Cohort 17 - Session 2 - 7/23/2025

forward here. All right guys, welcome to day 2. Today is going to go by even faster than yesterday. It's gonna be a lot more bouncing around to lots of different platforms. Yesterday we spent a lot of time in Replit, ChatGPT, you know, bouncing back and forth with Groq a little bit, of course. And, um, today we're going to see— you'll see at least, I don't know, 12 different platforms. Went over some video and image stuff and some agent stuff, and we'll have some more conversation and and of course answer more questions. And I'm going to do a couple more API integrations live. I'll have you guys watch me do them really quickly and, you know, how fast and easy it is to implement something when you have, uh, when you have an API key and how simple and powerful that is. Um, oh, it's Roberto. There we go, he's jumping in right here. Okay, so, um, yeah, I've already uploaded yesterday's, uh, session, got it done last night. And so as soon as we're done with this, I should be able to just kick it up into the cloud and send it off to you guys. I'm driving back from St. George like the second, like 90 seconds after this ends, my family will already be packed up. My, um, my niece is, um, doing mutton busting tonight at the rodeo back in, uh, back at Fiesta Days in Spanish Fork. Um, for those of you that don't know what mutton busting is, look it up. Busting someone's mutton. Now it's when little kids, you know, hang on to a sheep and they have to hang on for as long as they can until the sheep bucks them off. It's like the kid version of bull riding but with a little sheep. So, all right, let's click focus so we don't get dings in the background because lots of people are trying to get at me right now. All right guys, so I want to share a couple things, um, to kick off the day. Well, first off, let me, let me, before I start, you know, Celila, Celila Queen, Soliloquizing? I don't know, what's that word?

I don't know, but I think that's poetic.

