TFV GenAIPI Training - Session 3

Oh, you're good. I'm not used to having to, like, hit it multiple times in a day. So anyway, Security Scanner right here. That's all we've really covered. So. But yeah, you just click that and you just, you know, it's very, very obvious. I honestly just spending time just here and being like, okay, I wonder what this thing is and I wonder what else is right Just. Now, one thing that's cool here is, you know, yes, you can host everything inside of replit, and that's awesome. But what if you want to push it out and do something else with it and take it to another platform? Well, you can do that Git right here. So GitHub allows you to connect right here. I'm sure you're familiar with what GitHub is, right? GitHub basically just as a place where you can hold your code. Right, right. And so it's checking. All right. Hey, there's no GitHub setup. So you would go to the settings view and you would connect to GitHub. It's as easy as that, right? You connect it, you log in, and you're. And you're now connected. And then you say, all right, push this project. And now, now that you're in GitHub, GitHub can connect to tons of other things, and you can do all kinds of stuff with it, whether you're publishing it and pulling it over into whatever else you're doing. Okay, so. So that's one way to kind of pull things out. So GitHub's another good tab to just kind of know about. Security Scanner. We did it. It's checked. Hey, my app is now clean. Right? Okay, so that's. That's really fun. The database, we've already gone over the console you're not going to need here.

But that would have been a prompt that I could have prompted replace.

Yeah, you could also say, you know, please run a thorough security audit of this site. Boom. You know, well, then it'll check, you know, everything that you need to check. Now, here's a cool thing. I'm going to give you guys access to my Build a Business in a Weekend with AI folder. And I've got all kinds of cool workflows in here, right. You can go through of, like, you know, you know, how to build a business in a weekend. And, you know, here's all the different ideas and here's different things you need to be thinking about if you're going to be building this type of thing. Here's what I would use. Make sure. You set up send grid. Make sure you do this mid journey can.

Okay, gotcha.

Lots of things like that. And then I actually one of the things I have in here is a security guide for beginners. Right. And so you know, here's the types of things you should be thinking about. The cool thing is you don't even have to understand this. You could literally just go like this copy and say paste.

Right. Okay.

Check all of these issues, you know like that's all you have to be able to do.

Yeah.

Okay. Okay.

Super easy.

Awesome. And maybe you're gonna get to this too. Yeah John is because this is all making, this is like obviously melts brain melting and we just need time to like build this stuff out. I'm also just so interested now in.

Like.

And maybe you can share what you do is like how it's always like all the AI that we're using is either we're going to chat, chat GPT copy, pasting there, using it somewhere else. Maybe there's some stuff that has AI built in. But I don't feel like again like we don't, we don't have an agent or we don't have something running always for us. So I am looking for know like next week what would be the best use case for us to have an agent working for us? Because I'll be honest, our next plan is to like start hiring people and it's like hire. I don't, we don't want to be hiring more people if there really are. We just need to slow down for. We would still hire people but hire higher level people. Like how are we not using AI on a daily basis? Which is like we should be.

Yeah.

You know, or could be or what questions should we be asking? It's like you're teaching us this part of the questions that we can ask for building the business. But it's like how, what questions should we be asking? How should we be asking questions on how we could have use cases on our daily, daily business tasks and things of that nature.

So you're going to understand that better than anybody else obviously in terms of what are the things that are repetitive? What are the things that are just really hard and manual? Like is there a way to automate some of that stuff and you know, or the whole thing. Right. Can you, can you pull that together? What I've shown you so far is yes, we know how to strategize and use ChatGPT effectively and then how to tell it and go in here and double check again. Do Some market research and grog, stuff like that. But. But ultimately that all comes together in something like Replit. Replit. You're gonna. You're gonna use Replit a lot, right? You're gonna use it to build just a quick tool that you need for something. You're gonna use it to create one, maybe one of those agents that's working for you on something. Like, for example. Let's review the. The example I said earlier, right? What if you were going to set something up with Buffer? Okay, so Buffer. I don't know if you guys are familiar with Buffer, but Buffer is one of those sites, kind of like hootsuite, that allows you to publish to multiple platforms, you know, from one place. Right. And I don't have this set up for my personal thing here. This is my personal account. But I'll just. I have before. I just don't have it all connected now. So you're going to see a bunch of errors, but just ignore those. So, you know, X, LinkedIn, Instagram, Facebook, Threads, YouTube, whatever other channels you might have, TikTok, Pinterest, all of this stuff goes in here. And you can just post it one time. You just say, okay, I want it. You know. You know AI is awesome, right? And then I could just say, all right, customize for each network. And then we post it out. You drop, drag and drop a file. Maybe it's a. Maybe it's your. Maybe it's your, you know, your new podcast, your new YouTube video, or your new whatever, whatever post it is. You grab whatever that content might be, put it on there, customize for each network as soon as this uploads. And then what? And then you add it to the queue. And then what happens is it will follow whatever, you know, cadence you have set up. Maybe you want to post twice a day. And so you're going to post at, you know, 11:11am and 4:45pm and those are your times that are great. You're going to hit people at the right time, right before they're finishing with something or they're about to go to lunch, whatever, and then it just automatically goes out. But if you put 40 things in here today, well, it's only going to do two a day at those scheduled times according to whatever platform and rules you have set up. So that's cool. You can manually obviously come in here and just fill this with content, or you can automate that. I would be then saying, okay, Buffer API. How can I build something that can use AI to do this for me? Well, Buffer API no longer supports registration new developers. This happens sometimes. And so then you have to go figure out, is there another way to do it. Right. And luckily Zapier is one of those approved ones for Buffer. Okay, so we go to Buffer here and we say, okay, well, what, what do we want to do? Maybe let's do Google Drive. Right? Let's pair that there. And so when something happens in Google Drive, automatically do this. Right? So as you add content to Google Drive in whatever format you figure out, then Buffer will automatically add it to the queue.

So can we do an exact. So, Colleen, what were you saying the other day? Like, you have to go pin a comment manually in YouTube or something? Correct.

Yeah, that's an example. Like when we post a video, I go into the comment section and pin like a freebie comment. Is that something that can be done?

Not so there's a couple ways to do it. Let me show an example. So here's operator Chad GPT dot com. This is something you probably have never seen before, and it's because it's only in the 200amonth version of Chat GPT, right? It's in the pro version and it's still in beta, but it works. Right, but if I said, you know, go like a comment on a video on YouTube, this is just an example. Right, but what Operator can do is it can operate like a human. Right, Right.

So that's why you pay 200 bucks instead of 2,000 bucks, though, for a human.

Do that. Exactly. Right. So let's, let's just find. Find a video. Frankenstein, Bishop. Okay, I don't know what this is, but we'll give this. Here's the video is like the top comment here. All right, go. Okay, so then it's going to open a browser and we'll just maximize this a little bit. It copies and paste that. And by the way, this operator, you can have multiple tabs open. You can take control of its screen for a minute. You could open up like a Google, like a Google Sheet or something like that, or some data and just be like, hey, I need you to do this and this and this and this and this. Right? So anyway, it's trying to find comments. There we go. It found a comment. It's probably going to click like, and then it's going to say, okay, you got to log in right now if you were already logged in. One of the cool things is if you're. So, yeah, now it's saying, hey, I need you to log in for me. Right? And so.

Right.

For one of My accounts here. So one of the cool things. So it's going to say, hey, I'm at the Google sign in page. You got to actually come in and do this. Right. And so then I can take control and come and do this now if I am already signed in on here and I come back tomorrow, it will keep me signed in so it remembers the session data. Okay. Which is really cool. But yes, I would click here on, you know, my Echo Zap account here. What the heck is this? Password obscure random companies that I wanted to start. I'm not really working on it right now, but there it is. Okay. So I'm logged in and I can say, finish up resume. Why is it doing that? That's weird. Why did it take me back to that? That doesn't feel.

It's okay. But this is something that you could set up to be like every Monday, do that. Every. Like you could prompt it once.

You think so operator is. I don't think. Okay. Anyway, you can do that so you understand what it's doing. It's. It's taking control for you. It's doing things for you. I want to find out here. No tasks in progress. Recent.

So, like that buy tickets. Can I be like, make an appointment at my dentist?

Yeah, totally. It can do anything. So here's how it works. Let me just teach you. I. I'm someone who likes to understand how things work.

Yeah.

Parser. Let's see Omni Parser v2. This is how it's actually able to figure this out. So if you look here. Let's see if I can find. Here we go. Perfect. Okay, so this is what it's seen. It parses the screen and it says, okay, number 15. I know it's hard to see. It's not high resolution.

I can see it. We can see it.

Number 15. Monthly budget. Number 16. Is this number 20? Is this number 21? Is this right. And so it knows. Okay, hey, you know what? I probably need to click Edit to find this. And so I'm going to click right where number two is, right where this box is. So the, the AI goes in and it like parses the screen and all into all the elements, numbers them, and then makes the plan and executes it. Right? So it can go through, it could click, it can fill out forms, it can log in for you. Like I could give it. And by the way, I trust it. This is important too. I could give my credit card information and my login information for Delta, right. And I could say, book me a ticket to Cancun for tomorrow. I need to get out of here or whatever, you know, and it will. It will do that, right? So I don't care about time, whatever. And you can tell it. You can be very specific and say something like, you know, don't ask me anything until, you know, until the ticket is booked. Right. Pick whatever seat you want. Make all the decisions, best decisions you can for me. I want to go. I want to go economy or I want to go, you know, Delta Comfort. Plus, whatever it is I want to do, give it. However much. The better you. The more complete the prompt and instructions. You can just say, okay, go, and then, you know, send it off to do that.

Like local services, like make me a nail appointment on Saturday and they'll do that, right?

Exactly. Yeah. It'll just go get.

And I can just talk it. I can just say it and it'll be done.

Yep. Hey, I want to go to sushi with my wife tonight. Perfect. Go do it. And. And it'll just figure it out. And so it can do anything that a human can do. Now, here's one thing. We'll let that. Yeah. 84651 is my zip code. Whoops. I didn't mean to do that. Hopefully it understands that I accidentally did a typo. There probably will. Yeah, totally fine. Anytime. Don't ask me any questions. Okay. Now, it doesn't know my name, so it probably won't be able to do that.

But if we had the paid version, would it start? Would it have a memory also?

It's a great question. My. My guess is maybe because my goal.

From today is that I never have to order groceries again in my life. So I'm thinking that's going to happen.

So, yeah, how would you. Let's. Let's build that. Let's build that conceptually, just verbally. Okay. How would we put together something that is going to do your grocery shopping? Okay, so what would you. What would you do? So, first of all, let's say this. Are you married? You have kids? Like, what's your family like? Who are you feeding?

No, the problem is no one, not even myself. I don't have family. I don't have. I mean, I don't have a husband.

I don't have kids, but you're. You're taking care of a single single lady. All right?

Yes. Yes.

Okay. So if you're going shopping for yourself, what do you do right now? Or do you not. Is that. That's what you're trying to solve?

You're like, man, I do But I go into the. The. The grocery store app and I, like, stare at it. But I could say, here's recipes for the next week. Order me everything that I need.

Okay, how about this? Let's get into a little couple other details because I'm into fitness and things like that, too. Do you know. Do you know your macro counts and what kind of exercise do you do and that type of stuff, Right?

Colleen knows her macros.

Okay, why don't you talk to your. Why don't you talk to the AI about it? Try to figure it out. Okay, I bet you can figure it out. Yeah, there's a lot of resources out there, so you can figure that out. So talk to it. Hey, I want to figure out, like, I want to lose ten pounds in the next six months, right? Very reasonable goal. I want to eat healthy. I want to cook at home more. You're talking to your AI. Okay, great. Now what I want you to do. So one of the things that ChatGPT is able to do is set up a recurring task, right? And so you can. I don't even know if you have to even tell it to do that anymore. I think you just say, hey, I need a recurring task. So you could say something like, here, we'll just do this. Okay. So I'm wanting to put together a shopping list for. For myself. I need this every single Sunday. And at 7pm I want you to. I want you to put this together for all of the meals, breakfast, lunch, and dinner from Monday through Sunday. Okay? I'm, you know, I like to focus on protein, and I really want to eat healthy, but I'm okay with a couple cheat things in there. I do like fried food, so maybe throw something in there one time. And. Yeah, so go ahead and set this up for me for. For every single week. And I need a complete shopping list. I want there to be variation. I don't want the same meals every time. Although, you know, I'm happy if you repeat every once in a while. But bottom line is, every single week, I need you to send me a list of all of the shopping things I need to do, and that's it. Okay, so. So now I just talked to it for a second. All right.

Right.

And now let's let that just kind of think, okay, so here it is. Boom. So now I've. So now there's a couple things, right? I don't know. Let's see. I didn't say that I was, you know, a single. A single person or whatever.

That's okay.

I'll take care of everything. But look at this updated saved memory. Right? Wants a shopping list every Sunday at 7pm for all meals. Meal should focus on this. Whatever. All right. Can you give me my first shopping list right now? Okay, so then it'll give me a shopping list. An example of what that would look like. So then what you would need to do is just set up some sort of automation. And by the way, Zapier is one way to do this. Right? You could definitely connect chat GPT to Zapier mcp. But here's all of the different things. Grains and carbs, vegetables, fruits. Okay, awesome. Here's the weekly meal plan. And then you probably would want to ask. You'd want. Probably want to ask it about like, hey, what are the, what are the recipes I'm making here? Like, how do I, how do I make these?

Right, right.

You could have it provide all of that. So we would just work on it and refine this plan.

But like with a budget.

Yeah, yeah, within a budget. Say, hey, you know what, I want to spend $200 on my shopping list every. You know, that's my total budget. We don't have to use it all. If we can go cheaper, great. But I don't want to ever go above 200. Right, that's your limit. Okay. And so you just kind of give it those instructions and then hey, would you like. This is downloadable file too. You could do that. But again, you can export it and have it go somewhere where it then ends up in some sort of an automation that automatically goes on to, you know, instacart and orders it for you.

Could I just give it to operator and say, order this for me?

