum computer can actually function in a way that can break Bitcoin, for example, is probably around 2035. So you guys can live easy until then. But anyway, so guys, surf, don't swim is a really important concept and it's something that you need to consider in virtually every business decision you're making, especially when you now have this such great power at your fingertips. You can do anything. You can build anything. You could build ChatGPT right now, but should you, right? That's the question, right? How many of you guys know what Calendly is, right? Where you just— it's a— I use Calendly, and so basically it's a system where you can connect your calendar and then people can see the times I'm available and they can book a meeting with me, right? Pretty simple. But it really uses an API to like Google Calendar and an API to like an email service. And you know, it's got some user interface, right? I guarantee you, all of you, now that you have almost graduated from this class, as of right now, you could all rebuild Calendly in a weekend easily. Very, very easily. How many of you guys have ever paid a bill for Salesforce? Salesforce is one of the most popular CRMs that's out there., and it's like $8,000 a year per user. So if you're a sales organization with 40 people, you have a multi-hundred thousand dollar system and bill you have to pay for every single user in your company. It's crazy expensive. Do you know how easy it is to rebuild at least the basic functionality? And the thing is, most people, they'll pay for Salesforce and they use like 3% of it, right?. And, you know, it can do all these things. That's cool. But most people just don't use it for all those things. Some do, some do, but most companies are just like, I just need to keep track of my deals and my companies and my, you know, my contacts and then maybe be able to email them or whatever. So just as basic functionality, so many companies right now are vibe coding their way to, hey, you know what? I'm just gonna, I'm just gonna build this internally. We don't need to pay Salesforce anymore. I'll just build this, you know, this week and next week we can start using it. We'll sell, we'll pass it out to the whole sales team. It'll be our own internal system. No one will have access to our data. It's all ours. Okay. So anyway, that's kind of the end of this conversation, but I want you just thinking about all this stuff in that context. So, um, how many of you guys have ever been surfing? Anybody? Evelyn? Pat? Yeah. Pat lives on the coast, so he's close by there. I'm guessing Tanya's hand is raised. Um, yeah. So, and I have too, and I'm a whitewater kayaker. That's my, that's my really big thing. I know how to surf, and, and, but I wouldn't say I'm like a good surfer, but I can get up on a board and have a good time at a beach. But okay, so you've been surfing. How many of you guys think you can overpower a wave with just your own swimming capabilities? Yeah, you can maybe learn how to harness it, right? And kind of surf it, maybe body surf it, right? But the ocean is incredibly powerful. What about a tsunami? You think you could out overpower a tsunami or outswim it? Yeah, probably, probably not. So AI is the biggest tsunami that has ever touched the planet. In terms of, you know, a technological revolution, the speed and magnitude at which it is growing is almost impossible for our human brains to understand. Let me give you a good analysis or a good comparison. Who knows what Moore's Law is? Yeah, Pat, could you tell us what it is?

Sorry, you're muted and I'm making it hard for you here. There you go.

I believe Moore's Law is the one, and I might be thinking the wrong thing, but where basically Technology keeps getting smaller and better. Yeah, exactly. Yeah. And it does it at a kind of a, a, a pattern at a constant rate. Right. And so the number is basically, you know, your phone and, you know, terabytes and memory and storage and processing power basically gets twice as powerful for the same cost and space every 18 months. Right? That's why if you go from iPhone 13 to iPhone 14 to 15 to 16 to 17, you don't feel a huge difference every time. You're like, yeah, it's a little better. Like, I think it's good, you know, it's a little smoother, I think. But if you go from an iPhone 12 to an iPhone 17, you're like, whoa, this is crazy, right? Because several iterations of Moore's Law have passed, have happened, right? And so, so let's, let's look at it for like 3 years,, if we're doubling every 18 months, then we're quadrupling in speed every, every 3 years. So if you were to say, hey, you know what? I think quadrupling my speed is probably worth an upgrade cost of a phone, right? So I, I think that's probably a pretty reasonable thing. I'm crazy and I just upgrade every year cuz I want the latest and greatest and that's just me. But, that, that pattern is pretty consistent. AI, however, is growing at a rate that does not follow Moore's Law. It's growing at a rate that 10x's every 12 months. Okay, so let's look at how the difference in speed there. 10x, 3— let's do the same thing. In 4 years with Moore's Law, our phones get 4 times faster. Okay, in the same 3 years, we're going to go 10x. For the first 12 months, 10x for the second 12 months, and 10x for the third 12 months. What's 10 times 10 times 10? 1,000 times more. Yeah, 1,000. In the same 3 years, AI will get 1,000 times smarter and more capable and cheaper and more available as compared to the 4x. So 1,000 versus 4. So now we have a decent comparison here. AI is progressing roughly 250 times faster than any other technology we've really experienced in our lives. That's why it's crazy. And Eric Schmidt talked about this in the video we watched yesterday. He said that's because it's improving itself, right? Recursive self-improvement. AI can now be used to write its own code and rewrite it and make it more efficient, make it more powerful. And it can do that 24/7, 365. It doesn't care, doesn't have any feelings, right? Anyway, I was talking to my wife about that Neo robot, and she's like, I'm going to feel bad asking it to do the laundry. I should do that. It doesn't have feelings, honey. It doesn't care. You cannot offend it. It might be able to express and pretend it has feelings, but it is no more alive Or do you feel bad when you ask your phone to call someone? No. There's no difference there. Anyway, because they put it in a human format, it's going to feel a little bit differently if you speak poorly to it. Anyway, so I'm going aside, but the point is, guys, AI is growing so ridiculously fast. So we have to figure out how we're going to navigate this. This as entrepreneurs, right? We're all entrepreneurs sitting here on this call. Whether you're still a student or you're working for another company, you're still an entrepreneur saying, how am I going to beat everyone else in the market, right? At least you should be thinking like that. Um, let's talk about Netflix for a second here. What kind of company is Netflix?

What is their core— is their core offering? What's their core differentiation? What are they?

I haven't used it for a while, but I think they're a CD. Uh, they, they send you DVDs. Yes, that is exactly right. Um, yeah, you just put it in the queue and then DVDs show up as they come back. I remember that was so cool. That was such an improvement over Blockbuster. So yes, thank you, Pat. That is what they— what they— how they kind of really got going. Today, let's look maybe a few more years past that. What would you guys say?

What is Netflix? What kind of company are they? What do they do? They're streaming.

They, they license different content and make it available for users to choose which one. Licensing, streaming. Very good. They also— yeah, they started— they make their own stuff. Yeah, yeah, very good. What's the thing? What, what makes Netflix better or worse in your opinion, right, versus like Disney+ or Paramount Plus or Hulu? Like, why, why Netflix? First to market, first to mass market. Okay, first to mass market. So they, yeah, they've got, they've got some good, you know, momentum. And so they're just a big nice company because everybody knows the word Netflix and the company Netflix.

