Cohort 7 - Day 2 - 4-22-2025

Forget. All right, so welcome to day two. Hope you guys had a fun time thinking about all the things you learned yesterday. I had many of you text me or call me and, or email me or whatever and I appreciate those comments and, and I went back and forth with a couple of you on, on just, you know, little things. But, you know, today we're going to dive in deeper and, but I want to kick off the day actually with kind of more of this theoretical discussion around AI. And I want to kick it off by watching a quick video that is, you know, poor quality. I just literally screen recorded on my phone. But we'll just share it and well, you'll be able to see it. Let me, let me pull it up here. Okay. I think I need to start by. Right there and let's go. Hold on, let's go share and hopefully the audio comes through. All right. It should do pretty well. Okay, let's go like this and play. 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 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 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. It's 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. All right, okay. What do you guys think? Open discussion time, based on everything you saw yesterday and learned yesterday. And then, you know, I'm listening to the former CEO of Google. Is my mic working? Let me check that first. Yeah, it is, Brian. One of the things I have a legitimate concern, I think legitimate is censorship. We have, I think now it's kind of wild, wild west and you know, we have a lot of access to stuff. But I was even digging into Chat GBT the other day and I think eventually, you know, follow the money, right? And, and people with interest that are going to be undermined with the general knowledge are going to want to silo that information. So for example, LLMs have found some cures to some cancers, but the pharmaceuticals that sell the drugs to those are not going to want that, you know, being solved. So consulting with, with Chat gbt that basically what's happening right now is that a lot of that health information stuff is being siloed away and they're putting the, the data that's used to find those results behind paywalls, you know, really expensive paywalls. And so that is kind of a concern I have is I think overall we're going to be a lot better off. But I think depending on which way this goes, it could not be as nice as it should be. Maybe. Definitely. Other thoughts. I have a quick thought on this. It's interesting because if you look at China, right, we'll do a lot of R and D and then we'll ship the product up to China to be built. Then once it's proven out, China will just make it there in their same factories and sell it for cheaper, right? That's just a known fact. Whereas I look at it, all the software companies out there will this. My belief is one day it's not today, but that many people will be able to easily create the bill.coms of the world, right? And just all these different softwares that are out there pretty quickly and just compete a lot easier. Right. Like before, it Took us for our software company, 10 years to build. Right. And we're still building. But, but you know, what does the rapid development look like today? And how much competition does that introduce to all the software companies that are out there? Right. Because if any, just Joe Schmo that doesn't know how to code can build a software like say, a point of sale or whatever it may be within a week or two or a few days, it's interesting how that will change the software landscape and how everything runs right. So anyways, that's, it's an interesting thought to see where this will be in a few years and what the capabilities will be because that will disrupt the software industry like no other if it gets there. So. Agreed on all points. Thoughts, Guy? Go ahead, Travis. I have a quick thought, and this is just from, from clicking around yesterday, I'm not quite worried that all of our programmers are going to lose their jobs immediately, specifically because of the general knowledge base that they're able to develop over years of study and practice. Perfect example. Me and Replit, both of us were bashing our heads against this problem where my API wasn't working, just over and over and over again. And I used almost all my credits on this one thing. Yeah. And then finally I slept on it and said, well, I wonder if, you know, and I started asking questions outside of the Replit environment. So for someone like me, who's not a natural coder, I couldn't just open up and see, oh, I see what they're doing and then, and then know to ask the question I had to go to. I mean, ultimately I did end up going to Chat GPT, which did help me. So there's the AI there, still there. But if I had the general knowledge, like what Trevor has, I could have opened up the code, looked at it, seen something and said, I wonder if that's the right HTML that they're pulling, for example, because that's what part of the issue was. Yep. So we're still a few years out, I think, from that. But it is, I do agree that like you gotta, you gotta know these skills, whether it's for better or for worse, I think we can go back and forth on, but I do think we need to know the skills for sure. So, guys, you're all bringing up very valid points and I'll. I want to hear more. But just to comment for me on this, I, I think that one of the things that's hard to understand, one of the things that stands out that Eric Schmidt talks about here is People don't even really understand how to understand this. They don't even understand how to talk about it. There's no language around it. That's why he says it's under hyped. He says if you really get it or understood what I was seeing and what's really happening that isn't released to us yet, you'd be terrified. Right? I mean the reality is, and I believe this projection there will be somewhere between 300 to 500 million jobs lost globally by the year 2030 as a result of AI, right? That's a big portion of the world. That's like 12 and a half percent of the population or no, not 12 and a half percent, sorry, that's bad math. But it's like 8% of the population. It's a huge percent of the population. Imagine unemployment going to 8% all of a sudden like in the whole country. That would be crazy, right? I mean think about how bad things are when there's 8% unemployment, right? That hasn't happened for decades, right? Take Covid out of it, right? No, that doesn't count. It's, it's a real problem. It's a real problem now. It is an opportunity, of course. Right out of every problem comes opportunities. Right out of COVID many opportunities came out of every problem. There's opportunities out of a hurricane, there's opportunities to go in and build new roofs and help people and serve and you know, you can, you can find opportunities within problems and. But the fact that you guys are sitting on here means you're ahead of the game big time. But the reality is, um, where the, the, the speed, how many people here follow even one AI related social page on like Instagram, Twitter, LinkedIn, wherever. Okay, a few of you do I recommend that you go out and follow 10, all of you, right? Just throw it into your feed and just watch the types of things that pop up every single day. And I'm happy to, you know, suggest if you all show you a few, a few accounts later if you have some time or maybe I'll send them in a follow up email. But, but the bottom line is guys, it's crazy. Someone comes out and says, hey, we have this new model. Hey, we just hit that 100 million context. Hey, we just hit that billion context. Hey, we can now make movies. I saw the new Google, what's it called? I forget what it's called. It's not Sora, it's their new model, their new video model and some of the videos that it's making and it is unreal quality, like unbelievable quality. You can make full length feature films now with AI and no one would even know the difference, right? Other than the fact that you wouldn't recognize anybody and all the actors would be fake. But three days later China is going to announce, hey, we have something that beats that thing that Google just announced. And then four days later OpenAI comes out and says this and then anthropic and then some new company we've never heard of comes out and drops a model. And the speed at which it is improving is accelerating, right? Which means we're going like this, right? It's getting steeper and steeper and steeper in terms of the acceleration and. Or in other words the gamma, right? The acceleration of the acceleration is increasing. The gamma is increasing at a rate that is almost impossible to understand as a human. I don't even fully comprehend it. It is very, very difficult. Despite the fact that I'm very tuned in, it is very difficult for me to project what will be possible in 2030. I have no idea. I think that anything I can imagine is probably well under what will actually be able to happen. And one of you heard Eric say in that video is, you know, recursive improvement, right where it, it improves itself. That's where you start to get really, really crazy is when the AI that never sleeps, eats or cares about anything other than, you know, processing code can spin up, you know, a thousand agents, a million agents, whatever it is, and improve itself. Just like we're seeing with replit, with this chain of thought reasoning and multi agent systems. Devin, go ahead. I was watching a podcast with, with Elon. I can't remember who it was with whom exactly, but he was basically talking about how the, the challenge of the future is going to be in like fulfillment, like human fulfillment when you have robots doing everything for you. Like how do we find fulfillment? And I think that that's kind of the if just based off of what we were doing yesterday. I probably did, oh man, at least like $80,000 of dev work just by myself in a project that I've been waiting for. We're building out an app, it's taking for freaking ever. And I can't get the develop like I'm in such a complicated space that my devs don't really understand the type of work that we do. And so like what I've been able to build out now, will it be perfect? Absolutely not. But will it be the most, the most in depth rendering that they've ever been able to build From. Yeah, yeah, absolutely. So, and, and honestly. And all the code is, is usable. It's not just like a mockup on right. You can actually try it out. Yeah. And, and, and we'll also like the parts that work. They can just grab and stick right into our app and it's ready to go versus having to like then go code it based off of a bunch of PDFs. Anyways, all I'm getting at is efficiency gains are massive. But I think that human beings are going to throw the currency of relationships as a commodity is going to be the, the way that we transact value once, once we have automation for everything, it's going to be relationships. And it's interesting we talk about how in society we've lost a lot of our soft skills or there's going to be a resurgence of the soft skill. The philosophs, the thinkers, the contemplators, the relationship builders, the inner peace finders. That's going to be the. My prediction, you know, whether or not right is that's going to be where we, where we find our value is in our collective human capacity to experience, process, discuss. Because you won't, you won't need people to be grinding away at the keyboard anymore. Isn't that cool? Isn't that cool, Devin? Like yeah, that's what that means for humanity. Right? We've talked about some doom and gloom. What about the really beautiful side of it that all of a sudden we get to focus more on each other and we don't have to focus so much on figuring out that. Oh my gosh, there was know we, we were missing a, you know, a bracket here. We get those two hours of, you know, banging your head against the wall for that code. I know Travis, you experienced some of that yesterday. We're going to work on that today. We're going to work on how to, how to make replit get out of loops because it's, it happens. AIs get in loops and you got to know how to get out of them. But, but yeah, I mean what happens, you know, and, and how empowering is it? You know, Devin, Evan, others of you. I don't know a lot of you very well. Right. But Eric, I know you're an entrepreneur. Jacob, you're in a highly entrepreneurial thing, right? You guys are all entrepreneurs because you're just sitting here, right? How empowering is it to know, no offense to Trevor that we don't have to call up Trevor to see our idea come to life? For me, it's crazy empowering. I Am just like a kid in a candy shop where I can be like, I have an idea. I wonder. I wonder what it would look like 25 minutes later. I've got something that I can click through and look at and I could show somebody, even if it's not perfect, but I can immediately say, hey, guys, I have this idea. What do you think? Share link on LinkedIn. Get 20 people to comment on it. Oh, this is good. This is bad. I don't like this. I don't think this is going to work. Have you seen this? Oh, someone else is already doing this. Better. Okay, great. Boom. I've got my validation. I've got some feedback. I've spent a grand total of like 90 cents and I can now validate an idea and decide whether or not I want to go put more money, more time, more resources. Do I want to go raise some money? Do I even have to raise money anymore? I don't know. I don't think so. Right. I mean, there's definitely situations reacting. You want to go, you know, fund some crazy cancer research or build a rocket? Yeah, you might have to raise some money to do that because you might not have enough sitting in your bank account. But. But for software and digital products that a massive amount of the world is, you know, revolves around, you can make almost anything. You can make an internal tool. Gavin, in a meeting while you're sitting there with, with your employees and like, hey, I wonder what the calculation is on this. Oh, hey, let's do this right. We have an AI only fans account to make money. Yeah, exactly. There are plenty of only fans accounts runs by AI right now. I hope you guys know that there are tons of accounts, social accounts, only fans accounts, YouTube accounts, their YouTube channels that are 100 AI created videos, and they are some of the most viewed videos out there. Why? Because AI knows how to make it viral. They know how to make it, you know, catchy and jumpy and they put the right music on and they. And it's AI voices over it and you can't even tell, right? It's 100% AI generated content out there. And there's definitely downsides to that, obviously. Downsides to that. There's. There's the dead Internet theory, which some of you are probably familiar with, which is what happens when I can just produce, you know, a million times more than humans can. So everything you're reading is. Is AI generated. Like, where's, who cares? Like, why do you even want to read that content? I want to hear from other people, like Devin was talking about, I want to hear from the spiritual leaders. I want to hear from the people that can connect with my soul, right? I want to, I want to, I want to hear music from real people that, you know, express real human emotions and things like that. But, but I could show you music that is just absolute, absolutely beautiful. That's 100 AI generated and you would add it to your playlist because it's good music, right? So anyway, guys, AI is growing tremendously now. I'll just say this, you know, when you're wondering like, what are humans going to do? Well, this question has been asked before, right? You know, when the calculator came out, math teachers were so mad about it because they're like, everyone's going to forget how to use their brain and we're all going to rot away and die with just a little calculator, right? And then the computer came out. What about go back even further to the industrial revolution when all of a sudden factories and machines could do the jobs that humans were doing? Well, what are humans going to do now? We're all just going to sit around now that machines have taken all our jobs? Well, we figured it out, we adapted, right? Humanity is very, very powerful and resilient. And quite frankly, without getting too religious, I personally believe that humans are meant to do greater things than just sit and, you know, do mindless, or maybe not mindless, but repetitive activities over and over and over again. I believe we're meant to be creative and to create and solve problems and to do big things. And so I, I believe that humanity will find out a way, or find a way to do incredible things and maybe enter even a new, you know, renaissance of, of creativity and art and expression and things like that, that, that are real. At the same time, there will be societal problems when you have people, especially a lot of the, you know, lower class people that might do a lot of things AI can take really quickly, like fast food, a lot of fast food restaurants. Add a little bit of robotics in there and fast food employees are gone. So what are those people going to do? We're going to have to teach them, right? We're going to have to figure out, we're going to build systems that help them, you know, learn how to work in this. And that's really why I created the General AI Proficiency Institute that I created, you know, that with, you know, in three days. Let me share it a little bit. I want to show you guys around the product that I built because it is pretty awesome. Let's go, Jenny. API. So I built this entire thing, guys, with the exact same process that you guys are, are learning right now inside of Replit. Okay. It's a very professional landing page. This actually was the first draft. This, this header, I've spruced it up a little bit, but barely. Like I've moved a couple buttons and, and I added this kind of nice background that's got some dots and gray. Anyway, I kind of, it was just a plain gray background. So I barely, I've, I've changed it 5%. Right? It's almost exactly what it was now. I've changed a lot of things, right. I have it focused more on Enterprise solutions now and things like that. Right. But, and I have gone through and I've had it update a lot of these sections to be just a little bit more spot on. This image was created by AI. I uploaded it using AI, you know, using, using the system, right. And I had, so I've gone through and I've really thought through, okay, what does this do? Like, I had to update a lot of things. This number right here is driven by my database. I have a price for my AI proficiency certificate in my database of $39. If I want to change this price, I just go change my database and then everywhere in the entire app that references this price automatically updates. Right? Amazing. That's a database driven. That's the correct way to build something like this so that I don't accidentally have the wrong information. I say 59 in one spot, 39 in another spot, 42 in another spot. That's bad. Okay. And you know, corporate training, workforce assessment, you know, it kind of came up with this, I like this, this kind of, you know, outline type thing, this green outline and some other color, accent color outline, things like that. And so I, I, I, I saw it do it actually in one spot. It only did it once on the homepage. And I said, okay, but that's going to kind of be our theme throughout, you know, throughout the entire thing, right? So if you bounce over to enterprise solutions, you'll see that I followed. I said, hey, use that same formatting with some of these colors and these accents like that. And see, like, here's a little problem. I need to fix this problem. I can't see the most popular, right? And I don't even know why that's, that's there. I didn't even realize that's there. So it's not perfect yet. Right. And I need to go through and still do some things. It looks like this entire thing is actually Just higher a little bit. I don't know if that was an on purpose design that it decided like it wanted to stick out a little bit more or if it's just bad code. So it did that on my page too. It's the, the most popular one's bigger. Okay, there you go. So then, then that was a design decision that it made on purpose. And for me, I look at it and I'm like, I, I, I don't love that. Like kind of goes up. I don't know, maybe it's, maybe I'm not a designer. By the way, guys, I have to say that I'm not a designer. I'm not a code. It's like, this is not what I do. I'm an entrepreneur and I'm, and I'm. Here's the most important thing that all of you guys are, that is exactly the same as me. We are all consumers and that is how I design products and how I designed this entire thing is I said I am a consumer. So as I go through this, do I see anything that like, kicks off a little, you know, like alarm, like, oh, this is wrong. Or I don't like this, right? The goal of getting as fast as possible to a launchable product or a launchable site or whatever, right? Evan, if you want to launch Ohana Tours and get that up and running, right, what you're going to need to do and what we're going to work on a lot today, guys, is how to go through and make sure you're not dying by the death of a thousand paper cuts, okay? Because you're going to have some little mistakes there. But if you have two or three mistakes on your website, no big deal. I know of one that I haven't fixed and I keep meaning to do it, that this, this image for some reason stopped showing up. I don't know why, but you know, it's not showing up. So I need to fix that. I would call that a tiny paper cut. It's not a big one. It's not a big one, but it's there. It's there, right? And, and you know that other little, you know, on the enterprise solutions, this little, this little thing that's a paper cut. I don't love this one. I actually really don't like that one. I need to, I need to fix that. So maybe I'll fix that today while you guys are watching so you can see how I, how I work and debug stuff. But, but what you have to do is you have to act as the QA guy. Right. Why am I signed in? I didn't sign in. Maybe. Okay. Anyway, that's fine. I think I was signed in and it just didn't show. I was signed in on the main page. So there's another little paper cut, but not a big one because you wouldn't know that unless you were actually signed in. But you have to be able to click on this button. Okay, it works. Fantastic. This button. Okay, it works. This button. Okay, it works. This button. Okay, it works. Right, about us. This works. Is there anything weird on here? Nope. Looks good right now. Obviously you want to read through it and see and make sure. But, you know, quick look, quick glance, it's fine. Does that work? Great blog. Hey, this works. If I click here. Okay, this works. It looks good. If I click share. Yeah, that pops up and it works. It creates a nice share modal. Fantastic. If I click x.com does that work? Yep. Awesome. Right? You have to go through and got to click everything. You got to make sure that as a customer is coming through here through your site and saying, am I willing to, you know, put my credit card into this site? Right. Does anything set off an alarm? Right now this is, again, if you're doing something, you know, external facing, if you're like dev and you're building something internal, maybe, you know, maybe that doesn't matter quite as much. It does matter when you get to your final product, right? But. But if you're doing internal development, you need to say, this is what it needs to look like and you want to get it 90% right, and then let the developers go and build it, that's fine. Right? You can do that too. But everything's got to work. You got to read through this, right? If you, if you have AI write a privacy policy for you, which, by the way, it can do very proficiently, right? You got to make sure that it. It sounds right and it looks right. And there's no, like, you know, hey, the, you know, brackets, company brackets, you know, like fill in company name here. You know, you have to make sure that the AI actually finished it and put in all the details and made it actually work. Okay, this privacy policy is actually pretty good. What happens if somebody emails this privacynaipi.org well, is that set up. Did you set up email? Right. All of these little things, all of these little things matter when launching a business if you want to pay attention to those little details. And if you don't have the detail or you don't have this setup. Right. Or you don't know how to do it. Well, it's fine, just delete it, you know, or change it to something else. But what I did is I made everything on this site feel so professional, like an institute would be, right? An institute. A dot org. Like I made the whole, you know, image mirage, whatever you want to call it, right? Look like it was just super well put together and that it was put together by a huge team of very competent people. That's what AI was able to do. And I did the whole thing by myself. Right Now I pay attention to details. I really do. And so I looked at these things and I know to look for these things. But again, all I did was approach it as a customer. It's a huge thing that I talk about in my guides of how to do this. When you go through and you start reading all these different, where are they? All these different guides, all these overviews, you're going to find me. Talk about go through it like a customer. Go through it like a customer, right? Because that's where you're going to figure things out, right? Like if I try to click on this, it doesn't work. Oh, and so somebody could learn. Okay, well, this is just kind of a demo thing, right? This is obviously vague. This timer isn't doing anything. Okay, well, can I click on this? Yeah. Does that work? Like, I had one person take this test, you know, after I, after I revealed my, you know, after three days, I was like, okay, here's what I built, right? And I put it out on LinkedIn and I all of a sudden had like 25 QA people. Everyone is saying, hey, this thing doesn't work. And hey, this doesn't work. And it wasn't actually a lot, but, but I had one, one guy go in and be like, hey, I took the, the AI Mastery Predictor Test and I was really kind of confused by some of the questions that it asked me. Like, I didn't understand why you were asking me this kind of like kind of an IQ style question, like a puzzle, right? What does that have to do with AI? And I was like, hey, that's a great point. What if I add a little modal, right, A little dialogue, something that pops up before it that explains why we're going to do this, what kind of questions you're going to get, right? So first you're going to get some cognitive aptitude questions and you're going to get some AI and knowledge understanding. You get some prompt engineering questions, right? Start the test. Does it work when I click through, all the way through this entire thing, right? What happens at the very end, right? Does it actually display my score? Does it, you know, what is the, what is the process? Now I'm going to talk a little bit more about this tomorrow because we're going to talk about sales and marketing and how to get customers very quickly and how to get them to convert on your site, things like that. And by the way, using AI to drive the whole process. But what we, what we talked about, and by we, I mean me and my AI team talked about in my project in ChatGPT was figuring out how to get people to get involved and get emotionally attached to what they're doing. And so they take an IQ style test and it's, it's, it keeps signing me in. I don't know why it is signing me. And even though I'm signed out, it's not supposed to show my score right here when I'm signed out. That means my sign out isn't working properly, guys. So I got to fix something there. But, but when you do it, it's supposed to just say, hey, enter your email to see your score, right? Okay, perfect, right? That might be a little annoying to some people. Whatever. I don't care. A lot of IQ tests, if you just go and you're like, you know, free IQ test, right? And you go take the test, it will take you through the entire test, right? Get started and, you know, whatever. And then when you're done, it will say, okay, in order to see your test, you just have to sign up for our monthly, you know, whatever. It's 7.99amonth. And I'm just like, I hate that, right? I don't want, I don't want to do that, right? So I decided, hey, you know what, I'll just ask for their email. We'll email them the results and all of a sudden they'll have an account, right? And then as soon as they get that, that email, and also right here, it will take them right to their dashboard, which is where they see all of their, you know, all their different scores, everything they've taken. There's my 33 out of 100. That wasn't supposed to happen, but I need to double check on my sign out now. Or maybe I need to push an update. I've got this dashboard that's really nice. I can go to certifications, I can learn about that. I can take a more official test. And then, guys, I built an entire admin system for me so that I could go in and I could easily look up users, look up their emails, look at the roles. I could easily update. I can come in, I can edit my user here and I can change myself to a free or a paid user, an admin or whatever it is, I have the ability to do that. And if I need more features on that and say, hey, for admin users, give them this new ability, hey, for paid users, give them this new ability, whatever. And the system can build it out and I have to test it and work through it and understand what it's doing. But I can adjust course content, my live courses, the live course that you guys are in right now, I can change the price right here. When I change the price right here, it changes it in the database. And then when you click on live courses and you go to check out, the price is updated and it works, right? I. But I had to test all of this, right? I can add new schedules. I can look at all of the participants that are in this, right? I can see how many participants are in which class. I can see course content. I can add new content, which I haven't done yet, right? Anyway, test content. I can edit the test or configure the test. And it's pretty cool, like, pretty cool settings that it kind of thought of. Now here's an interesting thing. I bet you that half of these buttons don't work inside of my admin, but it doesn't matter because I'm the only one that sees that that was not a prerequisite for launching, right? At some point I'll go in and say, hey, let's make this button work. Hey, this looks like a really cool thing, right? I want to be able to change this. Hey, I can't change this. How do I change? Require purchase or how do I change? You know, Hey, I want to be able to allow retakes on this or how do I, how do I do that? Right now, the few things that do work, I actually can, you know, I can adjust. I can say, hey, you know what I want, I want the test to be balanced, a little bit different. You know, I want it to be like this. Let's get it back to 100% and put it more here. And I don't really care about ethics, right? That's, that's. I don't, I don't, I don't need to test that. Or maybe, hey, I want a lot of AI technical skills, right? And I can, I can build different tests actually with this topic weights. So I have hundreds and hundreds of questions made and they're all tagged by topic, right? AI thinking and cognitive, attitude, ethics, technical skills, all these different things. And so I can then design a test or a new test and I can just change it all around and say, hey, we're going to focus a lot on whatever and it will automatically balance it out and launch a new test for me, right? So I really thought through how can I scale these things? How can I build systems with relet that allow me to create content? Now when I create new content, I'm not going to save those things there. But when I come in here and I create a new course, for example, I have AI in business, AI in sales, AI in local government, AI for, for human resources. I can preview it, I can see the price, I can continue the course. There's quizzes that you can take and final tests and there's, you know, each module comes in here and I can work, I can't take the module assessment until I complete all the lessons first. Like I went through and I really, I built all of this guys in that one weekend. All of this worked in that one weekend. But now that I've built the system and the, and I thought through and I worked with AI to think through the structure and the type of content, right? And by the way, who do you think wrote all the content for all of my courses? I did. Right. And now I can just go into replit and I can say I need a new course for AI in, in marketing. I don't have that course yet. And I, and then I say I want you to follow the same format with the quizzes and module assessments and everything. I want you to make sure you have this much content in each course and just tell it to get to work. And it will take like 40 minutes, right? It takes a long time to write huge amounts of content, hours of content and boom. I now have a new. Now these are self guided courses. These aren't like the courses that I would teach live and things like that. But the content is good, right? The content really is good. When I go through and I learn about it, right? It gives you some key learning points and then objectives and I go through and read through the different stuff that's going on and, and it's actually, it's all right. And I've done, I've spot checked probably about 20 different spots in each one. I'll take it, I'll copy the text and I'll drop it into ChatGPT and I'll say please evaluate this for accuracy and it'll come back and say, yep, it's all right, right? And so you gotta check, you gotta make sure that it's not just making stuff up for you. But this is the power of AI, right? I mean, look at how much work this is, guys. It's a crazy amount of work to build this, to build the system, the software, right? Like Devin said, this is much more than $80,000. I bet this is easily in the range of probably $250,000 to build just the, just the system, right? And then to build the admin probably is another like, oh yeah, we can add that on for another 60k. And then to build all the content, right? You're paying a totally different type of person that builds, you know, educational content with learning outcomes and all these things. And maybe you're paying another 20 or $30,000 for content, maybe $5,000 per course or something like that. And then, you know, updates. And then I added a whole new blog feature just, just recently that I said, okay, I want to build this out so it looks really nice and I want it to be really. The purpose of the blog is SEO, right? I want, I want to be able to show up higher in search results. So we got to use keywords and things like that and categories, make sure it's all tagged really nice with structured data in the website so that it, so that it does serve as a great SEO. Now that I have it built, I just finished this. It was nine days ago, was the last time I worked on this specific piece. But all I have to do now is go in and say, okay, you built a general like, why are we, why do we exist right now? Let's build. Now let's build blogs or Write blogs for 20 other topics, right? And let's do one a day, right? Release them one in a day and I'll set a schedule so that it goes live on whatever specific day. And by the way, how am I going to schedule that out? I'm going to tell Replit to build a system for me that can build it out, right? So anyway, guys, I wanted to just take a minute this morning to show you all of these things that are inside of my actual product that I've built that is actually getting people to come in and purchase products and, and is generating revenue, right? And okay, so I got a couple questions here, Eric saying what, you know, which LLM is making the content for the course content? Yeah, it's, it's. So I, I did all of the. I'm going to tell Replit to build a system for me, that can build it out, right? So. So anyway, guys, I wanted to just take a minute this morning to show you all of these, you know, all of these things that are inside of my actual product that I built that is actually getting people to come in and purchase products and, and is generating revenue, right? And I did all of the instructional design piece inside of ChatGPT. Like, how should the courses work? What should they try to drive for what things? You know, what is the style that we should do? Right? And I think I built. No, I didn't. I built all the questions. I'll talk about the questions in a second. For the tests, I anthropic did it, right? So inside of replit, right? Replit is, you know, when you're talking to the agent, you're actually Talking to this LLM anthropic.com and you're talking to. Okay, so I got a couple questions here, Eric. Saying what? You know, which LLM is making the content for the course? Content? Yeah, it's, it's that. So I, I did all of the, I did all of the instructional design piece inside of Chat GPT. Like, how should the courses work? What should they try to drive for what? Claude? 