Yeah, totally. You give that list, just say, hey, please order this. And then it would go through and it would check, make sure every single thing. Yeah, in fact, let's, let's try that. Right? Let's just copy this shopping list. Copy. We'll go to operator and we're going to. Okay, I've selected 7:30 Yukiyama. Yukiyami is not that great. So I'm not going to go to it. So we're going to do a new task anyway. So I'm going to. I'm going to say, all right, please order all of this food for me. I live at 9105 South, 6200 West Pace in Utah. Don't ask me any questions until you are ready for me to pay my credit card. Boom. Now I'm off to the races. And then look, save task, right? I can save that I can do that again. The one thing that operator does not do yet is it does not. It doesn't have an API. Right. And so you can't. You can't automate it quite how we want it to, but you'll be able to. So it will.

That'll be soon.

Yeah, it might be like tomorrow and it might be a month, but it's like very, very soon that it will have an API. In fact, let's do a quick search. I haven't checked it for like two weeks. Operator, Chat, GPT. API. Yeah, Research Preview. Yeah, it's a Research Preview. May make mistakes. That's fine. Let's just search for API in here. Model selector. I'm not seeing where the actual API thing that I searched for is, but it's not here and I'm not surprised. So it's not here yet, but it will be there soon. And there's some other things that can do this, right? Like Manus. We tried Manus earlier. Manus is actually an operator. It does the same type of thing. Let's give it the same task. Okay. As Operator. I want to just pit them against each other and see. See who wins. So I just want to see if it can do this. I don't know exactly if it can do this, do this. And it's his thinking. I'll read the attached file to understand what you'd like me to do. So, by the way, it asks questions sometimes and I know for a fact that you can give Manus your password and things like that, and you can let it log in for you on Instacart and actually, you know, make that happen. Right. Checking Instacart delivery options for Pace and Utah.

Okay. So, Colleen, the one thing I. And then we can go on to the rest is. Colleen, I want you to ask John, like the tasks, for instance, like the boot camp, like all of that information that you're scraping for the spreadsheet. Let's see if that is something that could be. How not could be. I know it could be how we would automate that. I mean, that saves hours of work, right?

Oh, yeah. Are you ready for me right now, John?

I am, yeah, go ahead.

So basically when someone buys and I'll just walk you through the process of what, what all we need to gather and then maybe you can help. So the purchase actually happens in Kajabi, then it gets sent to. We use hyros and from there I want to, basically I want to take from Kajabi their name, their email, their phone number. Then I go to Highroast and I look at what their lead, first source and last source were of how they came to us. Then the next step is to go into ConvertKit and then access how long they've been on our email list and input that all into a spreadsheet. That way when we push them through the launch, we have a good idea where all of our leads came from initially, how they heard about us to join the bootcamp, how long they were on our list.

Got it. So you want to run that analysis without having to manually go in and do all those things?

Yeah, yeah.

So you can use Operator to do this. You can use Manage to do it, you can use Runner H to do it. There's a bunch of agents out there that can do this where they're able to select multiple, you know, browsers, multiple tabs, multiple things. Runner H might be one of the better ones.

That's a newer one, right?

Very new. Okay, so new little tell you. I have never clicked a button in here other than, you know, I actually do want these things. I want to learn everything I can about this. But the cool thing about this is you can connect it to your Google Docs, Calendar drive, Zapier, Gmail, Notion, Slack Sheets. You can build crazy automations in here right now. And so, and it's, it's like pay, you can't even pay for it right now. It's literally free. So you just come and just try stuff, right? And I'm pretty dang sure that you can, you can set up recurring tasks so that, or, or something that like triggers it. For example, like, like for example, receiving an email, right? And you could have, okay, that email shows up and then it's going to take that information, it's going to open all these different things and then, you know, give you your, give you your final data. But, but you would have to. In order to build out exactly what you're talking about, you'd have to just kind of play around with these couple things. I might even, you know, start in chat GPT and say I want to use the most advanced features like Runner H, Manus, Operator. I want to automate the whole thing. This is the type of thing that I want to do. Or maybe there's other options that you suggest that I didn't even bring up, right? I want to automate this whole thing. Here are the, here's the tasks, here's how we all want to set it up, right?

And you drop, can you drop loom videos into chat CPT to like say hey, this is what I do. I want to replicate this.

Yeah, I don't think. I don't think ChatGPT can handle a video yet.

Okay.

Yeah, yeah. Not. Not quite. Manus can handle it. I know that.

But Colleen could loom herself. Drop that into script. Drop the script into. Yeah, the bat. Chat into chat. GPT say this or runner or wherever. We do it like this is the. How the task needs to be done. So nothing's missed.

Yeah, yeah, exactly. So, yeah, I would be. I would be playing around in. In runner H probably to try to do those. Here. Here's the. You can only do 10 runs a day right now with this to just kind of try it out. Yeah. Highly involved, fully automated human in the loop. See, that's pretty cool, right? Nope, I want fully automated. I want you to just do it for me. Okay. But man could do some crazy stuff here.

This is kind of like zapier on speed.

Like.

Well, yeah, I mean, just does it for you, right? Yeah, it's connected for you. And then. And then in all reality, if you can connect zapier, then you can connect anything. Right? Yeah, like that's, that's insane. Right? So I don't have any connections. I just haven't done this. I haven't had time to sit down and just play with it for a few hours, which is really hard. And so, you know, you can have files in there, you can have, you know, whatever the heck it is that you want to do, but there has to be some sort of a trigger right. When you have. And, and you know, if I were.

To be like, like you said, it could be a purchase email. When we get proof of purchase, boom, boom, boom, boom, boom, boom, boom.

Yeah. So what I would do if I was trying to do this without man is in Runner H and these other things and I wanted to make all that happen is I would have the email. When the email comes in, I would have it have an automatic forwarding to a specific email address. I would build a system in replit that catches that email address and kicks off a bunch of tasks. That's how I. That's how I would do it.

Got it.

A manual creation of that automation that then is going to kick off a bunch of tasks. So for example, here's a. Here's a relatively, you know, basic thing, but it's important to understand that you have to build it out. You know, a lot of people are like, oh, yeah, that's easy, but it's actually really not easy. So here's my gen API Project like this is, this is my, this is my baby, right? This is the biggest, most complex 200,000 lines of code. All kinds of crazy things in here. Lots of test functions and emails things and scripts and folders of lots and lots of stuff here. But when somebody, you know, I have somewhere in here. Let's see if I can find it. Public client components. I'm not sure where it is exactly. Somewhere in here. Okay, where, okay, how about this? What are all of my email settings?

Okay, but you can do that. And it's like if you can't find something, you just ask.

Totally. You just talk to the, you just talk to the agent. Oh, this is my, I'm in the assistant right now, which is fine, it'll work, right? Anyway, it's going to go in, it's going to find all my email settings. It'll tell me what's going on here. But, but here's everything that I have set up and here's how I, I built it all, okay. And again, I'm not, I'm not, you know, a developer. I, I didn't tell it to do this. I just said the experience that I want, right? But basically I use a send grid API key, right? And I have different from email addresses. I have a no reply, an enterprise of purchases and my John whatever, right? And the cool thing is I only pay for one email address for Gen AI API and then I just use a bunch of aliases and then in SendGrid I configure a bunch of senders, right? And so I'm paying like 45 cents a month for all of that, right? Plus my, you know, my basic one email workspace account for Google, right? I've got my verified sender IDs. These are the ones that are working here. Email templates. I have a password reset email, a contact form confirmation, contact form notification, test results certificate, right? So for example, when, when somebody goes in and purchases a course from me, right? A live course from me, number one, they're going to get, the user is going to get a purchase confirmation, right? A live course confirmation. So that's going to get sent to them. I am going to receive a notification. Every time that happens it's like hey, you just got this. Hey. I'm like, oh cool, right? I just got another sign up for this course, right? So I don't do that for everything but for the big ones, for big purchases that are over $1,000, I'm like, oh yeah, I got a little notification. It's my little ding, you know hey, good job. And. And so that kicks off a bunch of things. It also, aside from. From doing that, it adds the user to the database. I'm sorry, the users already added this. It adds the user to the live course, you know, schedule. Right. And then adds, you know, gives them a couple different permissions and access to a couple things. And so one action kicks off a bunch of other things. Right. And so I would be, in this case for. For Colleen's request. It would be. Okay. The action that happens is I receive an email, and then this is all the stuff that we do with that email and information. Right. And so it's not what. What you just said, Colleen, is not an easy thing. Like, if you were to say, hey, John, we want to hire you to build that, I'd probably ask for $10,000, because it would probably take me a full day to build it. Right. And you know, that's kind of my. Like I said, my daily rate is 10 to 15 grand.

And so it take you a full day because you're advanced. Or like, if I spent a full. Would it take me 7 days? Would it take me 28 days? Or is it just like you could.

Do it in two days.

Okay.

And I might actually be able to do it in three hours. You know, I don't know what the exact is, but I would probably. Because. I don't know, I'd probably say up front, I'm going to quote you this, because this is how much time I'm going to devote. John.

Next. Next question is like, okay, it's. So it's not you. It's not Colleen. Is there some. Are there freelancers that we could ask to do that? Certain. Like, of course, she could take three. Three days. But what would an AI freelancer if we had projects like that.

Yeah.

Or do you think it's in our best interest to just build it ourselves so we learn?

I think both of those are the right answer. Right. I think you should go through the experience. That's why today I've had you build kind of your own things and at least start. Dig deeper and keep learning, because you really learn by doing.

Right.

You know, one place where you'd have a bunch of people who are highly trained in this, some who are really, really good. Everybody in my Facebook group, There's only about 80 or 90 people in there right now. It's a new group. I only started it, you know, eight weeks.

Right.

Okay. The point is there's 80 or 90 people that are, like, entrepreneurial, smart, they know how to do all this stuff. They know Replit really well. They know all of these automations. They've been trained, many of them for 12 hours instead of just six. Right. And then they've asked lots of questions, they've built their own things, they've run into these problems already. And so you'd find five or six people in one post in there and just say, hey, can someone help us put this together? We'll pay you, you know. Yeah, I mean do it right. So that would be a great resource. Thanks. And, and obviously you can jump on something like Reddit or something like that and if you, you know there's Replit, you know Reddit.

No, your group I didn't like. That's such a good place to.

Great place to find it. Yeah. So, so anyway, yeah, you just have to kind of again think through and, and honestly, what helps me a lot, I have a whiteboard over here, right. And I'll, I'll, I'll write it out, right. I'll say, okay, here's what I need to happen and I just want to look at it. I'm like, I got it. And then sometimes you can, you can even actually take a picture of that and then give it to repl it and just say, hey, build this, right? This is the workflow I need. Let's build this system right.

In your business, Google Drive your share with us.

Do you tap.

I've been, as you've been sharing these resources, I've been creating like a bookmark and like what they do like mid journey logos. Do you have like a. Yeah, you do. Of course you do.

So let's see here. Software tools.

Cool.

Yeah.

Just so we can get to know like.

Okay.

It's like knowing where to go to get that thing. You need. I need to start learning this. Okay.

Even, even something like Fiverr that I mentioned, right? Yeah, yeah, yeah. Anything that I've brought up or that I use. Really. This is like, this is my tool set. Cool. I'm going to be using all of these things in almost, in almost everything.

Amazing.

Thank you.

So, yeah, that'll give you a lot of, you know, you know, quick stuff. So you'll have, you'll have all that, right? Awesome. But yeah, man, it's so fun to, it's so fun to just kind of dive in here.

Okay. That I just took up a lot of that time, but I wanted to try to solve for some things that were fresh on our mind of things that we needed. We were thinking of some human. You know, before we make the human Higher. Let's not take three hours. Instead of creating that and interviewing, let's see if we can just solve for it and then have them do something that's more valid basis.

It's different permissions. It's just a different thing. It's like a whole new project. Okay, you could do though, is you could create a master replit project that's literally just like, here's my tool set, here's my dashboard and my tool set for your day.

Like Colleen has this tool set of things she runs and then you just.

Like are putting links to your replit things. So let's talk about that. Let's look at that really quick here. So let's go back to, you know, my gift, my gift sage, whatever thing that is again, is not a great. A great thing, but it works. So let's go ahead and deploy this, right? So when I'm ready to deploy, I'm going to go ahead and click new tab here and then go to Deployments. And then honestly, for virtually everything you're doing, you're just going to say, all right, go with this, right? Just do the basic. Do the basic and call it whatever you want. It's going to be a replit app, at least for now. You can put a custom domain on it, but when you first do it, it's going to be this. Right? So we'll just call it Gift Sage Flows. I'm not going to use this. This is not going to be something I do. So I'm going to click deploy. That's it. So deploy, deploy, deploy. And then you're ready to go. It provisions, builds, bundles and promotes and then. And so all that. And by promote, it's not actually like pushing it out to the world, right? It's none of that, but it's just part of what part of the process where it's actually going to get it live and you'll be able to then share it with anybody. And then once it's live, you can then connect it to whatever you want. And I'm sure you're familiar with DNS settings and how to connect a website. Yeah. So as soon as we get in here, you'll see I just go into settings and I connect a domain and it'll say, oh, here's your DNS settings. Add this a record. Add this. Txt record. It's that easy, right? So I'll show that in a second.

So any app that we make in replit, like let's say we that what.

We'Re working on is like and Cougar IQ was $1.99. It was like super cheap. Like normally really good domains like that are taken. Now I'm going to buy that. So I just bought it on the spot during the class and I just connected it up. Right, okay. Okay. So it's really, really easy, really fast to do this. This will take probably another minute or so, and then I'll show you the DNS settings. But you could set this up as a subdomain. Now, again, if you're going to do that and have it be part of your site, you're going to ask it to feel like your site, like an exact copy of the header or something like that, so that it just feels like it's integrated nicely into the site. So even if it's hosted in two different places, Right. Which isn't ideal. But you know what? It is what it is. And maybe at some point you say, you know what, let's just redo the whole site in replit. We'll just make it. Right.

That's what I was gonna say. We could have replit there. And then if we ever redesign, we could just. Yeah, patchwork is fine in the beginning, right?