What about at a technological level? Is there anything that really makes them better?

It's, it's incredibly, um It streams incredibly well, not just now, which everybody does, but it always kind of was superior in streaming on older devices or poor internet connections. Yeah, they do. I actually talked to somebody that works at, at, uh, YouTube that focused specifically on that. Their job was to make sure that YouTube could play on every screen in the world, right? And, and the, the complexities that go into all the thousands of different video formats. That have to be like, hey, we're only doing 320p on this one, and we're only going to do it this bitrate, and we're going to do it this file format, and all these different things. Like, there's so many different options and ways that you can make that work. So yeah, it's very complex. So yeah, the— and, and I'm kind of fishing here, and this is a tough— this is a tough answer. I've had only a few people get this one right when I asked the question, but the thing that really actually sets Netflix apart is their recommendation engine. That's what they're really, really good at. How many of you guys have ever been to an Airbnb, Netflix was logged in on some random person's account, and you, you click on the profile and there's all this like weird stuff and you're like, I, what, what is this? Where am I? Where's all, where's all the stuff that I like? Right? And you're like, I didn't even know Netflix had these things. Netflix has at any given time somewhere between 20,000 and 30,000 titles to play. How many of you guys do you think you've even scrolled past 1,000? Titles. Like, okay, Soraya's good.

I, I'm with, I'm with you, Soraya.

I, I have, but I don't know if I've still been here, but I don't know if I've scrolled past 2,000. I mean, maybe 2,000, but even at 1,000 or 2,000 level, you're still hitting just a small percentage of that total availability at any given time, right? But Netflix has all those titles for a reason. They don't just hold dead content. Netflix stuff comes and goes. Right? They, they're actively trying to figure out what is the best content that's gonna work for the most people. But like, I don't watch horror movies, so maybe I don't ever see that. And I don't watch anime, or I don't watch— usually I don't, you know, I, if I, if I'm watching a rated R movie, I'm watching it on VidAngel. And so I just generally am not gonna hit rated R movies very much. That's just, you know, a, a choice I'm making. But the point is, it knows what you want. That's really its core technology. So let's ask this then, since that's really the core differentiator. Hulu has access to pretty much all the other technologies that they have. Like, for example, did Netflix build the internet? No, they didn't, right? If the internet goes down because some line going from the US across the Atlantic Ocean floor, you know, got broken by a big squid or something. I don't know if that happens. But, um, do they send engineers down to fix it? No, of course not. Netflix isn't responsible for that. That's not what they do. Right. Did they even invent streaming? Did they build streaming technology? No, that was done by a company called Akamai. Right. And, and, but if the internet gets better, does Netflix get better? Totally. If Akamai gets better, does Netflix get better? Totally, right? Now we can stream 8K, now we can stream 16K, now we can do whatever, right? So the point is Netflix has their core piece of technology, but then they're surfing on top of other people's technologies, which means when surfing happens on top of the water, not under the water, okay? And so if the, the, the definition here of surfing versus swimming is If the technology gets better, does your solution get better also, right? And if the answer is yes, then you are correctly surfing on that technology. Surf, don't swim is the idea here, okay? So let's look now at AT&T. What kind of company is AT&T? What do they do? What's their core?

Let's figure out if they're a surfing or a swimming company, right? What do they do? Right now they're struggling. They started out building cables to be able to have telephone lines all connected, and people stopped using landlines so much, and that started getting so much more expensive to be able to maintain those same landlines with fewer users. They switched over to having cell technology as well, but then they were entering a fray that other competitors had already been in on. And so they're trying to hold their own with it, but it's—

they're just among the fray instead of like the big thing that had to be broken up as like according to trust law way back in the '80s because they were— So based on— that's a good analysis, Soraya. It's a good analysis. And they obviously do some other things, but you kind of hit some of the main points.

So what I want to ask then is, are they surfing or swimming and why? Right now they're surfing, not incredibly well, but they are surfing because they changed the tech that they were working with as a model. Okay. Why are they surfing?

Because they were able to take a lot of their landline people and put them into being cell users. Okay. Okay. I understand what you're saying. We're gonna review the definition of surfing versus swimming just again one more time.

Remember, if the technology upon which you are built gets better, if your user experience gets better, you're surfing. Okay? They added extra help. They added being able to have access to the internet and other things on your smartphones in a way way that allowed them to get more money. We pay like 10 times as much per month than we did back when we had just a landline with them.

That's definitely true. That's definitely true.

Um, question.

I think what you're trying to say is that they're the hardware of their, um, that your example of Netflix was they don't have to repair the cable, but AT&T in many cases do have to repair the cable.

Yeah, they have to go. What happens when we advance from 3G to 4G to LTE to 5G to 6G? Right. They're responsible for their infrastructure.

They have to replace that, right? They have to replace that stuff. So they have— I don't know. I don't even know technically if that's automatically true. Like, is T-Mobile and AT&T and Verizon, they all have 3 different towers?