3.7 Sonnet, which is a highly, highly capable. It's very good at coding. It's a highly capable LLM, but it can do everything else too, right? It can do high, you know, high level research and all kinds of great crazy stuff. It can access the Internet, so it does all that stuff, right? So I'm literally inside of replit. I don't even have to upload the content, Eric. That's the big thing, right? I don't have to up, you know, copy from one platform to another and say, here's the content, build it. I just say to the things, you know, what is the style that we should do, right? And I think I built. No, I didn't. I built all the questions. I'll talk about the questions in a second. For the tests, I anthropic did it, right? So inside of replit, right? Replit is, you know, when you're talking to the agent, you're actually talking agent, make new content and it's done, right? It just starts going. And then. Yeah, Travis. And you can even write your blog with AI too. Absolutely. You think I wrote this blog? No way, man. Right. Now, what I did do and what I do, what I do often is I will have the blog based off of. I'll do a lot of things based off of me, off of how I write off of how I think, right? And so how do I teach it how to do that? How do I, how do I teach an AI to write? Like talking to this LLM anthropic.com and you're talking to Claude 3.7 Sonnet, which is a highly, highly capable, it's very good at coding. It's a highly capable LLM, but it can do everything else too, right? It can do high, you know, high level research and all kinds of crazy stuff. It can access the Internet. So it does all that stuff for me. Can anybody in their mind, you know, or like speak up if you can think of a way? How do you make it think like you? So I, I said like I took 400 pages of LinkedIn posts that I've created the last couple years and then put that into a custom GPT and then have it analyze it and then that's my tone of voice. Perfect. Amazing. It's exactly what I did. Literally exactly what I did. I took every single thing I've ever written, which is a lot more. I'm a writer, I do a lot. I write a lot of articles. I've had a lot anyway, I like to write and so I took a ton of that stuff and I went into jet. So I'm literally inside of replit. I don't even have to upload the content. Eric, that's the big GPT and what he's talking about here, guys. And this is one of the things we're going to go over today. So I'll just go over it right now thing, right? I don't have to up, you know, copy from one platform to another and say, here's the content, build it. I just say to the agent, make new content and it's done, right? It just starts. But you can click explore GPTs and there's a lot of custom GPTs by a lot of different companies, outgoing. And then. Yeah, Travis. And you can even write your blog with AI too. Absolutely. There Canva Dolly. You know, this is your, you know, whatever, that's an OpenAI thing. But let's go to some more external ones here so we can find something Expedia. Here you go. Right? So you can, you know, Expedia has created their own custom GPT that will help you plan your trips better, right? And there's a lot of different things, right? You can coding experts and co pilots and all kinds of fun stuff, but right up here in the corner you have. You think I wrote this blog? No way, man. Right now what I did do. And what I do, what I do often is I will have the blog based off my GPTs that, you know, you can create your own. And it's super easy. Like any of you can do this in a minute, right? Just open it up, create your own GPT, name it, call it, you know, golf master Evan, right? And then you tell it what you want it to do, right? Teach me how to golf better and above. I'll do a lot of things based off of me, off of how I write, off of how I think, right? And so then you can give it instructions. Now, prompting, creating a good prompt is important and we could, we'll talk about that a little bit in a minute. I'll come back to it. But really, you're going to tell. And we did this yesterday when we set up our project, right? You told the AI, like the. My AI, right? You guys saw. How do I teach it how to do that? How do I, how do I teach an AI to write like says your highness every time it talks to me? And you know, okay, that's stupid, but I'm doing that at me. Can anybody in their mind, you know, or like, speak up if you can think of a way. How do you make on purpose to make a point that you can make it, you know, in. Follow those and get. Think like you. So I, I said like, I took 400 pages of LinkedIn posts that I've created the last couple years and then put that into a custom GPT and then had it analyze directions and it will then follow those guidelines every single time it talks to you or says something to you, right? So you give it instructions and your prompt, by the way, could be six pages long. It can, right? If you need, like, if you're creating maybe an AI therapist and you need it to understand like how you write and how you think and all these things, right? Not only do you want. And then that's my tone of voice. Perfect. Amazing. It's exactly what I did. Literally exactly what I did. I took everything to have a really in depth, you know, prompt on how to take different conversations and what not to do, what to do, how to speak tone and language and speak like me and all this stuff. Then you give it a few conversation starters, like, okay, you know, how do I, how do I get rid of a slice, right? And so you could, you could have some, you know, a few questions that automatically pop up. Thing I've ever written, which is a lot more. I'm a writer. I do a lot. I write a lot of articles I've had a lot anyway, I, I like to write and so I took a ton of that stuff and I went into JET as conversation starters, ideas on how to start talking to your GPT. But then here's the big thing. You can upload, just like Eric said, you can upload tons of files. Maybe, maybe you have your, you know, again, I'll go back to you Evan, just because I'm using your example of ohana, but maybe you have a bunch of different guides and a bunch of different GPT and what he's talking about here, guys. And this is one of the things we're going to go over today, so I'll just go over it right now. But you can click Explore GPTs and there's a lot of custom GPT by a lot of different companies, you know, blogs that people have written and a bunch of different things that are about out, you know, this experience, right. And, and you know, all the different cool tours you can go on and, and like you know what to bring, what to wear time, you know, what to do different during different seasons. You can create all kinds of different content and just drop it into a GPT or, or some sort of repository which I can show you how to build this there Canva Dolly, you know, this is your, you know, whatever, that's an open AI thing. But let's go to some more external ones here so we can find something Expedia. Here you go. Right. So you can, you know, Expedia has created their own custom GPT that will help you plan your trips better. Right. And there's a lot of different things outside of ChatGPT you can actually build one of these inside of Replic because you can't API, you can't pull, you can't API in your custom GPT. Unfortunately they just haven't done that yet. But it's really, really easy to do. Right? You basically just have a knowledge base and you say refer to the knowledge base when answering questions. Awesome. Right? Super easy. Third question, right. You can coding experts and co pilots and all kinds of fun stuff but right up here in the corner you have. Yeah, I don't know if it's, if it can do this yet, but could you upload for example in this scenario a video of your golf swing and say why am I slicing it my GPTs? You can create your own and it's super easy. Like any of you can do this in a minute, right? Just open it. Yeah. There isn't great video analysis yet with AI. That's one of the things that I think we'll probably see this year, I think we'll start to see that this year. It can generate up, create your own GPT, name it, call it, you know, Golf Master Evan, right? And then you tell it what you want it to do, right? Teach me how to golf better and video. But it can't. It can't really analyze video. Well, there are a couple things out there that do it, but it's not good enough yet to really be mainstream. You won't see it in any of the major LLM, you know, OpenAI. It's good enough. Once OpenAI is doing it, let's just say that. I mean, that's kind of. That's kind of your general benchmark is when the big companies are. Then you can give it instructions. Now, prompting, creating a good prompt is important. And we could. We'll talk about that a little bit in a minute. I'll come back to it. But really, you're going to tell it. And we, we did this yesterday, are saying, hey, upload your videos and you can get feedback now. And you could tell it to do all kinds of stuff. You could have it do, you know, very detailed analysis, Evan. So you could, for example, take a swing, you know, from behind of day when we set up our project, right? You told the AI, like the. My AI, right? You guys saw, you know, Tiger woods or, or Rory, whoever you want to look like these days, and then yourself or your customer for that matter, and then say, hey, compare these swings, right? Says your highness. Every time it talks to me. And okay, that's stupid, but I'm doing that at Measure angles, measure all these things. Use everything you know about measure everything that every golf pro has ever known about golf ever to give feedback on it. And it would be able to do that. It would be able to measure it, and it would be able to literally go in and draw lines and compare angles and all kinds of crazy things. So, you know, on purpose to make a point, that you can make it, you know, and follow those instruments. Not yet, but it's very, very close. So I think we're very close. I bet directions. And it will then follow those guidelines every single time it talks to you or says something to you, right? So you give it instructions. And your prompt, by the way, could be six pages long. We can make it tomorrow. And someone could say, hey, we can now analyze videos. You can now upload videos. And it can analyze, right? If you need, like if you're creating maybe an AI therapist and it can base things off of it. Right? Now you can upload videos to something like Runway ML, right? Runway ML allows you to upload a video and then have it continue the video or put something before that scene. It can make up content on either side of a scene and extend it, right? Even if you forgot you missed a piece of a shot or something, right? You're like, oh, we need this here and we need to connect or we need to connect to this, to this, right? Help us help you come up with some. And you need it to understand like how you write and how you think and all these things, right? Not only do you want content to connect these two scenes here and, and AI can fill in the gaps. It's crazy to have a really in depth, you know, prompt on how to take different conversations and what not to do, what to do, how to speak tone and language and speak like me and all this stuff. Then you give it a few conversation starter. It's crazy. What's cable doing? We'll do more, we'll do more of that today and maybe it's a video tomorrow. But it's like, okay, you know, how do I, how do I get rid of a slice? But, but anyway, guys, yeah, you upload all this. So what I did exactly like this. This is exactly what Eric did, right? I, I upload. I said, hey, this GPT writes like John Cheney does. I uploaded a ton of articles, right? If we go into my. And we can edit it, you can see what it does, right? And you know, I didn't give it a very good prompt here because I only use this internally, but I could make it a really, really nice prompt. Right? Right. And so you could, you could have some, you know, a few questions that automatically pop up as conversation starters, ideas on how to start talking to your GPT. But then here's the big thing. You can upload, just like Eric said, you can upload tons of files. But bottom line, you know, attached to this are articles written by John as well as some LinkedIn posts. And yeah, I uploaded a huge amount of shares. Massive amount. I post every single day. I posted for years and years and years and years and years. And at any time, by the way, I can come in here and upload more. I can update it. I can update its knowledge base by just dropping 20 more articles I wrote or my book. I'm working. Maybe, maybe you have your, you know, again, I'll go back to you, Evan, just because I'm using your example of ohana, but maybe you have a bunch of different guides and A bunch of different. I could drop it in and say, hey, here's a bunch of stuff that I wrote, right? And so not only that, not only will it write like me, it knows everything I've written about too. And so I could ask it to write a post about something that I'm writing my book know blogs that people have written and a bunch of different things that are about, you know, this experience. And it will, and it will pull a quote from my book, right? I mean it can do crazy stuff. So you could have an assistant, right? You could have an assistant of yours write your link, right? And. And you know, all the different cool tours you can go on and, and like you know what to bring, what to wear, you know what to do different during different seasons. You can create all kinds of different content and in posts if you want, I write all of mine. But. But maybe you don't have time or you don't want to, but you want to post on LinkedIn every day because it's a great way to get marketing done. Amazing. So you start talking about everything you're doing on LinkedIn, posting every single day. Okay, so bottom line is guys, GPTs are really, really awesome custom GPTs. Training your own custom GPT is really just drop it into a GPT or some sort of repository which I can show you how to build this outside of ChatGPT. You can actually build one of these inside of Replik. Cause you can't API you. Awesome. But it can be done inside of replit as well. So with that being said, guys, let's can't pull. You can API in your custom GPT. Unfortunately they just haven't done that yet. But let's dive into Replit and. And let's. I didn't, I didn't get to work on it really yesterday. Let me get this running here. Looks like we do have one question. Brian. Yeah, you mentioned that the custom GPT has. It's really, really easy to do, right. You basically just have a knowledge base and, and you say refer to the double. The core memory as the regular ChatGPT. Did you have something you wanted to add to that? Yeah, I was just gonna say so if, kind of. I think what Eric was asking essentially. So if I. Knowledge base when answering questions. Awesome, right? Super easy. Third question. I wanted to extend the memory. You said you can't do a API directly to it, but somehow I can use it as a reference, right? So say Google Doc. I could use that as my master hub. And then, yeah, I don't know if it's if I can do this yet. But could you upload, for example, in this scenario, a video of your golf swing and say, why am I slicing it? Refer to that so I can pull in more data? Or am I understanding that, yeah, there isn't great video analysis yet with AI? That's one of the, you know, things. I mean, if I'm in repl it, I'm going to be using one of the files over here. Remember how we made this progress MD or the readme MD or whatever. I'm just going to keep expanding that. That's what I'm going to tell it. And I'm just going to have it keep referring to that file as we keep expanding the project and making it better and bigger and whatever. So as you're going along, you might sense that I think we'll probably see this year. I think we'll start to see that this year it can generate. Hey, you know what? I came up with a totally new feature that wasn't in the original plan. Right. And it's not even in the business model. Like, it's just. It's a new thing we're going to do. I talked to some customers. They want this thing. Great. Tell readme. Here's what I did. Here's what we did. Here's why we did it. Please update the readme in the progress to reflect this. Okay, it'll update it. And now you're keeping that. The contact video, but it can't really analyze video. Well, there are a couple things out there that do it, but it's not good enough yet to really be mainstream. You won't see it in any of the major LLM, you know, OpenAI. It's good enough. Once OpenAI is doing it, let's just say that you're keeping it aware of where it really needs to be. Okay, so, all right. Why is it trying to go to an external page here on this? All right, so I started. I did ask it one thing. It was really funny. I took a screenshot right here of the header that I thought was really ugly. Right. Flow State retreats. Why Flow State experience is talking about. I mean, that's kind of. That's kind of your general benchmark is when the big companies are. And I just asked it, does this look good to you? Right. You can really ask at things like that that are very simple and, and sometimes it'll do something and it will say, hey, I fixed it. And I'll be like, really? Actually, like, look at this. This looks terrible. You just Made this. You think this looks good, and we'll say, I'm sorry. And then it will dig a little deeper so it looks like it, it, it move or saying, hey, upload your videos and you can get feedback now. And you could tell it to do all kinds of stuff. You could have it do, you know, very detailed analysis, Evan. So you could, for example, take a swing, you know, from behind of food flow state retreats to one line. I think that's pretty good. I mean, it's better we still have, you know, the tech. There's not, there's not enough room between the title here and this first item, by the way, guys, I'll show you this. You can look at the, you know, Tiger woods or Rory, whoever you want to look like these days, and then yourself, you know, you can look at the site using all kinds of different things. You, you absolutely, for every single thing, want to click on, you know, 14 Pro Max or whatever it is you love, right? If you like a Galaxy 8, you know, whatever you want to do. But you got to look at your mobile or your customer for that matter, and then say, hey, compare these swings, right? Mobile layout and make sure that it looks good. And by the way, it knows to really focus on a mobile, first measure angles, measure all these things. Use everything you know about, you know, measure everything that every golf pro has ever. And notice one thing, guys, this is pretty cool, right? And some of you might not understand house known about golf ever to, you know, to give feedback on it. And it would be able to do that. It would be able to measure it, but. And it would be able to literally go in and like draw lines and compare angles and all kinds of crazy things. So significant it is what we just saw. There's a hamburger menu right here that was not present on desktop, right? That's called conditional formatting, Right. It looks at a condition of, hey, am I on a desktop? Great. Well, then I'm going to display it this way. Am I on mobile? Well, great, I'm going to display it this way. Okay. And so just by switching between those, we have a totally different. Not yet, but it's very, very close. I think we're very close. I bet we can make it tomorrow. And someone could say, hey, we can now analyze videos. You can now upload videos, and it can analyze navigation menu, right? And so you can. You're going to want to talk to it and learn from it and ask it about it and, you know, say, hey, can we change this? Can we do that? You might have noticed in my. My Gen IPI that there's an admin button here. What is it? And it can base things off of it, right? Now you can upload videos to something when I sign out, assuming I actually sign out properly, which somehow it's like, I don't know why it was doing that. But like Runway ML, right? Runway ML allows you to upload a video and then have it continue the video or put something before that scene. It can make up content on either side of a scene and extend it, right? Even if you forgot you missed a piece of a shot or something, right? You're like, oh, we need this here and we need to connect or we need to connect to this. To this, right? Help us help come up with some content to connect these two scenes here. And AI can fill in the gaps. It's crazy. But look, the admin is gone, right? The admin button's gone. Now if I go click login again. Yeah, see, that appears to be working, right? I need to just ask it why I'm seeing that problem there. But that admin only shows up when I'm logged in as an admin. And how did I set that up? I said, hey, make an admin dashboard that only shows up for admin users. Okay. You know, and it. And it does that. Then I have to test it and, you know, make sure it works, right? You got to keep. Keep bouncing back into there. Okay, so guys, Replit is like any tool not perfect, right? It is made by, you know, great. It's crazy. What's cable doing? We'll do more. We'll do more of that today and, and maybe some video tomorrow. Bite Team that I'm actually in communication with, they were just emailing me back and forth last night. And so I'm really, really excited about this. I'm going to be able to get some really cool access to early stuff. And I'm going to. But. But anyway, guys. Yeah. You upload all this. So what I did exactly like this. This is exactly what Eric share it with you guys. You know, everybody that's taking the course. By the way, there is a Facebook group that I will send a link to all of you for that you can all join. For everybody that's taken this scripted, right? I. I upload. I said, hey, this GPT writes like John Cheney does. I uploaded a ton of articles, right? If we go into my. And we can edit it, you can see what it does, right? Course. And so people that, you know, if you're. If you're stuck or you're like, or you figure something out, or you finish a product, you want some feedback, or hey guys, I can't figure this problem out. You can go to that Facebook group, just ask your question and either one of the other amazing students that's taking the course that's now been working on it for months. And you know, I didn't give it a very good prompt here because I only use this internally, but I could make it a really, really nice prompt. Right. But bottom line, you know, attached to this are articles written by John as Well as some LinkedIn will say, oh, this is what I do to fix that. Or, you know, I jump in there. I, if I, you know, if you talk in there, I'm going to, I'm going to see it, I'm going to respond if somebody else hasn't. Or I'll say, hey, what he said is exactly right. Here's another way you can try it. Right? But I'm very active on that. I want to make sure you guys have, you know, so. And yeah, I uploaded a huge amount of shares, massive amount. I post every single day. I posted for years and years and years and years and years. And at any time, by the way, I can come in here and upload more important, at least at that level, you know, for as long as, you know, you want to. So keep, use that, use that tool. But, but guys, like any tool, it has its quirks. Some things work well, some things don't. You have to really learn how to conduct it. We talked about this yesterday and the, you know, the, the orchestra, the John Williams, you know, analogy here, right? You're the one up there. I can update it. I can update its knowledge base by just dropping 20 more articles I wrote or my book I'm working on. I could drop it in and say, hey, here's a bunch of stuff that I wrote. Right? And so not only that, not, not only you've got the, the batons or whatever, and, and you're, you're swinging your hands and if you say, will it write like me? It knows everything I've written about too. And so I could ask it to write a post about something that I'm writing my book on and it will, and it will pull a quote from my book, right? I mean, it can do crazy when your hands wrong, it's going to do, it's going to do it wrong. AI is usually I say usually because most of the time, pretty dang good at listening to you. It's going to do what you tell it to do. And if you, if you tell it to do something, it does something else. Well, maybe don't jump these stuff. So you could have an assistant, right? You could have an assistant of yours. Write your LinkedIn posts if you want. I write all of mine. But maybe you don't have time or you don't want to, but you want to post something to blame the AI, right? Maybe say, okay, what do I not understand? What do I not really, really know? And so you can have conversations with Replit, right? You can have conversations and say, I need you to explain to me. So what I would have done, I think it was Travis was talking about this. What I would have done when I'm stuck like that is I would have said, I need you to explain to me how this works every day. Because it's a great way to get marketing done. Amazing. So you start talking about everything you're doing on LinkedIn, posting every single day. Okay? So bottom line is, guys, GPT how it's supposed to work, right? And it will say, okay, you click here and you've uploaded this API we call the API. We get this hook back. You know, we get this here and this happens, and this happens, and this happens. Are. Are really, really awesome custom GPTs. Training your own custom GPT is really, really awesome. But it can be done inside of replit as well. So with that being said, guys, let's. And it will explain it in a way, guys, that again, back. You know, Jessica, who knows nothing about code, would understand this, right? She would be able to look at it and read it and say, okay, I can use my brain to understand. I'm just reading something. Okay, this talks to this. This talks to this. This talk. Let's dive into replit and, and let's. I didn't. I didn't get to work on it. Real to this. And then you could say. And then you could say, is. Is it yesterday? Let me get this running here. Looks like we do have one question currently actually set up to properly work that way. Don't write any code. Just tell me. Right? Yes. Great. Okay, awesome. Well, when I do this, I keep getting this error. Do you have any ideas about what might be caused? Brian? Yeah, you mentioned that the custom GPT has double the core memory as the regular ChatGPT. Did you have something you wanted to add to that? Yeah, I was just going to say this. Don't write any code. Just tell me. You can have a conversation with it and you can ask it to ask itself, right? You can help it to get out. You can Guide it to get out of a loop. Right. So even if you don't. Yeah, even if you don't. Yeah. So Trevor said this. Awesome. Even if. So if. Kind of. I think what Eric was asking essentially. So if I. You do understand code. That's still the blow a lot of the time, right? Trevor, will you talk about that for a second? What? You wanted to extend the memory. You said you can't do an API directly to it, but somehow I can use it as a reference. Right. So say Google Doc. I could use that as my master hub and then have it refer to that so I can pull in more data. Or am I understanding? Yeah, you do. Like, do you use AI and talk to it and things in and have a conversation about the code so you can understand what's happening. So I would say the company I work for is called 401 Go. Yeah, I would say we were one of the earliest adopters of as soon as. As soon as OpenAI released chat GPT, we. I mean, if I'm in repl it, I'm going to be using one of the files over here. Remember how we made this progress lhad accounts, so we've been using it since MD or the ReadMe MD or whatever. I'm just going to keep expanding that. That's what I'm going to tell it. And I'm just going to have it keep referring to that. Day one probably increased everybody's productivity at least 10x. Amazing. So. So yeah, we have. You will have conversations with the code. You'll copy in some code, be like, hey, I'm not understanding why this isn't working file as we keep expanding the project and making it better and bigger and whatever. So as you're going along, you might say, you know what? I came up with a totally new feature that wasn't in the original plan. Right. And it's not even in the business model. Mostly it's helped our junior developers. Like, it's just, it's a new thing we're going to do. I talked to some customers. They want this thing. Great. Tell readme. Here's what I did. Here's what we did, here's why we did it. Please update the readme and the progress to reflect this come up to speed like incredibly fast. Where maybe you wouldn't expect anything out of them for several months. They're. They're productive within weeks. Yeah. That's so cool. And guys, yesterday we did, you know, our mini coding boot camp that lasted all of about eight minutes. But just being able to understand things like UI and front end and a database and understanding what a post is or, you know, delete function is. You know, basic things like that, that's the beginning. But as you talk to your. Your. Okay, it'll update it. And now you're keeping that the context. You're keeping it aware of where it really needs to be. Okay, so, all right. Or agent. And you ask it questions and you have discussions with it, and if it says something you don't understand, you can say, I don't understand that. Can you help me understand it a little bit better? Can you simplify a little bit? I'm new to coding. Can you help me understand why this matters? Right. Why is it trying to go to an external page here on this? All right, so I started. I did. You will start to become a developer, right? You'll start to at least learn how to talk about code. Ask it one thing. It's really funny. I took a screenshot right here of the header so that you can understand a little bit more what you're. What you're doing now. Do this for, you know, spend 200 hours in here and really talk to it. All this time, now you're starting to really understand code. Can you write code on your own without AI? No, you can't. But that I thought was really ugly, right? Flow state retreats. Why? Flow state experience is talking about. But you get really, really, really, really proficient with it, right? And I have spent hundreds of hours inside of REPLIT now, and I just asked it, does this look good to you? Right? You can really ask it things like that, that are very simple. And sometimes it'll do something and it will say, hey, I fixed it. And I'll be like, really actually, like, look at this. This looks terrible. You just made. And so I'm getting much, much better at that. Go ahead, Travis. So one of my questions is. So it's suggesting. So it's giving me a suggestion. Says, basically, review outside documentation from, like, where I'm grabbing my API, make sure everything's correct from them, and then bring it back in. Can I just give it those reads? You think this looks good? And we'll say, I'm sorry. And then it will dig a little deeper so it looks like it. It moved. Flow state retreats to one line. I think that's pretty good. I mean, it's better. We still have, you know, resources. Here's the link, here's the documentation. I, you know, look in here, find the stuff. Can you find what you need? Could I do that? 100%. Yeah, just drop, just drop the thing. But here, the doc, here's the documentation. Figure out what's going on. What's going on. And it will take a minute to review it, to read it. And pull that in. I had this, the tech. There's not, there's not enough room between the title here and this first item. By the way, guys, I'll show you this. You can look at the. And that. That works for the most part. But I have this really stupid problem. There's a company called Cloudinary and Cloudinary is, you know, really good at handling images and videos and, you know, programmatically delivering them based on the need of, you know, whatever. So you can upload one picture to Cloudinary. You know, you can look at, at the site using all kinds of different things. You, you absolutely, for every single thing, want to click on, you know, I, you know, 14 Pro Max. And then it will give you 400 ways you can deploy it, right? Smaller square. I mean, you can see this image, you know, changing right here. It automatically, in the browser itself actually, or, sorry, in the link itself, gives instructions to how that image should change. It's really, really, really good. This company is worth about $2 billion. Awesome company. These guys almost acquired Max, or whatever it is you love, right? If you like a galaxy, you know, whatever you want to do. But you got to look at your mobile, your mobile layout and make sure that it looks good. And by the way, it knows to really focus on a mobile first. My company, like three years ago, right? And we ended up turning it down. It was, it was a little bit low and, you know, we probably should have sold to them, but whatever, we didn't. But Cloudinary, they handle these images and videos. And I was trying to. And notice one thing, guys, this is pretty cool, right? And some of you might not understand how to build a video upload feature. And I could get videos up to 100 megabytes to upload, no problem. But as soon as it got to 100.1 megabytes, it would just fail. I couldn't get it to work. I could not figure it out. And I was like, why is this not working? Why is this not working? And it would go through and it would say, it's working, it's right, it's set up properly. Like, we're doing this from this and this and this and that's perfect. And then I just got this idea. I was like, I wonder. And I jumped on and I, and I said, is there a Different API significant. It is what we just saw. There's a hamburger menu right here. Or a different way to handle large video files. With Cloudinary, it's Google it, right? And it said, yep. And there's a totally separate API for large video files. And somehow replit just didn't get that. So that's one of the consequences of me not being super familiar with APIs and Cloudinary's APIs specifically and maybe some bad documentation on Cloudinary side. Sometimes you just got to work the problem and eventually you figure it out, right. And oh, this