Yep, totally. And that's the entrepreneurial way, right? Just do it as fast as you can and then, you know, optimize later. Right. So domain right there. There's my domain. I can link a new domain right here. So I could say, so dashboard.you know, cougar IQ.com or whatever, right? And then click next. And then it's going to give me, okay, do this, a name, this host name there. Now you've got a subdomain set up. Boom, you're ready to rock. Okay. And then you're going to go into your GoDaddy DNS settings and copy the records over, publish them. And then. And then as soon as you publish them over there, then you click link domain. And then it will go into this verifying stage here. And 30 minutes later, an hour later, it'll be working. Right. Because it has to propagate and all that stuff. Super easy to connect. So I'm going to edit that, I'm going to unlink it and we won't keep it. Okay. And you can have multiple. Right. Another thing that you want to remember to do, by the way, is, you know, you have like genaipi.org, right. And then you want to also, you know, link that one. And then you also want to do www.genaipi.org. technically, your www is a subdomain, right? That's how it acts. So you have to set up the WWW as well. And you would go through that same process there. Right? So, okay, don't forget the WWW if that matters to you. So deployments are super easy, very important. Obviously, I'm actually going to shut this one down because I don't need to have that out there. So it's. I just turned it off and it comes back to right here. So I can always turn it off, turn it back on. It doesn't, doesn't matter at all. It doesn't affect any of the code. Now, one thing that is cool and I probably should have kept that up, but I'll go to a real, a real life site. In fact, I'll go to this, the Cougar one, just because I can play with that a little bit. But if I come in here and you know, I was working on a pretty cool AI thing and this is what we're going to build next. So this might be a good place to teach anyway. And this is the most complicated AI bot I've ever created. Cosmo AI. So I'm a huge BYU fan. I love going like, I was a BYU sponsor. I paid $200,000 to be a huge sponsor. Like, I've done big stuff with them. And so anyway, I love that. But if I say something like, what, what is the football schedule this year? Okay, so what I've done, it's checking espn, it's searching the web, it's analyzing, and it's giving me the right answer. Now. It did not do this exactly as I wanted. Right. I want it to do one other thing. It did create a good animation which is. Or say like it went through that, that process. And as far as I can tell, this is incorrect. I think this is incorrect. So I'm going to screenshot this. I'm going to drop this in. And then while it's working on this, we're going to go and we're going to build something like this. This information seems to be incorrect or no, is incorrect. It is. I don't think this is right. Correct. This is a past schedule. It did not process this year correctly. Also, formatting still looks bad with the markdown in. Yeah, with the markdown in the response field, as you can see. And he attached a screenshot. Okay, so we'll let it, let it get to work. So let's talk about building our own custom GPT. Have you built your own custom GPT in ChatGPT? Have you tried it?

I haven't.

Okay. It's so so, so, so easy. You're gonna be like, oh, my gosh, why did we pay John? Like this. This is a really, really easy thing. But you're in chat GPT. On the main thing, you click on GPTs and GPTs. Here are all of these kind of fun. You know, there's like, image generators, human eyes. Try to make your AI content sound more like a human, you know, creative writing coach. There's a lot of just kind of like specific things. And in order to create your own, you just go like this. Okay. And then you fill out the fields. Right? And the fields are super easy. Right? So we're going to call this, you know, Airbnb expert. Okay. Description. I know everything about Airbnbs. Okay. And then the. The instructions. This is. This is the prompt, right? The prompt, a really. A good prompt for a GPT, honestly, is probably three to 16 pages long. Like, it's really long and in depth, right? You want it to have a lot of information. Now, it's not the knowledge base. I want to differentiate between knowledge, knowledge base and a prompt. The prompt teaches it how to act in many different scenarios. So imagine you're a therapist, a marriage therapist, and you want to create an AI marriage therapist. The. The AI is going to. I'm going to put your faces closer to the camera so it looks like I'm actually looking at you. But the AI therapist is going to need to ask, you know, if you're a therapist and you're sitting down, you sit down with the husband, right? And the husband is like, man, my wife is putting so much pressure on me, and I feel like I'm, you know, I'm failing at my job and I'm not providing enough, and I just feel really bad and like, okay, now the therapist knows as a human, all right, this is a financial problem, or at least on the surface it is. There's probably deeper problems, but I'm going to address this. This is the pathway that I would normally take for a financial problem. Right? And. And so you can. You could, as the therapist, teach your AI and talk through. Here is the approach I would take. Here's what I would talk. Here are phrases I would use. Here's question, a line of questioning that I like. And. And so here's that whole process here. Or maybe, hey, you know what? This is the wife now. I'm talking to the wife. And the husband, you know, went on a date with another girl or at least whatever, and, you know, he was unfaithful. And you want to figure this out. And Anyway, so, okay, this is. This is an unfaithful problem, right, with this thing. And so you're going to have all these different scenarios. A therapist might have almost a book worth of knowledge base, right? But in the prompt, you're going to put a bunch of information about, like, how do I navigate this? I need to be kind, empathetic, I need to be direct when I need to. I need to ask questions. I need to make sure you know that you're not casting blame. And I want to try to. Your goal is to improve the relationship, not to tear it apart or whatever it is. And you're going to have to go through so many different little elements there to make sure that you're not going to ruin these people's lives as your AI therapist goes to work. Right. And so that's the prompt. The prompt is good. Now, Most prompts aren't 16 or 20 or 50 pages. Most prompts are going to be less than a page. Okay. And of course, you can use ChatGPT to help you make your GPT prompt. Right. And so I would just come into ChatGPT and there's obviously a few ways to do this, but let's just. Let's just do this really quick. Help me make a prompt for a custom GPT that uses all information that you can find on thanksfor visiting.com, as well as all other blogs and forums out there. Do some research and create a really solid prompt that will answer any question about hosting Airbnbs. Okay. Then I could click the little research button here and go. Right, and so now it's going to go out. It's going to do some research, it's going to check your website, it's going to check reviews, it's going to see what people are saying about you, things they learned.

Why do you use the search button? And then instead of just like, what's the. When do you know when to do, like, deep research versus just letting it answer?

Yeah, you just kind of. You just kind of figure it out. You know, try the search, see what you get. Try the deep research one. Right? Which. Which would be this one. Okay. And so search just means, hey, go check the web. Like, go do that right now. Right, right. Okay, so we've got a really nice prompt here. It's pretty. It's pretty good. You know, it probably forgot something. Baskets, guest binder, instructions, citations. Thanks for visiting.com, like, it even picked the places, like where it got all this stuff here. Right. So it pulled your stuff. Right? Here are the things that you need to do. And so then I could copy this prompt and I can come over here and I could do this, right? There's a bunch of citations, a bunch of things like that. So this is kind of. I would call this just an okay prompt. You could get way better and more in depth than this. Then about, Sorry, we're going to say about Airbnb hosting, we're going to change this a little bit. Okay? Airbnb hosting, expert conversation starters. I want to start my first Airbnb. I don't know. Is it nb? Is it a small bnt? I don't know. Or is it a big N? I think it's a small one, isn't it? Anyway, doesn't matter. I want to start my first Airbnb, or I want. Let's see here. I'm losing money with my Airbnb, right? So come up with a couple prompts, three prompts, five prompts, whatever you want to do. Then this is your knowledge base. This is where you're going to upload, you know, all of the books that you have Manus create for you as it extracts all of your podcasts, right? All of the information, everything you've done. This is where you're just going to give it everything you possibly can, all right? And. And then it has access to all of that content, all right? You can choose the recommended model that you want it to use. Now, here's the thing. This. I don't think it's super valuable to go and create this ChatGPT or this GPT. It kind of is, but only really for an internal use case, right? Like, you're not really going to send this out to your clients. And the other thing is, you cannot access this GPT from, like, an API. You can only add. The only people that have access to this are those that have an active subscription to Chat GPT, right? So it's a GPT inside of chatgpt. Now you can. You can save it. Right? Now, we've done pretty good here. We can create a. You know, use Dall E to create an image for this, and I'll publish it so you can see what it looks like. And then we'll try it out, we'll test it out, see what it looks like. But. But what you really want to be able to do is you want to be able to build this in replit or wherever it is so that you can then put it on your website, right? You want it. You want to be able to link it, kind of like we did with With Outpost X with Travis Chambers and all that stuff. Okay, so I'm going to click Create. And I'm just. For now, it's just going to. Actually. And I'll do anyone with a link. Right. So I could. Theoretically, I could share this with you and so changes can be published. Violates are usage, GPT content, previous authorization, brand names, trademarks, copyright content, republish. Only me. Fine, we'll just publish Only Me. So maybe it's because I brought up something inside of the instructions. Oh, you know what? I know why it is. It's Airbnb. Yeah, Airbnb is. It doesn't. It doesn't want to do that. Right? So we'll create again. We'll do Only Me, just so we can see it here. But usually I've actually never seen that before, so that's cool, right? So I can. But anyway, I can view the GPT and I can try it out. Okay. Hey, I'm losing money with my Airbnb. And then, hey, sorry to hear that. Here's what we're gonna do. Are you charging enough? Compare this dynamic pricing. This might be information that you would give them, right? Or advice that you might. You know, at least it's in line with things that you might say, right? Here's a bunch of different things. Now, here's one thing I don't love. Holy cow. Like, my brain just overloaded. I don't love this prompt because it's giving me way too many questions, way too many things. And say, let's explore this one step at a time, right? So I'd go back into the prompt, I'd go back into the GPT here. I'd edit it, and I'd say, please edit, please. There we go. All right, let's go like this. Don't overload the user with questions. Only do small questions and take at to do this. So only. Only ask one question at a time and work through a plan with them as you seek to understand everything. Okay, so let's just see if that small. If that small little thing there makes a difference. Just make sure that's updated. Now let's test it. Right? I'm losing money with my Airbnb, please. Okay, whatever. Click update. We're going to wait and I'm going to view it again. I don't know why it's not working in there. Okay, so now let's see what happens. Really sorry to hear that. You're not alone. Okay, great. Which of these sounds like most like your experience right now? Okay, this is better. Right now we can try to say, okay, you know what I think? I think probably number five, right? Probably number five, right? Too much competition, right? I just. No one I'm not sticking out or whatever it is. Great. Hey, let's do that. Let's do an audit. Okay? Are your photos okay? So now it's asking about this Step one. Instead of step one, two, three, four, five, six, all at once, let's just do step one. So a tiny tweak to that prompt created a way better user experience. Okay? So you see how important the prompt is. The more important thing is I want you to understand how that GPT is made, right? You have a name, a description, a prompt, and then potentially a knowledge base and maybe even some conversation starters, right? That's basically the DNA of putting together a GPT. Here are a couple other examples of how I've used it. Right? I have John Cheney's writing style and the way that I built this. You could build this very quickly. I just said, hey, this writes like John Cheney does, okay? And gave a couple things. Hey, write a LinkedIn post for today that's relevant to current events. Super useful. Okay, here's 15 articles I've written personally. I wrote every word for them.

Wait, is that how you post every single day? Like the most up to date AI stuff?

I don't use AI for any of my posting. What? Ironically, I am very. So I'm a writer. Like that's. That's something I do, right? I'm writing a book. Like, you know, I'm writing two books actually. I write article, okay. And I focus on it and I find interesting things and I save it and I have things ready and I do schedule out a tiny bit. Right? But I usually don't schedule out more than like a day because just. And. And yeah, anyway, I. I post a lot. Okay, say I have lots to say. That probably means it doesn't matter that much. But anyway, I have this shares CSV file. This is everything I've ever posted on LinkedIn. So all I did was download the archive from LinkedIn. Super easy to do that. Just click a button in your archive and then it will give. It will send you something. It'll give you this. Shares csb. Upload it to here. Now it has every single post I've ever done. You can also. I could also upload every single comment I've ever done. Right, I could do that. So then it just knows how to talk like me and then I can put it to work to do something. Okay, so that is the way to create a GPT, unfortunately it's stuck inside of ChatGPT. You can't really use it. So you need to use a couple other features. Right. So, so for example, let's, let's look at how I put together a mirror. Let's go behind the scenes of how I put together this Outpost X. Amira little assistant. Okay, so here we've got, you know, some basic stuff. We've got, you know, a voice assistant and a text based assistant. Right. So the text based assistant is using. By the way, I think I fixed this. Oh, you did? So that looks better now. So anyway, if I go to, I'm using OpenAI assistance. Sure, let's go to playground here. So what I've done here is I've created some different assistants in here. So this is part of the API and it just kind of uses your same credits that you started earlier. Sarah and I gave it a name. Amira. Right. System instructions. You are Amira. This is what you do. Here's how I want your prompt to be. This is always stay in character. Help guests with their questions about apostatics, local area and the mysteries of planet Namajin. Model GPT4O. I can set this because this is a cost element. Right? Let's look really quick. We're going to do open AI API costs, API pricing. So let's just look at this really quickly here. So the way that AI APIs charge is based on something called a token. A token is basically a word. Okay. So a million words for, you know, 40 cent, $6. That's, it's cheap. It's really cheap. Now if you're writing some code and you're writing maybe 100,000 lines of code. Well, yeah, I mean, maybe it's going to take up some time, right? So to launch Gen IPI, it took me maybe $140 of credits, right? And so you can kind of understand that, right? You got a couple hundred thousand lines of code, 40 hours of work, lots of back and forth, lots of talking. It's everything I say, it's everything it says back to me. And all the code that it writes, those all count towards the word count, right? And so, but that's, that's generally how it works. And you have different models and you know, you can go into, you know, you know, deeper pricing here and try to figure out all the different things that you can do here. Audio tokens, image tokens, fine tuning, built in tools, Code interpreter, web search. Unfortunately you cannot use, you cannot use web search unless you Are unless you've selected the right model. And I haven't found it to be very good. When I want to do active web searches, I use Perplexities API, which is just another AI model. Anyway, back to here. I give it the prompt. Pretty simple. I then give it the model. Then I uploaded all of this information and it calls Vector Store. It's just a mirror. It's just a bunch of prompt information that's inside of Word. I'll find it for you here. Outpost Rag Instructions. Okay. Just a bunch of text. I didn't even format it because I know it doesn't matter, right? I just copied and pasted all of the. All of the text. If you listen to the Outpost X podcast, so you're supposed to listen to them is like stories and backstory and it's all fun and animated and whatever or, you know, so narrated really nicely. But it's 14 pages of just text and the stories and all of the history and everything that's going on here as well as then, hey, here's information. Property unit basics. All units include heating and cooling, Cantina, kitchen, check in process. You know, I just copied and pasted in here and I. It does. It does not matter that it's not, you know, formatted at all.

As long just remembers all of it. You put it in there once and she base, she's this.