They do. They're not interconnected technology. Oh, there is, there is some interconnectivity and some sharing in certain agreements, but for the most part, they do control their own towers in their own connectivity. That's why I could be sitting here at my house with an AT&T line and my friend can come over with Verizon and I can have a great signal and they can have no signal. It's because they just don't have a tower in range, right? And so the reality is AT&T, Soraya, would probably be more of a, of a surf, or I'm sorry, of a swimming company because they are, they are responsible for their own technology. If it goes down, they have to fix it. And so that requires a huge amount of capital expenditure. And then here's another thing. If you can be passed up by another technology, that means you're swimming. Is there a technology you guys can think of right now that could make maybe cell towers obsolete? Is that the Starlink stuff that you're talking about? Yeah. Bingo. Exactly right, Sariah. Very good. So Elon could come out and say, I'm gonna make Starlink free for everybody this month or this year. And then after that, it's gonna be $5 a month per user. That would put AT&T out of business for sure. They'd be done. And every cell company in the world, right? Trillions of dollars worth of, worth of value, right? So how could, how could AT&T avoid that? Well, they could say, hey, you know what, let's, let's partner with Elon and let's make sure that our networks all work on top of his technology. So as he does make this available, we can, we can still provide phone numbers or whatever, right? So anyway, um, bottom line is AT&T is more of a surfing company— or sorry, more of a swimming company— because they are dependent on their own investment and they are threatened by advancement, fast advancement of other types of technologies, right? Global connectivity, right? They're a connectivity company that isn't in the lead for global connectivity, right? And by the way, neither is really anyone versus Starlink, right? Starlink is putting that same pressure on everybody, and that's good for the world, that's good for the market. That's why market economies work, okay? Um, you don't have to like Elon to, to know that his, his approach to the market is, is, uh, working, right? It's pushing innovation, it's pushing people to do better. Okay, so now let's bring this back to earth, right? Why are we talking about this? Why does this matter? Well, let's talk about my buddy who wrote a book called Surf Don't Swim. Okay, his name is Jeremy Witt. He was my chief strategy officer at my past company for many years. He invested about $400,000 in the company or so. Really, really sharp guy. He's about 55 or so, and He lives out in Florida, but I share his age because he was able to go through the dot-com bubble and that's where he kind of made his money. And then he bought a retirement home in Kansas City with about 200 beds. So good size, good size retirement community. And it's— sorry, it's like a care center, like intense, like everybody has beds and nurses and, you know, All that stuff. And he's also a technologist. He was a developer back in like the '80s and '90s and doesn't really do it as much anymore. He kind of does. He's still very, very involved, but not as much. He usually just hires people now at this point. But as AI started to come out, he said, you know what, I'm going to surf on this technology, right? He's very smart. He's like one of the people that I talk to in this world where I'm like, okay, Jeremy, you just said something. I think I understand all the words, but I don't understand why they fit together. Can you explain what you just said? I need— I don't understand that, right? And he'll, he'll walk me through it and, and whatever, and he'll be patient with my, um, slow understanding. By the way, guys, and I'm not trying to toot my own horn, I think I'm a pretty smart guy and I get things quick. I'm just trying to say Jeremy is very smart, super high IQ, and he just sees the world differently. But he, he, um, he built this system. Well, so because he's in the market, he's in the industry, he understands these senior care centers and their needs, one of the things that plagues these senior care centers is mental health, right? Diseases like schizophrenia and Alzheimer's and, um, dementia and things that, that really actually happen because you don't have people to talk to. That's where a lot of those diseases of the mind come from is you just don't talk to people enough. And so these senior care centers, actually insurance will pay for people to go in and talk to to them, right? Like, literally, you could get a job being a person who talks to old people, okay? And you could do that in person, you could do it over the phone or whatever. And so Jeremy says, you know what, AI is getting pretty good at talking on the phone. I've got an idea. So he calls up UnitedHealthcare and says, hey, if I came up with a solution that had AI call these people at specific times, or they could call it whenever they wanted just to talk to somebody about something, just to talk and use their minds, right? Would you guys cover that in the same way that you cover real humans coming in? They said, yeah, sure. Well, we'll pay a little bit less because it's technology, but yeah, we'll pay $90 a month per person who signs up for this, right? At, at this thing, right? So for a 200-bed place, that's an $18,000 contract. So what he did is he took a bunch of AI systems, one that could do emails, one that could write, you know, like a ChatGPT or something like that. One that could do voice agents, which we're going to do here in a minute, one that could do all these different things. And he kind of pieced together this little— this little, we'll call it a surfing system, right? Because he's not building any of the technology, he's just putting it all together. And, and then, um, what it does is it will automatically email the nurses, find a time that works, schedule it, get the name, learn about the person, whatever. And, and then it will connect it to an autodialer, right? A phone system. Connect that to a voice agent. And so then the voice agent will call and say, hey Betty, how are you doing? My name's Jon, right? I'm excited to talk to you today, get to know you, right? Tell me about your family. How are they doing, right? And then next week, now all that information is saved in the database. The entire conversation is saved and recorded, kept in the database and added to that specific agent's RAG knowledge, right? It's knowledge-based. So then when it calls next week or tomorrow, it says, hey, how was Lucy's concert, right? How's your grandson, right? How's your whatever, right? How are you doing with this thing? Is your hip feeling better? You know, whatever. But it can empathize, it can remember, it can actually have conversations that have continuity to them. And so he's now making about $2 million a year. Passively with this business. Okay. He does nothing. AI does everything. He pays a, about a 20-hour-a-week, you know, admin that helps with some of the billing questions and things like that. I'm just answering emails, you know, as needed. But most of the emailing and communication and coordination and everything happens automatically by AI. He just connected the right systems to it, and now he is surfing on that. If The voice agent gets better. Does his product get better? Absolutely. If the email AI gets better, does his product get better? Absolutely. If the internet gets faster and more reliable, does his product get better? Yes. He is sitting on top of those technologies, on all of them, so that as they get better, we get better. So you have to make this decision, guys, as you're looking at what kind of business am I going to create using AI? Do I have a risk of Starlink passing me up? And making me obsolete, right? Or are all of these next-gen technologies that are popping up and coming up, am I sitting on top of them so that as they get better, or even as— or maybe even, you know, ElevenLabs, we're going to look at ElevenLabs here as a voice agent system just right after, as soon as we're done with this discussion, that's the next thing we're going to do. Maybe there's another company called TwelveLabs, right, or whatever. Okay, if ElevenLabs or if TwelveLabs comes out with a voice agent that's even better than ElevenLabs Well, all he has to do is like switch out an API key and he's done, right? Switch up the documentation a little bit. It'll take an hour of adjustments and he's now surfing on that new technology. So you can swap, you can hot swap these things in and out. It's amazing. It's amazing what is possible if you build it right. Um, now he's making millions of dollars doing that. Um, back to the question I asked at the very beginning, should you rebuild Calendly for yourself?, right? Just an automated thing that helps you schedule appointments. Maybe, but maybe not. It's only $11 a month. If you build your own solution, you're now swimming. Okay? And every time something changes, like Google changes their API or somebody comes out with some feature, or you need to be able to have multiple people on the call, or you need to be able to do whatever, right? And, and Calendly comes out with these cool features and you're trying to keep up and you need some additional things. You have to develop that. You have to get into your code, your vibe-coded project. You have to update it. If something breaks, it's your fault. You have to worry about it. If some, you know, regulation changes and there's some new law and you're not following it, there's so many things. And for $11 a month, I'm just going to pay someone else to do it right now. Salesforce, when it's a $320,000 a month bill— or sorry, $320,000 a year bill for a company, Okay, now maybe it's worth it to put a resource on that and say, hey, you know, we're paying $320,000 to this. We could pay $180,000 a year employee to build our own thing internally and manage it. And now we own everything. It's all ours. It's completely ours. So there's always going to be— and this is, you know, some people might call this buy versus build. So this is a little bit of surf versus swim and buy versus build in there. But you need to make sure that you are surfing? Can you construct your business in a way that just gets better over time because the industry around you is getting better?

Let's look at GenAIPI really quickly for the last 60 seconds here of this discussion.

Is GenAIPI surfing or swimming? We are absolutely surfing.

Why?