feels so good to figure out a problem. Right. It's like a big relief. And you just, you feel good inside when you figure out a problem like that. So anyway, it's not super easy. That was not present on desktop, right? That's called conditional formatting, right. It looks at a condition of, hey, am I on a desktop? Great, well then I'm going to display it this way. Am I on mobile? Well, great, I'm going to display it this way. Okay. And so just by switching between those, we have a totally different navigation menu. Right. And so you can, you're going to want to talk to it and learn all the time. For the most part though, you know, I, I've actually kind of pulled away from using Cloudinary a little bit because I had some problems with it. So let's jump in here guys, and let's go over this. Replit tips and tricks. This is one of the documents that you have. I think it's super useful. A lot of the things that I do here, you've seen me do a few times and I've talked about a couple of them like the element selector or screenshotting something. Learn from it and ask it about it and you know, say, hey, can we change this? Can we do that? You might have noticed in. We'll go over it. We're gonna. This is again gonna be another, you know, five minute, maybe 10 minute discussion, but we're gonna, we're gonna kind of blast through that. All right. And then we're gonna get to some image creation. Where did you get this document? You sent it, right? Now this document, it hasn't been sent yet. But what I will do right here is I'll just send you the link. Let's see here. Yeah, if you guys want my, my gen AIPI that there's an admin button here when I sign out, assuming I actually sign up properly, which somehow it's like, I don't know why it was doing that, but, but look, the admin is gone, right? The admin button's gone. Now if I go click on, open it and read along. Feel free to. There it is right there again. I will send you this entire folder at the end of the course. You'll have a link to it, you'll have everything. I don't send it at the beginning because I don't want people to dive in and get lost. I really want to kind of guide you through it and then I'll, you know, give you the keys to the kingdom and you can. Right, so tomorrow after the course, you'll get, you'll get this. Actually it's probably going to be Thursday again. Again. Yeah. See, that appears to be working. Right? I need to just ask it why. Why I'm seeing that problem there. But that ad, if I'm being honest, it's. It's always the day after. When I met. When I finish a course, guys, my brain only shows up when I'm logged in as an admin. Okay, and how did I set that up? I said, hey is tired. Like I've given a lot of my energy to you guys. And so I usually don't get my emails, follow up emails till the next day. So. So we'll see. Make an admin dashboard that only shows up for admin users. Okay. You know, all right, so you guys know what replit is. I don't need to, I don't need to teach you that. But there's a few things that we've already, we've already gone over, right? And it, and it does that. Then I have to test it and make sure it works, right? You got to keep, keep bouncing back into there. Read me in progress. There's so many people, guys that don't do this, right? Especially developers or non developers, right. They don't know what a readme file is, right? Trevor's like, yeah, of course there's a readme file, right? Like, or at least a lot of good projects will have that and they'll tell you how this is put together, right? If you've ever been part of a code. Okay, so guys, replit is like any tool not perfect, right? It is made by, you know, great team that I'm actually in communication, but they were just emailing me back and forth last night and so I'm really, really excited about this. I'm going to be able to get some really cool access to early stuff and I'm an addition team, right? Where you're, you know, you're receiving code from Somebody else, another team that built it. There's a lot of documentation and things like that in there that tell you what, what are happening, and many times even have to have some meetings to really have that handoff happen here. And so if you understand that, hey, code doesn't just automatically explain itself. It can, and I can really try and figure it out. But if you're having it start from ground zero, every time you start a new agent chat, man, you're just. You're shooting yourself in the foot. So use these files. And by the way, there could be other types of files that you think of that you need. You can create anything MD file, you know, share it with you guys, you know, everybody that's taking the course. By the way, there is a Facebook group that I will send a link to all of you for that you can all join for everybody that's taken this course. And so people that, you know, if you're. If you're stuck or you're like, or you figure something out, or you finish a product, you want some feedback or, hey, guys, I can't figure this problem out, you can go to that Facebook group, just ask, you know, knowledge base. You can make a knowledge base file, right, that acts or a prompt MD and say, hey, ChatGPT, I want you use this prompt file that I'm going to update regularly. And I want this to be where you pull the prompt from. And we'll say, okay, great, I'll do that. And it will structure itself so that it pulls the prompt from the prompt MD file. It's not that hard. You create something, you point at it, right? If you feel like something needs to be updated all the time, just like I was talking about earlier with my. My. This $39 thing, right? Instead of hard coding or just typing it in so that it's always like that. I have it pull a variable. I have it pull something from somewhere else that I can then manage in, in a more reasonable way so that those things update automatically. Okay, so that's. That's a lot of what coding is, is understanding relationship building and systems, right? Get a whiteboard and say, okay, I've got this, I've got this, I've got this. Draw it all out. That's called architecting, right? You're architecting this solution about how does it all fit together? How can we make it better? Right? All right, so every time you start a question, and either one of the other amazing students that's taking the course that's now been working on it for months, will Say, oh, this is what I do to fix that. Or, you know, I jump in there if I, you know, if you talk in there, I'm going to, I'm going to, I'm going to see it. I'm going to respond if somebody else hasn't. Or I'll say, hey, what he said is exactly right. Here's another way you can try it. Right. But I'm very active on that. I'll make sure you guys have, you know, new chat. Use the command. Please reference the readme MD in progress in ME files and familiarize yourself with this, with the project. Right. This ensures continuity and prevents AI from losing track of where it was. Okay, I'm going to take one second here because I want to. I think this is very important to point out right here. Travis or somebody said that, that they used all of their, all of their AI credits yesterday. And that can happen. You can use them in a day if you just go meet important, at least at that level, you know, for as long as, you know you want to. So keep. You go for eight or nine hours like you're, you're going to use it, right? But when you click on this little button right here, it pulls up your usage, right? And it shows how much money you've spent in that month on all of the different things that replit does. One of the cool things that replit does for you guys, use that, use that tool. But, but guys, like any tool, it has its quirks. Some things work well, some things don't. You have to really learn how to conduct it. We talked about this yesterday. And the, you know, the, the orchestra, the John Williams, you know, analogy here. Right. You're the one up there is, is, you know, aside from the agent checkpoints, which you guys know about it, and by the way, I've spent $161. But that's because I, you know, this month alone, that's because I've, you know, I teach this class and so I'm constantly in it and I, you know, I run several things out of it. But you look down here at all the different things it does. It's got your, your databases right, your postgres equals storage, and your data, you've got the, the batons or whatever, and, and you're, you're swinging your hands. And if you swing your hands wrong, it's going to do, it's going to do it wrong. AI is usually, I say usually because most of the time, pretty dang good at listening to you. It's going to do what you tell it to do based compute, right? It's got your scheduled deployments and your auto scale deployments. Auto scale deployments are awesome. We'll talk about those. But basically, if you've ever tried to buy a ticket to a concert, you know, try to get into the Tabernacle Choir Christmas concert or try to buy a tailor when if you, if you tell it to do something, it does something else. Well, maybe don't jump a Swift ticket or BYU tickets or Utah tickets or whatever you like and you're stuck. Or it doesn't load or it's super slow and you feel like the site's broken. That's because they don't have great auto scale setup or their system just. There can be other problems. But a lot of times they don't have auto scaling things set up and it's just kind of broken. And then storing, right? Storing a video, storing a picture, storing a file, storing to blame the AI, right? Maybe say okay, what do I not understand? What do I not really, really know? And so you can have conversations with Replit, right? You can have conversations and say I need you to explain to me so that what I would have done, I think it was Travis was talking about this, what I would have done when I'm stuck, whatever, right? Now all of this stuff can be really, really cheap, right? I mean, cents per gigabyte per like that. As I would have said, I need you to explain to me how this works or a month, all right. Or $1.50 per gigabyte per month, right? Or whatever, right? It's very, very cheap. You can see how it breaks down for your specific project and the types of things you have set up. But again, the fact that I'm running dozens of different apps and deploying them and hosting them and all this stuff and I'm still only at $161 this month is just mind blowing, right? I mean you could spend so much money doing this. We'll talk a little bit later about how to use. If you want to use AWS or Azure or Firebase or other things you can hook up to other external systems that do these things. If you want, totally fine, right? But I personally love as a non developer not having to worry about that, how it's supposed to work, right? And it will say, okay, you click here and you've uploaded this. It's all here, it's all done for me. And so I don't have to worry about it. Now if I want a really API, we call the API, we get this Hook back. We get this here and this happens and this happens and this happens and it will explain it in a way. Guys, that again, back. You know, Jessica, who knows nothing about code, would understand this, right? She would be able to look at it and read it, say, okay, get serious and scale to a billion dollar company. I'm gonna have to graduate from Replit and I'm show you guys how to do that too. Right? And there's, there's ways to do that. But the you. I legitimately believe you can build $100 million business inside of Replit. Hosting inside of Replit. I really do. You can have 50,000 concurrent users on your product in Replit. It can handle that. It's set up to handle that. So, so anyway, all right. Replit isn't paying me yet to say that, but hopefully they are soon. They're going to give me at least some free credits, some fun deals. They're going to give me deals where I can. Well, sorry I didn't get it before your class, but they're going to give me some deals where I can, I can, I can have like a premium sign up so you get like 25 of free credits when you first, you know, when you take my class, stuff like that. Anyway, so they're doing, they're doing some nice stuff for me as I say some nice things about them. I can use my brain to understand. I'm just reading something. Okay, this talks to this, this talks to this, this talks that they'll put in their blogs and posts and things like that. So anyway, there's these, these prompts right here. These prompts are really, really awesome. I've used each of these many, many, many, many times as well as many other things. And again, I want you guys to realize that you're all smart people to this. And then you could say, and then you could say, is, is it you? I guarantee you you're all smarter than me in many things and use that intelligence to come up with good ways to talk. Right. It's just a communication thing. One of the things, one of the posts I've been. This is kind of rolling around in my head for LinkedIn is being good at AI at currently actually set up to properly work that way. Don't write any code, just tell me. Right? Yes. Great. Okay. Awesome. Well, when I do, directing an AI team is almost no different than leading a real human team. This, I keep getting this error. Are there. Do you have any ideas about what might be causing. You have to know how to communicate and lead. Right which means you as the CEO, seen this. Don't write any code. Just tell me, right? So you can have a conversation with it and you can ask it to ask itself, right? You can help it to get out, you can guide it to get out of a loop, right? Oh, all of you are the CEO. You have to understand how it all fits together. So you've got this picture and then you have to do your best to get that out into the prompts and into the overall plan, into the read me and the progress. Right? So that's why we went through that, that, that, that idea of hey, here's the whole plan. So even if you don't. Yeah, even if you don't. Yeah. So Trevor said this. Awesome. Even if now if you're really doing that well, you might take a couple hours having it ask you 50 questions, going deeper into whatever. Hey, I want to go deeper into this. Hey, you didn't ask this question. You do understand code. That's still the blow a lot of the time, right? Trevor, will you talk about that for a second? What you do, like do you use AI and talk to it and, and so things in and have a conversation about the code so you can understand what's happening. So I would say ask me a line of questioning about this thing, right? And then build that whole plan and make it really, really nice. We made a 1500 word business plan, make a 6000 word business plan. Whatever it is you need to do, you're in charge, right? So it's leadership and communication. So please create a plan to build this feature. Don't write any code yet, just outline the steps and I will approve before proceeding. So as you. The company I work for is called 401go. Yeah, I would say we were one of the early. As you understand how replit works better and better and you're getting a little bit more comfortable with it, say okay, we want to. I'm going to build this entire, I'm going to build a new blog section, right. I want the blog to, to operate, to pull from the database. I don't want you to hard code the blog, right. I want it all to pull from, from, you know, structured data and, and, and so that we can update and we can edit and delete and all those functionalities. Right? Just adopters of as soon as, as soon as OpenAI release chat GPT we, and there's, there's good reasons for all of those things. But I want to do this. I want to do this, I want to do this. I really want to focus on, you know, the reason I'm writing the blog is because all that accounts. So we've been using it since day one. Probably increased everybody's productivity at least 10x. Amazing. I want it to help with SEO. Right. And so please ensure that we're following best practices. Oh, so yeah, we have. You will have conversations with the code. You'll for that. Then you say, please create a plan to build this feature. Don't write any code yet. Just outline the steps. I know and then I'll approve before proceeding. So then it will take into account everything that you just told it. Here's what I'm going to do for SEO or here's what I'm going to do. Here's what I'm going to do. Here's what I'm going to do. And it's. And it doesn't say anything about SEO, then you can say, I asked about SEO. What is your plan to make sure that everything is tagged right and is set up right so that this will help with SEO? Anybody not know what SEO is? Anybody copy some code, be like, hey, I'm not understanding why this isn't working. Mostly it's helped our junior developers come up to speed, like incredibly fast, where maybe you wouldn't expect anything out of them for several months. They're productive within weeks. Yeah. I just want to make sure. Jessica, it's okay if you raise your hand. I'm just kidding. I'm sure you do. But Search engine optimization, right? You want to make sure you show up in Google, right? Hi. In Google. And there's things you can do to structure even a blog post to be done. Right. And so have it modify the plan. Okay. Hey, I like the plan now go ahead. Right. So instead of just letting it go, right. Ask it for its plan first and then use your brain to figure. That's so cool. And guys, yesterday we did, you know, our mini coding boot camp, the last shot, if it's, if it's accomplishing everything that you think it should. Okay. Limit AI scope for precision. Do not modify anything except the front end layout and the login page all of about eight minutes. But just being able to understand things like UI and front end and a database and understanding what a post is or you know, delete function is, you know, basic things like that, that's the beginning. But as you talk to your, your sometimes, especially when you're in a long, long chat. Guys, we're talking about long chats in a second here. When you're in a long chat, it Considers everything you've said in the chat every time you, you ask it anything. And so if it just randomly looks, you know, six hours ago at some thing and it's like, oh, I don't think we've done that yet. You know, you ask it to update a button and then it goes agent and you ask it questions and you have discussions with it and if it says something you don't understand, you can say, I don't understand that. Can you help me understand it a little bit? But can you simplify a little bit? I'm new to coding. Can you help me understand why this matters? Right? You will start to become a developer, right? You'll start to at least learn how to talk about code. Who's often starts doing that thing that you haven't talked about for six hours, right? So start new chats. We're going to talk about why right now. Okay, remember guys, what how Context works with AI, right? It considers everything at no, at least in ChatGPT, it knows everything in the memory, it knows everything in the chat, and it knows everything in the project itself. The way that replit is built is it will consider everything in the chat and everything in the project. Right? At least when it needs to, it will determine, okay, I'm working on the code so that you can understand a little bit more what you're doing now. Do this blog component or no, I'm working, let's say, let's say something else. I'm working on the menu, the nav menu for, you know, spend 200 hours in here and really talk to it all this time. Now you're starting to really understand code. Can you write code on your own without AI? No, you can't. But you get really, really, really, really proficient with it, right? And I have spent hundreds of hours inside of replit now. And so it then knows where to go. Look in here to say, okay, I need to find the menu and figure out where that is. And I know, I know I need the index HTML here and I know I need these things and I know I need this component. I know I need navigation, right? I need to look at all of these different pieces. And so I'm getting much, much better at that. Go ahead, Travis. So one of my questions is, right, so it's going to pull up the right files and then it's going to, you know, work on it. So it's suggesting. So it's giving me a suggestion, says basically review outside documentation from like where I'm grabbing my API, make sure everything's Correct from them and then bring it back in. Can I just give it those. If it's more complex than a standard setup, then you might have to ask it. Hey, I think there's some other fancy things we're doing with the menu. And you would know that because you built the project, but it's going to build it in as simple and straightforward a way as possible so it knows where to go look for the content that it needs. So then it's going to consider your context in your chat and your project. Now what if your chat gets super, super, super, super, super long, right? Number one, that confuses your chat and number two, it eats up your usage because you're, you're wasting tokens resources. Here's the link, here's the documentation. I, you know, look in here, find the stuff. Can you find what you need? Could I do that? 100 yeah, just drop, just drop the thing. Be like here the doc, here's the documentation. Figure out what's going on, what's going on with every single prompt, right? What's a token? Right, let's go here. OpenAI API pricing and it will take a minute to review it, to read it and ball that in. I had this, okay, a token. So you can see here, if you're using the open AI O3 model, it's $10 per million tokens, right? If you use an 04 Mini, it's A$10 per million tokens. So well, you got to choose which model you're using, right? I think by default right now if you ask it to set up an OpenAI, I think it sets it up with GPT4OH Mini and that, that works for the most part. But I have this really stupid problem. There's a company called cloudinary and cloudinary. But I would always ask Devin, I'd ask replit, hey, which model do you have set up with? OP is really good at handling images and videos and you know, program and AI, right? But 30 cents per million tokens, that's probably what it's going to do. And GPT 400 mini for medically delivering them based on the need of, you know, whatever. So you can upload one picture to Cloudinary. The vast majority of use cases is more than capable, right? The 4.1 is also getting a little bit cheaper for a lot of things you want to do for more. And then it will give you 400 ways you can deploy it. Right? Smaller square. I mean you can see this image, you know, changing right here. It automatically in the browser itself. Actually, or sorry in the link itself gives instructions to how that image should change. It's really, really, really good. This company is worth about $2 billion. Awesome company. Okay? These guys almost acquired bad stuff. So you don't have to know a huge amount about this other than the fact that you're getting charged for tokens. And a token, guys, is basically any input or output that happens, right? So if you ask it something and you give it a really long, you know, you upload a 90 page document, a legal document for it to review, well, that's a huge amount of tokens right now in the Chat GPT consumer version, right? If you're over here and you're asking it a question, it obfuscates all of this. It's just, hey, it's 20 bucks a month. Do whatever you want. It doesn't give you a limit right now. It might throttle you a little bit. It might throttle you a little bit if you do like my company like three years ago, right? And we ended up turning it down. It was, it was a little bit low and you know, we probably should have sold to them, but whatever, we didn't. But Cloudinary, they handle these images and videos and I was trying to breezy amounts of stuff and some of, some of the, some like, like if you've ever been on Chat GPT, the free version before, paying for it and you said, hey, you have to wait a few hours before you can ask another question. Well, that's because you've used build a video upload feature and I could get videos up to 100 megabytes to upload up your token limit for that time, that time frame. Okay? So bottom line is, guys, it costs money to run servers and to run GPUs that are in these massive warehouses with crazy air conditioning bills because they're so freaking hot because they're working so hard. As you guys ask, no problem. But as soon as it got to 100.1 megabytes, it would just fail. I couldn't get it to work. I could not figure it out. And I was like, why is this not working? Why is this not working? And it would go through and it would get to make all your studio Ghibli pictures, right? And say it's working, it's right, it's set up properly. Like we're doing this from this and this and this and that's perfect. And every time you ask it to do something, every time you say thank you. Like we were just talking about in the beginning here, right? That I think Eric. And then I just got this idea. I was like, I wonder. It costs money. It costs money to do that. It costs money to somebody to do that, right? So keep your chats shorter and you're going to pay less. You're going to use your credit slower and you're going to get better results. You're going to have tighter results because. And I jumped on and I, and I said, is there a different API or a different way to handle large video files with cloudinary? It's Google it, right? And it said, yep, and there's a totally separate API for large video files. And somehow as you're going to say, all right, let's finish this feature, update the progress MD and this whatever. I try to keep my chats under a dollar. Let me just put it that way. I'll go to $2 sometimes. Replit just didn't get that. So that's one of the consequences. Go to 7, sometimes I go to 14. I don't follow that religiously, but I try to go feature by feature by feature by feature. Now if I'm working on something that's maybe a little bit more complex like the blog, and I can't get something right and I'm trying to work on something, I'm probably going to keep it open unless I get stuck in a loop. If I get stuck in a loop, then I'm definitely starting a new chat, right? I'll still tell it to update and I'll say, hey, I need you to note that we're still stuck in a loop. This isn't working right yet. Note that in the progress md. I'll tell it that so that it knows of me not being super familiar with APIs and Cloudinary's APIs specifically and maybe maybe some bad documentation on Cloudinary side. Sometimes you just got to work the problem and eventually you figure it out, right? And oh, this feels so good to figure out a problem, right? It's like a big relief. And you just, you feel good inside when you figure out a problem like that. So anyway, it's not super nose. Then I'll pull up, I'll say, hey, please read this and this. And I want to focus in on this thing that we're kind of stuck on. I want you to take a fresh look at it. Are we doing it right or do we need to rip this out and start over right? Do we need to rip out this piece of this and start over? What's happening? What have you identified that we could probably fix here? Or what do you think the source of the problem is? Great. What else could the problem be? Easy all the time. For the most part though, you know, I, I've actually kind of pulled away from using Cloudinary a little bit because I had some problems with it. So let's jump in here guys and let's go over this replit tips and tricks. This is one of the documents that you have. I think it's super useful. A lot of the things that. Right. What else could the problem be? I mean, ask it, challenge it, follow up. Okay. All right, guys, is there a simpler way to implement this feature? Consider efficiency and speed, right? A lot of code is bad, right? A lot of AI code is bad. A lot of I do here, you've seen me do a few times and I've talked about a couple of them, like the element selector or screenshot in something. But human code is bad, right? And so you got to figure out, is there a better way to do this? As a non developer, you might not know how to ask that question, you might know, you might not know how to judge it, but you do know how to ask it. Is there a simpler way to do this? Right. You know, your, your proposed plan is 2000 lines of code. Could this be done in 200 lines of code? We'll go over it. This is again going to be another 5 minute, maybe 10 minute discussion, but we're going to blast through that. All right, then we're going to get. That's a good question. Let me think about that. Well, maybe if we did this, we can do this. We can't get it to 200, but we. Image creation. You sent it right? Now this document, it hasn't been sent yet. But what I will do right here is I'll just send you the link. Let's see here. Yeah, if you guys want to get it to 800, great. You just went from 2000 to 800. Now your app works faster, it has less logic that can break, it's smarter. Right? At least theoretically. Right. There are certainly times where that might not be the case, but challenge it, ask it to think outside the box, to do new things. And especially, you know, the Trevors of the world, right? I mean, go nuts on this. And I'm sure you do this all the time, right? Where you're just saying, hey, is there a way? That has never been done. Want to open it? Read along, Feel free to. There it is right there again. I will send you this entire folder at the end of the course. You'll have A link to it. You'll have everything. I don't send it at the beginning because I do it for where we could still accomplish this, right, with, you know, less code and can we, can we improve the speed? Right now it takes.07 seconds to load this thing. Can we get it to 0.2? Don't want people to dive in and get lost. I really want to kind of guide you through it and then I'll, you know, give you the keys to the kingdom and that's Right. So tomorrow after the course, you'll get, you'll get this. Actually, it's probably going to be Thursday. Like, how can we, how can we really do this? Right? All right. Okay guys, detect hidden bugs. This is something that, that I've had happen many, many times where you ask it to do something and then it uses its chain of thought reasoning, right? And all right, okay guys, detect hidden bugs. This is something that, that I've had happen many, many times where you'll ask it to do something and then it, it uses its chain of thought reasoning, right? And it goes, and it says, okay, for this to work, we need this and this and this and this and this and this and this. If I'm being honest, it's. It's always the day after when I met, when I finish a course, guys, my brain is tired. Like I've given a lot of my energy to you guys and so I usually don't get my emails, follow up emails till the next day. So, so we'll see. And it forgets to consider we already built that component over here and it will build it again the same thing when you actually want it using, you know, the same component across different places in the site so that we're more efficient. Right. You can also have it create conflicting code. So CSS guys is, is basically the part of a website or the language that, that determines, you know, how big your fonts are and the colors you use and you know what? All right, so you guys know what replit is. I don't need to, I don't need to teach you that, but there's a few things that we've already gone over, right? Fonts you're using and just formatting, right? It's basically your format. It's your visual part of, of the, of the website, at least part of it. And I can't tell you how many times. It's like going through the initial build and it's showing me my preview and then it keeps going, it keeps working, keeps working, keeps working, keeps working. And then all of a sudden it says, okay, I'm done. And now the website looks super ugly. Like it looked awesome in the beginning and then it just ripped it apart and destroyed it. I'm like, what the heck? Why did you, why did you go in and break everything? And I learned that occasionally it will wrap its own css, right? Which means it'll take CSS formatting and then it'll put other formatting around it. And then you're like, fix this thing. And it says, I fixed it. And. But the wrapped one is, is taking over, right? It's. It's eating the other one alive, right? And so it can't do it. So just ask it that kind of a question. I can tell you that with the V2, replied agent V2 that we're using right now, it is way better at avoiding those kinds of problems. I would say it's probably 500 better than V1 was. And I'm really excited for V3 because I know it's just going to keep getting better. Okay. And part of that relates to, by the way, Anthropic, right? Or wherever Anthropic was. I don't know where it is. Yeah. Anyway, when Anthropic comes out, when Claude 3.9 or, you know, it came from 3.5 to 3.7. And so I'm assuming 3.9 is next, but maybe 4.0 or 5.0 whenever, you know, every time the underlying models get smarter, then all the other things on top are going to get better. So I'm excited about those things for what it's going to make possible. Okay. Ask for conflicting code. Hey, you know, if you're stuck in a loop again, Travis, is there conflicting code or is there. Are there two? Are there are. Is there more than one part of the site that's trying to do this and that's why it's breaking? That can absolutely be the problem. Okay, read me in progress. There's so many people, guys, that don't do this, right? Especially developers or non developers, right. They don't know what a readme file is, right. Trevor's like, yeah, of course there's a readme file, right? Like, or at least a lot of good projects will have that and they'll tell you how this is put together, right? If you've ever been part of a code, take a step back, analyze what you just wrote. Are there any potential issues? Right. What's happening here? Right. Sometimes, guys, it will appear that the AI is stuck like this little thing right here where it says agent is waiting for your response. Or like, or maybe it's going. It says pause or whatever. Sometimes that disappears. Even if the transition team, right where you're, you know, you're receiving code from somebody else, another team that built AI, even if the agent is writing, you can refresh the page and it will not stop the agent. It won't mess it up. That's a good thing to know about. So just refresh it. The UI will refresh and your buttons will all be back. All right, this one might save you guys a ton. A ton of time. And I even copied the. The conversation I had with it and pasted it in here. I took a little screenshot of this. So sometimes replit will write in languages that are different from what it even expects, right? Or like, you open a new agent and it's like, oh, I didn't realize I was writing in that language. I was going in circles about a stupid fave icon issue. I could not. It took me two. There's a lot of documentation and things like that in it that tell you what, what are happening in many hours to get the favicon to show up, right? The favicon, for those of you that don't know, is this little icon up here on the browser, right? All these little tiny. These tiny little things. That's the favicon. And it shows up in a lot of different browsers in a lot of different ways, right? I know I have my new cool AI browser somewhere. Where is it? Maybe it's over here. No, I don't know where it is. Anyway, times even have to have some meetings to really have that handoff happen here. And so if those favicons show up in a lot of different places but I couldn't get it to figure out. And what happened was this, right? The error is directory name is not defined. This is because we're using ESM modules with TypeScript and directory name is only available in common JS modules. So Trevor probably could have figured this out in like three, if