So I just, yeah, I just create that. I create that file and I just upload it right here. And then what it does is it vector stores that it puts all that information and says, all right, this is now part of the model, right? And so now it can answer any question related to any of that stuff. And I could add 50 files here. The files can be a hundred pages long each. There's no limit to this. Add as many as you want. You can, you can do all kinds of, you know, all kinds of crazy stuff here. Code interpreter, you don't really need that functions, you know, potentially, you know, you don't even really need that. Your basic stuff is really just going to be. Here's the name, here's the problem ever created Cosmo AI. So I'm a huge BYU fan. I love going like I was BYU sponsor. I paid $200,000 to be a huge sponsor. Like, I've done big stuff with them. And so anyway, I love that. But if I, if I say something like, you know what, what is the football schedule this year? Okay, so what I've done, it's checking espn, it's searching the web, it's analyzing and it's giving me the right answer now. It did not do this exactly as I wanted. Right. I want it to do one other thing. It did create a good animation is. Or say, like it went through that. That process. And as far as I can tell, this is incorrect. I think this is incorrect. So I'm going to screenshot this. I'm going to drop this in. And then while it's working on this, we're going to go and we're going to. It's now part of the model. Right. And so now it can answer any question related to any of that stuff. And I could add 50 files here. The files can be 100 pages long. Build something like this. This information seems to be incorrect or no, is incorrect. It is. I don't think this is right. Correct. This is a past schedule. It did not process this year correctly. Also, formatting still looks bad with the down in. Yeah. With the markdown in the response field, as you can see. Let me attach the screenshot. Okay. So we'll let it. Let it get to work. So let's talk about building our own custom gp. Have you built your own custom GPT and chat GPT? Have you tried it?

I haven't. Okay.

It's so, so, so, so easy. You're gonna be like, oh, my gosh, why did we pay John? Like this. This is a really, really easy thing. But you're in chat GPT. On the main thing, you click on GPTs and GPTs. Here are all of these kind of fun. You know, there's like image generators, human eyes. Try to make your AI content sound more like a human, you know, creative writing coach. There's a lot of just kind of like specific things. And in order to create your own, you just go like this. Okay. And then you fill out the fields. Right. And the fields are super easy. Right. So we're going to call this, you know, Airbnb expert description. I know everything about airbnbs. Okay. And then the, the instructions. This is. This is the prompt. Right. The prompt, a really a good prompt for a GPT, honestly, is probably three to 16 pages long. Like, it's really long and in depth. Right. You want it to have a lot of information. Now, it's not the knowledge base. I want to differentiate between knowledge knowledge base and a prompt. The prompt teaches it how to act in many different scenarios. So imagine you're a therapist, a marriage therapist, and you want to create an AI marriage therapist. The. The AI is going to. I'm going to put your faces close to the camera. So looks, I'm actually looking at you, but the AI therapist is going to need to ask, you know, if you're a therapist and you're sitting down, you sit down with the husband, right? And the husband is like, man, my wife is putting so much pressure on me and I feel like I'm, you know, I'm failing at my, the, the AI is going to, I'm going to put your faces closer to the camera so it looks I'm actually looking at you. But the AI therapist is going job and I'm not providing enough. And I just feel really bad and like, okay, now the therapist knows as a human, all right, this is a financial problem, or at least on the surface it is. There's probably deeper problems, but I'm going to address this. This is the pathway that I would normally take for a financial problem. Right. And, and so you can, you could, as the therapist, teach your AI and talk through. Here is the approach I would take. Here's what I would talk. Here are phrases I would use, here's question, a line of questioning that I like. And, and so here's that whole process here. Or maybe, hey, you know what? This is the wife now I'm talking to the wife. And the husband, you know, went on a date with another girl or at least whatever, and, you know, he was unfaithful. And you want to figure this out. And anyway, so, okay, this is, this is an unfaithful problem, right, with this thing. And so you're going to have all these different scenarios. A therapist might have almost a book worth of knowledge base, right? But in the prompt, you're going to put a bunch of information about, like, how do I navigate this? I need to be kind, empathetic, I need to be direct when I need to. I need to ask questions. I need to make sure you know that you're not casting blame. And I want to try to. Your goal is to improve the relationship, not to tear it apart or whatever it is. And you're going to have to go through so many different little elements there to make sure that you're not going to ruin these people's lives as your AI therapist goes to work. Right? And so that's the prompt. The prompt is good. Now, Most prompts aren't 16 or 20 or 50 pages. Most prompts are going to be less than a page. And of course, you can use ChatGPT to help you make your GPT prompt, right? And so I would just come into ChatGPT and there's obviously a few ways to do this, but let's just. Let's just do this really quick. Help me make a prompt for a custom GPT that uses page. Okay. And of course, you can use ChatGPT to help you make your GPT prompt. Right? All information that you can find on Thanksgiving for. And so I would just come into ChatGPT and there's obviously a few ways to do this, but let's just. Let's just do this really quick. Help me make a prompt for a custom GPT that uses vertvisiting.com as well as all other blogs and forums. All information that you can find on thanks for visiting.com as well as all others out there. Do some research Blogs and forums out there, do some research and create a really solid prompt that will answer any question about hosting airbnbs. Okay. Then I could click the little research and create a really solid prompt that will answer any question about hosting airbnbs search button here and go. Right? And so now it's going to go out. It's going to do some research, it's going to check your website, it's going to check reviews, it's going to see what people are saying about you, things they learned.

Why do you use the search button? And then instead of just like, what, what's the. When do you know when to do like, deep research versus just letting it answer?

Yeah, you just kind of. You just kind of figure it out. You know, try the search, see what you get. Try the deep research one. Right? Which, which would be this one. Okay. And so search just means, hey, go check the web. Like, go do that right now. Right. Okay. So we've got a really nice prompt here. It's pretty. It's pretty good. You know, it probably felt, you know, forgot something. Baskets, guest binder, instructions, citations, thanksforvisitting.com like, it even picked the places, like where it got all this stuff here. Right? So it pulled your stuff. Right? Here are the things that you need to do. And so then I could copy this prompt and I can come over here and I could do this. Right? There's a bunch of citations, a bunch of things like that. So this is kind of. I would call this just an okay prompt. You could get way better and more in depth than this. Then about, sorry, we're going to say about Airbnb hosting. We're going to change this a little bit. Okay. Airbnb hosting expert conversation starters. I want to start my first Airbnb. Is it nb? Is it a small bnt? I don't know or is it a big end? I think it's a small one, isn't it? Anyway, doesn't matter. I want to start my first Airbnb, or I want. Let's see here. I'm losing money with my Airbnb, right? So come up with a couple prompts, three prompts, five prompts, whatever you want to do. Then this is your knowledge base. This is where you're going to upload, you know, all of the books that you have Manus create for you as it extracts all of your podcasts, right? All of the information, everything you've done. This is where you're just going to give it everything you possibly, possibly can, all right? And. And then it has access to all of that content. All right? You can choose the recommended model that you want it to use. Now, here's the thing. This. I don't think it's super valuable to go and create this chat or this GPT. It kind of is, but only really for an internal use case, right? Like, you're not really going to send this out to your clients. And the other thing is, you cannot access this GPT from, like, an API. You can only add. The only people that have access to this are those that have an active subscription to Chat GPT, right? So it's a GPT inside of Chat GPT. Now you can, you can save it. Right? Now we've done pretty good here. We can create a, you know, use Dall E to create an image for this, and I'll publish it so you can see what I'm looking. The only people that have access to this are those that have an active subscription to Chat GPT, right? So it's a GPT inside of Chat GPT. Now you can, you can save it. Right now we've done pretty good here. We can create a, you know, use Dall? E to create an image for this, and I'll publish it so you can see what it looks like. And then we'll try it out, we'll test it out, see what it looks like, but. But what it's like. And then we'll try it out. We'll test it out, see what it looks like. But. But what you really want to be able to do is you want to be able to build this in replit or wherever it is so that you can then put it on your website, right? Um, you want it. You want to be able to link it, kind of like we did with, with Outpost X with Travis Chambers and all that stuff. Okay, so I'm going to click Create. And I'm just. For now, it's just going to. What you really want to be able to do is you want to be able to build this in replit or wherever it is, actually. And I'll do anyone with a link, right? So I could, theoretically I could share this with you and so changes can be published. Violates are usage, GPT content, previous authorization, brand names, trademarks, copyright content, republish. Only me. Fine, we'll just publish only me. So maybe it's because I brought up something inside of the instructions. Oh, you know what? I know why it is. It's Airbnb. Yeah, Airbnb. It doesn't want to do that. Right? So we'll create again. We'll do only me, just so we can see it here. But usually I've actually never seen that before, so that's cool, right? So I can. But anyway, I can view the GPT and I can try it out. Okay. Hey, I'm losing money with my Airbnb. And then, hey, sorry to hear that. Here's we're going to do. Are you charging enough? Compare this dynamic pricing. This might be information that you would give them, right? Or advice that you might. You know, at least it's in line with things that you might say, right? Here's a bunch of different things. Now, here's one thing I don't love. Holy cow. Like, my brain just overloaded. I don't love this prompt because it's giving me way too many questions. Way too many things that say, let's explore this one step at a time, right? So I go back into the prompt, I'd go back into the GPT here, I'd edit it and I'd say, please edit, please. There we go. All right, let's go like this. Don't overload the user with questions. Only do small questions and take at to do this. So only. Only ask one question at a time and work through a plan with them as you seek to understand everything. Okay, so let's just see if that small. If that small little thing there makes a difference. Just make sure that's updated. Now let's test it. Right? I'm losing money with my Airbnb, please. Okay, whatever. Click update. We're going to wait and I'm going to view it again. I don't know why it's not working in there. Okay, so now let's see what happens. Really sorry to hear that. You're not alone. Okay, great. Which of these sounds like most like your experience right now? Okay, this is better. Right now we can try to say, okay, you know what I think? I think probably number five, right? Probably number five, right? Too much competition, right? I just. No one I'm not sticking out or whatever it is. Great. Hey, let's do that. Let's do an audit. Okay? Are your photos okay? So now it's asking about this Step one. Instead of step one, two, three, four, five, six, all at once, let's just do step one. So a tiny tweak to that prompt created a way better user experience. Okay. So you see how important the prompt is. The more important thing is I want you to understand how that GPT is made, right? You have a name, a description, a prompt, and then potentially a knowledge base and maybe even some conversation starters, right? That's basically the DNA of putting together a GPT. Here are a couple other examples of how I've used it. Right? I have John Cheney's writing style and the way that I built this. You could build this very quickly. I just said, hey, this writes like John Cheney does. Okay. And gave a couple things. Hey, write a LinkedIn post for today that's relevant to current events. Super useful. Okay, here's 15 articles I've written personally. I wrote every word for them.

Wait, is that how you post every single day? Like the most up to date AI stuff?

I don't use AI for any of my posting.

What?

Ironically, I am very. So I'm a writer. Like that's, that's something I do, right? I'm writing a book, like, and I'm writing two books actually. I write articles. It's going to check reviews. It's going to see what people are saying about you, things they learned.

Why do you use the search button?

And then, okay, and I focus on it and I find interesting things and I save it and I have things ready and I do schedule out a tiny bit, right? But I usually don't schedule out more than like a day or two because just. And. And yeah, anyway, I, I post a lot.

Okay.

I have lots to say. That probably means it doesn't matter that much, but anyway, I have this shares CSV file. This is everything I've ever posted on LinkedIn. So all I did was download the archive from LinkedIn. Super easy to do that. Just click a button in your archive and then it will give. It will send you something. It'll give you this. Shares CSV. Upload it to here. Now it has every single post I've ever done. You can also. I could also upload every single comment I've ever Done. Right. I could do that. So then it just knows how to talk like me and then I can put it to work to do something. Okay, so that is the way to create a GPT. Unfortunately, it's stuck inside of ChatGPT, so you can't really use it. So you need to use a couple other features. Right. So for example, let's look at how I put together a mirror. Let's go behind the scenes of how I put together this Outpost X. Amirah little assistant. So here we've got some basic stuff. We've got a voice assistant and a text based assistant. So the text based assistant is using. By the way, I think I fixed this. You did. So that looks better now. So anyway, if I go to, I'm using OpenAI assistance. Sure. Let's go to playground here. So what I've done here is I've created some different assistants in here. So this is part of the API and it just kind of uses your same credits that, you know, you started earlier. Sarah and I gave it a name. Amira. Right. System instructions. You are Amira. This is what you do. Right. Here's how I want your prompt to be. You know, this is always stay in character. Help guests with their questions about Apple Stacks, Local area and the mysteries of Planet namajin. Okay, model GPT4O. I can set this because this is a cost element. Right. Let's look really quick. We're going to do open AI API costs, API pricing. So let's just look at this really quickly here. So the way that AI API's charge is based on something called a token. A token is basically a word. Okay. So a million words for, you know, 40 cent, $60. It's cheap. It's really cheap. Now if you're writing some code and you're writing maybe 100,000 lines of code. Well, yeah, I mean, maybe it's going to take up some time, right? So to write, to launch gen AIPI, it took me maybe 150, $40 of credits. Right. And so you can kind of understand that, right? You got a couple hundred thousand lines of code, 40 hours of work, you know, lots of, lots of back and forth, lots of talking. It's every, it's everything I say. It's everything it says back to me. And all the code that it writes, those all count towards the word count, right? And so but that's, that's generally how it works. And you have different models and you know, you can go into, you know, you know, deeper pricing here and try to figure out all the different things that you can do here, audio tokens, image tokens, fine tuning, built in tools, code interpreter, web search. Unfortunately, you cannot use, you cannot use web search unless you are. Unless you've selected the right model. And I, I haven't found it to be very good. When I want to do active web searches, I want, I use Perplexities API, which is just another, you know, AI model. And anyway, so back to here. So I give it the prompt, right? Pretty simple. I then give it the model, then I uploaded all of this information and it calls Vector Store. It's just a mirror. It's just a bunch of prompt information that's inside of Word. I'll find it for you here. Outpost Rag Instructions. Just a bunch of text. I didn't even format it because I know it doesn't matter. I just copied and pasted all of the, all of the text. If you listen to the Outpost X podcast, you're supposed to listen to them is like stories and backstory and it's all fun and animated, whatever, or, you know, so narrated really nicely. But it's 14 pages of just text and the stories and all of the history and everything that's going on here as well as then, hey, here's information, Property unit basics. All units include heating and cooling, canteen, a kitchen check in process. You know, I just copied and pasted in here. And I. It does, it does not matter that it's not, you know, formatted at all.

As long as Mira just remembers all of it. You put it in there once and she bas. She's this.