Because we're creating new things that are— yes, we're, we're swimming in a sense that we do depend on other systems. However, we are producing them, them in a way that is actually adaptive and ahead of what they're currently doing. We're putting them all together and being able to put it into other companies or help other companies to actually move them forward as well. So they don't just have to swim. Let me ask it this way. Tanya, you're right that we are surfing, but I think you're just a little off on why. If AI technology gets better tomorrow, do people want to work with us more or less? More. Yeah. Because most, they're all swimming. Exactly. Because AI, as AI gets better, we're sitting here as AI gets better. We are an AI enablement company. We help companies adapt and adopt and learn. You guys, you know, whatever. Even though we are building some of our own solutions and things like that, for the most part, I've spent this course teaching you guys about Replit and Manus and ChatGPT and systems I didn't build, right? I had a little delivery system to get you guys into the room with me, but for the most part, I— our product gets better and more desirable as the industry increases. So, or, and improves. So anyway, guys, think about your own businesses. Um, have these discussions, you know, text me about it. Um, ask it on the Facebook group, you know, whatever. Just say, hey guys, I'm trying to figure out how I can make my business surf better. Everyone who's ever taken this course will understand that and they'll be able to help you, you know, figure that out. Um, don't be afraid to talk about your business ideas. Also, that's another just kind of entrepreneurship thing. No one's gonna take your idea. And if they do, they're going to do it differently than you, and you're going to do it better than them because it's your idea. People get worried about people stealing other people's ideas, and the reality is it just doesn't happen in the real world. I don't know why people get worried about that.

Okay, we've got a little less than an hour here. I want to make a note just so I don't offend anybody in the future. I don't participate in Facebook, so I'll never be on a Facebook group.

Okay, so if there's another method of communicating, I might, but the Yeah, that's really all we have. Um, you will all have my cell phone number at the end, so you can certainly text me individually and say, I've got a question about something, right? If you want to work with me, like long-term, you want me to help you build your business and stuff, it's going to cost you a lot of money, right? But, um, but if you just have like a quick question and like, hey, I, you know, what do you think about this? Or, hey, look at my new thing I just finished, right? I'm totally happy to look at your website and be like, hey, what do you think? And I'll click on it. I'll be like, hey, make sure you focus on this. And hey, this is really cool. Have you thought about this? I'll give you a little feedback, you know, little tidbits. No problem. You guys are my friends now. You're in my circle and you'll be able to reach out to me whenever you want. But yeah, or, you know, if you're in town and you want to go to lunch or something, great. But anyway, right now all we have, Pat, is the Facebook group in terms of, in terms of, you know, having a circle to turn to. So you can make a Facebook account and just, you know, have it only be that purpose. If you want. So up to you. Okay guys, let's go over voice agents really quick and then we'll finish up with some images and video. And then I think I'm gonna say, cuz we're doing this 2-hour session next Wednesday, um, I think I'm gonna save most of the sales and marketing stuff for that. I'm gonna, I've talked a little bit about marketing, about how you could automate social posting and things like that. That's certainly important. Um, but there's some more sales and marketing stuff I'd like to go over. And some of the voice stuff actually here that we're about to do kind of does apply there. So we're gonna start, we're gonna start moving that way. Um, but, but we're not gonna finish the discussion, um, by design because we're cutting this an hour short and you guys are getting 2 hours for 1 hour. And I was actually talking to my wife about this last night. I think, I think the format of a 4-hour, 3-hour, 2-hour session, you know, course is actually better than probably 4 and for, right? Because you just, you get a week to kind of work on your project, come back with more questions, think through things, and then you get 2 full hours to really discuss with the group. Anyway, I think it's cool. So anyway, um, all right, let's dive into ElevenLabs. So I'll share the screen here. Um, ElevenLabs is super fun. Trying to remember which account. Is it this one or is it, is it this one over here? I think Probably this one. All right. Let's log in. Yeah. Okay. Cool. So, 11Labs, guys, is, in my opinion, the leading AI voice agent system. There are lots of systems out there that will, for example, right? Like, Tanya, you might look around and someone might reach out to you and say, hey, you know, we can deploy AI voice Robots to help you guys sell more at GenAIPI. You know what they're doing? They're actually just wrapping something over 11Labs, right? 11Labs is the technology that's underneath that's doing everything. Okay, so we're in the agents platform right now. Let's come over to the creative platform first. So in here you have voices and there's all kinds of voices. We'll listen to a couple of them here just so we can see what's up. Let her go now. That'll be the end of it.

I will not look for you. Even the smallest steps create ripples of happiness.

Keep moving forward.

Hey everyone, welcome back to the channel.

Today we're talking about all the latest news, trends, and celebrity gossip. Imagine you are alone on a cool dark night. As you lie in the damp grass, you drink in the majesty of the sight before you. Anyway, so you've got different types you can go in there and there's professional voices on there. There's people that have spent a long time putting together something really nice. You can also go in and you can clone your own voice, right? And a little progress each day adds up to big results. Just like flowers need sun, you need kindness to grow. By the way, I didn't say those words. It just takes my stuff and then it makes up some random phrase. But gardens remind us of the kindness of the earth. So some of those are okay. Now all of those were created with a very quick clone. You can do something called an instant voice clone with only 10 seconds of audio. That's how long it takes to clone your voice. And it's incredible. It's really fun. Here's a fancy, or I don't know, maybe kind of an out-there example of some things that people have done. You could take old home video from your grandparents and kind of bring them back to life a little bit, right? Have them narrate a video, right? And, you know, for your family reunion or whatever it is, right? I mean, there's cool things you could do with this that It could really have a lot of meaning to people if you do it in the right way. But sometimes people just like to hear their own voice. You can clone your own voice. You can also do a voice design, design entirely a voice from a text prompt, or you can do a professional voice clone where it's going to take 5 minutes of audio, and then you can transform existing voices as well. You can also, by the way, take Tom Cruise's voice if you want, or let's take a better voice, Morgan Freeman or something like that. You could say, hey, I'm going to make my own voice and use Morgan Freeman for my own video. Now, can you do that? Yes. Are you legally allowed to do that? Probably not. Okay, so you're responsible for ethical and responsible use of AI, but it's not very hard to find someone's voice on YouTube, download 5 minutes of it, and upload it. And I mean, you could make Again, think about this from the bad side, right? Take— I'm gonna go with a religious example, right? Take something that a current religious leader, a prophet or something says, right? Or a president of a country or something like that says something. You can make Donald Trump say anything. You could make, you know, President Oaks, right? The president of the church say something and then put it over, overlay it over a video that looks incredibly real. And tons of people will not be able to tell the difference, right? And I can tell the difference today, but maybe in a year or two I won't be able to tell the difference because it's only getting better. It's getting more realistic. And so there's always, with any tool, you can do amazing positive things with it. And then of course, plenty of negative things with it. So use it, use it wisely, use it well.

Now, once you've got a voice and you can find some that you like, um, hey, I'm Zara.

Let's create something that feels real, sounds human. Anyway, you find things you like, then you can do what you want with it, right? So maybe you just wanna go text to speech and you want to, you know, BYU is the best, right?

I don't know. Sure. Where's my Zara girl?

Let's do that and let's generate the speech. BYU is the best. All right. I don't know if I love that. But it's kind of slow, slow excited, right? Anyway, but you can try that. You can edit it. You can give it, by putting it in brackets, you can tell it how you want it to act, right? So instead I could say whispering, right? Generate speech. BYU is the best. Right? So it changes it obviously.