you understand that, hey, code doesn't just automatically explain itself. It can in AI minutes, maybe, or maybe not, right? I mean, it's just not all debugging is the same. Sometimes you just don't can really try and figure it out. But if you're having it start from ground zero every time you start a new agent chat and where to look. And sometimes an AI agent is just like a human. Sometimes it doesn't know where to look or it didn't consider one little thing. But after A while. After a long time, I finally figured it out. And so several times when I'm in a loop and I'm like, you're saying that you're updating this, but you're not actually updating this. I will literally grab this image, right? I have it sitting on my desktop somewhere. It's right here. Okay. I will grab this image and I will drag it in to the thing and I'll say, look at this image. This is not from this project, it's from another one. But can you. Is it possible that the problem you're dealing with is related to something like this? Right? And. And then it will say, oh, man, you're just. You're shooting yourself in the foot. So use these files. And by the way, they're. Oh, yeah, maybe I should look and consider. Maybe this is being done a different way because we're using different types of language or different module or different. Whatever it is, it'll go in and we'll figure that out. So I've found could be other types of files that you think of that you need. You can create anything MD file. You know, that is really, really helpful. Just ask it. Hey, are you sure you're using the right language for this? Right? Are you. Are you understanding what's going on? This is a big one. I will sometimes pause, you know, when I'm. Especially when I'm in one of those longer. When I'm in one of those longer chats, I will literally pause it, right? I can pause the agent and I'll say, why are you doing this? Right? Explain what you're doing. What's. What are you trying to do here? Explain the logic. Right? Another thing you want to be careful of, guys, is making sure that you never accidentally overwrite your database. I've had that happen a couple times. You know, knowledge base. You can make a knowledge base file, right, that acts or a prompt MD early on. Yeah, I haven't had it happen with V2, which is another thing I think they fixed. You can also have it make backups of your database, right? Just say, hey, make a backup of the database. And then let's work on the database and say, hey, ChatGPT, I want you to use this prompt file that I'm going to update regularly. And I want this. You know, you're going to be like, really changing things in the database. Just say, hey, make it. Make a quick backup of the database. Then let's work on something. Okay? APIs are. Are. Can be tough as, as Travis, you know, discovered. You know, really try to understand what it's trying to do. And. And by you trying to understand what it's doing, it's going to have to explain itself. And oftentimes as it explains itself, it says, oh, by explaining this to you, I. To be where you pull the prompt from. And we'll say, okay, great, I'll do that. And it will structure itself so that it realized that, you know, and then it fixes it. Right? So sometimes just ask it to do that. The debugging tool, right, there's. The screenshot. Debugging tool is amazing. Just screenshot stuff all the time. Drag it in, see what you can do. Okay, let's see here. AI Getting stuck in a loop pulls the prompt from the prompt MD file. It's not that hard. You create something. I don't think this is actually written out nicely here. I don't love this, really. I'm using some of these other tools to get out of loops, right? Make a plan. Explain what you're doing. Is there a simpler way? Check for duplicate conflicting code. Like this. This is how I'm getting it out of a loop. Okay, so, guys, I think that's really all we need to cover on that, but I hope that's helpful. I mean, these. These types of things are really. You point at it, right? If you feel like something needs to be updated all the time, just like I was talking about, are really important. Understanding how the system works and understanding how to talk to it, understanding some of its quirks, hopefully help you guys avoid a lot of trouble. Okay. All right, Brian, you asked a question here that says at the checkout earlier with this $39 thing, right. Instead of hard coding, we're point things different than the progress MD Absolutely. They're very, very different. So, yeah. And thank you, Trevor, for saying, generally speaking, one word is one token. That is about right. And anyway, so they're very, very different, these checkpoints. And this is an important thing to teach as well. So if for some reason, you know, the latest change that it did just broke everything instead of having to go and fix it, or if you're just like, oh, my gosh, this is like, so bad, or maybe you've been working for an hour and you can't get out of a loop. I've done this many, many times. I'm just gonna say, you know what, Whatever. I paid three bucks for that hour, and I don't care. Like, I'm gonna roll back and so you can roll back to a previous point, right? And say, okay, let's Go back to where I know this was working, right? And then we'll try a different approach. And then I'll tell it. I'll say, hey, I just rolled back to here. Just typing it in so that it's always like that. I have it pull a variable. I have it pull something from somewhere else that I can then manage in, in a more reasonable way. So that there. And we tried to do this and we got lost in this loop, right? Let's try to redo this again and build it in a way that won't get us stuck here, right? And if you've been very active in talking to it, you probably. And reading what it's saying, you'll probably understand what's going wrong. Like even those of you that are brand spanking new to coding, I was brand spanking new to coding and within a few hours of working with replit, I was telling it things to do because I was reading those things update automatically. Okay? So that's, that's a lot of what coding is, is understanding and like, oh, I understand what that's called now. I understand what this. You know, I just started learning, I just started learning different things and I could actually direct it much, much better. And that's just, you know, I'm a kayaker and one of the things, one of the best ways to get, you know, really good at kayaking really fast, any relationship building and systems, right? Get a whiteboard and say, okay, I've got this, I've got this. I get. This is simple. It's time and boat, right? If you want to get better at golfing, you know you're going to need more swings, right? If you want to get better at anything, just step up to the plate, right? Get, get at it and try it. All right, guys, so I'm going to pause for a second here. Any questions? Brian, I see your hand is raised. Was it related to that progress? Yeah, just one more question on that. So am I explicitly then at the end of the day, always telling it to update the progress MD or how's that relating? I do that. I tell it to update the progress MD almost every time I build a new feature now, right? Like right in the middle of chat, I'll say, hey, please update the progress. So it knows this. Okay, great. Now let's work on this thing. As long as I'm still working on something related and not moving to any, just draw it all out. That's called architecting, right? You're architecting this solution about how does it all fit chat, right? But I'm moving to new chats much, much faster now. Guys, I've just found that I get cleaner context, right? It's just, it's better right now if I'm building. Yeah, to clarify, the MD is basically like a save spot where you save all your work so that you can always go back to that if you screw up. No, no, no. Your progress MD does not allow you to roll back to anything. Okay? Your progress MD is a guide. Okay. You can click on the progress MD file, you can look at your own progress MD file. It says, okay, here's what we're trying to do. We're launching a mobile first fast loading, full featured platform for post API get together. How can we make it better? Right? All right. So every time you start a treats, right, it just kind of tells it what it's done, what it still needs to do and what maybe some, you know, next immediate, you know, priorities are. If you look, I'll open up my new chat. Use the command please reference the readme MD in progress enemy files and familiar API. I'll show you what it looks like. Okay, first of all, I got a billion attached assets here. Familiarize yourself with this, with the project. Right? This ensures continuity and prevents AI from losing track of where it's so many different things going on here and lots of test things. And yeah, I've done a lot of showing this is enough. But order progress md. I even have a backup of my progress md. Okay, but here are some of the latest changes I made when I did the blog, right? Comprehensive blog system implementation. I just had it update all those things. Homepage certification cards and enterprise page visual enhancement. Right. Navigation and site wide improvements. Enterprise enhancement. I did a lot of structured data here, by the way. When I did the structured data, I broke my Facebook implementation. My. What do you call them? Pixels. My pixel setup security enhancements. I did a complete security audit and implementation on April 6. Right? Here are things that I've done was I'm going to take one second here because I think this is very right and what's happened here, I have this progress MD file and then I've got my readme MD file. And here are all the different things that we're doing. Here's what we're building. Here are different things that I wanted to understand and why we're going to use. We're going to use Enterprise for this. We're going to use no reply for this. I'm giving it all kinds of instructions about how to build and communicate with users and with the system. No, the progress MD file has nothing to do with rolling back to a previous state. You can look at your deployments, right? So I've deployed a million times here, right? Many, many, many, many, many, many, many, many times. Keep it paginated. So it's important to point out, right here, Travis or somebody said that, that they used all of their, all of their AI credits yesterday. And that can happen. You can use them in a day if you just go, if you go for eight or nine hours, like you're going to use it, right? But when you click on this, it'll keep loading. I can kind of roll. Really, guys, I never. Here, this is, this is my general rule for this. Unless the site is in a good working state, I will not start a new chat because after you leave the, this existing chat, you can't roll back to previous chat states, okay? So rolling back is, is really what you want. You want to keep it open and say, okay, this is working well. Nothing's really broken. Everything's good, okay? Because I, I can't roll back, okay? If you, if you move on and you've broken something now, you're going to have to fix it, right? In the new chat when you get a little bit better. And this is not that hard. It's really not that hard. But you can connect to GitHub, okay? And I highly recommend doing this. This is, this will help you do what's called version control or versioning, right? You can, you have the ability to build it out and then you connect to GitHub, which is free at this level, right? And what it will do is, by the way, just, this is GitHub for anybody who might not be familiar with it, right? Those of you that have built software, it's like, this is basic 101, right? This little button right here, it pulls up your usage, right? And it shows how much money you've spent. And there's even AI tools inside of GitHub, right? But basically what it does is it's just a repository. It's a place where you keep your code, right? Your app. Everything that it is is right here. It's just a bunch of files with, with code, right? Like in that month, on all of the different things the replit does, one of the cool things that replit does for you guys is, you know, aside from the agent checkpoints, which you guys know about it. And by the way, I've spent $161, but that's because I've in this month, this is your ip, this is what you've built with AI. If you copy this code somewhere else, you can use it somewhere else. It's not just running inside of Replit, it's building real code that can be deployed anywhere. You set up GitHub, you just log in, create a free account, and then you'll come in here and you'll push to GitHub and say, all right, here's my latest state. The way that you do it, it's super, super easy, guys. All you do is, is say, please push the current project to get. You just tell it to do it and it'll run it for you. Okay. Now I don't think I've actually set it up with gen ipi. I've set it up with some other ones here. Yeah, no remote repositories configured, right? So I'd have to go to the settings view. Oh no, maybe I have. No, no, no, no. I haven't, I haven't done this one yet. Yeah, it's executing git add, but it's, it's going to fail doing this. That's because I've, you know, I teach this class and so I'm constant and that's fine. So I'll stop it. Right, but you have to connect to GitHub. But it's super, super, super, super easy to do it. Okay, I have a really good system for how I manage it on here. But one of the things that you can do, guys, when you connect to GitHub is you can then use other AI tools. For example, maybe you want to connect to cursome and I run several things out of it. But you look down here at all the different things it does. It's got your databases, your PostgreSQL storage and your database compute. It's got your scheduled deployments and your auto scale deployments, auto AI. Some of you maybe have used cursor AI. This is another system that's not cursor AI, it's cursor. This is another AI code editor, right? And it can do a lot of things. Now granted, this is probably more for developers that know what they're doing a little bit better. Now you are editing a natural language, but just know that it's way, way higher, right? If replit is level one, cursor's level three and developing on your own is level 10, right? So you're, you're, you know, steps a couple steps up from here. So you can try it though, right? Doesn't hurt to jump in and try it out. Jump in and see if you can do something, see how hard it is, play with it, learn. Okay. Now one thing that I really, really, really, really love. I've done this with another project. Not going to show it right now, but augment code. This is for the, the big time players, right? This, this, the SW event Bench verify. This is basically a measurement of how good at coding these different systems are. Augment auto scale deployments are awesome. We'll talk about those. But basically if you've ever tried to buy a ticket is amazing. This company just raised $220 million. They're based in the Bay Area and they're built for, you know, enterprise level huge companies that are, let's see here, where's, where's some of their. To a concert, you know, try to get into the Tabernacle Choir Christmas concert or try to buy a tailor, you know, this thing. Here you go. Like huge companies are using this to manage millions of lines of code across hundreds of developers and you know, big sites. I mean, you know, do Swift Ticket or BYU tickets or Utah Tickets or whatever you like and you're stuck or it doesn't load or it's super slow and you feel like the site's broken. That's because they don't have great audience. And what augment code does guys, is it cleans your code, it cleans it up in really, really, really good ways, right? It's SOC 2 compliant, right? You can do crazy stuff. You can store people's credit card information and whatever else you need to, to be able to do with that type of, with that level of security and plugs and so what you could do though. Now here's the problem. Augment code does not connect to replit but like I auto scale setup or their system just. There can be other problems but a lot of times they don't have auto scaling things set up and it's just kind of broken and then storing, right? Storing a video, storing a picture, storing a file, storing whatever. Right? Now all of this stuff can be really, really cheap, right? I mean cents per gigabyte per month or $1.50 per gigabyte per month or whatever, right? It's very, very cheap. You can see how it breaks down for your specific project and the types of things you have set up. Said it connects to GitHub so you can go from REPL.it into GitHub and then you have augment code pull, pull from GitHub does all its stuff and then you push to GitHub and then you pull from GitHub back into Replit, you can build new features, Augment code, won't build new features like Replit will. Right? But it will fix the code and it will. But again, the fact that I'm running dozens of different apps and deploying them and hosting them and all this stuff and I'm still only at $161 this month is just mind blowing, right? I mean you could, you could spend so much money doing this. We'll talk a little bit later about how to use if you want to use AWS or Azure or Firebase or other things. But again, the fact that I'm running dozens of different apps and, and deploying them and hosting them and all this stuff and I'm still only at $161 this month is just. You can hook up to other external systems that do these things if you want. Totally fine, right? Stand it up and make it, what's called harder. It'll make the code much, much stronger so that it can handle, you know, the hacking and security and all. But I personally love as a non developer not having to worry about that. It's all here, it's all done for me and so I don't have to worry about it. Now if I want to really get serious and scale to a billion dollar company, I'm going to have to graduate from Replit and I'm going to show you guys how to. Mind blowing, right? I mean you could, you could spend so much money doing this. We'll talk a little bit later about how to use, if you want to use AWS or Azure or Firebase or other things, different things that you want to work, work on. Now I'm not saying Replica can't do that. Replica can do a lot of those things quite well if you know how to just go through the questions. But, but ultimately version control do that too, right? And there's, there's ways to do that. But the you I legitimately believe you can build a hundred million dollars, you can hook up to other external systems that do these things if you want. Totally fine, right? Business inside of Replit. Hosting inside of Replit. I really do. You can have 50,000 concurrent users on your product in Replit. It can handle that. It's set up to handle that. So anyway, all right, Replit isn't paying me but I personally love as a non developer not having to worry about that. It's all here, it's all done for me and so I don't have to worry about it now if I want to really yet to say that but hopefully they are soon. They're going to give me at least some free credits, some fun deals. They're going to give me deals where I can go back to Evan's question. Version control is probably best done. Well, sorry I didn't get it before your class, but they're going to give me some deals where I can, I can, I can have like a premium sign up so you get like $25 of free credits when you first, you know, when you take my class, stuff like that. Anyway, so they're doing, they're doing some nice stuff for me as I say some nice things about them through GitHub. That's my personal opinion. Okay. All right, guys, we're cranking along here. We're, we're doing a lot. We're about, we're about half done with the day. I want to spend the next little bit here working on images. Okay. Get serious and scale to a billion dollar company. I'm gonna have to graduate from Replit. I'm going to show you guys how to, they'll put in their blogs and posts and things like that. So anyway, there's these, these prompts right here. These prompts are, are really, really awesome. I've used each of these many, many, many, many times as well. Right. And there's, there's ways to do that, but I legitimately believe you can build a hundred million dollar things. And again, I want you guys to realize that you're all smart people too. I guarantee you, you're all smarter than me in many things and use that intelligence to come up with good ways to talk. Right. It's just a communication thing. One of the things, one of the posts I've been. This is kind of rolling around in my head for LinkedIn is being good at AI, at directing an AI team and is almost no different than leading a real human team. You have to know how to communicate and lead. Right. Which means you as the CEO, business inside of Replit. Hosting inside of Replit. I really do. You can have 50,000 concurrent users on your product in Replit. It can handle that. It's set up to handle that. So anyway, all right, Replit isn't paying me, so I'm going to pull up, I'm going to pull up my app that I was just working on here and I'm going to update a few images that we create with AI. Now let me, let me ask a question before we, before we dive in. All of you are the CEO. You have to understand how it all fits together. So you've got this picture. And then you have to do your best to get that out into the prompts and into the overall plan. Yet to say that, but hopefully they are soon. They're going to give me at least some free credits, some fun deals. They're going to give me deals where I can and do the read me and the progress. Right. So that's why we went through that, that, that, that idea of hey, here's the whole plan where I'll sorry I didn't get it before your class, but they're going to give me some deals where I can, I can again. Now if you're really doing that well, you might take a couple hours having it. Ask you in here how many people have not created an image with AI. Okay. Jacob and Jessica. All right, that's cool. It's surprising to me when I even meet people that haven't done that right. It's so, so, so, so easy and so powerful. You can do unbelievable things. 50 questions going deeper into whatever. Hey, I want to go deeper into this. Hey, you didn't ask this question. Ask me a line of questioning about this thing. Right? And then build that whole plan and make it really, really nice. We made a 1500 word business plan. Make a 6000 word business plan. Whatever it is you need to do. You can have like a premium sign up so you get like 25 of free credits when you first, you know, when you take my class, like you're in charge. Right. So it's leadership and communication. So please create a plan to build this feature. Don't write any code yet. Just outline the steps and I will approve before proceeding. So as you. Anyway, so they're doing, they're doing some nice stuff for me as I say some nice things about them. And there's about, I don't know, 900 tools you can use to make an image with AI. Okay. There's so many out there and I encourage you to go experiment with them a lot as you understand how replit works better and better and you're getting a little bit more comfortable with it. Say that they'll put in their blogs and post things like that. So anyway, there's these, these prompts right here. These prompts are, are really okay. We want to, I want to build this entire. I want to build a new blog section. Right. I want the blog to, to operate, to pull from the database. I don't want you to hard them are free to try and then you pay for if you want to use them more. But we have one that's pretty dang good that you guys have all paid for and that's called chat GPT. ChatGPT is quite good at making images code the blog. Right. I want it all to pull from structured data so that we can update and we can edit and delete and all those functionalities. Right. And there's good reasons for all of those things. But I want to do this. I want to do this. I want to do this. I really want to focus on the reason I'm writing the blog is because really awesome. I've used each of these many times as well as many other. I want it to help with SEO. Right. And so please ensure that we're following best practices. A lot of you have seen images on your social media and all kinds of stuff. But let's go into our for that. Then you say please create a plan to build this feature. Don't write any code yet. Just outline the step things. And again, I want you guys to realize that you're all smart people too. I guarantee you you're all smarter than me in many things and use that intelligence to come up with good ways to talk. Right? It's just a communication thing. One of the things, one of the post ops I know and then I'll approve before proceeding. So then it will take into account everything that you just told it here my adventure retreats thing. And I'll just say create an image. Whoops. Create an image of a kayaker in a rapid. Okay, here's what I'm going to do for SEO or here's what I'm going to do. Here's what I'm going to do. Here's what I'm going to do. And it's. And it doesn't say anything about SEO. Then you can say I asked about SEO. What is your plan to make sure that everything is tagged right now? Again, I usually start with an okay prompt or a bad prompt and is set up right so that this will help with SEO. Anybody not know what SEO is? Anybody? And it's kind of rolling around in my head for LinkedIn is being good at AI at directing an AI team is almost no different than leading a real human team. And you know, I can obviously get much more descriptive and and possibly get better results. Yeah. I just want to make sure Jessica, it's good if you raise your hand. I'm just kidding. I'm sure you do. But search engine optimization, right? You want to make sure you show up in Google, right? Hi. In Google. And there's things you can do to structure Even a blog post to be done. Right. And so have it modify the plan. Okay, Hey, I like the plan now go ahead. Right, so instead of just letting it, you have to know how to communicate and lead, right? Which means you as the CEO, all of you are the CEO, you have to understand how it all fits together. So you've got this pixel now. ChatGPT has gotten way better at creating images lately. I didn't give it any instructions. I didn't say make this look like a cartoon or make this photo go right. Ask it for its plan first and then use your brain to figure out if it's, if it's accomplishing everything that you think it should. Limit AI scope or precision. Do not modify anything except the front end layout and the login page Sometimes, especially when you're in a long, long chat. Guys, we're gonna talk about long chats in a second here. Picture. And then you have to do your best to get that out into the prompts and into the overall plan. When you're in a long chat, it considers everything you've said in the realistic. Or make this look like this animation or make it look like, you know, whatever. I didn't give it any instructions. So chat every time you, you ask it anything. And so, and into the read me and the progress. Right? So that's why we went through that, that, that idea of, hey, here's the whole plan. If it just randomly looks, you know, six hours ago at some end. Now if you're really doing that well, you might take a couple hours having it ask you. It's just going to make those decisions on its own for me and, and hopefully it uses preferences out of my project and maybe preferences in my memory. In the memory and all the sets of things thing. And it's like, oh, I don't think we've done that yet. You know, you ask it to update a button and then it goes. But anyway, so this is one system that you can use to make images while that is loading 50 questions. Going deeper into whatever. Hey, I want to go deeper into this. Hey, you didn't ask this question. Goes off and starts doing that thing that you haven't talked about for six hours. Right? So start new chats. We're going to talk about why right now. Okay, remember guys, what how context works with AI. We're going to bounce over to here to Mid Journey. Mid Journey is another one. And there's an awesome. Ask me a line of questioning about this thing, right? And then build that whole plan and make it really really nice. We made a 1500 word business plan. Make a 6000 word business plan. Whatever it is you need to do, right? It considers everything it knows, at least in ChatGPT, it knows everything in the memory, it knows everything in the chat and it knows everything in the project itself. The way that replit is built is it will consider everything in the chat and everything in the project, right? At least web based interface. Now you used to have to use Discord when it needs to. It will determine, okay, I'm working on the. You're in charge, right? So it's leadership and communication. So please create a plan to build this feature. Don't write any code yet, just outline the steps and I will approve before proceeding. So as use a blog component or no, I'm working, let's say, let's say something else. I'm working on the menu, the nav menu. And so it then knows where to go. Look in here to say, okay, I need to find the menu and figure out where that is. And I know I need the index HTML here and I know I need these as you understand how Replit works better and better and you're getting a little bit more comfortable with it. Say okay, we want to, I want to build this entire, I want to build a new blog section, right? Things and I know I need this component, I know I need navigation, right? I need to look at all of these different pieces that and Discord can be a little bit confusing if you're not familiar with that platform. But in here you're basically just going to say, you don't say create. You don't. In ChatGPT you'll ask it, you know, create an image that does this or whatever. You'll kind of give it a little bit of an instruction here you're just going to, right? So it's going to pull up the right files and then it's going to, you know, work on it. And I say, you know, an image of a kayaker. If it's more complex than a standard setup, then you might have to ask it, hey, I think there's some other fancy things we're doing with the menu. You would know that because you built the project. But it's going to navigating a class file. I want the blog to operate, to pull from the database. I don't want you to hard build it in as simple and straightforward a way as possible so it knows where to go look for the content that it needs. So then it's going to code the blog, right? I want it all to pull from, you know, structured data and consider your context in your chat and your project. Now what if your chat gets super, super, super, super, super long, right? Number one, that confuses your chat and number two, it. And, and, and so that we can update and we can edit and delete and all those functionalities, right? And there's, there's good reasons for all of those things. But I want to do this. I want to eats up your usage because you're, you're wasting tokens with every single prompt, right? What's a token? I have rapids, cool colors, jungle. Yeah, jungle environment. Dslr. Right, let's go here. Open AI API pricing. Okay. A token. So you can see here, if you're using the open AI03 model, it's 10 with bokeh effects. This is just like photography stuff, right? So I'm saying make it dollars per million tokens, right? If you use an 04 Mini, it's $10 per million token. Do this. I want to do this. I really want to focus on, you know, the reason I'm writing the blog is because I want it to help with SEO, right? And so please ensure that we're following best practices for that. Then you say, please create a plan to build this feature. Don't write any code yet, just outline the step. So you got to choose which model you're using, right? I think, I think by default right now, if you ask it to set up an OpenAI, I think it sets it up with GPT4 mini. But, but I would always ask Devon. I'd ask replit. Hey, which model do you have set up with ops? I and I. And then I'll approve before proceeding. So then it will take into account everything that you just told it here. Like it was taken by a DSLR camera with a bokeh effect, which means it's going to have kind of a blurb and AI, right? But 30 cents per million tokens. That's probably what it's going to do. And GPT4 mini for background. And we'll go, we'll just tell it to go, right? So I could obviously give it more instructions. I can get pretty detailed in my prompt. But. But we'll start with that. By the way, guys, the vast majority of use cases is more than capable, right? The 4:1 is, is also getting, is also getting a little bit cheaper for a lot of things you want to do for more advanced stuff. So you don't have to know a huge amount about this other than the fact that you're getting charged for tokens. And a token, guys, is basically any, here's what I'm going to do for SEO or here's what I'm gonna do, here's what I'm gonna do, here's my do. And it's. And it doesn't say anything about SEO, then you can say maybe you don't know how to do that or how to write a good prompt for mid Journey. Can anyone think, hey, I asked about SEO. What is your plan to make sure that everything is tagged right and is set up right so that this will help with SEO? Anybody not know what SEO is? Anybody input or output? That happens, right? So if you ask it something and you give it a really long, you know, you upload a 90 page document, a legal document for it to review, well, that's a huge amount of token right now in the Chat GPT consumer version, right? If you're over here and you're asking it a question, it obfuscates all of this. It's just, hey, how they could make a good prompt for mid Journey. Travis, use Chat GPT to ask it to make one. There you go. That's exactly right. Okay, so let's bounce back. It. It's 20 bucks a month. Do whatever you want. It doesn't give you a limit right now. It might throttle you a little bit. It might throttle you a little bit if you do like crazy amounts of stuff. And some of, some of the, like, if you've ever been on ChatGPT, the free version before, paying for it. And you said, hey, you have to wait a few hours before you can ask another question. Well, that's because you've used ChatGPT and you know, sorry, we were doing it on the browser. It's right here. Yeah, I just want to make sure. Jessica, it's good if you raise your hand. Just kidding. I'm sure you do, but search engine optimization, right? You want to make sure you show up in Google right? High and up your token limit for that time, that time frame. Okay, so. And there's things you can do to structure even a blog post to be done. Right? And so bottom line is, guys, it costs money to run servers and to run GPUs that are in these massive warehouses with. Have it modify the plan. Okay. Hey, I like the plan now. Go ahead. Right, so instead of just letting it crazy air conditioning bills because they're so freaking hot because they're working so hard, as you guys ask. All right, here's my, here's my image of a kayaker. You know, I've asked go, right. Ask it for its plan first and then use your brain to figure it to make a lot of. I've asked AI to make a lot of whitewater kayaking images. Get to make all your studio Ghibli pictures. Right. And, and every time you ask it to do something, every time you say thank you, like we were just talking about in the beginning here, right. That I think Eric brought up. It's. It costs money. It costs money to do that. It costs money to somebody if it's, if it's accomplishing everything that you think it should. Okay, limit AI to do that. Right. So keep your chats shorter. And there's a few things wrong with this one, but this is one of the better ones. The paddle's a little bit too short. It looks like he's using his son's