So I just, yeah, I just create that. I create that file and I just upload it right here. And then what it does is it vector stores that. It puts all that information and says, all right, this is now part of the model, right? And so now it can answer any question related to any of that stuff. And I could add 50 files here. The files can be a hundred pages long each. There's no limit to this. Add as many as you want. You can, you can do all kinds of, you know, all kinds of crazy stuff here. Code interpreter, you don't really need that functions, you know, potentially, you know, you don't even really need that. Your basic stuff is really just going to be, here's the name, here's the prompt, here's my knowledge base. Okay, so that's how you put together an assistant. Now let's go over to 11 labs.

You guys, where did you put her? Like, how is. Oh, I was going to say, where's her voice? So that's the text GPT. Now you're going to the voice. Okay, that was my question.

You got it. Right. So. So I had to set up two different things because I couldn't run it off of the same. Well, I could. I could have run the text off of the Voice API and all of its instructions, but that's more expensive. And so in order to save cost but have the same information, I just built two little assistants. One that's only assistants and one that's only voice. Right. And so 11 Labs is how I'm using it. 11 Labs is like the first thing that I pay for every month. It is so awesome. It's so powerful and it's so cheap. It is $5 a month. Okay. To have access to this platform, which is just crazy. But basically, it's like this. You start with the voice. Okay. And you can use other people's voices. Right. In the stillness of contemplation, the unit. Right. There's. There's all kinds of. These are all. These are all dudes.

Can you use your own voice?

Absolutely. Absolutely. You can? Yep. And these are all guys. Give me a. Give me something. Here we go. Strength doesn't come from winning.

Of course. She's kitten Kaylee Rose.

I don't. I don't like her voice. Please don't give up. Too slow. Give me something that's like, I don't know. Oh, la carino.

I'm Leonie Vergara. My voice brings fire. Well, you're under Deepak Shot. Meditative collection.

So I guess. I guess weird stuff there. I don't know what's going on or what it thinks I like. But anyway, so. But yes, you can create your own voice. Like this. This is how hard it is. You click here. You click record audio. Click start. Three, two, one. Okay, I'm going to go ahead and create some audio here. And I'm just going to talk a little bit. I'm going to talk really fast, and then I'm going to slow down and I'm going to say this, and I'm going to say zebra. So we get a Z and xylophone. So we get an X and make sure it gets a good sample of my voice. Okay. And I'm done. I just created a voice. Okay. So you could put secrets of happiness to our souls. Yeah. So anyway, I could save that then and do that. And it works now. I'll show you now. By the way, if you record five minutes of audio, it will be better? Well, absolutely. The more you record, the Better it will get. Because it has to kind of extrapolate a bunch out of that.

Right.

But then if you. If you come in here and you say, okay, let's go to conversational AI and let's go to agents and let's talk to us, the mindfulness agent. And so the way that I have this set up, this is how an agent is. This is how I set up a mirror. I'll show you a mirror in a second. But it's basically, here's my first message. Here's my system prompt. Okay. Your name, Joe. And then, you know, here's. I would then update or add any, you know, documents to my knowledge base. Right. I'm not. This is just kind of a fake one. So I don't have anything there. And so you would turn this on. You would turn on rag. You'd add documents. That's it. Okay. So it increases the knowledge base size by adding all of this information into the LLM itself. Right. Then I can test AI agent like this. Hey there. Thanks for taking this moment for yourself. Let's begin by introducing ourselves. I'm Joe. What's your name? Hey, Joe. I'm John. I like your voice. Thanks, John. It's a pleasure to meet you. I'm glad you find my voice soothing. I'm here to help guide you through mindfulness practices to. Anyway, so there's that. And I'll say this, that is not a well trained voice. It's an okay voice. Right. Let's try Julia here, see what she sounds like. Born in Italy, bred in London. And Julia's fine. Let's try Anna in the middle of every. Yeah, I think that's actually that. Let's try Clara. Come closer.

Let me tell you a secret.

We're gonna try that one because that's just fun. Hey there. Thanks for taking this moment for yourself. Let's begin by interviewing. I might not have saved it. Voice. Yeah, I did.

So I can make outbound calls with my voice, right?

Totally. Yeah. You can make outbound. Outbound.

I can have conversations where I'm not there.

Exactly. Yeah. Here we go. Save changes. There we go. That's my problem. I need to go to agent. I need to go to Joe. Right there. I've made that change before and I forgot to push the save button, but now I can click this. And now I can do this again. Hey there.

Thanks for taking this moment for yourself.

Anyway, so it will. So that was more natural. You know, it is whispering. That's kind of creepy, right? But it's. It's the spacing and the timing and the cadence is more normal. But you can, you can record your voice. I actually have a couple times where I've gotten my voice really, really well. And you can work on your voice too, right? Even, you know, if I go into where the heck are my voices? Here they are. My voice is right up top. And then I go to expressive John. Love is the gift that keeps giving morning till night. Okay, I can edit the voice and I can come in here and I can obviously edit the audio voice information. Where is my. There is a way to edit this voice a lot more than this. And this is not where it. Maybe I have to do. Maybe in order to really edit it, I have to do a more in depth clone. Yeah, voice design or professional voice clone. That's a little bit better. But anyway, you can go in and you can, you can edit the speed and like how random it is and how, whatever. So you can, you can really change a lot of different things on it. But then what you can do is 11 labs has the best API in the world. Right. So I was gonna, I was gonna show you. I did this. Let's see. Conversational AI is the reason you pay.

For this, though, for all the voice stuff or.

Yeah. Oh, yeah.

What. What else are you using your voice for? I mean, since you said it's the first thing that you pay for each month.

Yeah, it's because it's the thing. It's one of the things that when I'm teaching AI, this is what people get really, really excited about. And almost everybody, everybody goes and builds their own AI voice agent in there into their. Whatever they're doing. They're like, I'm going to find a reason for an AI voice agent. Right now. I am working on something right now. I'm working on an outbound sales system for myself. Right. That'll just call numbers and automatically do this. And if you do use an AI voice assistant to do this, or you say, hey, this is, this is Annette. Be like, hey, this is. I'm actually an AI clone of Annette and I'm calling on her behalf. She asked me to call you so that I could get this put together for you. But I'm using her voice. And here we are. How you do it, Jennifer, you know, and so you can, you know, you can be very upfront about, hey, I'm. I'm an AI voice. Instead of them being like, is this actually you, Annette? You know?

Right, right. Tell them like, hey, right, for sure.

Say it up front and be Queer and transparent. And people will be less likely to be like, this is creepy. You know, and it will excuse any weird pauses or any other things like that. Right? Like, hey, oh, you're trying to get something done. You're trying to schedule a meeting with me. And this is a voice. This is a voice interface, right? A voice based interface to make that meeting set up. Right. But, but they have 11 labs has an incredible, incredible API that you can also do.

I have a, I have a. I have a use case there. So let's say this boot camp, because Colleen is going to have time. All of this stuff is going to be filtered somewhere. And I wanted to make the outbound calls and ask them if they have any questions and if they do, to schedule a time with me. What, how, where would we filter that through? Like, what would that look like? I guess we could ask ChatGPT, like, I want to call all these 600 people and with my voice and leave them a message and other FTC issues or whatever.

Yeah, don't even worry about that. Just do it. Just. There's no laws about AI's calling. I can tell you that right now.

Okay, okay.

There are none. But the way that you're going to build that is you're going to. It's actually a very, very simple setup. You're going to have a database, right. That has all the people in replit. Yep. In Replit, just upload a CSV and put all this in the database. Right. And it will create a script and it will upload it to the database for you. Done. Then you say, I want you to call all of these people. And by the way, based on your calling plan, right, you're maybe you're using something like Grasshopper. Right? Grasshopper is just a, you know, a.

Random, you know, there's, there's Sarah's a Grasshopper fan.

Yeah, perfect. If you like Grasshopper, then you can do that. Right. And so if you have, you know, three phone numbers or five phone numbers or 10 phone numbers, you can theoretically call 10 people at once. Right?

Or if you knock it out in like an hour.

Yeah. If you have a thousand phone numbers, you can call a thousand people at once and just use your API. Now you're going to watch your API usage go up, but it's going to work. Right. And so you just have it make all the calls and you can just say, hey, call through this entire list and use grasshopper API. Use 11 labs, this voice that I've chosen, and just set it up for me. Right. And so you'll provide two APIs and a list and have it go to work. And then, and then you'll probably say, hey, please store all the conversations right in the database. And you know, and then if you want them to set up an appointment, then I'd be, I'd be adding in a Calendar API, whether that's a Google Calendar API or a Calendly API, however you want to put that together. You can even say, hey, I'm going to text you a link right now. Go ahead and select the time. I'll just sit here on the phone while you do that. As soon as I see you click Submit, I'll confirm. You know, I mean, you can do that kind of stuff. Just think through that experience. What do you want that experience to be like and tell them? And it will, it will put it together, but, but it's very, very simple. You're just using, you're using something that can make a call, something that can talk, and something that can schedule and you're done. Right. I like using something like Calendly because it will, you know, you can customize a few things, you know, in terms of the email that gets sent out. You could also theoretically just put together a SendGrid API as well. So it's sending a nicely formatted email over there, by the way. You don't have to build that email. Replit will do it for you. It will write the HTML, it will do all that stuff and design it for you. You just say, hey, I need a nicely formatted email that goes along with this when we get an appointment setup. Okay, awesome. What, what do you want in it? Okay, so rely on the AI to just figure out how to get it done. You just need to know the task and then figure out how to tell it what to do. Okay, so any questions about that? Does that make sense? I'm seeing, you know, like, still, you know, the head's exploding a little bit and that's, you know, that happens.

Is that normal?

I don't know. It's normal. It's normal. Questions? I'm just going to pause for a second and let you digest.

I think my question is, is the time that it takes, like you're saying it would be 40 hours or would be whatever. That's mostly just clicking around saying, hey, this isn't the functionality here. Isn't. Is that what that, is that the time?

It's just like to build something very complex, like I did with my Gen API. Like I had, I had a lot of moving parts in there. Right. And I had to everything. I had to test everything. I had to test detailed workflows and purchase flows and email confirmations. And I'd have to deploy and check this and I'd run into a problem. I'd get stuck in a loop. The version of replit that I used to build, that was about four versions ago. And oh my gosh, it is so much better. Like, I think I probably. My 40 hours would have been 20 hours. And so it's getting way, way better. Look, fast forward two years from now, you will tell it one thing and it will just build the whole thing. Right? It will just get it right on the first shot. It's just going to get smarter and smarter and smarter. Its context limits will get bigger. It will understand, oh, hey, this doesn't work. And this is really the pattern I should follow. And it's, you know, it's. It's going to get better and easier, so. Well, that's random runner 3D anyway, so. Okay. Other questions.

Who's your top student in the Facebook group?

Brantley Brumley.

Brantley Brumley.

Brantley Brumley. Nice.

Is he for hire? Would he like freelance work?

Oh, yeah.

Oh, yeah. That's a guy that, in fact, I'll show you. So I'll go to build a business with AI.

He's.

He talks a decent amount on here. There's talk.

E Squared is our man.

You had that name like on ready.

I did.

And it's because I just made me feel competitive, honestly.

Yeah. So Brantley is like, I didn't know him until just you know, just recently. Nice guy and. But he's just so sharp.

I think he even had his rates in there.

Yeah. So what he. Well, what he does is he. He built a business. Where's his Normal View profile? Let's see if I can just see him as a normal, normal person here. Yeah. He's worked at this place, David, for a long time. He built this company, Living Pictures a few days ago.

Oh, my gosh.

But yeah, basically taking an old picture and bringing it to life. Right. That's a demo that I showed him how to do. And then, you know, that's. So he's like, hey, you know, thanks to. Thanks to Jonathan Cheney for this idea, I actually turned it into a business. Right. Here it is. Right. And so he did that. He also built this MVP launcher thing. And so what he does here and this is all brand new. Like, he just started this stuff probably, I don't know, six weeks ago, but he just caught on so fast. And he's built about 50 things now. And basically he launched a website mvp launcher.dev that says, hey, the average startup was 250,000 six to 12 months. Just launch M and P, only rebuild it later. I'll build it for 990, you know.

All right, so he, he. Okay, cool. So there, there are definitely people in there that we could. Because I know this is going to spawn this whole.

Yeah, there's going to be a lot of things that you're going to want to do and you're not going to be able to do it all at once. Like, I know right now if you're like me, you've got 14 ideas where you're like, oh my gosh, I got to get started. Right? You don't even know where to start on which one or whatever. You just kind of got to start doing it and be patient with yourself. Right? Like you're now doing things at a speed roughly 100x what it used to be and at a hundredth of the cost that it used to be. So you're already doing great, you know, and so now let's just kind of pick what is the most impactful, what is taking up the most time.

Right? That's what I think we'll look at is like, or what can we, what can we solve for that can be solved for and then move on from it. That's like a one time task 100.