I could be, you know, yelling, yelling excitedly.

Let's just try that again. BYU is the best.

Again, it's kind of slow. I'm going to—

yelling excitedly and quickly.

I don't know. This is, of course, a marketing campaign for BYU. BYU is the best. Yes, it is. It is partially. If you are not a Cougar fan by the end of this course, then I have failed. Anyway. Yeah. Subtle, just little things, right? This isn't so subtle, but anyway, point is you can, you can modify how it works. So you can go there. So a lot of people will use this section for maybe creating a voiceover for a video, right? They're maybe doing an ad or something for 30 seconds or 60 seconds, or maybe a 3-minute long infomercial ad. And you've got this script that needs to be read over it. And so So you create the nice voice, you go and you get all your AI video stuff created and everything, and then you just overlay it and put it all together in CapCut or Premiere or whatever it is you want to use. Okay? So text-to-speech is interesting. You can also do a voice changer, which is really cool. So maybe when we create video with Veo or something like that, when we create an AI video, you can hear when you see a lot of AI videos sounds like the voice sounds weird. You're like, it should sound normal, but it just, it sounds a little off. And so you can come in here and you can say, all right, so I can do an example like this. Okay, I'm just gonna do a little bit of testing here and then we'll see what I sound like as a woman. Okay, so now we've got our live recording, and now I can switch to, you know, this girl Zara, right? We'll just do her, and they'll just generate that speech.

So you can take a recording from a video or do a recording right then and there, and we should be able to then listen to this.

Okay, I'm just going to do a little bit of testing here, and then we'll see what I sound like as a woman. Okay, so it took my exact performance and just changed the voice, right? So it can be super useful for all kinds of things, like for those of you that might be creative and, you know, you, you could, once, once I teach you how to do some of this AI video stuff and show you kind of some of the workflows, you could make your own animated series for your kids or for yourself or for other people or whatever. Right. And you could do really neat things and you could be all the different voices and you can have, you can change the voices, but you could be the voice actor for everybody. Um, they have some sound effects stuff and some other things and voice isolator and some tools for processing audio. Really, most of the time when I'm in here, I'm using text and voice. That's really it. Image and video. Looks like they're doing— they're trying to do video in here too. Switch seamlessly between image and video, access popular models, upscale, lip sync, and more all in the same flow. All right, well, that's pretty cool. That is pretty cool. Um, well, we're going to keep these private, but yeah, being able to— so there's other tools out there that you can use to lip sync something. So you could basically show an image like this, this girl right here. I want to use in lip sync speech. Let's try this one. BYU is the best. I don't know if this is going to even sound good, but we're going to try it. Okay. Generate. So this should theoretically Yeah, I don't know why it failed. Theoretically, what it should do, what you can do on other platforms, is you would be able to— while she's walking, she would be saying, BYU is the best, right? Unexpected error. I don't know why. I don't know why. So I'd have to— obviously, it's new. They're trying to figure it out. All right, here's a render image. Yeah, I don't know. I don't know. So I'll have to play with that. I won't play with it right now live because I'm going to keep moving forward. So What I use this for the most is agents. Agents are amazing and they work almost exactly the same as building a GPT, which we already did earlier. When you want to create an agent, you can just come in here and you can just say, all right, let's just create a new one. You can start with a blank agent or you can start with something else, whatever. Let's just start with a blank agent today and we'll just call it Lucy like we were working on earlier. Chat only. That's good. Create agent. And then putting this all together is actually quite simple. Okay, here's the system prompt, right? We already know what to do with this, right? We're going to come in and we're going to grab our system prompt from, from here. And you know what? I'm going to go into my GPT creation because I had the right thing pulled up. Here it is. Okay, so here's Lucy and we'll go over to the system prompt here and we'll just paste in Lucy. We got it. Okay. And we've got a default personality. Of course, we can change some things if we want. We can set a time zone if for whatever reason we need it.

Lucy's a girl, so we're going to switch the voice. We'll go to—

let's listen to Eve. Cherish the Earth and its infinite—

you got to be careful if you don't know where you're going because you might not get— in the middle of every difficult—

cherish the Earth.

Hey, I'm Zara, born in Italy, bred in London, and married to an American.

You've I've never heard a voice like this. All right, we're going to do that. Okay, and then here's the first message, like, hey hero, what can I do for you? Okay, so now I've got my personality there. Then I can come in here, this is pretty cool. I can say knowledge base, right? I can add documents just like I could to, you know, with any of the other GPTs in the RAG systems, right? Just drop in your folder, drop in your Google Drive, Configure RAG. That's how that would work. You can go through and you can look at all these different crazy things, lots of tools. For example, does it have the ability to end the conversation or is it always going to keep trying to talk? Do you want to detect the language? Do you want to skip the turn? You can teach it, by the way. That's pretty cool. Detect language. You can have it speak multiple languages. You could literally give it the ability to speak English and Spanish. And then say, hey, I want you to teach me Spanish. And it just talks to you and says, okay, here's where we're going to start today. And then you say it back to it. How did I do? Oh, okay, your pronunciation was a little off here. And let's try this one again here. I mean, you could build your own custom AI language tutor in seconds, right? So you can give it the ability to skip a turn, to transfer to another agent, right? So you can say, hey, once you're done with this agent, move over to this agent that has a different purpose, right? So maybe you build an agent that receives phone calls for your company. And, you know, press 1 to do this, press 2 to do this, press 3 to do this. And if they push 3, it transfers to an agent, which is sales, and it's a sales agent. And now that sales agent, its job is to set up an appointment with a real salesperson for tomorrow or whatever, you know. So you can give different agents different tasks and then connect them together in a little web, a little web. You can also have it transfer directly to a number, right? Like maybe you have an emergency hotline and we're, you know, we're working on You know, it's a, we, we actually did this specifically for a company, but, um, they, they had a, a hotline where if you get hurt, you call this number and you report it and you have to work through that process. Right. And so they had people staffing that number 24/7. Now it's an agent staffing it 24/7, but if there's an emergency, you can always transfer to a number. Right. But most of the calls are like, what's the status of my claim? And hey, you know, I, I twisted my ankle and I had to go to the doctor and now I need to submit my receipts and what do I do and how do I open a claim? You know, it's, it's that kind of stuff. It's a lot of busy work that an agent can actually do, but you can transfer a number if you need. So it has all kinds of really cool capabilities, even voicemail detection. So it doesn't just call something and just sit there forever. Okay. So you can give it all kinds of different tools and when you're done, you publish it. Right. And now I can, and actually I don't even need to publish it, but Should be good.

That's saved. I'm going to preview this now and I'm going to call the AI agent.

Hey, Hero. What can I do for you? Yeah, I'm just trying to figure out what I should focus on today.

Should I go kayaking or play the piano?