paddle and you're going to pay less. You're going to use your credit slower and you're going to get better results. You're going to have tighter results because you're going to say, all right, let's finish this feature. Update the progress MD and this whatever I try scope for precision. Do not modify anything except the front end layout and the login page. Sometimes, especially when you're in a long, long chat. Guys, we're going to talk about long chats in a second here. To keep my chats under a dollar, let me just put it that way, I'll go to $2. Sometimes it needs to be a lot wider, but I just isn't a good kayaker yet. Let's just say that. Just not there. But it's still pretty good, right? It's not bad, but go to 7. Sometimes I go to 14. I don't follow that religiously, but I try to go feature by feature by feature by feature. Now if I'm working on something that's maybe a little bit more complex, like the blog and I can't get something right, I'm trying to work on something. I'm probably going to keep it open unless I get stuck in a loop. If I get stuck in a loop, then I'm deaf. When you're in a long chat, it considers everything you've said in definitely starting a new chat, right. I'll still tell it to update and I'll say, hey, I need you to note that we're still chat every time you, you ask it anything and so I can loop. This isn't working right yet. Note that in the progress md. I'll tell it that so that it Know if it just randomly looks, you know, six hours ago at some nose, then I'll pull up, I'll say, hey, please read this and this and. And I want to focus in on this thing. And it's like, oh, I don't think we've done that yet. You know, you ask it to update a button and then it goes, let's do this. Please create a good prompt for me to get a good. A nice kayaking image from mid journey. Okay, so now I'm using AI again. We did this yesterday going back and forth between chat GPT and grok. Right? We're using AI to create a prompt for me. Absolutely, your highness. Here's a mid journey ready prompt that will give you a vivid. Who's often starts doing that thing that you haven't talked about for six hours, Right? So start new chats. We're going to talk about why right now. Okay, remember guys, what how context works with AI. Cinematic kayaking image. A kayaker plunging through raging whitewater rapids. Dynamic camera angle from river level, dramatic lighting, V6, whatever. Okay, right. It considers everything at. No, at least in chat GPT, it knows everything in the memory. It knows that we're kind of stuck on. I want you to take a fresh look at it. Are we doing it right or do we need to rip this out and start over? Right. Do we need to rip out this piece of this and start over? What's happening? What do you, what, what have you identified that we could. That we could probably fix here? Or what do you think the source of the problem is? Great. What else could the problem be? So I'm going to copy this and I'm going to paste it in. We're going to see if it works. Okay, so here's what. Everything in the chat and it knows everything in the project itself, the way that REPLIT is built. Right. What else could the problem be? I mean, ask it, challenge it, follow up. Okay. All right, guys, is there a simpler way to implement this feature? Consider efficiency and speed. Right? A lot of code is bad, right? A lot of AI code is bad. A lot of human code is bad. Right? And so you got to figure out, is there a better world? Is it will consider everything in the chat and everything in the project, right. At least when it needs to, it will determine. Okay, I'm working on the blog component. Or no, I'm working. Let's say, let's say something else. I'm working on the menu, the nav menu, what it created with my prompt. All right, that one's not bad. That's not bad for the right use case. This one looks like a miniature, like what do you call that? What's that style? You know what I mean? But it did create this bokeh effect where it's kind of a word background a little bit, but kind of cool. The way to do this, as a non developer, you might not know how to ask that question, you might not know how to judge it, but you do know how to ask it. Is there a simpler way to do this? Right. Your proposed plan is 2000 lines of code. Could this be done in 200 lines of code? That's a good question. Let me think about that. Well, maybe if we did this, we could do this. We can't get it to 200, but we. So now let's prompt or let's. Let's paste the prompt that Ch B gave us. Awesome. Get it to 800. Great. You just went from 2000 to 800. Now your app works faster. It has less logic that can break. It's smarter, right? At least theoretically, right there. There are certainly times where that might not be the case. But challenge it, ask it to think outside the box to do new things and, and especially, you know, the tr and so it then knows where to go. Look in here to say, okay, I need to find the map. Let's see, let's see the difference in the output here, right? It'll take another minute to do that while that's loading rivers of the world, right? I mean, go nuts on this. And I'm sure you do this all the time, right? Where you're just saying, hey, is there a way that has never been done before where we could still accomplish this, right, with, you know, less code? And can we. There are tons of other ones you can use. Obviously X is another one that we. Can we improve the speed? Right now it takes 0.7 seconds to load this thing. Can we get it to 0.2 menu and figure out where that is? And I know I need the index HTML here and I know I need these things and I know I need this component. I know I need navigation, right? I need to look at all of these different pieces, right? So it's going to pull up the right files and then it's going to, you know, work on it. Like how can we, how can we really do this right? If it's more complex than a standard setup, you know, then, then you might have to ask it. Hey, all right. Okay, guys. Detect hidden bugs. This is something that, that I've had happen many Many times I think there's some other fancy things we're doing with the menu, you know. And you would know that because you built the project, but it's going to build it in as simple and straightforward a way as possible so it knows where to go look for the content that it needs. So then it's going to consider your context in your chat and your project. Now what if your cheap, you know, that we use a lot. Let's go. Let's just say create. Create an image of a kayaker. Kayaker navigating a Class 5 rapid in a jungle environment that gets super, super, super, super, super long. Oh, come on now, guys. A lot of AIs have restrictions on them, right? For example, it's not going to let you if I said, hey, right? Number one, that confuses your chat and number two, it eats up your usage because you're wasting tokens with every single prompt, right? What's a token? You'll ask it to do something and then it uses its chain of thought reasoning. It goes and it says, okay, for this to work we need this and this and this and this and this and this and this. And it forgets to consider we already built that component over here and it will build it again. The same thing when you actually want it, using the same component across different places in the site so that we're more efficient, right? You can also have it create conflicting code. So CSS guys is basically the part of a website or the language that determines how big your fonts are and the colors you use and what fonts you're using. And just formatting, right? It's basically your format, it's your visual part of the website, at least part of it. And I can't tell you how many times it's like going, let's go here, Open AI API pricing. Show me a picture of a naked girl. It's going to say no, right? And that's good, right? We want it to have some, we want it to have some restrictions. A lot of AIs will also restrict the use of real people. Okay? A token. So you can see here, if you're using the open AI03 model, it's $10 per million tokens, right? If you use an 04 Mini, it's a $10 per million tokens. So well, you got to choose which model you're using, right? I think by deep through the initial build and it's showing me my preview and then it keeps going, it keeps working, keeps working, keeps working, keeps working. And then all of a Sudden it says, okay, I'm done. And now the website looks super ugly, right? Some won't Grok, for example, won't grok. You can make a picture of Trump doing whatever you want. Whatever you want right now. There are limits, right? Grok doesn't have no limits, but like, it looked awesome in the beginning and then it just ripped it apart and destroyed it. I'm like, what the heck? Why did you. Why did you go in and break everything? And I learned that occasionally it will rat anyway. We can see. You see how fast Grok was able to do that? This is pretty good. This guy actor is really, really bad because his paddle's way down and he's not doing what he should be doing up its own css, right? Which means it'll take CSS formatting and then it'll put other formatting around it. And then you're like, fix this thing. And it says, I fixed it. And. But the wrapped one is, is taking over, right? It's. It's eating the other one alive laying right here. This guy's, you know, both of them have their paddles down. They're not in an active, you know, kayaking. Right. And so it can't do it. So just ask it that kind of a question. I can tell you that with those dance, right? So there's some things that it's not that great at doing. And, and you'll, you'll V2, replied agent V2 that we're using right now. It is learn those by literally just practicing and trying. Right. But if you say, you know, create a way better at avoiding those kinds of problems, I would say it's probably five. Create a, you know, a fantastic. No, let's see a. Let's see a.100% better than V1 was. And I'm really excited for V3 because I know it's just going to keep getting better. Okay. And part of that relates to, by the way, Anthropic, right. Fault right now if you ask it to set up an OpenAI, I think it sets it up with GPT4O mini. But. But I would always ask Devin, I'd ask Replit, hey, which model do you have set up with OpenAI? Right. But $0.30 per million tokens, that's probably what it's going to do. And GPT 400 Mini for the vast majority of use cases is more than capable. Right. The. Wherever Anthropic was, I don't know where it is. Yeah. Anyway, when Heavenly Scene of Anthropic comes out, when Claude 3.9 or you know, it can remember 3.5 or 1 is, is also getting, is also getting a little bit cheaper for a lot of things you want to do for more advanced stuff. So you don't have to know a huge amount about this other than the fact that you're getting charged for tokens. And a token, guys, is basically any 3.7. So I'm assuming 3.9 is next. But maybe 4.0 or 5.0, whatever the, you know, violent weather. Okay. It's going to create something pretty. Every time the underlying models get smarter, then all the other things on top are going to get better. Dang cool. Like it can be so creative. You could also say here, let's go back to mid. So I'm excited about, about those things for what it's going to make possible. Okay, Ask for conflicting code. Hey, you know, if you're stuck in a loop again, Travis, journey here and check on our kayaking pictures. Okay, so here's what I got with my little boring prompt. They're conflicting code. Or is there. Are, are there two? Are there, are there, is there more than one part of the site that's trying to do this and that's why it's breaking? That can absolutely be the problem. Okay, and then ChatGPT gave me a much better prompt. Right, Take a step back. Analyze what you just wrote. Are there any potential issues? Right, what's, what's happening here? Right. Sometimes, guys, it will appear that the AI is stuck. Like this little thing right here where it says agent is waiting for your response. Or like, or maybe it's going. It says pause or whatever. Sometimes that disappears. Even if the AI, even if the agent is writing, you can refresh the page and it will not stop the agent. It won't mess it up. Okay, that's a good thing to know about. So just refresh it. The UI will refresh and your buttons will all be back. Okay. All right. This one might save you guys a ton, a ton of time. And I even copied the conversation I had with it and pasted it in here. I took a little screenshot of this. So sometimes replit will write in languages that are different from what it even expects, right? Or like you open a new agent and it's like, oh, I didn't realize I was writing in that language. I was going in circle. Input or output. That happens, right? So if you ask it something and you give it a really long, you know, you upload a 90 page document, a legal document, for a trans review. Well, that's a huge amount of tokens right now in the chat GPT consumer version, right. If you're over here and you're asking it a question, it obfuscates all of this. It's just, hey, it's 20 bucks a month. Do whatever you want. It doesn't give you a limit right now. It might throttle you a little bit. It might throttle you a little bit if you do like crazy amounts of stuff. And some of. Some of the. Some like. Like, if you've ever been on Chat GPT, the free. These images are better now. This one is kind of weird. It's like. It's like it's being taken. Yeah, that's actually kind of interesting. This. There's two kayaks about a stupid fave icon issue. I could not be. It took me twos here. Like, this is my point of view. I'm sitting in a kayak right here looking at this other kayaker coming, right? Flowers to get the fave icon to show up, right? The fave icon, for those of you that don't know, is this little icon up here on the browser, right? All these little tiny. These tiny little things. That's the fave icon. And it shows up in a lot of different browsers in a lot of different ways, right? I know I have my new cool AI browser somewhere where at me, right? It's like we're going down the rapid together at the same time. It's kind of cool. It didn't add version before paying for it. And you've said, hey, you have to wait a few hours before you can ask another question. Well, that's because you've used it. Maybe it's over here. Nope, I don't know where it is. Anyway, those favicons show up in. In a lot of different. In a lot of different places. But I couldn't get it to figure out. And what happened was this, right? The error is directory name is not defined. This is because we're using ESM modules with TypeScript and directory name is only up your token limit for that time. That time frame. Okay. So it's good to do that. That's kind of interesting. All right, here. I don't know what the heck is sticking available in common JS modules. So Trevor probably could have figured this out in like three. Bottom line is, guys, it costs money to run minutes. Maybe or maybe not, right? I mean, it's just not all debugging is the same. Sometimes you just don't have the front of this kayak. I've never seen A kayak with that maybe. Yeah, we'll just move on from that one. Okay, so. But where to look. And sometimes the AI agent is just like a human. Sometimes it doesn't know where to look or it didn't consider one little thing. But after a while, after a long. Anyway, guys, you can see, you know, this person's head is not even there. So that. That's fine. But you. Long time. I finally figured it out. And so several times when I'm in a loop and I'm like, you're saying that you're updating this, but you're not actually updating this. I will literally grab this image, right? I have it sitting on my desktop somewhere. It's right here. Okay. I will grab this image and I will drag it in to the thing and I'll say, look at this image. This is servers. And to run GPUs that are in these massive warehouses with crazy air conditioning bills because they're so freaking hot. Because they're working so hard. As you guys ask it to make all your studio Ghibli pictures, right? And you keep editing it, modifying it, say, hey, I want it to be a little bit different here. And I want it to be, you know, But I do like this. It's a splashing with water droplets frozen midair, right? It did that really well. In each of these images, you've got all kinds of water patterns and different things showing up all over the place here. Not from this project, it's from another one. But can you. Is it possible that the problem you're dealing with is related to something like this? Right? And. And then it will say every time you ask it to do something, every time you say thank you. Like we were just talking about in the beginning here, right? That I think Eric. Oh, yeah. Maybe I should look and consider. Maybe this is being done a different way because we're using different types of language or different module or different. Whatever it is, it'll go in and we'll figure that out. Okay, so I've found product. It's. It costs money. It costs money to do that. It costs money to somebody. That is really, really helpful. Just ask it. Hey, are you sure you're using to do that? Right? So keep your chats sure to write language for this, right? Are you. Are you understanding what's going on? Okay, this is a big one. I will sometimes pause, you know, when I'm. Especially when I'm in one of those longer. When I'm in one of those longer chats, I will literally Pause it. All right, this is so is Mid Journey just a lot better than Chat GTP coming up with pictures, just more options. Is that why you subscribe to it? Absolutely. I think Mid Journey creates the mo the highest quality images by far. Now there are some things, right, I can pause the agent and I'll say why are you doing this right, Explainer. And you're going to pay less. You're going to use your credit slower than what you're doing. What's what, what are you trying to do here? Explain the logic. Right? Another thing you want to be careful of guys and you're going to get better results. You're going to have tighter results because you're going to say, all right, let's finish this feature. Update the progress MD and this whenever I try making sure that you never accidentally overwrite your database. I've had that happen a couple times early on. Yeah, I haven't had it happen with V2, which is another thing. I think they fixed the Mid Journey really, really, really sucks. That and the chatgpt sucked at up until. You can also have it make backups of your database, right? Just say, hey, make a backup of the database and then let's work on the database too long ago, right? So let's go back to this, this image here. Actually, you know what? Sorry, we're using. If you know you're going to be like really changing things in the database, just say, hey, make it. Make a quick back to keep my chats under a dollar. Let me just put it that way. I'll go to $2 sometimes I'm using the browser here. Look at that, guys. That's pretty cool. Backup of the database. Then let's work on something. APIs can be tough, as Travis discovered. Really try to understand what it's trying to do and by go to seven. Sometimes I go to 14. I don't follow that religiously, but I try to go that's a pretty nice image in the right circumstance. However we want to do. If you're a creative person going to have a lot of fun, you trying to understand what it's doing, it's going to have to explain itself and off feature by feature by feature by feature. Now if I'm working on something that's maybe a little bit more complex like the blog and I can't get something right, I'm trying to work on something. I'm probably going with AI creating. Creating a lot of stuff. But let's, let's do this oftentimes as it explains itself, it Says, oh, by explaining this to you, I realize that you know, and then it fixes it. Right. So sometimes just ask it to do that. Let's do. The debugging tool. Right there's the screenshot. Debugging tool is amazing. Let's try something. Try something cool here. We'll. We'll say keep it open unless I get stuck in a loop. If I get stuck in a loop, then I'm def. Just screenshot stuff all the time. Drag it in, see what you can do. Okay. An image of a woman holding a cake. Definitely starting a new chat. Right. I'll still tell it to update and I'll say, hey, I need you to note that we're still stuck in a loop. This isn't working right yet. Note that in the progress md. I'll tell it that. So that it. Let's see here. Getting stuck in a loop. I don't think this is actually written out nicely here. I don't love this. No. Then I'll pull up. I'll say, hey, please read this and this and, and. And I want to focus in on this thing. Okay. All right. I like this one. Really. I'm using some of these other tools to get out of loops. Right. All right, let's see if we can save this image and save. Make a plan. Explain what you're doing. Is there a simpler way? Check for duplicate conflicting code like this. And then we're going to upload right here. Nope, I don't want it. Yeah, this is how I'm getting it out of a loop. Okay, so guys, I think that's really all we need to cover. You know what, that's fine. We'll just go into a new chat and click this, run that. But I hope that's helpful. I mean, these types of things are really that we're kind of stuck on. I want you to take a fresh look at it. Are we doing it right or do we need to rip this out and start over? Right. Do we need to rip out this piece of. Of this and start over? What. What are really important? Understanding how the system works and what's happening. What do you. What. What have you identified that we could. That we could probably fix here? Or what do you. And understanding how to talk to it, understanding some of its quirks, hopefully help you guys avoid a lot. I think the source of the problem is great. What else could the problem be? Of trouble. Okay. All right, Brian, you asked a question here that says that the check. And then let's say, let's see an image Of a, of an apple. Now let's see of a picture of a monade, right? What else could the problem be? I mean, ask it, challenge it, follow up. Okay. All right, guys, is there a simpler way to implement this feature? Consider efficiency and speed, right? A lot of code is bad, right? A lot of AI code is bad. A lot of human code is bad, right? And so you got to figure out, is there a better. Okay, so let's take this one right here. It's a little bit way to do this. As a non developer, you might not know how to ask that question. You might not know. This is kind of cool. This is a good challenge, right? Because it's got a few other things in it too. Let's go like this. And then we'll go like this. I'm going to say this. I'm going to say, please swap out the cake in the woman point. Things different than the progress MD Absolutely. They're very, very different. So yeah. And thank you, Trevor, for saying, generally speaking, one word is one token. That is about right. And anyway, so they're very, very different, these checkpoints. And this is an important thing to teach as well. So if for some reason you know, the latest change that it did just broke everything instead of having to go and fix it, or if you're just like, oh my gosh, this is like so bad. Or maybe you've been working for an hour and you can't get out of a loop. I've done this many, many times. I'm just gonna say, you know what, whatever. I paid three bucks for that hour and I don't care like I'm gonna how to judge it. But you do know how to ask it. Is there a simpler way to do this? Right. Roll back and so you can roll back to a previous point, right? And say, okay, let's go back to where I know that, you know your, your proposed plan is 2000 lines of code. Could this be done in 200 lines of code? This was working, right? And then we'll try a different approach. And then I'll tell it, I'll say, hey, I just rolled back to here. That's a good question. Let me think about that. Well, maybe if we did this, we could do this. Now we can't get it to 200, but we hand for the pitcher of lemonade and we tried to do this and we got lost in this loop. Right? Let's try to redo this again and build it in a way that won't get us stuck here. Right. And if you've been very active in talking to it. A picture. Hey, mid journey can't do this. Mid journey cannot do this chat gbd, can you probably in reading what it's saying, you'll probably understand what's going wrong. Like even those of you that only recently do this. But if you have a product, right, like, like you want to sell, get it to 800. Great. You just went from 2000 to 800. Now your app works faster. It has less logic that can break. It's smarter, right? At least theoretically, right? There are certainly times where that might not be the case, but challenge it, ask it to think outside the box to do new things. And especially, you know, the water bottles, right? You can literally take a picture of your water bottle that you want to sell and you're brand spanking new to coding. I. I was brand spanking new to coding and within a few hours of working with Replit, I was telling it things to do because I was reading and like, oh, I understand what that's called now. I understand what this. You know, I just started learning. I just started learning different things and I could actually direct it much, much better. And that's just, you know, I'm a kayaker and one of them get a picture of someone holding anything, right? And you can drop it in and it will of the world, right? I mean, go nuts on this. And I'm sure you do this all the time, right? Where you're just saying, hey, is there a way that has never been done. It will swap it out and it make it look pretty dang awesome. Let's see. Let's see what happens. I wonder what it's going to do with this hand. I told it to swap out the cake, so I want to see. I want to see what it does. The things one of the best ways to get, you know, really good at kayaking really fast is simple. It's time and boat, right? If you want to get better at golfing, you know, you're going to need more swings before where we could still accomplish this, right, with, you know, less code and can we? All right, if you want to get better at anything, just step up to the plate, right? Get. Get at it and try it. All right, guys, so can we improve the speed right now? It takes 0.7 seconds to load this thing. Can we get it to 0.2? Like, how can we. How can we really do this right here? We'll see. This is a fun test, by the way, guys. I hate waiting for AIs. Oh, I'm going To pause for a second here. Any questions, Brian, to see your hand is raised. Was it related to that progress? All right. Okay, guys, detect hidden bugs. This is. And so when I was building my business in three days, there's no way I'm sitting here waiting. I could say go and I bounce back over and I, I jump back into mid journey here and I'm saying, okay, let's go here and let's do this and let's do that. One of the things that you have to create for your business. Yeah, just one more question on that. So am I explicitly then something that, that I've had happen many, many times where at the end of the day, always telling it to update the progress md you'll ask it to do something and then it uses its chain of thought reasoning, right? And it goes and it says, okay, for this to work, we need this and this and this and this and this and this and this, right? Or at least the smart thing to do is to create a logo, right? And so you can try, you know, creating a logo with a lot of different things. You could try with, you could try with ChatGPT. In fact, I will try that. Yeah, I use this here. Oh man, I'm excited for this. It looks like it's changing her face. Just. No, it's doing pretty good. And it forgets to consider we already built that component over here and it will build it again. The same thing when you actually want it, using the same component across different places in the site so that we're more efficient. Right. You can also have it create conflicting code. So CSS guys is pretty dang good anyways, that's coming along. But creating logos, guys, you want to try different systems, right? So let's go into X here and let's say please create a logo for. And by the way, I'm giving it no ideas or prompts or whatever for this adventure. So that, for this is basically the part of a website or the language that, that determines, you know, how big your fonts are and the colors you use and you know, what fonts you're using and just formatting, right? It's basically your format. It's your visual part of, of the low states retreat company. Okay. Of the website. At least part of it. And I can't tell you how many times it's like going through the initial build and it's showing me my preview and then it keeps going, it keeps working, keeps working. So we're going to see what Grok comes up with. We will go to chat GPT, please create a logo for the Flow State Retreats company. Okay, I've got that started. We'll come to mid journey. Keeps working, keeps working. And then all of a sudden it says, okay, I'm done. And now the website looks super ugly. Like, it looked awesome in the beginning, and then it just ripped it apart and destroyed it. I'm like, what the heck? Why did you. And then, let's see here. Let's go Back to the ChatGPT here. Oh, my gosh, look how cool that is, guys. Look at that. So it took this image right here. Why did you go in and break everything? And I learned that occasionally it will wrap. Okay, this woman with the cake takes this picture of Lemonade and combines them together. App its own css, right? Which means it'll take CSS formatting and then it'll put. Now I would prefer that she's holding the picture instead of holding it like that. And it's kind of weird that there's a lemon there. So I could get other formatting around it. And then you're like, fix this thing. And it says, I fixed it, but the wrapped one is taking over. Right. It's eating the other one alive. Right. And so it can't do it. So just ask it that kind of a question. I can tell you that with the v2replit agent v2 that we're using right now, it is way better at avoiding those kinds of problems. I would say it's probably 500% better than V1 was. And I'm really excited for V3 because I know it's just going to keep getting better. And part of that relates to, by the way, Anthropic. Right. Or wherever anthropic was to do. I don't know where it is. Yeah. Anyway, when Anthropic comes out, when Claude 3.9 or, you know, it came over 3.5, give some updated instructions. But that's pretty dang amazing image composition. Right. It was able to take, you know, it mirrored this basically, and it just dropped it in. So I could say, get rid of the cloth. Make it. So she's holding the picture by the handle. And in fact, we'll do that and I'll go to another platform to have it do it. Get rid of the cloth and make it. So she is holding the picture. Picture by the handle. Also get rid of the extra 3.7. So I'm assuming 3.9 is next, but maybe 4.0 or 5.0, whatever the, you know, lemon that is outside of the Picture. Okay, so let's spot where you save all your work so that you can always go back to that if you screw up. No, no, no. Your progress MD does not allow you to roll back to anything. Okay. Your progress MD is a guide. Okay. You can click on the progress MD file. You can look at your own progress MD file. It says. Okay, here's what we're trying to do. We're launching a mobile first fast loading, full featured platform for Flow State. See how it does there. We'll bounce back over to ChatGPT here. All right. Flow State retreats. So it kind of knows what I'm. Every time the underlying models get smarter, then all the other things on top are going to get better. So I'm excited about, about those things for what it's going to make possible. Okay. Ask for conflicting code. Hey, if you're stuck in a loop again, Travis, it's treats. It just tells it what it's done, what it still needs to do, and what maybe some next immediate priorities are. If you look, I'll open up my Jenna API. I'll show you what it looks like. First of all, I got a billion attached assets here and so many different things going on here and lots of test things. Yeah, I've done a lot of showing. This isn't enough, but order Progress md. I even have a backup of my progress md. Okay, but here are some of the latest changes I made when I did the blog. Right. Comprehensive blog system implementation. They're conflicting code. Are there two. Are there, Are there, is there more than one part of the site? This presentation. I just had it update all those things. Homepage certification cards and enterprise trying to do this and that's why it's breaking. That can absolutely be the problem. Okay, take a step back. Analyze what you just wrote. Are there any potential issues? This page, Visual enhancement. Right. Navigation and site wide improvements. Right about. It's a decent logo. What do you guys think of this logo? Let's see a couple, A couple nods. Yeah, it's pretty good. That's decent. Yeah, I, I think, I think I'd wear a hat. Yeah, that's, that's kind of what I like to ask. What I. Would I put this on a hat? Would I throw this on a T shirt? Right? Would I, would I put my little logo right here? There's enterprise enhancement. I did a lot of structured data here, by the way. When I did the structured data, I broke my, my Facebook implementation, my. What do you call them? Pixels. My pixel setup, security enhancements. I did it Completely. Right? What's. What's happening here? Right, Reef. Sometimes, guys, there's other things you want to think about as well. Guys, you want to think about the favicon, right? What is your little image? Security audit implementation, you know, on April 6th, right. Here are things that I've done. It will appear that the AI is stuck like this little thing right here where it says agent is waiting for your response here, right? And so if you get it, maybe, maybe something that's really complex. Well, it's going to look terrible. Small, right? How is this going to look as an app? Do you like the color scheme? Do you like whatever right now? Or maybe it's going. It says pause or whatever. Sometimes that disappears. Even if the AI, even if the agent is writing, you can refresh the page and it will not stop the agent. It won't mess it up. Okay, that's a good thing to know about. So just refresh it. The UI will refresh and your buttons will. Right. And what's happened here? And so I have this progress MD file. And then I've got my readme MD file. And you know, here are all the different things that we're doing. Here's what we're building. Here are different things that I wanted to understand and why we're going to use. You know, we're going to use Enterprise for this. We're going to use no Reply for this. Right. I'm giving it all kinds of instructions about how to build. I'll be back. Okay. All right. This one might save you guys a ton, a ton of time. And I even copied the. The conversation I had with it and pasted it in here. I took a little screenshot of this. So sometimes communicate with users and with the system, right? So. So no, we're not going to spend this class talking about design of logos partially because it's so personal, right? If we were to ask every person. We have nine people on this, on this meeting right now. If we were to ask everybody their opinion, we would get nine opinions or maybe 10, right? Everyone's going to have their own opinion on color schemes and logos and what you like and what you don't like. You're the CEO, man. Just go for it. Do what rep will write in languages that are different from what it even expects, right? Or you open a new age you want. Okay, now I do want to talk a little bit more about the technical aspects of it. Right. I can come over here to mid journey. Now, what I'm going to say is this. I'm going to say, whoops, we're going to bounce back over here. Hold