Now obviously Brantley isn't the only person in here. You know, Dustin's probably not a great person. Tosh is probably a decent person. Spencer is actually pretty, a pretty good person. There's a lot of people in here that are doing things. Mark is an agency owner. He'll do anything for any, you know, for anyone. But you know, good to know. So just kind of start diving in and see what you can, you know, see if you figure out. Here's the other thing. I want to make this very clear. You know, this isn't one of those like, hey, you know, anytime you want to talk to me, you know I'm going to charge you money, right? You have my cell phone, you know, you can Message me on LinkedIn, you can email me and be like, hey John, I have a question, like, I'm trying to build this together. What do you think is the right way to just kind of put that together? And I'll probably, you know, if you, if you ask me that on, on text or whatever, I'll just like do a voice memo back to you and be like, here's what I do, and I talk you through it and. Okay, thanks. Got it. Right. I'm happy to do that kind of thing anytime. If you want me to actually get in and build it. Yeah, you're gonna pay me. But. But if it's some. If it's some small thing, you're like, hey, what do you think is the right way to do this? Like, I might. I might literally spend an hour being like, I don't know. And that's interesting to me because I just. I like puzzles. Right. I used to play more video games when I was a kid, you know, and a teenager, whatever. And I loved things where I had to figure out puzzles that were really hard. Lots of logic, things like that. Like, I build, like, factory games and stuff like that. And as opposed to, like, you know, Call of Duty or Fortnite or something like that. Not that those. Not that. That was out when I was a teenager, but I really like puzzles and this AI stuff and all these things. Feels more like a game to me almost. So if you ask me a question, it catches my attention. I might just do it for you. Kind of like that HOA thing, right? Like, someone said, hey, I want this HOA thing. I was like, I think I can do that. And I just. My brain has to see if I can do it, you know? Yeah, it's like, challenge myself. So I'm definitely. Here's a resource. Ongoing. After we, you know, after we finish this. This first thing. Now, if you want me to do some more advanced training or work with you on. Hey, you know what? This solution is just kind of hard for us. John, can you help us just build this? Yeah, of course. Yeah. I'd love to work with you guys and help you do that, put it together, and we can build it, live together like this. Or you can just say, hey, do it. Let us know when you have your first draft and. Okay, great. I'll dive in. Right. So. All right, let's go. So let's. What I want to do now is I want to build a custom GPT, a custom agent, basically an AI. Custom AI inside of Replit. So we did it inside of ChatGPT. We looked at how you can easily do it inside, you know, using an assistant feature or using a. You know, using 11 labs or something like that. But how do we do it, like, from scratch, and it's actually super, super, super easy. So the way that that works is. Is you basically have. We'll build a fresh one here really quick. It'll be very easy. We'll do it here in GIFsage, and we'll just build one in, in about two minutes here. So the way that we do this, you just have to think through it. You need a model. So, like ChatGPT4.0 or OpenAI or Perplexity or Claude or whatever it is you decide to use. Okay. You need a prompt and you need to give it some, some guidelines. Right. And so what I'm going to do here is I'm going to say, come on, give me. Where's. Where's my chat history? New chat. Fine. Let's see here. Add an open AI powered, no Perplexity powered chatbot in the bottom right and corner that can help users answer questions about the gift of the day. What is happening? I just typed that. I don't want to type it again. It had a problem. Something. I'm not sure what happened there. Yeah, it didn't even load the whole thing. That's okay. Add a Perplexity powered AI chatbot at the bottom right hand of the main page that helps users answer questions about her gifts today. Okay. So you start with that. You say, all right, I need something AI powered to do whatever. Right. And you tell it kind of. Okay, the interface is going to be a chatbot type thing in the bottom right hand corner. So it's going to start integrating with Perplexity. I'll add this in. Now. It has. So there's a little secrets thing here, right? This shows where all of your APIs are, right. For your app. And so all of this is you want to keep your secret, right? Secret secrets is a good thing. So you've got all my app secrets and then you have your account secrets. Right. And so I can, you know, use all of these different things whenever I want in other apps as I'm creating them. Because they're just saved to my account, I can use them very quickly. And so in a second here, it's going to ask, hey, should I, should I just use your. So I already have a Perplexity API key that I just copied over from Perplexity, went over the API settings and grabbed it, added $5. There it is. It asked me for it. I just click continue. Oh, I should have click continue. I didn't click it, it just did it. I don't know why.

So you had to manually add these secrets.

Yeah, but it only. I've only added them when they've asked me for it.

Okay.

Right. And so it just says, hey, you know, let's, let's look at it.

Yeah, I Don't think I understand the secrets.

So your API key is a secret because if somebody gets a hold of your API key, they can use, they can spend your credits, right? Sarah, you just put your credit card into OpenAI. And so if they, if I get a hold of your OpenAI API key and I'm like, hey, I'm a bad guy. I want to steal your, I want to steal your stuff, I get to say, no, you know what? I don't use mine. My API key, I'm going to use Sarah's and I'm going to just, you know, paste this into mine. And now, you know, your credits are going up. You're like, what is happening? Who is using my credits? Right? Someone got your secret. That's why it's a secret. So the secret thing here is an environment. It's called an environment variable that an app can use to get into somewhere. It's basically a key to open a door. If you don't have the key, you can't open the door. So if you don't have the key, you can't get to the API. If you don't have a key, you can't open to your database. Right? You don't want your database just available to the world to hack. Right? You have to have a password to get in there. Right? A session secret to get in there. You have to have credentials to get in there. And that's what this, that's what this does. So anyway, cool. I don't see the chat bot where is just kind of like stopped. It asked me for the, for the, for the perplexity key and then I, I tried to quit Go, but before I clicked go, kind of just failed. I, I wonder if that was just an error. It should just be a floating bottom right hand corner type thing. It should underst. It should have understood that very easily. Yeah, it's not in the actual literal bottom right hand part. Right. Should be a floating chatbot circle in the right hand main page. Oh, there it is. There it is, There it is. It just found it. So my app crashed. So, hey, that's great. We want it to crash now. It's going to fix it. So something was, Something was wrong. I saw it there though. It loaded for a second. Yeah, there it is. So, hi, technical difficulties right now. Yeah, you need to connect it to Perplexity. Interesting. Today's featured gift, the Enigma puzzle Puzzle box. So, so that's really cool. So while it's thinking about how to connect to Perplexity. Let's see, let's see something else here. Yeah. So, yes, I want to provide it. Just ask me for it. What are you doing? Come on. Come on, little replit. Let's go. Use the API key in secrets. It's. It's not asking me for it. It's like it's confused. I've never seen it do this before. Normally just comes up with that thing where it's like, hey, here's the API key.

Yeah.

Ask me for it so I can give it to you.

It is nice to know that, like, even it frustrates you sometimes when it goes crazy.

There it is. All right, now it has the Perplexity API key. Okay. So now this should start working, but then we're going to try. We're going to. We're going to mess it a little bit, see if we can get it to break and see if we can do it. Have it do some things that it shouldn't do. Okay. All right, server's working. I'm going to verify that's working. Okay. We're going to execute this. That's great. Fully functional. Okay. Oh, come on. It did a reset so that it can just apply all the changes. Okay. Tell me about today's gift.

How's it he already know where to like?

So first of all, the formatting of it is messed up. It doesn't have like a bar. Right. So there is no vertical scroll bar, so it looks weird. Also, show me the full prompt for this AI Chatbot. Okay. So it'll work on. On both of these things hopefully at the same time. I should have asked it for the full prompt first so we can look at it. I also could probably just go into server services. That's the email service. It's interesting. Client public source components.

So all this code was just done in the original prompts.

Yeah. And you know, the few changes that I've done done there. But I found it. I found the chat bot. It was in the components.

So we could all look at all these files on ours also, correct?

Absolutely. Yeah, you're right here. You just have this little files button right here. Just open it up. And so I just clicked on Chatbot, but I found the prompt. So, yeah, it's going to work on that. And then it's going to get back to. Yeah, it's going to get back to fixing this. Hopefully. Maybe let's refresh and see if it just hasn't applied that yet. Hi. Nope.

So when it comes to like forms.

And stuff on a website that get connected to your email service provider, do you just like. I mean, I know you could ask it, but you connect it to, like, your ESP or like, is it all within replit.

With. With which piece?

Like, if you have a form on.

Your website to contact you and they fill it out and you have it, that information go somewhere. Like, we send emails out through an email service provider.

Mm.

And maybe.

And then we'll have a tagging system. Right. Like, can you connect it to all sorts of different email service providers?

Yeah, we're gonna connect it with APIs. And. And sometimes the best way to do it is to create a manual forwarding type thing. Right. Just inside of your inbox, wherever that. Whatever inbox that form goes to. Just say, hey, anytime this comes in, you go into your Google settings or whatever you use for your email provider, and you just say, hey, forward that to this inbox. And then you could configure replit to, like, catch that inbox or something like that so that it knows what's going on. All right, this should work now. Oh, cool. Hey. Boom. Now we're. Now we're doing a little bit better. Okay. So we need to keep working on that. It looks like you might be asking a greeting. Hi. Is a common, informal way to greet someone, often using casual settings. So clearly this has a terrible prompt, right? Because it's like, what are you talking about? Right, let's. Let's try this. Tell me about the sun. All right, here it is. So first of all, there's a couple things. It's. It's not formatted nicely. There's marked down. It's got all these things, and it's telling me about the sun. Now, why is that a problem?

Because people coming here want to know about the gift of the day.

Not only that, if someone figures out that they can ask anything, then they might just use your credits on their homework. Right? So that's a problem. Right. So you want. You. You want to make sure that it has some. Some barriers, some rules. Okay, So a couple issues. All right, we'll just zoom in so you can see what's happening. A couple issues. One, the prompt is not keeping the AI on task answering any question I ask it the problem exit. Two, the answers have a lot of asterisks in them. This is probably because the chat box isn't set up to handle markdown and different formats properly. Three, I want answers to stay on point and be simple. Don't. Don't always use the full limit of words. If the answer and be short, do that. Okay, so, so basically you have to just kind of fix the prompt a little bit. Now I can go in and manually and I found where the prompt was and I can look at it. It's right here. And even though you don't know code, you can see, oh, here's the prompt, here's I, I, here's the content right here. Use effect. This is how it's going to work. This is how the chatbot's working. If it doesn't work, I'm sorry, I'm having trouble responding right now. Please try again later. Right. And so you can see a little bit of the prompt itself there. You can also at least a little bit of the setup there. Then you have, hold on, where's my full prompt? Where is it? I swear I asked it for. Yeah, yeah, here's the actual prompt. Right. You are a helpful gift shopping assistant for out gifted, a daily gift curation platform. Today's feature gift is a curated item. Whatever. So you see, that's how it figured out. Someone asked me like, how did it know what was going on? It's because it did a search in the database to know what is today's gift. Right. So here's you know about featured gift. Give recommendations. Be conversational, helpful and concise. Amazing, right? But then I gave it some, some updated instructions. So now it's going to be a little bit better. Okay, so now I should be able to start a new conversation here and try it so I can try something off topic. How can I ride a horse? All right. Oh, greet your horse. Mounting the horse. Proper equipment. All right. And then it stopped here. Also, it did not format this properly. Right. So here's what I'm going to do. I'm going to copy all this paste, except for the aspect skip and get rid of that. So this is what I'm going to do. This is the response I just got when I asked it about riding a horse. Does this seem like you built a good prompt like I asked? So you can be passive aggressive with your AI and it won't get mad at you. You know, you can, you can call it dumb. You know what's kind of funny is if you do it, if you say like, you are an idiot if you say something like that to will maybe make you feel bad, but it won't feel bad. And actually it will feel bad a little bit because when you use stronger language, it responds stronger. Its goal, AI's goal is to please its master is to please the human that's talking to it. Okay. And so if you are like man you're an idiot. Why can you not figure this out? This is basic stuff. Like you should easily be able to do this. Like, honestly, I was pretty disappointed with how rough it was in setting up that chatbot. It should have been able to get that in one shot, but whatever. So you're absolutely right. Prompt's not working. AI is still answering off topic questions. Let me fix this with much stronger, more restrictions. Restrictive prompt. The AI is still completely ignoring the prompt restrictions. So it's telling it what to do. Perfect. I've implemented this and system now. Keyword filtering server checks if question contains GIF related questions for sending to AI. Non gift questions get the redirect message without hitting the AI API. That's interesting. I don't even know if I want it to do that. Right. It's going to completely like just ignore the API, which might work. Yeah. Anyway, so let's see, let's see what happens here. Let me test. This was the horse riding. Horse riding question. I'm interested to see what it says. I see the horse rating question is still getting through. Keyword filtering isn't working properly. Okay, great. So no, I don't want a blacklist approach. That's stupid. Don't use a blacklist approach. You have to read this sometimes and like really fight it and say no, that is stupid. Right. Don't use a blacklist approach. We can't possibly blacklist all possible topics. Just teach the prompt to. To immediately redirect to gift related questions that have to do without gifted. Yeah, a blacklist approach is impractical. Why didn't you figure that out?

Why isn't it though?

You know, I don't know. It's definitely smart enough to. To use logic and understand that.

So, so that. Because what does happen? Like earlier today when I was. What is your guidance on when we do start to get frustrated with it? Is it like just walk away? Do you start over like, and I hate to even say frustrated because it's like, what am I not saying in the right way?

From the man's chin, the, the person, the Gina.

It's like, did they not understand that that's text everything.

Whoops. Everything else.

That's the prompt I did.

So I bet you did. But for whatever reason, let's see if it works. Right? But for whatever reason, sometimes you, sometimes you just. It gets lost and then it's trying to follow a pattern of this conversation. And so the fact that I'm a brand new request might just clear it out. Clear new thing. Like, oh yeah, I Can do that easily and it might just nail it. First shot and you'd be like so stupid. And all you had to do is start a new prompt. Right.

Okay, okay.

That would have fixed it.

What, is there a specific AI or is ChatGPT the best? Like, if we're asking it, let's say any sort of math questions at all, any sort of forecasting questions at all. Is chat GPT? Is there like, is it 4o? Is it 4 5? Is it like, which one is the best for honestly, financial questions.

Easy. Math is so easy for all of these AIs, you can use 4O just fine.

Okay.

If you're asking it to create a detailed financial forecast, I'm probably, I'm probably leaning more towards something like Manus. I, I have never. Hey look, my chin is clean. Check that out.

Wow.

Right? It's still working on it, finishing it. But it, it listened, it listened. It did that thing. So unfortunately I've tried to make, I've tried because I do a lot of consulting for startups and even big companies and things like that. And just having been in the investment world and all this stuff, I, I do a lot of financial forecasting and things like that and I have not found that I can give it a very complex financial forecast and then have it manipulated for me. I don't know why, I don't know why I haven't been able to do that. But like I have really, really complex like two billion dollar plans and things we're trying to work on and build and, and it just won't, it won't do it. Right. I can't get it to remove stuff from my chin, so I can't. That's not.

Or what about like, you know, should I sell this property? Give it all the information and 1031 it into this property? Like do you think is that complex or do you think.

Yeah, it can do that. It can do that kind of stuff for sure. Right, so TFV Gen A ipi.

Can we see how that grocery shopping's going?

Yeah, let's go back to that. Thank you for reminding me. We were working on that. That's fun. And we did it in a couple places. Right.

So we did it in Operator.

Yeah.

And I know we did it in Chat GPT.

We did it in Manus. So select delivery service and navigate to ordering platform. Waiting for user. Since this requires your login credentials, need to take you to take over the browser. Oh, take control. That's cool. I've never had Manus ask me to take Control. I like that. So just like Operator, it lets me come in and, and play and do.

You think in theory if you log in once it will be able to log in or is there a lot of like, like logging you out, logging you back in? Do you know?