This is a choice between two different paths for your day to help you decide. Consider what— I don't like her voice, but you can see I can just start having a conversation with it. That one wasn't very natural. In my opinion, but I can go in very easily and open up Lucy wherever she is. Where'd she go? Lucy Wilson. Here we go. There's Lucy.

And I could switch the voice and just say, nope, I didn't like that one.

Right?

I'm gonna come in here and I'm gonna switch to Zara. Hey, I'm Zara.

Right? And so let's create something that feels real, sounds human, and connects every time. Yeah. So that's a little bit better. So now we'll click publish again.

To make sure that we're using the right thing.

Then we'll preview again and we'll try it again. Hey, Hero, what can I do for you?

Yeah, I'm just trying to figure out what I can focus on today. That's a good question.

To help you figure that out, could you tell me what kind of person you are working to become? Anyway, so you can see that's much more natural. I like that a lot better. I find the voice, the right voice for the type of thing I'm trying to do, and I put it out in the wild, right? So we have— there's so many cool use cases for this. For example, we've got one here Any of you guys ever heard of Travis Chambers? You guys ever met him? So he's got his— he has something called Outpost X. Outpost X, right there. Yep. Outpost X is this place down in southern Utah. It's amazing. I've been there. I've taken my kids. It was the coolest thing in the world. Now, we got to go there when Ryan Trahan was there too, so that was cool. We got to hang out with him for a bit. If you know who Ryan Trahan is. If you have kids, you know who he is. If you don't, you might not. But anyway, really cool spot. But he built this, or we built this thing.

I built it using Replit and 11Labs. And let's enable the microphone here.

Allow visiting. Welcome, traveler. I am Amira. How shall I guide you? Yeah.

Can you tell me about these speeders? I want to I want to ride on them. Ah, you must be referring to the sand cruisers. They are quite the experience. These high-performance relics from the Namajin resistance are available for daytime use.

Keys are in the cruisers themselves. First come, first served. So it's been trained on, you know, if you go into all the training here, you can see, um, the knowledge base is going to have information. So it's Signs with audio lore. So there's like a story, there's backstories and stuff that he all made up, right, for this whole place there. And it's kind of a Star Wars feel, but it's its own story because he doesn't have Star Wars IP, right? But, um, but there's a big long podcast. And so this is like 12 pages long of just like stories. And then you've got all these signs with more lore and then all the basics of Outpost X.

So I could also jump on there and just say, hey, what's the Wi-Fi password?

Welcome, traveler. I am Amira. How shall I guide you? Yeah, can you tell me the Wi-Fi password?

And also, I'm hungry. Where can I get something to eat? Of course. The network is Outpost X and the password is a capital O for Outpost followed by a capital letter X and then followed by the number 1. Outpost X1. As for your hunger, the cantina kitchen is always open and stocked with cookware. It is also BYOB.

For a more curated experience, you may consider the culinary experience, which includes 3 self-cooked meals pre-stocked in your unit. So, it knows information about this destination. It knows how to whatever. Now, if you don't want to talk to it, you can text it. No big deal. Right? You can just talk. You can just type to it. Okay? But, but we set up this nice little system here. It did that. And by the way, this is two separate systems. This is actually ElevenLabs connected right here via the voice. And then OpenAI runs the text, right? But it's been trained on all the same information, all those same, those, those three documents. We put all the information into those and just uploaded them, put the prompt in. Took me maybe an hour to build this, right? And anyway, just kind of gets used.

So, pretty cool. Any questions about this voice agent and ElevenLabs? How do you connect those together? Like, did you just tell Replit to pull from ElevenLabs?

Was there— do you have an API key that's connecting the two?

Absolutely.

That's exactly right. So you've got— you go into the developer section, you've got your API keys here. Here they are. You can grab, you know, whatever. You can create a new one just like we did over there. So you just create a unique one for your project, put it in there, and then you'll have agents. And so I could click on like— in fact, here it is. Here's my Amira. There's Amira. And I can go over to the agent here, and there should be somewhere— they've changed this. Let's see, maybe it's advanced. There should be an agent ID somewhere. Oh, here it is. It's right here. Yeah, I would just copy this agent ID, and I would say, hey, here's my ElevenLabs API key, here's my agent ID, please set this up.

Done. Very simple. Okay? Does that make sense? Does that sound simple to you, Sariah?

Somewhat.

I've sometimes getting API keys in has been a bit of a chore figuring out where exactly they go and making sure we get it straight, but I, it just takes practice, I'm sure.

Yeah. It's, it, it shouldn't be in Replit. If you're using Replit, Replit will will literally, like, like, watch this. If we go over to, um, to the project I'm working on here, right, this one right here, I could say, um, I want to, um, add a voice agent from ElevenLabs. I have an API key for you and an agent ID for you. Okay. So I didn't give it a lot of information, by the way. That's kind of a bad prompt. And it's probably going to say, hey, I can just use my own API key, or you can use your own. And I'll say I want to use my own. And that would allow me to then put my specific ID in there, which we, which, which you would need. I, you really, you really have to kind of do this. I don't, I bet Replit doesn't have an automatic automatic implementation because you have to get the agent IDs. And they can't just save all those agent IDs over there. Searching the web, it's thinking, it's thinking way harder than it needs to think, probably because I have it on, you know, high thinking mode. If I had it on low mode, it would probably have already asked me for it. But I want to— I'm glad you asked this question because it's good to just see the interface and how this works here. On how easy it is to actually set up the, set up the agent. Okay. Very straightforward to add. Boom. Here you go. So this is what it does. It says, hey, please provide your ElevenLabs API key and agent ID so I can integrate it. Right. So it's already filled in some things here because I have some on my account. Right. If I go to my secrets and I go to my account secrets, do, do, do, do, do more docs. Where's my, oh yeah. Link account secrets. Yeah. So I have all these other secrets. I've got 11Labs keys and things I've used on other projects. I could pull those over, but I'm going to say, no, I don't want that. Let's, let's get rid of these and let's just put in our own, our own things here. Right. And so I don't want to do this. Maybe I have to unlink that. I don't want, I don't want that. Just let me delete it. Anyway, that's what you would do. It's normally going to just ask for it, right? You're going to paste in your key and then you'll paste in a new key and You're done. Okay. So, um, and then you just say go and then, then it just, it, it knows how to read all the documentation and put everything together and figure out what's going on. Nevermind. Let's not add voice right now. Okay. So, um, it's that easy. And then what happens is it just stores in the secrets here and you can see all the different things you have for the project. Um, and it stores it in a way that is safe. It's not, you're not gonna be able to hack into it and see it. It's not gonna be exposed and left out somewhere. What you don't want to do, Soraya, and anyone else, is you don't want to take that agent ID or that API key and give it to the agent in here in the normal chat. Because if you do that, now you're telling the chatbot what it is and it might just make a note of it somewhere and put it somewhere in some open whatever document. It kind of knows not to do that now, but it used to happen a lot more., and people were having their API keys just used and getting waking up to $1,000 bills with other people using their API keys. Okay. So, but if you use the Secrets Manager correctly, where you just say, hey, look, I have an API key. You can even say, hey, I have an API key. Please use the Secrets Manager to store this. Like if you want to be explicit, you don't really have to do that, but you could, and it will come up and it will say, hey, tell me the API key. And then you just paste it in there. Save it and then it will reference, it will reference that key in safe ways. Okay. Other questions about ElevenLabs or voice agents? There's another company, I think it's called Agora. Yeah. Conversational AI platform. Yeah, I'm right. So yeah, anyway, you can, you can, this is maybe an alternative to ElevenLabs and there's several out there. If you just ask ChatGPT, hey, what are the best, voice agents out there, right? It'll tell you, it'll tell you whatever. But, some of them are more set up so that you could say, hey, you know what? I already need, you know, I, I just need something that, that does, you know, outbound sales for me or something like that. Gaming, conversational AI, telehealth, metaverse, whatever, right? You can just ask it, you know, whatever you need and it will, it will help you figure it out. So, I personally think ElevenLabs is is the best by a long shot. I've never used Agora, but I actually just— you know what's funny, Tanya, is I just got contacted by somebody at Agora that said, hey, we're having an event during, um, during Utah Tech Week and we're co-sponsoring it with Replit. So that's kind of fun, kind of random. So anyway, maybe they're doing multiple events and spreading wide, which if I were them, that's what I'd be doing too. Um, so no big deal, no big deal. Our event's going to be cooler. For sure. But, uh, but anyway, so I heard about Gore yesterday because someone reached out to me on that. Okay. Um, all right guys, we have about 15 minutes left here. Um, you guys all got access to this, um, to this folder here. Um, obviously over the next week, if you have questions, um, you can go into overviews and you can look at different things. You can look at some of the guides I have for here's how you go. These are some of the processes and things that you do. I actually don't go over in these guides as detailed as I did in the course, but it should give you some additional ideas and maybe even some additional tools to go after to figure that out. Um, there's a few different kind of variations on that. Um, the software and tools folder I think is really useful. Um, just different, just different things that you can do that, that I've kind of compiled and put together, just actual like mastering Replit, like this document is really, really helpful. These are actual prompts and things that I've copied and pasted into, you know, into the workflows to make sure that we can, we can work through problems.