on, Agent. And it's like, oh, I didn't realize I was writing in that language. I was going in circles about a stupid fave icon issue I could not figure out. It took me two. The progress MD file has nothing to do with rolling back to a previous two hours to get the favicon to show up. Right? The favicon, for those of you that don't know, is this little icon up here on the browser. Right. Oh man, look what it did. Look how cool guys. Boom. Right now I don't like her teeth. They're a little bit off but for the state. Okay. You can look at your deployments, right? So I've deployed a million times here, right? Many, many, many, many, many, many, many, many times. We'll keep it's paginated so it'll keep. It'll keep loading. I can kind of roll back to here really little tiny. These tiny little things. That's the fave icon. And it shows up in a lot of different browsers and a lot of different guys. I never. This is. This is my general rule for this. Unless the site is in a good working state, I will not start a new chat. Most part this image is completely usable, right? I could tell it. I say make teeth look a little bit more natural and you know, make it so she's not smiling so big or whatever. Make her have a natural smile. Because after you leave this existing chat you can't roll back to previous chat states. Okay. So roaming ways, right? I know I have my new cool AI browser somewhere. Where is it? Maybe it's over here. Nope, I don't know where it is. Anyway, those fav icons show up in. In a lot of different. In a lot of different places back is is really what you want. You want to keep it open and say okay, this is working well. Nothing's really But I couldn't get it to figure out and what happened was this, right. The error is. But we completely changed an image and think about this as you're a broken everything's good because I can't roll back if you. If you move on and you've broken marketer right. If you're selling physical products so you want to show something and you don't have the picture the time now you're going to have to fix it right in the new chat when you get a little bit better. And this is not that hard. It's really not that hard. But you can directory name is not Defined. This is because we're using ESM modules with typescript and directory name is only and connect to GitHub. Okay? And I highly recommend doing this. This will help you do what's called version control or versioning, right? You can time or the money to go to a photo shoot AI in 30 seconds, man. It's available in common JS modules. So Trevor probably going to figure this out in like three crazy. It's crazy that we can do that. So now we're going to do is we're going to bounce back to this. We're going to say please create a mid journey minutes maybe or maybe not, right? I mean it's just not all debugging is the same. Sometimes you just don't know where to look and sometimes the AI agent is just like a human. Sometimes it doesn't know where to look or it didn't consider one little thing. But after a while, after a long prompt for a logo for flow state, long time, I finally figured it out. And so several times when I'm in a loop and then we're gonna give it a little bit more instructions, I want you to focus on the V shape of you have the ability to build it out and then you connect to GitHub, which is free at the, you know, at the, let's see, at the entry of a rapid. And I'm like, you're saying that you're updating this, but you're not actually updating this. I will literally grab this image, right? I have it sitting on my desktop somewhere. It's right here, okay. I will grab this image and I will drag it in to the thing and I'll say, look at this image. This is not from this project, it's from another one. But can you. Is it possible that the problem you're dealing, White water rapid, just be really clear, all right? So it's gonna understand what I'm talking about. I know what I want this level, right? And what it will do is by the way, just this is GitHub for anybody who might not be familiar with it, right? Those of you that have built software, it's like this basic 101, right? And there's even AI tools inside of GitHub, right? But basically what it does is it's just a repository. It's a place where you keep your code, right? Your app, everything that it is is right here. It's just a bunch of files with, with code, right? Like this is your ip, this is what you've built with AI, okay? And so if you copy this code somewhere else, you can use it somewhere else, right? It's not just running inside of replacing with is related to something like this, right? And. And then it will say, right? Whenever you're going into a rapid. Guys, you look for a V or the tongue, right? And. And that creates this cool kind of image, right? So let's see. Well. Oh, yeah, maybe I should look and consider. Maybe this is being done a different way because we're using differento here and we're going to go over to mid journey. We're going to paste it, we're going to tell it, do something. Okay. Different types of language or different module or different. Whatever it is, it'll go in and we'll figure that out. Okay, so I've found that that is really, really helpful. Just ask it. Hey, are you sure you're using the. It failed. Said style should be raw. Okay, so I'm going to copy this error building real code that can be deployed anywhere. So you set up GitHub, you just log in, create a free account, and then you'll come in here and you'll push to GitHub and say, all right, here's my latest state. I'm going to take it back. Sorry, I keep bouncing back and forth to that. Guys, sorry. And then I paste here the right language for this, right? Are you understanding what's going on? This is a big one. I will sometimes pause, you know, when I'm. Especially when I'm in one of those. Failed to submit. Style should be raw. Okay, sorry. I'm going to correct the prompt. So if you get an error with something that you get, just say, hey, this is longer. When I'm in one of those longer chats, I will literally pause it, right? I can pause the agent and I'll say, why are you doing this? Right, Explains the error. I got no big deal. It'll come back, it'll give you a new one and it'll fix it. It's the error. I got no big deal. It'll come back, it'll give you a new one and it'll fix it. Copy, paste, go. All right, so now we're creating a logo in mid journey. We'll see how it. We'll see how it does. Grant, please create. And the way that you do it, it's super, super easy, guys. All you do is say, please push the current project. And what you're doing. What's what? What are you trying to do here? Explain the logic. Right? Another thing you want to be careful of, guys. Copy a logo for this Flow State Retreat company. Those aren't logos. Those are pictures of kayakers past Go. All right. So now we're creating a logo in mid journey. We'll see how it. We'll see how it does. Grock, please create a make me a logo. I'm surprised by Gro. That is a really bad. Is making sure that you never accidentally overwrite your database. And I've had that happen a couple times early on. Yeah, I haven't had it happen with V2, which is another thing I think they fixed. You can also have it make backups every database. Right. Just say, hey, make a backup of the database. And then let's work on the database logo for this Flow State Retreat company. Those aren't misunderstanding. That's okay. No big deal. I'm a bad. I'm a bad director. All right. So we're working on this logo creation now. These are different. These are cool. What do you guys think about this one as compared to this one? Right. What do you think about if you know you're going to be like really changing things in the database? Just say, hey, make it. Make a quick. These guys go ahead and speak up. Right. So I can hear. Right. It's like premium. Yeah. It feels a little bit nicer. Right. This kind of looks like a whale tail. Logos. Those are pictures of kayakers. Make me a logo. So probably not going to go with that. The other one better. You like the other one better? I prefer the first one I got better too. Yeah, you just tell it to do it and it'll run it for you. Okay. Now I don't think I've actually set it up with Gen API. I've set it up with some other ones here. Yeah. No remote repositories configured right back up in the database. Then let's work on something. APIs can be tough as Travis discovered. Really try to understand what it's trying to do. And by so funny guys for what it is. Yeah. Like I said, we just heard 10 opinions from nine people. Okay. This creation, it's all. It doesn't matter. It doesn't matter. Guys. Get something awesome and roll forward. You can change the logo at any time. And you know what that means. You might get jackets. Two different jackets. You trying to understand what it's doing. It's going to have to explain itself. And off gets with two different logos for your closet. Cool. You got some cool swag. All right. That's awesome. So I'd have to go to the settings view. Oh, no. Maybe I have. No, no, no, no. I haven't. I haven't done this one yet. Super easy to take one of these logos and throw it on something. Now again, back to just a few, a few things. One of the, one of the things I don't love about this logo is that I think that the lines are a little weak. I just think I want them to be a little bolder. Right. So I might, I might do that. But again, that's just a, that's just a preference thing. These, these images actually would look awesome as fave icons. Okay. Right. Regardless, in times as it explains itself, it says, oh, by explaining this to you. I realize that. And then it fixes it. Right. So sometimes just ask it to do that. The debugging tool right there, the screenshot debugging tool is amazing. Yeah. It's executing Git add, but it's going to fail doing this and that's fine. So I'll stop it. Right. But you have to connect to GitHub. But it's super, super, super, super easy to do it. I have a really good system for how I manage it on here. But one of the things that you can do, guys, when you connect to GitHub is you can then whether or not you like the logo itself, they would actually look quite good as favicons and they're not bad logos themselves. Now, one thing I don't love is that it doesn't really use other AI tools. For example, maybe, maybe you want to connect to Cursor, just screenshot stuff all the time. Drag it in, see what you can do. It Flow state. Let me show you guys what my Flow State logo actually looks like. So I, I don't need to do this because I've already had it designed. But that's my logo, right? It's the V going into a rapid. Right. And so this is a great logo because it's very recognizable. It's very easy to, to, you know, it's a great fave icon. It's very. It works as a social profile image. It works great on, in different colors. You know, I, I have. This is an Adobe Illustrator file, so I can, you know, Infinite zoom is important. We'll talk about vectors in a second. But, but back to the, back to the logo itself, right? It. This is what I use, right? AI, right? Some of you maybe have used cursor AI. This is another system that's not cursor AI, it's cursor.com. this is another AI code editor, right? And it can do a Lot of things. Now granted, this is probably more for developers that know what they're doing a little bit better. Now you are editing a natural language, but just know that it's way, way higher, right? If replit is level one curse. Again, it's just an opinion thing. Now I'll say this. When I built my genaipi, I didn't really have the time to, you know, get everything exactly the way I wanted it. I just needed a quick logo that was going to work really, really nicely. And I had somebody actually text me yesterday that hasn't taken this class. She's a good friend of mine that owns a business, said, hey, I'm making a new logo where what do you recommend for AI? And I. And I sent him actually to design. Let's see here. Getting stuck in a loop dot com. Okay, design dot com. I actually found this by typing in first. Level three and developing on your own is level ten, right? So you're, you're, you know, steps a couple steps up from here. So you can try it though, right? Doesn't hurt to jump in and try it out. Jump in and see if you can do something. See how hard it is. Play with it. Learn. Okay, now one thing that I really, really, really, really love. I've done Free AI logo generator. Right? I did that and I found this. And it allows me to come in here and say, all right, flow state. Flow sorry. Flow state retreats. I don't think this is actually written out nicely here. I don't love this, really. I'm using some of these other tools to get out of loops, right? Make a plan, explain what you're doing. Is there a simpler way? Check for duplicate conflicting code. Like this, right? And I'll click create logos and you know, tell us about your business. Meditation. Wellness retreats, Adventure. This is how I'm getting it out of a loop. Okay, so guys, I think that's really all we need to cover this with another project. Not going to show it right now, but augment code this we're going to get rid of meditation because that's not as much leadership. Whatever. Okay, so I give it a few ideas. Flow state retreats, Wellness retreats, Adventure. Leadership. And I'm going to say water. I'm actually going to update that so that we get a little bit update there. And then I can come in here and then I can just choose, right? Like, I don't know, maybe I like this one. I don't love these water droplets quite as much. You know, maybe instead of water, I say you Know, river, like, I don't know, maybe I like this one. I don't love these water droplets quite as much, right. And so then I can come down here, like, see, it's giving me this nice complex thing. It's not bad. I mean, maybe, maybe it's good for a T shirt, right? It's kind of, it's kind of fun, you know, maybe instead of water I say, you know, river, right? And so then I can come down here, like, see, it's giving me this nice complex thing. Some of these are great, some of them are not. But this uses AI to come up with a bunch of different things and really to categorize. There's a huge amount of data that it processes in here. This one's pretty good because we've got a river, right? It's kind of a little bit abstract. I kind of like this angle. This will look really good. It'll be very recognizable up here. It works pretty dang good as a profile picture. So anyway, this is what I did guys, for Gen ipi, right? I just jumped in here and let me see, this is for the, the, the big time players, right? This, this, the SW Event Bench Verify. This is basically a measurement of how good at coding these different systems are. Augment is amazing. This company just raised $220 million. They're based in the Bay Area and they're built for, you know, enterprise level huge companies that are, let's see here. Where's, where's some of their, you know, this thing? Here you go. Like huge companies are using this to manage millions of lines. And let's see. Did I log in here? Did I log in on my other browser? Sorry, Let me find this. I want to actually pull up the, the Gen AIPI logo so you can see a code across hundreds of, of developers and you know, big sites. I mean, you know, here it is. Okay. Design.com. yeah, yeah, yeah. Here we go. All right, here's my logos. So, so this is what I did. I chose this one right here, right? So I can edit the logo. And you can edit so fast in here, right? Once you find something. The reason I like this guys is my idea that I worked with Chadis and what Augment code does, guys, is it cleans your code. It cleans it up in really, really, really good ways, right? It's SOC 2 compliant, right? You can do crazy stuff. You can store people's credit card information and whatever else you need to cover on that. But I hope that's helpful. I mean these, these types of things Are really. Are really important. Understanding how the system works and understanding how to talk to it, understanding some of its quirks, hopefully help you guys avoid a lot of trouble. Okay. All right, Brian, you asked a question here that says, are the checkpoint things different than the progress md? Absolutely. They're very, very different. So, yeah. And thank you, Trevor, for saying, you know, generally speaking, one word is to be able to do with that type of. With that level of security and plus GPT on, by the way, my idea was to have something that kind of was a brain, because I was focusing on iq. I was focusing a lot on the. On the concept of IQ at the time when I first started it. But I also wanted it to be kind of integrated into, like, computers, right? And so this looked to me like circuits and a brain kind of. Right. And so I really liked that combination of like, AI plus the brain. Like, that's just one token. That is about right. And anyway, so what we're trying to do, right? And then, you know, I clicked here and I said, well, I have a green color, so let's do this. Hey, anybody seen that logo before? Right? That's how long it took me to make my logo, right? And I had to pay for this. It cost me something like 30 bucks or whatever. So I paid for it and I've got it. And but what this did, one of the cool things that it let me do here is when I click download, I can. They're very, very different, these checkpoints. And this is an important thing to teach as well. So if for some reason, you know, the latest change that it did just broke everything instead of having to go and fix it, or if you're just like, oh, my gosh, this is like. And so what you could do, though. Now here's the problem. Augment code does not connect to replit. But like I said, I can choose the icon only. I can have a buffer or transparent background, things like that. I can come into the full logo. I can grab the EPS file, the vector file, which is super important. If you want to blow this up and be on a billboard or be on, you know, the new skin board, it said it connects to GitHub. So you can go from Replit into GitHub and then you have augment code pull from GitHub, does all its stuff, and then you push to GitHub and then you pull from GitHub back into Replit. You can build new features, BYU games, or, you know, where, you know, Times Square. You need a so bad. Or maybe you've Been working for an hour, and you can't get out of a loop. I've done this many, many super, super high resolution image, right? Because if I come in here into ChatGPT and I grab. Sorry, I come in here to ChatGPT and I grab this logo right here and I download it. Augment code. Won't build new features like Replit will, right? But it will fix the code and it will stand it up and make it, what's called, harder. It would make the code much, much stronger. Right? Let's open it up really quick and let's look at it. Let's see the quality of it. Now, is it going to work as a. Why is it not working? Come on, download so that it can handle, you know, the hacking and security and all times. I'm just gonna say, you know what, Whatever. I paid three bucks for that hour, and I don't care. Like, I'm under different things that you want to work, work on. Now, I'm not saying Replica can't do that. Replit can do a lot of those things, and JGBD doesn't work right now, so I can't even look at it. But I can tell you this. If I were to zoom in on this image a lot, man, roll back and so you can roll back to a previous point, right? And say, okay, let's go back to where I know this was working, right? And then we'll try a different approach. And then I'll tell it. I'll say, hey, I just rolled back to here. It's like, okay, we're gonna just refresh and see if it's just lost connection to a socket or something. What's going on here? All right, let's try. Let's try. Yeah, yeah, I. That worked. That worked quite well. If you know how to just go through the questions. But. But ultimately, version control, back to Evan's question. Version control is probably best done through GitHub. That's my personal opinion. Okay. But anyway, if I download this and I zoom in, you know, what do I get here? I get all this crazy pixelation. Now for a favicon, you know, now we're looking at the favicon. This is what you're going to see up here, right? Does that look good? I can't really tell. All right, guys, we're cranking along here. We're doing a lot. We're about. We're about half done. What it is, it's just not very bold right here. I can tell. I can see Facebook. I could see Amazon and we tried to do this and we got lost in this loop. Right. Let's try to redo this again. Like, I know what these things are. I even can see my Gen API right here. Right. I can see my green logo and my little green thing. And if I recognize that, I recognize it now, I'll recognize this over time. Not everybody has perfect favicons, but anyway, the problem with this is it's not. And build it in a way that won't get us stuck here. Right. And if you've been very active in talking to it with the day, I want to spend the next little bit here working on images. Okay? So I'm going to pull up. I'm going to pull up my app I was just working on here, and I'm going to update a few images that we create with AI. Now let me, let me ask a question before we, before we dive in. Not an EPS. I don't have that. I can't even ask ChatGPT to create an EPS in here. How many people have not created an image with AI in mid journey? I can't do the same thing either. I can't say, hey, please generate these, you know, in this. Now what I can do is I can say, you know what I love? I love this one. Let's vary it. And I can come up with ideas and I can, I can actually upscale it to be, to be pretty good. To be, like, good for, you know, publishing. All right, guys, we're cranking along here. We're, we're doing a lot. We're about, we're about half done. You probably, in reading what it's saying, you'll probably understand what's going wrong. Like even those of you that social media or whatever, maybe even, maybe even on some print, you can, you can upscale it a few times so that it looks good, but it still doesn't have the infinite scalability that a proper logo does have. Right. Where you can zoom in forever and it will never, ever, ever pixelate. All right, so once you with the day, I want to spend the next little bit here working on images. Okay. Okay. Jacob and Jessica. All right, that's cool. It's surprising to me when I even meet people that haven't done that. Right? It's, it's, it's so, so, so, so easy and so powerful. You can do unbelievable things. So I'm going to pull up, I'm going to pull up my app I was just working on here, and I'm going to update a few images that we create with AI now let me, let me ask a question before we, before we dive in. We've got an image guys, you know, in the right format. And let's just, let's just say we're going to use this flow state one, right? You can jump into Photoshop or you can jump onto canva, Canva or whatever tool you happen to be familiar with. You create a little thing here and you just drag your image in and there's about, I don't know, 900 tools you can use to make an image with AI okay. There's so many out there and I encourage you to go experiment with them. A lot of them are free to try and then you pay for if you want to use them more. But we have one that's pretty dang good that you guys have all paid for and that's called ChatGPT. ChatGPT is quite good at making images. They're brand spanking new to coding. I was brand spanking new to coding and within a few hours of working with Replit, I was telling it things to do because I was reading and like, oh, I understand what that's called now. I understand what this. You know, I just started learning. I just started learning different things. A lot of you have seen images on your social media and all kinds of stuff, but now let's go into our Click Enter and then you could just export this as a PNG and now you've. And I could actually direct it much, much better. And that's just, you know, I'm a kayaker and one of got a logo, right? Flow state retreats logo. Okay. Then I can just come into Replit and I can upload it. Now what I'm actually going to do is I'm going to give it one of my good ones, right? I'm going to give it one of my. In here. How many people have not created an image with AI? Okay. Jacob and Jessica. All right, that's cool. It's surprising to me when I even meet people that haven't done that. The things. One of the best ways to get, you know, really good at kayaking, really fast. SL logos. Come on. I've got this little button here from Zoom that's in my way. No, that's not what I wanted. I want this. There we go. All right. And I want. Okay, come on, come on, come on, come on, come on. Drag this over to the next in here. Okay, so let's give it a PNG is simple. It's time and boat, right? If you want to get better at Golfing, you know, you're going to need more swings. I think this is what I want. Black icon, FSL icon. That's kind of a bluish one. Let's use the black one. Okay, so please use this file as the logo and fave icon. Please set it up throughout the site and as the favicon. Okay, so now I'm installing a picture, I'm installing a logo. So theoretically it should put it right here. I still need to work on some spacing things right here, but that's how easy it is, guys, to set up. To set up a picture, add a picture, add an image, add a logo. You can tell it, hey, you know what? I want the logo to show up in the footer. I want a logo, it to show up wherever. It's probably naturally going to put it in the footer. And it's probably. And it's going to put it probably in the menu. I didn't actually give it instructions. I just said, please set it up throughout the site. And it's going to say, okay, well, let me consider this. What does he mean by that? Well, most websites put a logo up in the top left and that's the, you know, it's usually a button too. And that button, right. If you want to get better at anything, just step up to the plate, right? Get, get at it and try it. All right, guys, so I'm going to pause for a second here. Any questions? Brian, I see your hand is raised. Was it related to that progress? Yeah, just one more question on that. So am I explicitly then at the end of the day, always telling you to update the progress. MD takes you to the home page, to the index page, and we probably want to show it in a couple other places. And if we're using this, maybe we need to adjust this so it goes through all of that thinking and it's going to make its own judgments and, and it's going to bring it to life for you. So dropping a, you know, dropping a logo in, dropping an image in, how's that relating? I do that. I tell it to update the progress. MD almost every time I build a new feature. Now, right, like right in the middle of chat, I'll say, hey, please update the progress. Is that simple, guys, okay? You go in, you create the logo, you create the image using one of these amazing tools. Get your image and then drop it over. Brian, question. Can you like, save, for example? Let's say you didn't like the one, so it knows this. Okay, great. Now let's work on this thing. As long as I'm still working on something related and not moving to a new. From the design.com. can you like say generate something with chat GPT like chat. Right. But I'm moving to new chats much, much faster now. Guys, I've just found that I get cleaner that more. Can you pull it into the design.com and have it create that the EPS or the vector file? EPS file? Yeah, yeah, you can't. But there are systems out there that can do that. So there, there are some you may be able to Google, you know, like PNG to vector file. Now if you can't context, right. It's just, it's, it's, it's better right now if I'm building. Yeah. To clarify, the MD is basically like a save spot where you save all your work so that you can always go back to that if you screw up now, find that it's actually kind of a hard task. But there are some people that are really good at it. Does anybody know how they would do something that they don't know how to do that AI can't do, but they want it done quickly and cheaply? What would you guys do? You need to get something done. You don't know how to do it, but you still want it done cheaply. No, no. Your progress MD does not allow you to roll back to anything. Okay. Your progress MD is a guide. Okay. You can click on the progress MD file. You can look at your own progress MD file. It says, okay, here's what we're trying to do. We're launching a mobile first fast loading, full featured platform for post. You guys are all thinking like entrepreneurs. That's exactly what I would do and I have done many, many times. You just bounce over to Fiverr, right? And you just say, all right, here we go. We need, you know, PNG to vector, right? And I search for that and treats, right? It just kind of tells it what it's done, what it still needs to do and what maybe some, you know, next immediate, you know, priorities are if you look, I'll open up my gen and There you go. 10 bucks, 15 bucks, 20 bucks, 5 bucks. You know, go for it, right? What we have to do is says I will retrace logo. You actually have to go in and like trace it out. And it's just, it's a little bit time consuming. It might take you. If you don't know how to do it, it's going to take you forever and you're not going to do it. Well, if you do know how to do it. It's still going to take you 30 minutes maybe for your loading API. I'll show you what it looks like. Okay, first of all, I got a billion attached assets here and so many different things going on here and lots of test things. And yeah, I've done a lot of. Sorry, let's. This is enough. But order progress md. I even have a backup of my progress md. Okay, but here are some of the latest changes I made when I did the blog. Right? Comprehensive blog system implement API. I'll show you what it looks like. Okay, first of all, I got a billion attached assets here, so pay someone five bucks and get it done. And you know, forget about it, right? It's off your plate. Move back there. And so many different things going on here and lots of test things. And yeah, I've done a lot of. Sorry, let's listen up. Order, progress md. I even have a backup of my progress md. Okay, get back to the project and get going. Right? And so when I'm trying to launch something fast, and that's what we're talking about here, fast iteration, fast launch, fast development of a product internally, whatever it is, get it done fast. Spend $90 homepage certification cards and enterprise. Who freaking cares, man? Like, get it done. Get it done fast. Launch it, get it out into the public, get it in front of the team, get it in front of whoever you need to get it in front of and let it go. That's what we're trying to. That's what I'm trying to help you just kind of redo here. So I'm still black on this image here. So I should probably update that. This page. Visual enhancement. Right. Navigation and site wide improvements. Right. Enterprise enhancement. I did a lot of structured data here, by the way. When I did the structured data, I broke my Facebook implementation at some point. Let's find an image of me. I wonder if I have it. John profile. I don't know if I have a picture just saved on here. Here's a few. Okay, here we go. This is good. It's a good one. Let's grab this and just drag it right here. And then I'm gonna go like this. What do you call them? Pixels. My pixel setup. Security enhancements. I did a complete security audit and implementation on April 6th. Right here are things that I've done. Element selector right here. Please replace the Right. And what's happened here? And so I have this progress MD file and then I've got my readme MD file. And here are all the different things that we're doing. Here's what we're collected image with this one. Perfect. Yeah, go ahead. So say that. I mean after this changes the picture, right? Yeah. Do you see the left houses John Cheney founder. Then it's got the wording underneath the content. So if I just wanted to like not have to go and do it on the left side and just. Is there a way that I could just go into that and just change the language myself? I think it's just to the database. But can I just see you do that? Building here are different things that I want it to understand and why we're going to use. You know, we're going to use Enterprise in a second. Yeah, good question. That's a great question, Evan. And actually brings up a few things rise for this. We're going to use no reply for this. Right. I'm giving it all kinds of instructions about how to build. So number one, this right here is not actually driven by the database then communicate with users and with the system. Right. So. So no, this is driven. It's called. It's called what? It's hard coded. Right. And so we'd have to come over here and we'd have to find the actual page. Right. Which might be index HTML. Nope, it's not. It is navigation form. The progress MD file has nothing to do with rolling back to a previous forms UI sections hero section experience state. Okay. You can look at your deployments. Right. So I've deployed a million times. Where? Where is it Here. Outcomes testimonials Y flow here. Right? Many, many, many, many, many, many, many, many times. Keep it's paginated. So. So I have to, I have to find it right now. I could also do something where I could ask the agent, I could select it and say what is this? Where is this file? Right. And then it could tell me. Okay, yeah, go to client component. It'll keep loading. I can kind of roll back to here really. Sections about. Yeah, about founder section. Here it is. John's profile picture. John is world class adventure. Guys. I never here this is. This is my general rule for this unless you're a successful entrepreneur. Right. This is what it's saying here. Right. And great. So I can come in here in the code Now I'm actually editing code. Right. And I can say, you know, editing is disabled while the agent is working. That's true. I have to wait till the agent as the site is in a good working state. I will not start a new chat. It's done. But tell Me, how would you make the change? I would just go to the agent. Because after you leave the. This existing chat, I would say, I'd say right here, update this. You can't roll back to previous chat states. Okay. So rolling back is, is really what you want. You want to keep it open and say, okay, this is working well. Nothing's really to say. Yeah. And then paste whatever I've changed it to. Right? So, so I mean that, that's literally all I'm going to do. And so what I could do with this, I could come in here and I could say, copy, come over here to chat. GPT go into my adventure retreats. It knows everything. ChatGPT knows everything about me. And I'd say, broken, everything's good. Okay. Because I can't roll back. Okay, if you, if you move on and you've broken, rewrite this so that it's the same length something. Now you're going to have to fix it right in the new chat. But more inspiring and awesome, right? So it then rewrites it and when you get a little bit better and this is not that hard. It's really not that hard, but you can wait for it. Copy. Okay. Then I could bounce back over to here like this and connect to GitHub. Okay. And I highly. I select. I'm going to give it the whole section, update the text to say this. Recommend doing this. This is. This will help you do what's called version control or versioning, right? You can. Boom. So now again, I could have gone in and just pasted it in here, but I would have had to do it one paragraph and you have the ability to build it out and then you connect to GitHub, which is free at a time. Don't mess up my paragraph and all these different things here, right? It's giving this level, right? And what it will do is. And by the way, just, this is GitHub for anybody who