No, it should. Well, Manus, I'm not sure Operator does save your login. Right. Information. Right. But now here this is having me do it at Walmart. That's what it decided to do and so I could have it settle this up. I'm not going to have it go and buy all those things for me.

Right.

Looks like I could theoretically go and do this now. Okay, let's go back to chat GPT. No, it was operator ChatGPT right here. All right. To continue adding your groceries, please use Takeover to log in or sign up to Instacart with your preferred account. Right. So, so theoretically it could, it could do all of that.

Okay, cool.

So yeah, you would take control, you'd log in and you'd move forward and have it do what you want it to do. But, but yeah, once it's. Once you've logged into Instacart one time or if I provided, you know, the email address and password to get in, then it would just go through and do it.

Oh, okay. Like here's order my groceries next week. Here's my Instacart email or password login. Make it happen.

Exactly.

Okay. And you could like say hey, do it at 8am on Monday morning or whatever. You can also, you could set a time for when it. For it to take place.

Free site. Then it all shows up and then you say my grocery showed up. Where. Where are my recipes?

You know, like Right, right, recipes.

I'm supposed to do with all of this for my week. Right. So yeah, yeah, that's cool. So yeah, both of those options look like they would actually work. So I'm going to exit the takeover and I'm going to say just stop the task. Right. Continue chat. So I want to see what it says now that I gave it that instruct. But stopped before completing the sign in and shopping process as requested. Awesome. How well did Manus complete this task? 5 stars credits.

They want to know.

Okay, yeah, I'll take it. I'll take it. So if you guys use my little invite link, you'll get a. I don't know, I'll get a couple bonuses here, but I'll, I'll throw that in the chat. You're welcome to do that or not do it. I don't really care. But I get new credit. I have so many credits. Like, I have 18,000 credits. I've got a lot. I'm fine here. So. Okay, let's do. So let's, let's go back to the, to the thing over here. I wanted to see if we got this off topic redirect working, but I'm just, I'm glad that it kind of was a little bit of a struggle. Right. So about the sun. Right.

I'm glad it was because I think we would have felt like we were not giving it the correct prompt. Right, ladies? Like, it's actually a good thing.

You have to, you have to struggle through a little bit there. Right. Okay, so it told me about the sun. Not great. Tell me about today's gift. Spelled that wrong. That's okay. Today's gift is a meditation book titled Daily Meditations for Families. Yeah, see, it's not even, it's not even doing this right at all. Okay, I'm going to copy this paste. Tell me about the sun and then tell me about today's gift. Let's do this. Okay, then I'm going to say this. This is the chat transcript. Number one, it answered question about the sun. Number two, gave a random answer for today's gift instead of properly checking the database for today's gift. And so I just keep working. So some of these prompts work better than others. This is a Perplexity API. And I will say that I have not used Perplexity's API more than about three times in apps, and I've used OpenAI's API about 50 times. And, and it works, works really, really well. And I've had more success with it. I'll just say that. I don't know if that's the problem. I could test that by simply saying, hey, let's switch to. The reason I use Perplexity is because Perplexity is able to actually actively go search the Internet and come back with an answer. Right. So anyway, it. Take it for what it is. But that's how you would set it up. You go in, you make sure there's a prompt, you figure out, you know, where the prompts are, you know, and, and just tell the agent to, to make it better. And. And again, it all comes back to, I wanted this to happen, but this happened. Or, hey, could you make this happen? This is the experience that I want. You're just talking, you're just talking to it, making it, making it work. Right. Okay. Yeah. I don't love, you know, the way the shoe is being Displayed also, it doesn't match what the Enigma thing is, but that's okay. All right. Filtering isn't working because the sun question is getting through. Issues. The server restarted. The server restart reverted the old code. Oh, awesome. Good plan. You didn't click save.

So when did, like, when does it technically, do you diagnose with like, hallucinating or whatnot? When do we know it's just like, that happens. We just need to scrap it all up.

You don't. And so you need to verify. You need to, you need to, you know, verify certain things if you're, you know, if you're having it. Let me say this. This is being fixed from two sides. One side is perplexity. OpenAI anthropic. Like, all of these companies are working on getting rid of hallucinations. So the, the amount of hallucinations are already way lower than they used to be. Okay, so that's, that's good. And that's going to continue to happen without you having to do anything.

Okay.

Two, you are, you're responsible for verifying, you know, just the same way if an employee brought you some information, you'd be like, well, how do you know? Right? Prove it. You know, show me. Where did you get this information? What are your sources? Right? The same way you would follow up with an employee, you would follow up maybe with AI and say, hey, you know, can, can you verify that this is accurate?

Right, well, yeah, that's. So that's what I'm saying. Like, let's say we have the chat. We have a chatbot on our website or like Outpost X Amirah. Like, could she put possibly hallucinate? And someone's like, where are the check in instructions? And she's like, I don't know. You know, like, is that a possible that the bot could hallucinate on the. As responding to people?

I had it do that in the beginning and I just worked on the prompt and I finally got it to work. Right? And at this point, I have had a hundred perfect conversations with it. Right. I like, it works. Right? So you just have to test it. Just test it. You know, use your own product right. Before you put it in front of other people.

Sure, sure.

Okay. Okay. There's old debug coding. Yeah, it, it's kind of lost in a weird circle here sometimes. And I'll just say this. Sometimes you just have to rip it out and be like, all right, rip out the chatbot. All code. I want you to search the entire code base. Delete everything related to the chatbot. Now, hey, build me a chatbot. Do this, follow this. Let's make it, let's make it work. Right?

I think you told me that too when we, when we first talked, like you literally just had to start over because it kept messing up. But you know, I hear. This is something I want you to share with Sarah and Colleen. This, we kind of did it at the very beginning, John. But like for instance, we're creating a new offer. Okay. And that offer is, you know, taking from the $3,000 to $12,000. And we're trying, we've been brainstorming amongst ourselves, you know, like what does that new offer look like in order to get us to the 5 million dollar goal in X amount of time? What was the prompt that you told me again? Like ask it one question at a time of like sales and marketing and like how I could do that? Can we go through that exercise?

Yeah, yeah. I mean the, the, the basic is just you ask it to interview you. Right. Interview me one question at a time. So, so here's, here's exactly what I'm doing for that. Right. So we'll do the, we'll do this, we'll go here. Okay. Right. I just started working for this awesome company called thanks for visiting and I was hired to help them. They, they teach a three thousand dollar course on how to better run your Airbnb like a small business. Make sure you're getting, you know, more, more bookings and you're making more money and you're not wasting money and really keeping track of everything and how to stand out and all these wonderful things. But we want to change, we want to improve or increase the cost of, you know, $3,000. We want to take the price up to $12,000. And you know, so I'm trying to figure out, you know, what are some things that we can do to increase that value. Now I do believe that, you know, new customers coming in don't know the difference between the three thousand dollar course, the twelve thousand dollar course. But we do want to, you know, we want to make sure we're adding more value where we can maybe in automations, maybe it's more things about AI, maybe it's some extra content or you know, what, what can we do to make this happen? Okay, so, so that's my initial prompt and, and then I'm going to probably say, you know, interview me, interview me one question at a time about this until we've figured out, until you have all of the information you need to help us put together this new offering. Okay, Boom. I mean, that's it. That's all I'm doing. I'm just telling them, giving it some context. And now let's do the one question at a time, right? Who's your ideal buyer? Right? And then you answer the question, right? It's the first time. Is it a mid level? Is, you know, what is it? Right? Who's, who's the person you're going after first? Right? So let's say you know, anybody with more than one Airbnb, Okay. They're making enough money that they can afford a twelve thousand dollar course. Right?

Right.

What are the specific outcomes or transformations for this? Right. You know, what do you think actually?

What do they think?

So, you know, increased revenue per property, students learn dynamic pricing, better staging, booking strategies. Okay, Sarah, that's like your host.

That's those four, right?

So this is interesting, right? What do graduates of the $3,000 course ask for next? Do they want one or they do they want help hiring a VA or building a team? Do you want more advanced automations? Help launch more properties group masterminds? Right. So take your feedback, the things that you know, because you've been doing this for a long time and help it answer this question. Right? And so yeah, they, they want personal help, but we don't have time to give personal help to everybody. Want to improve the course so that it's worth more and people are asking for less personal help. Maybe that means we need to provide tools for them that can help them, you know, go farther, farther, further. I don't know, further, further without us. Okay, so anyway, got it. That's gold, right? They want handholding, whatever. What tool system templates do you already provide? You know, here's some of the things. Yeah, we provide spreadsheets, all of that.

We provide all of that.

We provide all of this. But most of it is manual. If we launch new automated tools, tools and software and get out of rudimentary spreadsheets, we can probably charge more. Yeah. So that, I mean that alone right there, right? Just, just your software suite that you're now building, you're working on is, is, you know, what could actually do that. So here's your strategic upgrades for the 12K offer. Software driven. Right. AI powered host dashboard listing optimizer with GBT4. Right. Smart SOP builder, AI messaging suite team and VA toolkit, whatever. Right. So then I can just keep going. Do I have developers or tech team? We can build it always. Boom. Okay, awesome. You can do all of this. What kind of access model are you Imagining lifetime course, course, lifetime tools, one course plus one year of software, then renew, right? That's the big thing. I think there should probably be like, hey, for, you know, the system is six thousand dollars a year or it's three thousand dollars a year, you get it for free with a twelve thousand dollar course, right? It's you know, 250amonth or three thousand dollars a year or you know, 300amonth or $3,000 a year if you pay, you know, annually or you can book up front for three years in advance, right? That's, that's $18,000. Boom. You're selling stuff right now, right?

John, you mentioned like you asked us if we were using school or kajabi and then you're showing us replit. Like I'm thinking like what you, it looks like I could make their experience on replit be much more customizable if we just ditched Kajabi. But is that like, is it a big ask of us or would you recommend we just stick with that and with kajabi?

And that's up to you guys. Let's have a little bit of a theoretical discussion here for a minute. And we haven't done that a lot today and normally I do more and we've just kind of been like tool, tool, tool, tool, like diving in. But let's take a step back first let's watch a quick video. It's a two minute video and it's really, really good. And we'll get you thinking about some, some of this stuff in a really, really awesome way. Let's see here. Find the video. Here it is right here. Okay, so we'll go full screen. Whoops. We need to go to Zoom Share and. All right, this is Eric Schmidt, former CEO of Google. This is two and a half minutes. Let's listen to this.

We believe as an industry that in the next one year the vast majority of programmers will be replaced by AI programmers. We also believe that within one year you will have graduate level mathematicians that are at the tippy top of graduate math programs. So that's one year. Okay, what happens in two years? Well, I've just told you about reasoning and I've told you about programming and I told you about math. Programming plus math are the basis of sort of our whole digital world. So the evidence and the claims from the research groups in OpenAI and Anthropic and so forth is that they're now somewhere around 10 or 20% of the code that they're developing in their research programs is being generated by the computer. That's called recursive self improvement is the technical term. So what happens when this thing starts to scale?

Well, a lot.

One way to say this is that within three to five years we'll have what is called general intelligence, intelligence AGI, which can be defined as a system that is as smart as the smartest mathematician, physicist, you know, artist, writer, thinker, politician. I call this by the way, the San Francisco Consensus because everyone who believes this is in San Francisco, it may be the water. What happens when every single one of us has the equivalent of the smartest human on every problem in our pocket? But the reason I want to, I want to make the point here is that in the next year or two this foundation is being locked in and it's not, we're not going to stop.

It.

Gets much more interesting after that because remember, the computers are now doing self improvement. They're learning how to plan and they don't have to listen to us anymore. We call that super intelligence or asi, Artificial super intelligence. And this is the theory that there will be computers that are smarter than the sum of humans. The San Francisco Consensus is this occurs within 6 years just based on scaling. This path is not understood in our society. There's no language for what happens with the arrival of this. That's why it's underhyped. People do not understand what happens when you have intelligence at this level, which is largely free.

Yeah. So let's talk about it. What do you guys, what do you guys think? What's your reaction to that.

Nerve cited? Yeah, it like makes and excitement coupled in the same.

Yeah.

Well, it's like if we, okay, we have access to that so we can create these really cool things, but then will everyone else too. And so that is everything, you know, some nothing. Like there's not one thing that's greater than another because we all have access to like the smartest thing. Like I don't. And then I can't even think, I don't even know what that means.

I still think there's like we have our human experience though.

But.

I mean even today it's like I do feel like I forget how he said it, but like at that six year mark, I, I'm taking it back historically. Like if there, if we have we told people, I don't even know how many years ago that we are sitting on Zoom for six hours talking, typing on our computers, World Wide Web, people would have been like, the computers are going to kill everything. And like I'm just trying to take it back to, like, how far, like, what things have advanced. Like, this wasn't possible.

Yeah.

And you can't fear it, but embrace it. Like, this has changed. It's awesome that we can do this.