Okay. Actually, I don't have access to that Google Drive.

I've been watching for an email and I didn't see the link to it. Yeah, I dropped it. I dropped it yesterday in the chat. There it is again. So go ahead and copy paste that. Um, It normally goes out in an email at the end of the course. And since the end of the course is next week, I don't know, maybe, maybe I'll be able to get it. I'm not gonna be able to get it out today because I'm gonna be traveling. I don't know if, if it is today, it would be like late at night. Cuz what I have to do is I have to upload yesterday's course and today's course to YouTube, which takes forever cuz they're 4-hour videos. And then I put it as a private link and I share it with you guys. Um, but then we're gonna have the third course next Wednesday. Wednesday.

So anyway, you'll get it all by next Wednesday. Thank you for doing that because I know if you just do the Zoom link itself, which is faster, when you have other events on top of that, then the older ones go away, where YouTube will keep it if you have the right level of membership on YouTube.

Yes, which I do, and I've made sure that I've done all the verifications so these videos never go away, and they're all saved on my computer. I have the raw video files on my computer. Wow. That takes a lot of space. That's why I have the 8 terabytes. Okay. Anyway. Um, anyway, so yeah, guys, um, there's yeah, lots and lots of opportunities for, for continuing to learn here, um, beyond the course itself. Um, you know, a lot of people, again, I'm not in the, the course guru game. Like that's not my business. Um, I never really have been. I didn't even want to do a course because I was like, I don't want to just be like a course guy, you know, and, and, you know, but, but I've enjoyed teaching these courses. Um, but I don't have like the next thing to sell you. Like, here's the next course. Right. Um, I might, I've had, I've had more and more requests for kind of a how to build a business with AI in a weekend with AI 2, you know, advanced workflows and just kind of picking up where we left off. And you would have to, you really wouldn't want to take that without having already taken this course. I may or may not do that. My guess is what I'm going to do, guys, is I'm going to put together recordings of like, I'll just record 100 different modules of like all these different things. And so it'll be a lot of the same content, but taught in just like modules, like, hey, we're going to go over this thing first and watch this 8-minute video. And then, hey, here's image generation here. Like, like we just did this section on ElevenLabs, right? That would be a module and maybe there'd be, here's the introduction to 11Labs and here's 5 things you can do with it. So that might be 6 videos, right? So my goal was to put all that together sometime this year.

And then because you guys have taken the Build a Business in a Weekend course, you would all get free access to it.