might not be familiar with it, right? Those of you that have built software, it's like, this is basic 101, right? Get instructions here so I can't mess up the HTML or we're in a problem, we're going to have a problem. So I would have to go through now you can see, see it, you're watching it, you know, do that live. It's just. And there's even AI tools inside of GitHub, right? But basically what it does is it's just a repository, just updating that text live. So now it is, it has updated that to what we Gave it. So there's two fast ways. It's a place where you keep your code, right? Your app. Everything that it is is right here to figure that out. Right? You figure out where's the file and then go edit it manually or click it, tell it what to do. And I used AI to make that update in seconds. Right. It's just a bunch of files with code. Right. Like, if I didn't love what it gave me here, I could change a few words and then, you know, move it over there. This is your ip. This is what you've built with AI. Okay. And I mean. But faster than I could have typed that or made the changes myself, I was able to go, if you copy this code somewhere else, you can use it somewhere else. Right. It's not just running inside of Replit, it's copy and paste. That in a way that was, you know, a lot more efficient. Okay. So we're starting to make improvements on this right now. Okay. It actually looks like John Cheney. This is written a little bit better game apply for the next retreat, but building real code that can be deployed anywhere. So you set up GitHub. It happens quick. Okay. Application. Oh, I have to create a user account for this. I don't want to create a user account before I do that. Maybe I do, but let's try it. Let's say John Cheney and we'll say password is my password. Okay. And login. Oh, login. Create a free account. Okay. And then you'll come in here and you'll push to GitHub and say, all right, here's my latest state. Login failed. Okay, let's. Let's move forward. And the way that you do it, it's super, super easy, guys. All you do is say, my login failed when I was trying to apply to the retreat. Okay. Is it because. Is it because you were. Please push the current project run log instead of register. Possibly. Yeah. So, okay, great. Maybe it's going to find that out, but I didn't recognize that. Right. I do get you just tell it to do it and it'll run it for you. Okay. Now, I don't think I've actually set it up with Jenny API. I've set it up with some. I wasn't even thinking. I was just kind of typing, let's see if it works just for fun. Don Cheney. John Cheney. Don cheney7gmail.com here's this. And let's make the password password and create account other ones here. Yeah. No remote repositories configured right okay, well, it did something here. On authentic authentication, I'd have to go to the settings view. Oh no, maybe I have. No, no, no, no, I haven't, I haven't done this one yet. We're going to fail. Let me check the backend authentication system to fix the problem. So even that didn't work. It should have done something and forwarded me somewhere and it, it didn't work. So guys, this does not mean that Replit did it wrong. I want that to be really clear. It just hasn't built this part of the system yet. Yeah, it's executing Git add, but it's, it's going to fail doing this and that's fine. So I'll stop it. Right, but you have to connect to GitHub. But it's super, super, super, super. Okay, this is where you, this is where you start to turn it from a. Easy to do it. Okay, I have a really good system for how I manage it on here. But please push the current project from an idea into a real product. And it just takes talking to it. Right? I'm going to update an image. I'm going to update this. I'm going to use my things to get it out of a, out of a loop. Right, but now we just have to sit here and wait. One of the things that you can do, guys, when you connect to GitHub is you can then. Right? Which sucks to do in a class. So I'm not going to sit here and watch it do this, but use other AI tools. For example, maybe, maybe you want to connect to Cursor. But we're going to go, you know, so we can back, you know, go back into some more image creation stuff and maybe even some video creation stuff. That's kind of fun. But yeah, it's trying to figure out what's going on. Hey, we need to make sure that we have better error handling and we're fixing the login issue. That's great. Okay, we're checking the routes. It's checking all kinds of different things. Things AI. Right. Some of you maybe have used cursor AI. This is another system that is not cursor AI. It's cursor.com. hey, it made my. It saved my password for this site. So we'll do that and then we'll try it again in a minute. Now another thing I can do, this is another AI code editor, right. And it can do a lot of things. Now granted, this is probably to see if it created a log or a login is I can Click on See my data. Wait for it. Users. Hey, more for developers that know what they're doing a little bit better now. You are editing a natural language, but did create it. John Cheney. My password is hashed automatically, right? That means the system it won't allow you as the user to see other people's passwords. You want that? That's great. Just know that it's way, way higher, right? If replit is level 1 curse v1 of replit did not do that. It just showed plaintext passwords. So I had to. There's level three and developing on your own is level 10. Right. So you're, you're tell it. Hey, set up, set this up so that we don't see that role user. I can see all these different things here, right? You know, steps a couple steps up from here so you can try it. So I can. I'm starting to learn. But now I know hey, a user was actually created correctly. So though. Right. Doesn't hurt to jump in and try it out. Jump in and see if you can do something. See how hard it is Play with being happened correctly. When we did the registration, but it didn't forward me to the next page. Now maybe the next page where it's supposed to go doesn't even exist yet. It probably doesn't even exist yet. Like I haven't even told it what's what should it learn. Okay, now one thing that I really, really, really, really love. I've done this with another project. Not going to show it right now, but augment code. This could happen. Hey, when I click, when I click register, I want it to bring me to a dashboard, right Where I can then apply for retreats or I can see past retreats or I can see other teammate information, other people or classmates. Whatever. Whatever I wanted to do. I just have to tell it. Okay, so now it's had a couple problems here. All right. Authentication system is now working as the API response shows logged in. This is for the big time players, right? This, this the SW event bench verify. This is basically a measurement of user. Let's also enhance the registration feedback for a better user experience. So now it's going to keep doing stuff. This is one of those times where I might be like okay, let's just, let's pause for a second now at this point I'm probably going to just let it work on things because it's going to start finding a lot of stuff that isn't working. It isn't connected yet. Okay. I still haven't fixed how Good at coding. These different systems are. Augment is amazing. This company just raised $220 million. They're based in the Bay Area and they're built for, you know, enterprise level huge companies that are, let's see here. Where's, where's some of their. My visual issue here, but I'll do that later. I'll do that at the right time. I wonder if that visual issue fixes itself. I make it wider. Yeah, it doesn't. Yeah, those are just overlapping. That's just bad. It's just bad code. So now it's fixing some, you know, tags and some other things and. All right, there's a DOM nesting issue, right. You don't need to know what that means. Doesn't matter. Right? You've got, you know, this thing. Here you go. Like huge companies are using this to manage millions of lines of code across hundreds of developers and, you know, big sites. I mean, you know, there's different problems here. All right, I'm analyzing and making queer progress by systematically addressing issues in the code base. I've resolved DOM structure problems with the HERO section by properly implementing scroll behavior without invalid nesting. This looks great. Let me run some tests, make sure it's running properly now. I mean, and what Augment code does, guys, is it cleans your code. It cleans it up in really, really, really good ways, right? It's SOC 2 compliant, right? You can do. Those of you that have paid developers, guys, they do this, they do each of these things as they go. And you're paying them thousands of dollars a day maybe to do this type of stuff, right? And now it can just work through it fast and get to a finished product. And then if you need to, you know, pay some. By the way, guys, I, I, I want crazy stuff. You can store people's credit card information and whatever else you need to, to be able to do with that type of, with that level of security and make it clear, in one of my companies, I still have four developers. That's awesome. I don't think developers. And so what you could do, though. Now here's the problem. Augment code does not connect to Replit, but like I said, it connects to GitHub. So you can go from Replit into GitHub. They're bad. I think they're awesome and they're super useful. And they're so good at using AI, just like Trevor, right? Developers can be very, very, very good. But I raised $3 million to be able to pay for those people, right? If you don't have $3 million. Well, you don't have to do that now for gen AI PI. And then you have augment code pull from GitHub, does all its stuff and then you push to GitHub and then you pull from GitHub back into Replit and you can build new features. I haven't hired one developer. This is, this is all John, right? This is all me. And that's changing things, right? But we, we're even changing things at, at my company, Velocity Access, because of all this AI stuff. And we're gonna talk about that in about 15 minutes. We'll jump into that conversation. Okay, but, but guys, the way what you do from here, Augment code won't build new features like replit will, right? But it will fix the code and it will stand it up and make it, what's called, harder. It would make the code much, much stronger here all the way through from this point on, right? Is you just say, okay, what's next? To go to the full application. Okay, hey, this worked. Now. So I must be logged in. Yeah, I'm logged in because of this. Now here's a problem. I don't see any login profile button or state of login. Like I can't log out. So let's. So that it can handle, you know, the hacking and security and all the different things that you want to work, work on. Now I'm not saying Replica can't do that. Replica can do a lot of those things. Let's fix that. Right? I don't see a login or a logout button anywhere. Can we add that to the main menu? Quite well, if you know how to just go through the questions. But, but ultimately version control also, we're just going back to Evan's question. Version control is probably best done. Still having some formatting issues with the menu through GitHub. That's my personal opinion. Okay. The logo and flow state retreats title is still. All right, guys, we're cranking along here. We're. We're doing a lot. We're about, we're about half done over mapping with menu items. Okay. With the day, I want to spend the next little bit here working on images. Okay. So I'm going to pull up. I'm going to pull up my app that I was just working on here and I'm gonna up. So I gave it two tasks that are kind of closely related. It's going to be working on a menu. It's going to build a login log out feature for me and some buttons that do those things. It's going to use some APIs, you know, to tell the system, hey, login, log out with the authentication system. But yeah, and while that's going, I'm going to look at this. Okay. Hey, you know, do I need a few images that we create with A.I. now, let me, let me ask a question before we, before we dive in here. How many people have not created an image with AI? I like this text. Okay, great. Hey, I actually kind of love it, right? Current occupation? CEO. Why are you interested? Previous adventure experience, if any? What upcoming dates are you interested in? Okay, there's no okay, Jacob and Jessica. All right, that's cool. It's surprising to me when I even meet people that haven't done that, right? It's, it's so, so, so, so easy and so powerful. You can do unbelievable things. There's no data here, so this is interesting. Why is there no data here? Is it broken? I want to ask the gallery and Trevor doesn't get the answer because I bet he can figure it out. Or Travis. I'm going to say no, Trevor or Travis on this question. Why. Why is it not showing any dates right here, even though it's asking me? Brian, go ahead. And there's about, I don't know, 900 tools you can use to make an image with AI. Okay. There's so many out there, and I encourage you to go experiment with them. A lot of them are free to try and then you pay for if you want to use them more. But we have one that's pretty dang good that you guys have all paid for, and that's called ChatGPT. ChatGPT is quite good at making images. A lot of you have seen images on your social media and all kinds of stuff. But let's go into our my adventure retreats thing and I'll just say, create an image. Whoops. Create an image of a kayaker in a rapid. Okay, I'm guessing it hasn't made that yet. It's not. It doesn't have dates to select yet, so it hasn't made it designed yet. Why else? Why else could it be obvious now again, I usually start with an okay prompt or a bad prompt. And you know, I can obviously get much more descriptive and possibly get better results. Obvious answer. Thank you, Brian. Anyone think of another reason why maybe it's not there yet? Like, it's a smart system. Why would it make something and then not put any dates there? Because it doesn't know your availability. Yes, that's exactly right, Jacob. So Now Chad, GPT has gotten way better at creating images lately. I didn't give it any instructions. I didn't say make this look like a cartoon or make this photo. What I would do now is I'm going to ask, hey, when are the dates, location, start date, end date? It's probably pulling from an empty table realistic. Or make this look like this animation or make it look like, you know, whatever. I didn't give it any instructions so it's just going to make those decisions on its own for me. And hopefully it uses preference. It's pulling data from an empty table and so it receives nothing back and therefore it renders nothing. It has nothing to render. Okay, all right. Do we have a. Yeah, so now. Yeah, I can't see anything on this. Now let's go to. Whoops, sorry, doing the wrong thing here. I want to click here. Boom. Okay, now preferences out of my project and maybe preferences in the memory and all the sets of things. But anyway, so this is one system that you can use to make images. While that is loading. We have a log out and apply now. See this, this looks okay. We definitely don't want a double logo here. So we'll just tell it to get rid of this. The get rid of. We're going to bounce over to here to Mid Journey. Mid Journey is another one. And there's an awesome. There's a double web based interface. Now you used to have to use Discord to do in the mobile drop down menu. That in Discord can be a little bit confusing if you're not familiar with that platform. But in here, get rid of the one that I selected and then you're basically just going to say, you don't say create, you don't. In ChatGPT you'll ask it, you know, and I'm going to say. Main menu on desktop also still doesn't work. Create an image that does this or whatever. You'll kind of give it a little bit of an instruction here. You're just going to say, you know, an image of a kayaker. Let's just fix this by making the Hamburger menu functionality happen on desktop view as well. Okay, so now I'm just going to drop everything into the menu for now and then later I can go back and fix it and I can maybe move something out and say, hey, let's put this heading and this heading here. But you know, navigating a Class 5 rapids, cool colors. Clearly, somehow it's just building something wrong. And so I'm going to have it kind of Just do a totally different thing and then I might go back and fix it later. Okay. I also, one of the problems is I don't love how much space there is. Jungle. Yeah, jungle environment. DSLR is right here. Why isn't. Why is this all the way over here? Let's move it over. Right. So again, this is just a constant state of almost debugging, right? But what I'm doing with bokeh effects, this is just like photography stuff, right? So I'm saying make it look like it was taken by a DSLR camera with a bokeh effect, which means it's going to have kind of a blur. Don't think of it as though I'm just a QA engineer now. Right. Although that's a lot of developers and a lot of people now that we're using AI, that's a lot of what you do. But think about it as you're a customer experience background. And we'll go. We'll just tell it to go, right? So I can obviously give it more instructions. I can get pretty detailed in my prompt. But. But we'll start with that. By the way, guys, engineer, you're a user experience engineer, and you're going through as your customer and saying, what do I think about this? Is this button where I think it is when I click this button, does what. What. What happens? Did I expect that? Was that surprise? Maybe you don't know how to do that or how to write a good prompt for mid journey. Can anyone think, amazing. Was that different? Do I want it to go somewhere else? Whatever. And you could see, I just used normal how they could make a good prompt for mid journey language to say, hey, there's a double logo here. I selected this. Hey, let's move this over here. Hey, this isn't working right. So now all of a sudden we've got a better, you know, a better working thing. Travis, use chat GPT to ask it to make one. There you go. That's exactly right. Okay, so let's bounce back on. All right. Well, we should. We almost did. Now the hamburger menu is broken. I've completely redesigned the navigation menu. Great. Yeah. It's not visible chat gbt. And you know. Sorry. We were doing it on the browser. It's right here. All right, here's my. Here's my image of a kayaker. You know, I've asked. So this is. This is called a Z Index. Index problem. I saw that there was like this little thing here. It looked like a scrolling thing. Okay. So I'm gonna Go like this. I'm going to drag this in to make a lot of. I've asked AI to make a lot of whitewater kayaking images and there's a few things wrong with this one, but this is one of the better ones. The paddle's a little bit too short. It looks like he's using his son's paddle. Needs to be a lot wider. But AI just isn't a good kayaker yet. Let's just click this. There we go. I think there's a Z index issue here. Say that just not there. But it's still pretty good, right? It's not bad. But let's do this. Please create a good prompt. Z index, guys, is how close it is to the screen, right? Or how close it is to the user. So like it's the stacking of all the different elements. Can't see the menu for me to get a good. A nice kayaking image from behind the page content. I still don't know if I'm mid journey. Okay, so now I'm using AI again. We did this yesterday going back and forth between chat, GPT and grok. Right? We're using AI to create a prompt for me. Absolutely, your highness. Here's a mid journey ready prompt that will give you a vivid. Love it. Right. I don't know what it's trying to do in of front. Full size. Yeah, it's making it kind of a full screen even thing here. I. I just don't love. So anyway, it's going to fix it, by the way, if you don't know what Z Cinematic kayaking image. A kayaker plunging through raging whitewater rapids. Dynamic indexes. It's fine. It doesn't matter. You can just say I can't see the menu. Where's the menu? I'm clicking on it, but I can't see it. And then it will look and it will say, well, I see it because it can see my camera angle from river level. Dramatic lighting, V6, whatever. Okay. It can't see what you see. So do it. Take a screenshot, right? Click on it and say, okay, this is what happens when I click the menu, right? It doesn't. I can't see. So I'm going to copy this and I'm going to paste it in. We're going to see if it works. Okay, so here's what. What it created with my prompt. All right, that one's not bad. Let's see it. Right? And then you'll learn and it might say something. Oh, this is a Z index. Index issue and then you'll learn what that is. It will teach you, right? You'll learn as you go. Okay, not bad. Kind of like for the right use case. This one looks like a miniature. Like what do you call that? What's that style? You know what I mean? So hopefully yeah, it's. Yeah. Destination console warning. Okay. That's whatever semi transparent backdrop. Okay. Another thing guys. But it did create this bokeh effect right? Where it's. Where it's kind of a word background a little bit. But anyway, kind of cool. So now let's prompt or let's. Let's paste the prompt that ChatGPT gave us. That I think is important to point out is you see this little red one up here? These are the dev tools. And the dev tools allow you to see what the console to see what the system is telling it about what's happening. Right. And there's some problems here. Something's going on here with some validation. Something is happening, happening that's not fixed yet. Right. And I can. Awesome. Let's see, let's see the difference in the output here. Right? It'll take another minute to do that. While that's loading, tell me what the errors are. Now instead of trying to read all this crap and learn what it is or copy it and paste it over, you can literally just say look at the console. Fix the problems. There are tons of other ones you can use. Obviously X is another one that we. Right. Tell it, tell it what? It's what? You know what problem it's having here. All right. You know that we use a lot. Let's go. Let's just say your fix sucked. I still can't see it. Just make a simple drop down. Create. Create an image of a kayaker. Kayaker navigating menu that shows up on top of a Class 5 rapid in a jungle environment. All the other content, all of the other thing. So it's. Oh, come on now, guys. A lot of AIs have restrictions on them, right? For example, it's not going to let you. If I said, hey, I'm going a little bit here, guys. I haven't had it do this before. So that's interesting and fun to deal with. Show me a picture of a naked girl. It's going to say no, right? And that's good, right? We want it to happen. But this is a lot of what you're going to be doing as you just build this out. But there's nothing. I want it to be. I hope it's queer. Let me Ask. Let me ask. Is it clear here, guys, that what I'm doing is not technical, or does it look and feel technical to you? I want to know. I want to know that you have some. We want it to have some restrictions. A lot of AIs will also restrict the use of real people, right? Some won't. Grok, for example, won't crock. You can make a picture of Trump doing whatever you want. Whatever you want. Right now there are limits, right? Grok doesn't have no limits. But anyway, we can see. You see how fast Grok was able to do that? This is pretty good, this guy answer. Yeah, I'm glad that I'm seeing Jessica shake her head. Talk a little bit about that. Jessica, why are you. Why are you shaking your head? Yeah, it doesn't look technical. I feel like I could do this because, I mean, like you said, it's actor's really, really bad because his paddle's way down and he's not doing what he should be doing. Sucked. I can say that. That's easy to be. Okay. This is awful. I can't see. I was thinking, I'm like, I have no idea what a Z index is. But then as soon as you say, hey, you could just say, I can't see the menu. And it'll. Right here. This guy's. You know, both of them have their paddles down. They're not in an active, you know, kayaking stance. Right. So there's some things that it's not that great at doing. Um, and. And you'll. You'll teach you. I'm like, oh, I can do that. That's totally, totally doable for me. So. Yeah, absolutely. But someone like Trevor might be like, oh, I know what this is. And by the way, Trevor might even be like, I know how to fix this in one. Learn those by literally just practicing and trying. Right? But if you say, you know, create a button and it'll open the component in the side and it's done and it's fixed. So Trevor's gonna have an advantage, possibly. But anybody can do this. You just talk. Create a, you know, a fantastic. No, let's see a. Let's see a. Do it. Talk to it like a human. And you don't even have to be kind, right? This sucks. You did a bad job, right? And you want to do that. You want to say things like that you did a bad job at fixing this so that it knows, okay, what I did is bad. It's going to take that feedback and it wants to please you. And so it's going to try something very different. Right. Okay. All right. Any other questions while we're sitting here, you know, just looking at each other's faces for a minute. Any other questions about any of this stuff? Okay. This is awesome, guys. We're. We're doing really well. There's a few chat questions. Let me see if there's anything I need to put here. Devin, have a question about working on large projects. I've tried to write a few books. They always seem to shorten the content. Heavenly scene of violence, weather. Okay. It's going to create something pretty dang cool. Like, it can be so creative. You could also say here, let's go back to main big time. Yep. Also getting to export into a single document is always problematic. We'll be covering this during the course. I think it's a great question, Devin. And I'm also writing a book journey here and check on our kayaking pictures. Okay. So here's what I got with my little boring prompt. And then chatgpt gave me a much better prompt. Right. These images are better now. This one is kind of weird. It's like. And I'm actually not using. I don't use AI to write. I use AI to be my editor. Right. I ask it to give me suggestions. I ask it to fix a typo, you know, things like that. But I say, don't change the content. It's like it's being taken. Yeah. That's actually kind of interesting. This. There's two kayaks here. Like, this is my point of view. I'm sitting in a kayak right here, looking at this other kayaker coming right at me. Right. It's like we're going down the rapid together at the same time. It's kind of cool. It didn't add. Only give me suggestions and tell me where I can improve it. I won't even have it. Like, I somehow I just want to keep that as something that I do. Right. I like to write. Now you can also say, hey, you know what? It's all my ideas. I want AI to do the writing for me. I want it in this style. And you can. You can do it really, really, really fast. But yes, what you're running into there, Devin, is context lens. Get to do that. That's kind of interesting. All right. Here. I don't know what the heck is stickiness. Right. You're running into context on the input and the output. And so you might need to switch to out of the front of this kayak. I've Never seen a kayak with that. Maybe. Yeah, we'll just move on from that one. Okay, so, but anyway, guys, you can see, you know, this person's head is not even there. So that's fine. But you can another tool to do that or you might need to just break it down and have it do it one chapter at a time. I would use the project tool. I really like the project tool in, you know, in Chat GPT, right? Where you're able to add files and you're able to give it all the context of the whole book, keep editing it, modifying it, say, hey, I want it to be a little bit different here. And I want it to be, you know, can say, all right, we're working on this chapter right now in this chat. Okay, now we're going to work on the next chapter in this chat. And it might shorten. It might try to do a lot of things. Sorry. Okay, so you could also Devon. Create your own book. But I do like this. It said splashing with water droplets frozen midair. Right? It did that really well. In each of these images, you've got all kinds of water patterns and different things showing up all over the place here. All right, this is. So is Mid Journey just a lot better than Chat GTP Writer tool in replit? That's actually what I would probably do if I wanted to solve this problem, like for real. Because OpenAI, the OpenAI API coming up with pictures, just more options. Is that why you subscribe to it? Absolutely. IPI won't have the same context limit that like Chat GPT does or that GROK does. And so yeah, you're going to pay your own API credits. Do you guys know this is cool. This is pretty cool. Did you think Mid Journey creates the the highest quality images by far? Now there are some things the Mid Journey really, really, really sucks at. And the Chat GPT sucked at up up until not too long ago. Right, so let's go back to this, this image here. Actually, you know what? Sorry, we're using reusing the browser here. Look at that, guys. That's pretty cool. That's pretty nice image, right? In the right circumstance, however we want to do, if you're create, if you're a creative person, you're going to have a lot of fun with AI, creating, creating a lot of stuff. But let's, let's do this, let's do. No, that Chat GPT, sorry, right here is just a UI with an API to OpenAI's API. Does that make sense? It means that any of you let me, let me be clear here. This right here is a user interface where you can ask questions and you can act. Let's try something, try something cool here. We'll, we'll say activate different functionalities that uses the OpenAI API key. And guys, we haven't gone over API keys. We're going to have to do it tomorrow. And I want to do it a little bit more. But all it does here is, is it. It used an image of a woman holding a cake. All right, I like this one. All right, let's see if we can save this image. And save. Uses this API key, whatever it is, right? And it connects to a system, right, that is called an LLM or a large language model. It's connecting to one of these AI. And then we're going to upload right here. Nope, I don't want to. Yeah, yeah, Models. And then right here, it's just a fancy prompt. You could rebuild Chat GPT. The product Chat GPT is different from OpenAI. OpenAI is the brains Chat GPT. You know what, that's fine. We'll just go into a new. New chat and like this. And then let's say, let's see an image of a. Is the software solution interface that people buy to connect to that AI and talk to it. We have an interface that allows you to type and it types back. Okay. Of an Apple. Now let's see, you could, you could say to repl it, right? Create app. Create a Chat GPT clone. You can build your own version of Chat GPT using a picture of Lemonade. Okay, so let's take this one right here. It's a little bit. This is kind of cool. It's a good challenge, right? Because it's got a few other things in it too. Let's go like this. And your own API key. And it. It will even look exactly like Chat GPT. We'll go like this. I'm going to say this, I'm going to say. And you can call it Jessica GPT or John GPT or Evan GPT or Jacob GPT. Go whatever the heck you want, right? And now you've rebuilt the entire product Chat GPT. Now do you want to do that? No. And I'm going to talk about why in a minute. But you can, right? Understanding that this is just please swap out the cake in the woman's hand or the pitcher of lemonade. The other picture. A very simple API setup means you can actually do it yourself. Mid Journey can't do this. Mid Journey cannot do this. Chat GBD can, right, build it, implement different AI model selection options. Perfect. Approve plan. Let's just watch it. Let's see what it does for a second here. I think it'll be fun to, to just see if it, if it pulls from, you know, Chat GPT. I only recently do this. But if you have a product, right, like, like you want to sell com and tries to, you know, follow a similar water bottles, right? You can literally take a picture of your water bottle that you want to sell and you can get a picture of someone holding anything, right? And you can drop it in and it will, it will swap it out and it make it look pretty dang awesome. Let's see, let's see what happens. I wonder, I wonder what it's going to approach. I told it to make a Chat GPT clone, so it should theoretically build something that's kind of like that. Now it's got its own logo. That's not using the ChatGPT logo to do with this hand. I told it to swap out the cake. So I want to see, I want to see what it does here. We'll see. This is a fun test by the way, guys. I hate waiting for AIs. And so when I was building my business in three days, there's no way I'm sitting here waiting. I could say go and. I bounce back over and I jump back into mid journey here and I'm saying, okay, let's go here and let's do this. In a minute it's probably going to ask me for an OpenAI API key and I think since I have one on my account already, I'll just be able to click add and it'll be done. But it just says message chatgpt. It's just this and let's do that. One of the things that you have to create for your business, right? Or at least the smart thing to do is to create a logo, right? And so you can try, you know, creating a logo with a lot of different things. You could try with, you could try with jgpt. And right now it's creating. Remember guys, what it does at first is it creates like a high level, like, okay, this is the conversation. This is kind of what it's like right now. We haven't created a login or authentication system that's going to save our chats and our chat history and all of our context and memory. And those are all things that ChatGPT spent a lot of money and time thinking through and trying to make. And so that's part of why you might decide, hey, this is worth 20amonth for me. Whereas maybe you say, you know what? I only use chat GPT $4 a month and I really just want to chat. I don't need any of these other features. I just want to chat with it. Okay, great, we'll make your own and use it for 4 bucks a month or 3 bucks a month or 27 bucks a month if you use it a huge amount. And then, you know, you're set up. Now you're gonna have to host it on Replit, which will cost you, I don't know, a buck 35amonth maybe plus, plus the cost of Replit. Assuming if this is your only project then that might not, that might not make sense. Maybe just pay for ChatGPT. In fact, I will try that. Yeah, I use this here. Oh man, I'm excited for this. It looks like it's changing her face. Just. I know it's doing pretty good. It's doing pretty dang good. Anyway, so that's coming along. But creating logos, guys, you want to try different systems, right? So let's go into X here and let's say please create a logo for and by the way, I'm giving it no ideas or prompts or whatever for this adventure so that for this low states retreat company. Okay, but, but the point is guys, a lot of these really magical. Like I had somebody get pretty mad at me yesterday. They got pretty mad at me. They messaged me on link. So we're going to see what grok comes up with. We will go to Chat GPT. Please create a logo for the flow state retreats company. Okay, I've got that started. We'll come to mid journey and then let's see here. Let's go Back to the ChatGPT here. Oh my gosh, look how cool that is, guys. And they showed me this really cool tool. We'll, we'll. I'll