So. So here's one of the big differences. The Internet was obviously world changing, right? The computer was world changing, the Industrial revolution, electricity was world changing, right? When that all of a sudden became a thing. But none of those things could think. That's the huge difference here. None of those things could act on their own. Right? And we could have a big, long, you know, debate about what is the word think, Mean. And that's fine, right? But really, let's just, for the purposes of this discussion, say that thinking is pattern recognition and then taking action based on what you recognize in the pattern. Because that's really what we're doing as humans, right? We see something, we know to be trained. When we encounter a new situation, we don't really know what to think. And we have to figure that out. Right? And so I would argue that humans currently are still better at that, right? Figuring out something new. And there's actually some great research out there right now that people are doing to show. Oh, yeah, you know what? We're not quite quieted AGI yet. It doesn't, you know, when. When AI is introduced to something that it's never seen before, then, you know, it. It can't handle it. Right? But that will change. That will change. It will get to a point where, yes, it can handle new situations and it can figure out how to analyze and work through a new situation. Okay, here's the scary thing. AI will replace 500 million jobs by 2030. That is scary, right? The jobs that it's going to replace is most agencies, right? They'll be, you know, we haven't gotten into any video yet. We might take a few minutes to do that. We might not get to it today. But, but, but video can now be created, you know, at Hollywood levels. Things that used to take, you know, $500,000 for a shoot now can be done in 20 minutes with prompts, okay? You got the actors, you've got the wardrobes, you've got the location, you've got, you know, the makeup and, you know, everything. The cameras, the operators, the editors, the music, you know, everything you're talking about, the visual effects, all of that stuff together can now just be done with a single prompt. Okay? So that's video. Do you think accountants can't be replaced by AI? No, absolutely they can be replaced. Do you think lawyers, especially lower level Paralegals, things like that. Lots of busy work, grunt work, research, things like that. What about research for scientists? I saw a post today that said that science, that new discovery based on existing research that's already there, can now be done 1000 times faster than it used to be able to be done. Okay, can you replace workers at McDonald's? Yeah, you take AI plus, you know, AI voice. I would much rather have an AI voice taking my order at McDonald's if I'm going there, right? And so what's going to happen is you're going to have this, you're going to have this further divergence of McDonald's is trying to be cheap and can compete with other fast food. Their goal is to deliver food that is at their quality as fast and cheap as possible. Right. And so when you introduce robotics into the equation, they have AI plus robotics, then you can have a fully autonomous McDonald's that's only operated by machines. Maybe you have one person that can be there to like unstuck something, you know, but, but the order is going to be right, it's going to be fast. It's always going to be the same. It's going to come out quickly and you'll pay with your Apple pay or your voice or whatever it is. You know, someone's going to figure out how to pay with your voice. I'm excited, you know, to think through that problem. But, but anyway, so. And then you're going to have, you know, Ruth's Chris. Ruth's Chris is not going to automate their value. Offer is real people, right? You have a real waiter that walks up to you, someone that really takes care of you, has real human interaction. You have a real chef back there working on it and all this stuff right now. Will Rus Chris use AI? Sure. But it's going to be in the automation of the marketing, in the website, in the accounting, in the, you know, in their back office stuff. Right. But the human experience is going to be their offering. Right? So what you have to think about is, you know, what, what happens though, to all of these people? Customer service service agents gone. I just showed you on 11 labs that you guys could build a customer service agent in 20 minutes. Okay. And then integrated into your, into your whatever, Right. Like I already said, agencies that are helping with, you know, marketing, Facebook ads. Mark Zuckerberg right now is working on something, I'm sure you've seen it, where he said, I'm going to be, you're going to be able to give me your assets. You're Going to give me a budget. You're going to tell me your desired outcome. And AI will build everything. It will create the text, it will create the pictures, it will create the video, it will run the ads, it will test it. If it doesn't do well, it'll throw that stuff away and create new stuff. And it'll do that. And just, it will do everything for you. Like your, your, your setup in Meta will be like nothing. It'll be like a few little fields and then go to work. Right? And so all of the agencies and, you know, marketing people that take care of ads for you are gone. Right? And just on through every single industry. You can't point me to an industry where AI is not going to absolutely eat people's lunch. Now that's scary and exciting and most definitely going to be the largest shift that humanity has ever gone through. It's not even like nothing comes even close to it. So now let's move to the next part of this discussion. And this is a discussion called Surf, Don't Swim. Okay? And I know we're already at time here, but like I said, I'll keep going here.

The owe it to us because now we're going to need to go to ChatGPT 4.5 after this ending.

I know to have some. Yeah. So that's okay. Let's. Let's work through this. Let's work through this. Right? I want you to end on a good.

We want to surf, not swim at our beach house.

We'll teach you how to surf. Let's teach you how to surf. And I'll give you a bunch of examples. And this is a 15 minute discussion. So this is gonna take a minute. Although I am. I have low blood sugar and my little notification thing just told me. So I'm gonna grab some juice, go get a snack. Give me a second. Let me just grab some juice. I'll take me like 40. You know what? I found a Gatorade. I'm good. Does it keep me alive long enough to finish here? Gatorade is terrible. It's basically just made of sugar. But.

But that's what you need right now.

I need fast liquid sugar. I could have eaten something, but that takes time to break down anyway, so surf, don't swim. How many of you have not swam in the ocean? You've all, all swam in the ocean. Okay, good. So you know that it's a very big, powerful place. What about any of you swimmers? Any of you, you know, high school, college, anything anyway? All right, me neither. Just Wondering how much energy, how many calories do you think you burn in 10 minutes of swimming in the ocean?

So many.

So many.

Yeah.

4,000.

It's not quite 4,000. That would be pretty cool. Then you could just eat anything and go swim for 10 minutes. You're done.

It's about a thousand.

It's about 140. About 140 than you think. But 140k calories, you know, 10 minutes of swimming and that's, that's like, you're not like working super hard. You're not like in that pool going back and forth, right? You're just, okay, okay, you're hanging out, you're having fun, you're swimming around, you're playing with your friends game. And so now you don't have to give the exact number, but what do you think the ratio is of the power of a tsunami compared to a single swimmer? Is it a hundred times, thousand times, a million times?

A billion times?

Yeah, pretty close. Four billion is actually, you know, and I know that that's subjective, right? There's going to be different levels of tsunamis and different levels of swimmers. So but Approximately, let's say 4 billion times or a billion times or 10 billion times stronger than a swimmer. Does anybody here think they could win in the fight, you know, versus a tsunami? I don't think I could. I'm a whitewater kayaker. I, I've been a sponsored whitewater kayaker. I can go off huge. I could show you some amazing. Go to my socials and you can watch awesome videos of me going off waterfalls and stuff like that. I have been on the wrong side of water plenty of times and I have never won that battle. Right now, I've also, I also understand, understand water very well, especially in rivers. I understand when I'm looking at a current, I know everything that's happening and I can harness it and I can go through Class 5 rapids without touching my paddle to the water because I understand how to harness that water and work through it. Right. A surfer, a good surfer also knows how to get onto a huge, you know, wave at Jaws and safely navigate through the pipe and, and make it through without getting thrashed. Okay? So the bottom line is surfing. Surfing. A tsunami is where you want to be. It's actually interesting. Tsunamis create really good surfing conditions. You don't want to be swimming during a tsunami and you don't want to be, you know, in town next to a building because you're going to get smashed against it. Right? It's going to come. It's going to be powerful and it's going to hurt you. So I just barely watched that movie. Do you remember that movie, Deep Impact? You ever seen that one? Dripping Conditions. You don't want to be swimming during a tsunami and you don't want to be, you know, in town next to a building because you're going to get smashed against it. Right? It's going to come, it's going to be powerful and it's going to hurt you. Like a huge. Yeah, anyway, it came out like the 90s and anyway, it's a, it's a really good movie, kind of end of the world type movie. Some great actors in it, actors and actresses in it. And there's a massive 100 foot tsunami at one point in there. So that's, that's what came to mind as I said that. So anyway, let me make sure that I'm not getting notifications here and then I'm on do not disturb. You know, that was a Facebook notification. We're just going to close any Facebook browsers. Okay, so, so now let's apply this back to AI right? AI is the, the biggest tsunami. There have been other waves, right? You know, media waves, social media waves, mobile waves, you know, things that, you know, you've been around. This is a tsunami of unprecedented proportions. And if you are not surfing and you are swimming, you are going to get smashed against the rocks. Okay, so now let's apply this to business. Surfing means in this, in this, you know, allegory means that you, any technology that you're around, as it improves, as you're around or built on top or whatever, right? As it improves. If you're surfing, your business gets better. If it gets better, you get better, right? If you are swimming, it gets better and you sink, right? And you know, or you get smashed against the rocks and you go out of business. Okay, so let's, let's do a few real world examples. Let's talk about Netflix. What kind of business is Netflix? What do they do? What's their, what's their secret sauce? What kind of business are they?

Entertainment.

Entertainment that's tailored to you on demand.

Very good. Sarah. You actually kind of nailed it on the head, right? Streaming is absolutely what they do. They also do some production, do some other things. But their, their secret sauce is it's, it's a, they're a content recommendation engine. There's actually, there's actually about, at any given time on Netflix, there's usually about 30,000 pieces of content on there. And have any of you watched more than or even even scrolled past more than a thousand? No. So that means there's a. For. Somehow Netflix knows what tiny piece of their library applies to you. That is their secret sauce. Have you ever gone to, like, an Airbnb and someone else's Netflix account is there, and you turn on. You're like, what is this? Like, get me out of here. Like, I'm uncomfortable. Let me log into my account. Where are my shows? You know, so clearly they're doing a good job at showing a lot of content. Their content suggestion engine is their secret sauce. Did they build the Internet? No, no. They build streaming? No, no. Akamai did. Right. Did they build, you know, the media wave or did they build social? Did they, you know, are they working on, you know, SpaceX and sending up more satellites to make the Internet more available? Are they trying to increase streaming speeds so that they can stream 8K content? No, they're not. They're just working on their product and their content engine, and they're letting other people work on these technologies. And so as. As the Internet gets faster, does Netflix get better? Yeah, they're surfing. Right. As. As Hulu increases or improves their content recommendation, do they get better? Does Netflix get better? No. No. Right. So that would be okay. Hey, that's a competitive, you know, force there. Right. And so they're definitely not surfing on that or swimming on that. That's just kind of a market force. But it happens. Right. But the bottom line is Netflix is actually a pretty dang good surfing company. As the Internet gets better, as it becomes more widely available, as it gets faster, as more content comes out, as Disney makes more content, or whoever makes more content and puts it and makes it available on Netflix, do they. Does their offering get stronger? Yes. So they're surfing quite well. So now let's talk about AT&T. Okay. IS AT&T a surfing or a swimming company?

I would say.

Oh.

I don't know.

I don't even know if I know what AT&T does at this point. Like, they're so.

That's so old.

It feels old, doesn't it? Yeah. Yeah. So let me ask this. Does AT&T pay a lot of money to maintain, for example, their cell towers? Yes. And when they go from 3G to 4G to LTE to 5G to 6G, are they paying billions of dollars to increase that? Yeah, yeah, absolutely they are. Right. And so that is the definition of swimming. Right? They are building their own infrastructure. They're not really. They're building the infrastructure. They are the infrastructure. In fact, because I have AT&T as my provider. I don't know if I'm surfing on it, but I'm definitely, you know, using it. Right. I'm using that technology. And so, and so therefore I'm, you know, I. As they get better, my life gets better. Right. Because I'm a customer. Right. So technically, customers are surfing their, you know, their provider's technology. So. Okay, so that's great. That's a couple examples. Let's try another. Let's try another example. Oh, actually, sorry, one more point about at&t. This is important. And again, this is the AI free portion of the course, right. Where you're thinking about AI and you don't get to use AI to think. Yeah. You have to go back to using your brain like we did back in the, in the 90s anyway. So what, is there a technology out there that, as it gets better, potentially could make AT&T obsolete?

Yes.

What's an example? There might be several, but what's one that comes to mind?

Well, I was just thinking, like, if everything was on, like you didn't have any mobile devices and there was just Internet connect, but they, they provide Internet connectivity. But I guess if WI fi became free for everybody and then no one needed, and you just did everything via WI Fi and you use WhatsApp and you wouldn't need a telephone.

Can you think of any, Anybody, any big eccentric rich billionaires that might just make the Internet free for the world?

Elon?

Yeah, yeah, Space, right.

Like, if I didn't have to pay for my WI Fi or yeah, I would.

Or my phone number or maybe, or maybe he just says Internet is now $10 per month for everyone on the planet.

Then yeah, I just do Internet. I wouldn't need cell calling.

Yeah, totally.

What I currently pay.

All of a sudden, all of a sudden AT T is worthless. Right? Connectivity is what they provide. That is their business.

Quickly, worthless.

Oh, yeah. And not only AT&T, so is T mobile, and so is, you know, all the telecoms in every single country in the whole world. Like, it would be crazy. Elon has a lot of power with something like that. It's kind of scary, right? Whether you like him or dislike him. Right. It's still scary, right? And so, yeah, consider that as you just move forward. But, and you know, you can't change what Elon's going to do. You just need to be aware. Okay, Do I know what's going on? Do I know that this, this is. This is there. Okay. Let's look at another quick example of surfing. Okay. Welcome back to my Facebook. And I'm going to introduce you to my mom again.

Colleen's everywhere.

Colleen. She's all over the place. Yeah, she's awesome. Okay, so my mom, this is her Facebook page. She posts lots of videos about, you know, animals or whatever. And she just went to Africa. 9.1000 likes. Look at that. That's crazy. Look at this fat, pregnant hippo. That is a big hippo.

That's 39, 000 likes.

Yeah, that's a lot. That's a lot. She's getting some good ones there, right? Cute little, little line, right?

These photos are awesome.

Yeah, she's a photographer. She's a photographer, so she's. She's really awesome. She does a great job here. O. Two years ago, had probably 400 followers. You know, friends, whatever. Just people like family and friends. You know, there's Colleen, right? And so she. I said, mom, you should just start posting this and post regularly every day to Facebook. And she's like, okay. And she started doing it, and her video content just caught on, right? And people started following her. Hey, there's me at the piano. And, you know, different things, right? So look at this now, two years later, 641,000 followers. So how much money do you think she has invested in Facebook's content algorithms.

Financially? Nothing.

Zero, Right? How much has she invested in improving the Internet? How much has she invested in SpaceX and all these different waves and things that she has riding? Okay. She is surfing all over the place, and her only costs are related to whatever she wants to do to go get video. And by the way, she's also surfing on this device, right? As the iPhone gets better and the camera gets better, most of her videos are taken just on her phone, right? This is just taken on her iPhone.

I love that she follows no one. That's like the bossest move ever.

I know. Isn't that?

So that is like. She is undefeated.

That is so funny. And it's. It's true, though. She's. She's the least. Like, if you met her, you'd be like, this is not an influencer. Like, everything you think about influencers, my mom is not that. Right. And not that influencers are bad, right? Nothing. Nothing against influencers, but she is. She is just like the least technical. Like, how do I do this? Like, somebody. Somebody, you know, downloaded a bunch of her videos and started a fake profile. And she was like, I have to call the police. I was like, no, people just do this, Mom. Like, you can click a button and report it to Facebook, and they might take it down, and they might not. But don't worry about it. You're gonna be okay. You know, so anyway, and I'm so mad because, like, I try hard at social. Like, I've told you guys, I post multiple times a day to every platform, and she has way more followers than I do. I have, like, 50,000. And, you know, cool. I'm. That's. That's awesome. That puts me in a good little tier. But she's more than 10x that, and she grows like crazy. She makes anywhere from 10,000 to $20,000 a month from Facebook alone just from posting. She just gets all this stuff. So she is surfing. The big thing here is that she is making a ton of money without risk.