So that's what— would you want us to be testing and like giving you feedback on it then? Potentially. Yeah. I mean, I, I always welcome feedback. If you have something and you're like, hey, this was kind of confusing or this didn't work well or whatever, or, you know, you accidentally showed your, you know, password on this thing, you know, I, you know, would love to, would love to see those kinds of things. And please, please give feedback on it. But yeah, I'll probably launch that. And then, and then we may, you know, we may sell those as a company kind of in a more automated fashion so that I don't have to sit here for, you know, 8 or 9 hours, um, in the same way. Uh, but you miss out on the, you miss out on the live interactions and the questions and things like that. So anyway, so I might go into more advanced content in those videos eventually. So you could kind of look, look for that. If you want to keep learning and keep, you know, working with this community, there are again a lot of people on that Facebook group that are very active. And if, if you're like, hey, what's this new thing? Like that one we looked at earlier today, that, that X post that showed like, hey, we're not vibe coding, we're vibe businessing now, you know. Um, it's like, hey, has anybody tried this? Somebody might be like, no, but let's, let's do it and compare notes, you know. So anyway, you can you can meet other cool people that are, that are trying things on there. One guy that's very active, his name is Brantley, and he's actually my chief AI officer, right, at GenAIPI. He's the guy that does that. And I, he took this course in March, I wanna say in March of this year. So he is one of the first people to take it. And then he just like took off. He made about $10,000 with his first idea and about $50,000 with his second idea. And I was like, man, you are killing it over here. Could you help me with a couple of clients, things like that? So I started kind of just using him on some things because he was just sharp and he just got it. And eventually I was like, yeah, man, you got to come, you got to come work with me. Let's, let's do this together. So he's, he was my first hire, my first official hire. And now, you know, we have a few more people. Excited to have Tanya on the team. But yeah, guys, I, I could spend the next 10 minutes zipping into, you know, like images and video. Um, I'm not going to do that. I'm going to spend just a few minutes doing that. Um, but I'm just going to show you the platforms and we're going to go into them next time because the image and video, like, to do it well is really a good solid hour. Um, and maybe even a little bit longer than that if we're experimenting with different things. But here are my go-to image platforms. If you want to just, you know, take notes on this stuff. Midjourney is number 1, hands down. And actually, I'm going to go over to this profile because this is where all my accounts are. So Midjourney is incredible. Okay, this is the one I had up yesterday where you saw some treasure images and things like that. This was an album cover I was working on for some music. You can also animate things in here, right? I was trying to have it, you know, work on some cool ideas with things. This was— I don't know if I animated this. I kind of want to animate this. This looks cool. Animate. See what it does. Anyway, so that just started a new job here. So now it'll animate this. But you can just ask it things and you actually have to, when you prompt Midjourney, you have to ask it in kind of a special way. So you can use ChatGPT to help you build prompts, right? I do that often and we'll do that when we get over to the image and video stuff. This one was kind of weird. This was kind of a cool image that I I came up with, which was just this girl holding a mini elephant. And then I asked AI to animate it and weird stuff started to happen. I think she eats. Yeah, she's eating it over here. Anyway, so it's kind of fun. It's really made— Midjourney originally was made just for images. They've added this animation thing more recently, but you know, you'll see, you'll see all kinds of cool images that come out of this that end up on, you know, on my website or on blogs or on different things. If I need a cover image for something, if I need logos, all of— there's lots of logos in here. This is actually a really cool one that I've, I've meant to use and I just haven't used it. But it's like somebody just kind of absorbing and figuring out AI and working with it anyway. But, but the quality of the images in here is just absolutely off the charts. Like, like, it's amazing, right? This was a post for Pioneer Day and how pioneers and entrepreneurs have a lot of similar similarities. So you've got Utah mountains and pioneers and then tech. And anyway, but the quality of what it can produce is just amazing, right? This was a Balrog, right? Now it's obviously a little bit different than maybe a traditional one, but the quality of these images is mind-blowing. So I love this one. I really love Midjourney. ChatGPT can create images. Grok can create images. The best one out of kind of the big ones right now is Gemini. Gemini is— and Gemini has something called Nano Banana, right? There's the little mini banana right there. But Nano Banana is the best at creating images, right? So if you need a photorealistic shot of a 27-year-old female sipping a cup of hot chocolate in a cafe, um, with, you know, by the window with it snowing outside. It's not great English, but it works. It'll get the job done. But Gemini creates the best photos for photorealism, hands down. Midjourney creates very creative things and really high quality stuff, but it's not as good at photorealistic things. Okay? So these are things you just kind of learn over time. You're like, okay, Gemini's gonna be better at this. So there's our girl drinking hot chocolate. Although you might not know it's hot chocolate if you see that it's snowing outside, she's by a window. Did a pretty good job, right? Think she's 27. She looks about 27. It's pretty good. So, Yeah, amazing. Now, just a note on this. A lot of these systems, most of these systems, especially the big ones, will not let you generate things that are inappropriate, right? So you don't have to worry about it generating pornography or anything weird like that, or even things that it might consider offensive, right? Like if you want somebody to be killing somebody in a picture, it's, it's not going to generate that for you. Um, there's there's some, some guardrails, right? They just want people to be able to, um, they don't want liability for crazy stuff like that. In fact, there was kind of a big deal the other day. Grok created an image that was kind of sexualized. It wasn't completely nudity, but it was of, of minors, right? Of like two 14-year-old girls, right? And it was kind of a provocative image. And people are constantly trying to break the systems. They're trying to get past ask the things, and then obviously, you know, Elon's like, nope, we, we patched this, it's not going to be able to do that in the future. So, so there are systems that— there are systems out there though that have no guardrails. So just if you venture off into other maybe lesser systems, just know that, you know, that kind of thing can happen. But if you're talking about ChatGPT and Grok and Gemini and Claude and any of these kind of major systems, you're not going to run into that. Another system is Runway ML. And Runway is really image and video, but you can do a lot of video with it. Um, it's quite amazing at what it can do. Um, but you can, one of the cool things you can do here is, um, is you can give it a shot idea and you can even go into, if you get into like, um, let's say it's this one right here. But yeah, this one you can, you can take an existing video and completely change it. So you'll see examples of this. Uh, let's see, let's see, uh, Runway ALIF examples. This is another good way to learn. Just go to YouTube or whatever. Um, but when Runway, you can take an existing, um, video and then change you see this, right? It's, it's going like, like here's this video and then now here's a different angle of it. Or yeah, so there's a different angle of the same thing. Now we're switching. So you could change one thing. So for example, maybe you get some film of, you know, you're walking through Times Square and it's sunny. You could say, change this to a night scene and it will do that, right? It'll keep you the same. It'll keep your family the same and all the other details the same, but it will change everything so that now it's nighttime and all the billboards will be lit up and stuff. And it's unbelievable what it's capable of doing. So we'll get into a lot of that, next Wednesday. But, you know, Midjourney and Gemini are really fun to play with. And then when you start bringing in video and then you start bringing in the voice stuff, like we were playing with ElevenLabs, you can start to make full-on commercials, that are really, really impressive. In fact, we'll end today by showing you guys a video I got, Tanya, you weren't able to meet— oh, no, I've frozen here. Yeah, you weren't able to meet Eli on that video training we did, but he texted me last night.

It's like, hey, check out this video I just made. So cool. Watch this, guys. This— think about how many people you communicate with every single day. Friends, coworkers, family. Now imagine how different that communication becomes when those people grew up in different cultures. At its core, learning cultural intelligence is not about memorizing facts about other countries. It is building a skill that helps you navigate any environment where people are different from you. We could spend hours trying to teach you every fact about every country in the world, but that is not our goal, and honestly, it is not even possible. Instead, our goal is to teach you something far more powerful. In this lab, you will learn how people think, communicate, and make decisions across cultures. You will start to see patterns that show up again and again. Once you can recognize those patterns, you can adapt anywhere. As you go through this experience, reflect on your own culture. How do you communicate? What do the people around you expect? You may not visit every country you study here, but every day you interact with people shapes by different cultural experiences.

Because culture is not just international, it is everywhere. Welcome to the Cultural Intelligence Lab. Completely created by AI. Probably took him a couple hours. I think that style of painting is really, really cool. And then I asked him, I said, hey, how did you do this? Right? This is the prompt. Used this in Midjourney and did this thing. And this is This is like exactly the prompt, right? He copied that and then he inserted a couple prompts and did this and I plugged it into here and I did that and whatever. Anyway, so learning these workflows, I, I learned by seeing what other people do, getting ideas, and then if there are prompts, that's great. But again, you can find these on X, you can find 'em on Instagram, you can do that. So, um, good luck out there guys this week. We'll, uh, we'll talk to you again next Wednesday at 9:00 AM.

Mountain Time.

Okay, have a wonderful day and we'll be in touch. Thank you. Thanks, guys.

Thank you so much.

Enjoy your trip. Thank you. See ya. Thank you.