come back to this in a second. I'll share the screen as soon as it's done, you know, doing more. Let's talk to you guys for a second. But I had to message me and they showed me this tool that basically you put in a business. You type in a business idea. It was a tool for entrepreneurs, but you put in this business idea and in fact let me pull it up. I'll show it to you. I'll show. Look at that. So it took this image right here. Oh, to you I think this is, I think this is cool. I like show and tell. And yeah, this Guy. Let's see here. Let's see here. Where is it? Okay. This woman with the cake takes this picture of lemonade and combines them together. Okay, here it is right here. Now, I would prefer that she's holding the picture instead of holding it like that. And it's kind of weird that there's a lemon there. So I could get Flyway. Flyway to. Okay, let's share my screen share. All right. Flyway to take your ideas from what if to a winning plan. Okay. You give some updated instructions. But that's pretty dang amazing image composition. Right. Put in your idea. You just type to it, and you talk to it a little bit. It's kind of like what we did at the beginning with chat, GPT and GROK and going back and forth trying to validate a business idea. Right. And then it, you know, evaluate this idea's potential, market viability, feasibility, profitability. All these different things it was able to take. You know, it mirrored this, basically, and it just dropped it in. So I could say, get rid of the cloth. Make it. So she's holding the picture by the handle. And in fact, we'll do that, and I'll go to another. Another platform to have it do it. How many of you guys think you could build this tool now? Anybody? Jessica's kind of a maybe. I see Travis. Yep. Jacob thinks he could do it. Brian thinks he could do it. Trevor could definitely do it. Yeah. I'm seeing all you guys raise your hand. You could absolutely build this. Not only that, you could just say, hey, I want you to go to Flyway to. And I want you to build it. Get rid of the cloth and make it. So she is holding the picture pitcher by the handle. Also, get rid of the extra lemon that is outside of the picture. Okay, so let's product for me, right? I responded to the guy on LinkedIn. I said. He said, look at this really neat tool for entrepreneurs. This is really cool. You could maybe integrate this into your course. And I responded to him. Maybe I shouldn't have. Maybe I was just being stupid. But I said I could rebuild this Entire app in 60 minutes. This entire company in 60 minutes. And I really believe I could see how it does there. We'll bounce back over to ChatGPT here. All right. Flow state retreats. So it kind of knows what I'm about to. Decent logo. What do you guys think of this logo? Let's see a couple. A couple nods. Yeah, it's pretty good. That's decent. Yeah, I. I think. I think I'd wear a hat. Yeah, that's. That's kind of what I like to ask. What I. What I. I could take all the screenshots in the videos it's showing me its whole product. I could rebuild this whole thing. I take a few screenshots and I say, here's what I want you to build. Exactly this. This is what I want. You click on here, you evaluate. You evaluate this idea. You do this and make it stronger, make it better. How can we make it better? And I could call it Flyway. Better Flyway to, you know, and I could put this on a hat. Would I throw this on a T shirt, right? What I. Would I put my little logo right here. There's other things you want to think about as well, guys. You want to think about the fave icon, right? What is your little image here, Right? So if you get it, maybe. Maybe something that's really complex. Well, it's going to look terrible small, right? How is this going to look as an app? Do you like the color scheme? Do you like whatever? Right now message it to the founder of whoever sent it to me. Anyway, they were really excited about this thing and I kind of made them feel dumb and I apologize to them afterwards. So that's a different story. That's just my fault. That's my problems. But now they've got a nice website here. It's pretty cool. They've got some nice little things here and they've got some good design. I wonder how they designed this or who they paid to design. Hopefully they used AI because I guarantee you, they used AI to make this. And all this has behind it, guys, is a chatgpt or an OpenAI API key that we're not going to spend this class talking about design of logos partially because it's so personal, right? If we were to ask every person. We have nine people on this, on this meeting right now. If we were to ask everybody their opinion, we would get nine opinions or maybe 10, right? Everyone's going to have their own opinion on analyzes this stuff. And then they just have it, you know, ask a few prompts, right? Hey, give us some market viability, give us some feasibility, give us profitability, right? And then categorize that and put it over there. And you wouldn't even. You wouldn't even have to tell it to do that with re guys, just show it the dang screenshots and it would build it. It would be done, right? Yes. This is pretty cool. All right, continue. Okay. Color schemes and logos and what you like and what you don't like, you're the CEO, man. Just go for for it. Do what you want. Okay, now I do want to talk a little bit more about the technical aspects of it. Right? I can come over here to mid journey. Now what I'm going to say is this. I'm going to say, whoops, we're going to bounce back over here. Hold on, I want to. You can see my screen here, right? So OpenAI API key. Here's my value. I already have my OpenAI key in here. So I'm going to click, continue. It looks like we're, we're almost ready to do this now. Oh man, look what it did. Look how cool. Guys. Obviously the formatting, it doesn't look exactly like ChatGPT, but we've created a chatbot, right? It works. So it's going to check the implementation of our OpenAI API integration. Now I need to check our routes, see how we're handling chat requests. I mean, in just a couple minutes here, I will be able to sit here and talk to this. Boom. Right now, I don't like her teeth. They're a little bit off. But for the most part this image is completely usable. Right? I can tell it. I say make teeth look little bit more natural and you know, make it so she's not smiling so big or whatever. Make her have a natural smile. But, but we completely changed an image and think about this as bureau, just like it's Chat GPT. That's amazing. And all of you can do what I did. All I did was say make a chat. Create a Chat GPT clone. This is all I've done so far on this project to create this. Right? Adding APIs in are amazing. Okay, so let's see if it's working properly. Marketer, right? If you're selling disco products, so you want to show something and you don't have the picture or the time or the money to go to a photo shoot AI in 30 seconds, man, it's crazy. It's crazy that we can do that. So now we're going to do is we're going to bounce back to this. We're going to say, please create a mid journey. How far is earth from the sun? Oops. I clicked Enter, by the way, and it went enter instead of sending it in, which I expected it to do. I could tell it, hey, when I click enter, automatically send the message. But I haven't programmed that yet. Okay, I got an error. Right? We gotta, okay, check, check the logs and fix the problems. Okay. It Might take a couple prompts here, and then we'll have it working. Okay, now while this is still building that out, it's time for an awesome discussion to finish out the day. And this discussion, guys, is called Surf Don't Swim. And my. My. The purpose of this discussion is to completely have you rewire your brain in preparation for the incoming AI tsunami. And how you go about doing. You're running your existing businesses, starting up new businesses. A lot of you are doing, you know, all of you are doing one or two of those things, right? One of those two things. You're starting up something new, or you're thinking about starting up something new, or you already have started up something new, or you have existing things, and you need to make sure that you don't get smashed by the tsunami on the rocks. So let's ask a question. How many people here have never swam in an ocean? Anybody? All right, so everybody knows how powerful the ocean is. How much energy do you think you expend in. In, say, like, joules, right? Measurement for energy, right? How much energy do you think you expend as a swimmer? Just a single swimmer. Let me ask chat gtp. I'll get back to you. All right, Brian's gonna get back to us with the real answer, right? Say in joules. And it might say, oh, I like K calories better. Whatever. I get it. Right? Just see if you can find the answer in joules. Okay, so let's say it's 4,000 with flow against flow. Depends on the resistance, how much joules you're expending. Absolutely. It's going to. It's going to depend on a lot of things for sure. How long are you going? 10 minutes, 1 second? An hour? You know, there's a lot of variables there, right? But. But let's say. Let's say it's 4,000 joules, okay? Or 1,000 joules. I don't know. We'll use 1,000 joules. Now, I want you guys to think about a tsunami, okay? How much energy do you think relative to the swimmer? What do you think the ratio is right, between the tsunami and the swimmer in terms of how much energy is created is like 100 times bigger. A thousand times bigger. Trillion times kilojoules, multiple. I'm with Devin. I'm with Devin. I think it's. I think it's. And I think I did the math one time. I think I actually did go through with chat GPT and talk to it for a bit. I think it's 4 billion times more Right. So, okay, nobody's surprised there, right? So who here now thinks they're a really good swimmer and can win against the tsunami? Nobody. I can't. I've said many times I'm a kayaker, right? I've fought against just a few CFs, a few dozen CFs, maybe 30 CFs, and lost. I'm not going to win. I can't paddle upstream in a lot of situations where there's just like a little channel coming through. I can't get up and over it or I'm stuck and getting pushed against the rock. I've got plenty of stories I can tell you later, but water is going to win, right? It's relentless. It's never going to give up. The incoming tsunami. We talked about this a little bit at the beginning of the course, or, sorry, the beginning of the class today, about how there's this revolution, I mean, very, very similar, not dissimilar, at least from the Industrial Revolution that completely changed how humanity operated. And AI, in my opinion and in many other people's opinions, is 10 times, 100 times more impactful than any other revolution that's ever happened. Right? Now, there's been other tsunamis that have come in, right? The Industrial Revolution revolution, the computer, right? Personal computer, the things that you guys have lived through, the Internet, coming out, social media, coming out, streaming capabilities, satellite, you know, whatever, right? There's a lot of different waves and tsunamis of different, Varying heights and varying powers. But the point is, those waves and these tsunamis exist, and if you fight against them and you don't fight with them or you're not aware of them, you're going to get smashed against the rocks. Now, what's cool is, even though swimming in a tsunami is incredibly dangerous because you can get sucked to the bottom and pulled out to sea and just crazy rip currents and things like that. Surfing is actually awesome in tsunamis, right? Because you're sitting on top of everything and, and you just. For some reason, surfers flock to tsunamis. If they think there's going to be a tsunami, the surfers are heading it out there. They want to be there for it. They want to be. They want to catch the waves, all right? They're, you know, they know what they're doing and they're safe and they still have rip current, you know, problems to deal with. And anyway, so they, they'll wear crazy life jackets and things to make sure they can. They can do that safely. But we don't need to get the technicalities of surfing. What we need to do is we need to understand how we can surf the AI tsunami instead of trying to swim against it or even around it. Right? So let's talk about some real world examples. Let me, let me ask you guys a question. Who here can tell me what Netflix's actual product is? What do they. What is Netflix? What kind of company are they? Not a trick question. Just what do you think? Host company. Hey, Hosting. Devin, Travis said media. Both of you are correct. What else. What else is Netflix to my understanding, a streaming service. Streaming service. Very good. Yeah. Production company. They do their own productions. Fantastic. Jessica, they are now also producing products, right? They're producing their own stuff. So they are. They're doing a lot of different things, right? I mean, we all know what Netflix is, right? They tell you, okay, you know, we've got. We've got this movie here. We want you to see it, click the play button and you watch it. So we heard streaming company, we heard hosting, we heard. We heard creation or sorry, we heard media company, and we heard, you know, production. Okay? Netflix's actual product is a content recommendation engine. That's it. And, and I will add, Jessica is correct. They are production. That's a separate. That's a separate division. And by the way, it loses them a crazy amount of money, right? Which is fine. They've chosen to do that. They're like, we can afford to lose a billion dollars over here because we're making all this money over here. And that is about what they're losing per year. They're losing about a billion dollars. But Netflix is a content recommendation engine. Do you know how many pieces of content are in Netflix? Does anybody know? Without searching the Internet or whatever, it's about a hundred thousand. Okay, how many of you guys, how many. How many pieces of content do you think you've ever actually seen in Netflix? Like, watched? Not, not, not, not even just watched, but like, even scrolled past a thousand? Maybe, maybe 500, 800. I mean, probably somewhere in that range, I would guess, for most, most normal users, right? If you're like a crazy power user, maybe you've seen 2000. So how do they take this much content and only show you this much content? Right? Have you ever. Have you ever gone to, like, an Airbnb and you log into someone else's, you know, Netflix account and you're like, what is all this stuff? Like, I have never seen these movies before. Like, this is weird. I need to log out and get into, like, the real part of, like, where's my Netflix? You know, that's their magic right now, they've done this in varying ways throughout the years. They started out with DVDs and you'd mail them and get it back and forth. And they had that cool model which was very expensive and difficult. Then they switched to streaming. Okay, so here's how you tell whether or not a company is surfing or swimming. And then we're going to examine Netflix and try to figure out which one they're doing. Okay, if. Here's the question you have to ask yourself. If the underlying technology upon which you are built or the technology that's around you is built improves, does your company get stronger or weaker from a competitive standpoint? Okay, I'm going to say that again. If the underlying technology upon which you're built improves or even is around you, meaning competitors can use technology to pass you up. Right. Do you improve or do you stay the same? Okay, so with that definition, what do you guys think Netflix is? Are they a swimming or a surfing company? Irfing. Why? Because they. If the underlying technology improves, they're just riding on top of the wave. What waves are they riding? What are some of, like, as the streaming was like, you know, like you said before, they were like mailing out DVDs. Now that they're able to stream, they just are riding that wave of streaming. Right. Even though that's not their main product, which is. What did you say it was? Was the. The production. Yeah. No, they are making that their main product still. Is this streaming service. They are a streaming service. None of you guys were wrong, by the way. No, but what was the thing that you said it was the. Where they recommend content generation. Yeah, the recommendation. Yeah, there we go. Yeah. Yeah. So let me ask this, guys. I'll ask this in a different way. How much money do you think Netflix has spent laying fiber across the Atlantic Ocean to improve fiber, Global Internet connectivity? Zero. They could use Starlink. That's exactly right. We're going to get there in a second. Devon. Yeah. Zero. Right. How much money have they spent building the underlying technology that makes streaming possible? AKA the flux capacitor of streaming. Right. Like Akamai and different content delivery networks and things like that. Right. What are the. You know what, how much money you think they've spent on that? Zero. Right. How much money have they spent on. On exactly what Devin was saying on, you know, satellites, you know, by Elon Musk. How much did they invest in SpaceX to send up satellites just for Netflix? They haven't. They haven't spent a dime on it. Right, so they're surfing on other people's billions and maybe even trillions of dollars in improving. Improving the Internet as a whole. As the Internet gets stronger, as streaming speeds increase, maybe they'll be able to actually stream a UFC fight without, you know, crapping out every 30 seconds. Right? They will get stronger. Their offering will get better. Okay, perfect. Netflix is surfing. Now, are they surfing on the production side or swimming? I would think they're swimming because eventually we'll be able to create our own movies just for ourselves. Perfect. I love the answer. Eric, you're exactly right. You're applying. As technology improves around them, how are they going to compete with AI? Like, that's crazy. Like, the new Google models for video are just insane. Right? Not Sora, but the new one, whatever it's called, I haven't played with it yet, but, oh, my gosh, I've seen some. Some outputs, and it is just mind blowing. Okay, very good. All right, test number two. You guys passed. Passed the Netflix test at and T. Surfing or swimming? I think they're sinking. Okay, Devin's good. They're sinking. All right, fair enough. I think that means that by definition, they're swimming and they're not keeping up. Very good. Okay. Why are they. Why are they sinking or. Or swimming? Why are they. Devon, I want to just follow that up. I don't know, just from the. I don't use AT T, but just from the advertising side of things, I feel like they're always chasing down. At least out here in Utah, where we're at, they're just always chasing down Verizon. Their costs are always lower. Like, I just. I don't feel like they're bringing anything new or exciting or revolutionary. Maybe I'm just unaware of them and. But if I'm unaware of them, considering all things that I'm plugged into, they're probably not doing anything too groundbreaking. Okay, so I'm gonna. I'm gonna say back to you what you just said. The reason you believe they're swimming or sinking is because they're. They don't seem to be keeping up with competitors around them. Basically. Right. From a marketing standpoint, product standpoint, whatever it is that you think. Okay, good. Let's. Let's. I'm not saying that's wrong. I just. I want to clarify what we're saying here. Evan, looks like you're unmuted. Do you want me to say something? Or maybe you're just unmuted. No, I'm trying to remember, does A and T own any Type of Dish network. They have their own TV network, right? Of course, yeah. They have like you. They have U verse, I think, I think is what it's called. If I'm correct. I'm not an AT expert, but anyway, I thought there was something else, but it requires like having a dish on your TV or on your house. I don't know if they do. I don't know if they do. I don't know if they're competing with Direct and Dish like that. I'm not sure. I thought they owned Direct, some type of. But anyways, if you talk about Dish or Direct, right? Like even YouTube TV is just way easier. I don't have to have a stinking like satellite on top of my house. And as like the Internet gets better, YouTube TV gets better. Right? And it's. It's a month to month. I mean, I would consider that one definitely like a surfing for YouTube TV, Google on that sector of their business compared to like a Dish or, you know, a direct tv, which I thought was owned by atm, but I was wrong. So I'll say back, I'll say back what you just said. You're saying that they're swimming, I believe, because their user experience isn't as good as other. Other competitors. Okay, so Evan's talking about user experience. Devin is talking about, you know, relative to, you know, just everything about them just doesn't seem as good as other companies around them. Okay. Relevance factor. Like just becoming less relevant. Yep, becoming less relevant. But let's talk about, let's go back to the definition. This is really critical and important and it's hard to understand. Guys, I've had. There's a guy writing a book called Surf Don't Swim, and I've had four hours of conversations with him about this exact topic. Just to have my brain wrap, like, wrap my tentacles in my brain around this concept. Right. It is not easy to understand. Okay. And that's okay. Remember the question is, are, does their company get stronger or weaker when technology around, Relatively speaking. Right. When technology around them or underneath them, you know, grows. Right. So let me ask this, does AT and how much money does AT and T. So I asked the question, how much money does Netflix spend on laying fiber zero hardcore surfing? They're awesome. How much money does AT T spend to build a cellular network? Tons. Billions of money. Billions of dollars to build those dang towers. And every time you go from 4G 3G to 4G to 5G to 6G, you got to upgrade them again and you got to maintain them and they break and there's windstorms and there's solar storms and there's all these things that can just screw up their network and their infrastructure. And so they're constantly spending massive amounts of money building their own infrastructure. Right? Whereas we here are just sitting here. We're all, all of us are surfers. All nine of us are surfers here on this call because we have cell phones. We don't build cell towers. But we accomplish crazy things by using our cell phones, right? Because we're surfing on other people's technology that. Where people have spent billions of dollars to do that. I'm going to show another example of surfing. I want this to be really clear. So, yes, AT&T is hardcore swimming. Okay? We're going to bounce over to Facebook and we're going to look at my mom, okay? My mom, for those of you that know her, she's not an influencer. Like, she's not, right? Anybody that knows her is just smiling and laughing, thinking about whether, you know, considering her to be an influence. She's not. She's like the most down to earth, whatever. She. She follows zero people on Facebook. She just doesn't even know hardly how to use it. Okay, But I want you to look at this. Look at how many followers she has. Holy crap. 592,000 followers. She. She makes over $20,000 a month posting videos of her travels, right? She goes to Africa and she goes to wherever and she'll go on maybe three trips a year to see some animals. She takes her super expensive camera and she pushes record and then she goes home and she has hundreds of videos from each of these trips. So she has over a thousand videos that she takes a year and she just posts them a couple times a day or maybe once a day, or maybe she'll miss a day and then she'll post or. Looks like she got a bunch on April 16th, right? April 15th. April 15th. April 15. April 15th. So sometimes she'll sit down and she'll do a bunch, but some days she'll miss it. And she has some videos, guys, that have 500 million views. It's like crazy, crazy numbers. She has over a billion views on. On Facebook just from posting these videos. So is she surfing or is she swimming? Hardcore surfing hardcore. Why? Because it costs her nothing. Facebook's free for her. She gets paid 20 grand a month to post videos. Exactly. She's even. She's even surfing her. Her travels because Facebook's paying. Like her whole thing is free, right? It's just this Self perpetuating engine that builds for her. Okay, you're absolutely right. She is not spending a dime improving Facebook's content algorithms. But does she get richer if the algorithms improve? Yeah. What if the Internet improves? Does she get richer? Yeah. What if Elon gives global access? You know, everybody in the world now has Internet now. Everybody in the world can access Facebook now. There's a bigger audience. Does she get richer? Yes. She's surfing on all these different things and she's spending.00 on, on putting that, that thing. So influencers kind of by definition are surfers. They all are right now. We can't all become influencers. We can, but it takes a long time. Can't do it overnight. We'll talk about that tomorrow. We're talking a lot about social media tomorrow and how to automate, how to do like crazy scale production and auto scheduling and auto commenting and all kinds of really neat things stuff, right. That can help you get to your first customer. Because you got to get to your first customer. Otherwise we're just building cool ideas. Tell you have customers, they're just cool ideas. Okay. We're building a business in three days. Okay. So we got in order to build a business you need customers. Let me give another example. My buddy that is built is writing this book Surf Don't Swim. He owns about seven AI companies right now. He, he, he's a product of the dot com bubble. He, he, you know he was a developer in the 80s and in the 90s and built some really cool stuff and ended up in a good spot and made, made a lot of money, right. Had about $250 million exit around the dot com time and so did really well. His name is Jeremy Whit and if you go to Surf don't swim.com you can sign up for his book right when it comes out. It's not out yet but you'll be, you'll be on the list and you'll get his, you'll get his notifications about it and I highly recommend it. But anyway, he is right, one of his businesses, he told me about one of his AI businesses that is a pure, pure surfing model. Okay, this is going to blow your mind guys. He's making 60, I think he's actually up to about $65,000 a month now in pure profit right on top after his expenses. He has two like entry level like secretary type staff people that, that just kind of, you know, do a little bit of communication. Call it customer service. Right. But here's, here's what it is. Back when he made all his money, one of the things he did is he went out and he bought a retirement home. He bought a, you know, about 200 person retirement home, right? He's like, people are always going to get old and need care in a senior care center, right? It's in Kansas City. He bought this thing. And so he's been in the industry for, I don't know, 20 years, 25 years almost. And he's, you know, he just kind of learned the tricks of the trade. So he's familiar with what's going on there. Now one of the things that almost everybody knows is that it's a lot cheaper to prevent illness than to treat it. Right. And one of the things that seniors deal with is their minds degrade. Why do their minds degrade? Because they aren't talking to people. They're not having conversations. They're alone and that's why they're in a home. Right. Like whatever, for whatever reason, Right. They just don't have the ability to get up and move around and they need some extra help to live. And so these senior care facilities play a very important role. Jeremy is an AI expert and he says, well, I'm having some pretty cool conversations with AI, like with voice. Why don't I just build a system that calls these people and talks to them all the time? Right? And so he uses a few technologies. Right. I'll walk you through them really quick. Okay. He uses something called 11 Labs. Okay. 11 Labs I.O. 11 Labs allows you to, let's see here, go to app, Let me in. But allows me to talk to like an AI agent. I could talk to somebody. I could use this support agent, whatever. Right. Test AI agent. Hey there, I'm Alexis from 11 Lab Support. How can I help you today? Yeah, could you just tell me about your pricing model? Our pricing model includes a free plan with limited characters and various subscription tiers offering more characters and features. All right. By the way, did you notice that that was my voice? Right. I trained it to talk like me. It took me about 30 seconds to teach it to train to talk like me. I could teach you guys that later. Okay. But it created a system or Jeremy created a system where he uses 11 labs. He uses an AI like sales bot that he built and. Or not that he built that actually, that there's, there's ones out there. He use. He uses a CRM that can, that has APIs that can accept all these different things. And he put all these different AI systems and just automated automations together. And so what he did is this. He called up UnitedHealthcare, which, you know, manages their insurance for this, for this facility. And he said, hey, I have this solution. I, you know, if I can, I sell it to you guys at this center. You know, full disclosure, I own it, right? But I want to sell it to you so that we can prevent more, you know, more Alzheimer's and, you know, other diseases of the mind that happen over time. And they said, okay, yeah, hey, great, we love this. You know, we're always looking for things like this. You know, there's. There's actually other services out there that people pay for where it's actual humans that will call these people and, and set this up, right? And so he just said, I'm going to make a, a faster, better version of this, right? And so he got United Healthcare to agree to pay $90 per month per, Per. Per person that's in those places, right? So now he's got an $18,000 contract with his own facility. And he says, okay, I can now expand this to other facilities. So then he goes and he talks to his brokers that sell him services to his retirement, whatever. He's like, hey, I have a product I want you guys to go sell to all of your people, right? I'll pay you commission. And not only this, you can tell these senior care centers will pay them $10 per month per person that lives at that senior care center. And so, you know, if you have 200 people there, you can add an extra $2,000 to your operating budget, right? Just by setting this up so it doesn't even cost anything for them, right? He takes some of that profit and gives it right back to the center itself, okay? And so he's now built a system that automatically he. That automatically messages these senior care centers. And by automatically, I mean someone else does it. The broker talks to them, okay? Then they get it set up. That's what his little customer service people do, right? They just get it all set up. But then he has a system that automatically contacts each nurse, schedules a time. It's all A.I. he does nothing. His assistants do nothing. On this part. The AI contacts the nurse, schedules a time, gets it in the calendar, tells the system. By the way, if all 200 people all want their phone call to be at 2pm on Thursday, great. No big deal. There's just 200 calls to the API. Who cares, right? And then he built a system that keeps and manages those conversations, manages those relationships, manages those memories. And so when they get on the phone, hey, hey, Bertha, how are you? Right. Hey, how are your kids? How is your daughter's track meet? Your granddaughters tracked me last week. Hey, it sounds like you had another baby. That's awesome. Right? Whatever. It remembers the AI records the entire transcript, puts it into their profile, and when they call them up every single time, it's like you're talking to your old friend. And. And the conversation is actually very, very enjoyable. It's very natural. It's gotten to a point where it's good. Now, here's the cool thing. If AI voice agents get better, does his solution get better? Absolutely. What if a competitor to eleven Labs comes along and builds a better solution? What does he have to do? Swap out an API key? Boom. He's got the latest and greatest. What if something gets better at writing things? What if there's a better power or an AI powered CRM? Whatever. He has six systems all connected together that are on complete autopilot. He can get on Virgin Galactic and go into space with Katy Perry, and his system is still working. Okay, that's pretty dang cool. That is pure surfing. So we've gone over a few examples, guys. I'm going to share one more example. I have a company called Velocity Access, and my idea for my company is so good, right? It's a venue management system that helps venues, like a complex one, like Sundance Ski Resort, for example. Sundance Ski Resort has, you know, restaurants. They've got. They've got a couple bars, they've got a ski lift on all the operations tied to that, they have a concert venue, they have lodging, they have rental places, they've got all kinds of different things, right? And they have about 10 pieces of software that they use to run their venue, and they have to try to keep it all together, right? Now, at Sundance, the only place that all of the data from customers comes together is in mailchimp. So whether you spend, you know, a hundred dollars a year at Sundance or a hundred thousand dollars a year at Sundance, you're all in the same marketing funnel. You're all in the same communication channels. You. There's no personalization. There's nothing. So my idea was, hey, let's build a system that does all of those things in one, right? You will handle the restaurants, we'll handle this, right? And so when Devin shows up and he's a big spender, it knows, hey, Devin always orders this filet mignon. Hey, he always orders this drink, he always orders the calamari, and he always orders this specific salad, whatever, right? So when the waiter walks up says, hey, Devin, welcome back. Can I bring you your usual drink? Yeah, that's perfect. Hey, can I suggest something different? Sure. Hey. Okay. I already know you want your steak medium rare. Perfect, right? It knows Devin. And when Devin checks into the hotel, it knows his favorite drink. And so they say, hey, we've got a big spender showing up. Let's take a drink and put it on his nightstand. It's say, welcome. Welcome to the Sundance Dev. Right? That was my. That's my idea. Right? That's what we're building. We're building this amazing connected piece of software that really understands the customers because it understands every single touch point that happens. Right? Who else think that's a good idea? I think it's a good idea. Right? It sucks. It is the worst swimming idea in the world. Why is it swimming, guys, by the way, I've spent $3 million building it and I've scrapped it. I've deleted every line of code. We're not using any of it. We're rebuilding the whole thing. That's the end of the story. Why is it swimming? I'm glad you're all kind of struggling to see this one because it's important. It's really important. We're going to go just a couple minutes over here again today and I want to finish this discussion. We'll be done in four minutes. We'll be done by 12:05. Okay, you can get to lunch. But. But this is what it is. I am competing with like 50 companies in every single thing that I do. Restaurants. I'm competing with Toast and all those guys ticketing. I'm competing.