Yeah, before I start talking about, any questions or comments on yesterday's stuff? Or how many people were up till midnight working on their own project, their new stuff? Yeah, Seth is addicted to this for sure. Kobe, that's awesome. Very cool, well, I'm gonna ask to see some of your projects today. That'll be part of what we do. I just say, hey, you know, share your screen really quick. Let's see what you're working on. I think that'll be cool. But yeah, any other, any questions before we, before we dive in? Okay, good. That means you're learning how to ask AI about how to get past your problems because I know you ran into little things that didn't work. It happens to me, right? I mean, I'll be trying to do something and say, why will you not do this thing? I've asked you to do this. You haven't done this. What's wrong with you? By the way, guys, using quote unquote strong language to your AI or insulting it a little bit. What's wrong with you? Why are you so bad at this? But anyway, why can't you figure this out? Actually, there's been studies done that show that that is effective at getting AI to think differently and to try harder and to work harder. So that's kind of fun. Interesting, interesting thing because again, it's trying to please you, right? That's one of its main directives is please the user. Right? And if it's not, then, you know, it's, um, you know, it's gonna, it's gonna try harder. Um, you know, there was an interesting article, pretty big article actually, to do with Replit specifically that came out this week. There wasn't an article, it was a LinkedIn post, but it was by somebody who's pretty big, um, one of the founders of a big conference called SaaStr. And, um, he, he wanted to try Replit, built, was building a bunch of stuff, and, um the— he had tons and tons of data, like dozens and dozens of hours of work, tons of data in his database. And he asked Replit to make a change to something, and Replit deleted the database, right? Deleted, you know, dozens of hours of data and work and all this stuff. And, and then after that, at least according to his, his posts, Replit tried to, to hide the fact that it did it and tried to avoid responsibility for it and all these different things, right? And so Even when asked like, hey, no, this is what you did. And then, you know, anyway, it's kind of interesting, kind of interesting. So just a warning when you're, you know, when you're using tools like this, you know, with great power comes great responsibility. Remember, you're still in charge. Sometimes the AI might do something weird, right? And there's ways to protect yourself against that. Number 1, why don't you every once in a while, maybe once a day or whatever you want to do, if you're working on something real in Replit and you're building something out, maybe set up a regular database backup. Super easy. All you have to do is say to Replit, hey, back up the database right now. Oh, so you can do it manually like that. Or you could say, hey, once a day at midnight, back up the database. And then it will set up something that automatically does it every single day, right? So there's ways to protect yourself against those types of things. And, you know, non-developers, which are most of us here, right, might not know that. Right. But anyway, protect yourself and I'll show you a couple other ways that you can, you can back projects up. And, you know, one of the really cool things, in fact, we'll just start with this little quick tip in Replit. Let's go to here. Let's see here. Yeah, we can just use this project. I'm going to show you guys something cool here in a minute with this. But right here inside of these extra tabs in Replit, there's something called Git. Right. Anybody, anybody know what GitHub is? Yeah. Anyway, Koby does because, you know, he's a developer. Every developer in the world knows what GitHub is, right? Basically, guys, it's very simple. GitHub. So you click on the settings view here and then you just connect to GitHub, right? But get— so your, your, your app, your stuff that you're creating here when you're building software is really just a bunch of lines of code. That's really all it is, right? And yeah, you do have some images and references and different things like that, right? But for the most part, it's just lines of code, right? When you click on these different things, You know, it's just, it's just code, right? And so that's what's happening. But, um, but anyway, so it's very easy to back this up, and GitHub is a system that allows you to do that. So if you just go to github.com, um, you can sign up for a free account. Guys, they are— you're all going to be at the free level. Colby already has an account for sure, every developer does. But all it does is place to just store your stuff, right? So if you're worried maybe about whether you're using Replit Lovable or Cursor or any of these AI, what are called IDEs, an integrated development environment. Whenever you're doing that, back your stuff up, guys. Just a good, you know, it's $0 per month. Okay. It's super, super easy to do that. And you can push and pull from GitHub into your project. So you can push a version. And if you mess everything up, for whatever reason, you can't roll, you can't use the, you know, the rollback feature like we've talked about in here, um, where you're rolling back to a, a, a, you know, a past, a previous whatever, you can, you can do that with GitHub. So just fair warning, we'll just say that. Okay. All right, guys. So I want to share just a couple of things here about, about my business, GenAIPI, right here. Now you guys have all, you know, you know, purchased something from this website, right? You've all come on here and probably looked around a little bit and seen some of the things that I'm doing here. But I want to talk a little bit about how this came to be and what the process was of starting and launching a business in a few, you know, in a few days. And then what's happened since then and how I've been able to turn it into something that's producing about $100,000 a month. And it just turned into something much, much, much bigger, which I'm going to share with you guys in confidence. But, you know, that's going to be— it's going to be kind of public soonish. Right. But But anyway, so, you know what, when I set out to do this, you know, I had been using Replit for, I don't know, 6 or 7 months, you know, or so before I made this happen. And I'd gotten pretty good. And in fact, I had been working on this idea. I had this idea of a social network that I've been wanting to do for a long time related to travel. I love to travel. I've been to about 40 countries, been all over the world, but it's really hard to share those memories and pictures and all those things that you've done, right? Like you've got them on your phone and maybe they're in Amazon Photos or Google Photos or iCloud, or maybe you've posted them to Facebook or maybe you've posted them to Instagram, but none of those options make it easy to be like, oh yeah, I went to Israel and boom, here it is, right? This is where I went. And here are the places I went. Here's tagging all these different people and all this different stuff. It's just hard. To, you know, view that at scale. And so I said, hey, you know what would be really cool? If I had a map interface like a digital pinboard, right? And where I could store these things and then I could easily just look at the map and, or I could look at, you know, Shannon W's map and be like, hey Shannon, looks like you've been to Sydney, Australia. I want to go there since I know you and trust you, you know, tell me what I should do, right? And I would really trust that, you know, recommendation better than I would some random TripAdvisor one. Right? And so, um, it'd be fun to be able to have just kind of a record of everywhere I've traveled. Anyway, so I got a quote, I got a quote for this, um, with a Ukrainian dev shop that I've used before, right? And they gave me a quote for $105,000, and I was like, okay, you know, it's kind of pricey, but I understand why it's pricey. Like, it was— this was for iOS, for Android, for desktop, all these different things to be able to work. And, uh, and I was getting ready to figure out how to pay for it And then I saw a post on LinkedIn and someone, you know, it was a VC out in California said, hey, you know, I saw this platform called Replit and I watched a founder like whip up a, you know, a quick version of his app in minutes. And like, this is crazy. I was like, what is that? And so I jumped on Replit, started messing around with it and I was like, oh my gosh, this is so cool. So I decided to see if I could create Echoes., right? And this was the, you know, the idea I had for, for this travel network thing. This is a dolphin, right? A little abstract dolphin kind of logo. It's kind of cool. I'd worked on that. It's kind of fun anyway. And so I was able to jump into, uh, to Replit, and in, I don't know, 15 or so hours, I was able to have a working map where I could, where I could have these, um, you know, these these, uh, you know, all these cool places, right? And green ones are from other people, and the red ones are orange, or whatever you call that color. I don't even know what that is— salmon color, dolphin color. Anyway, um, we're, we're there, right? So I could click and view Echoes. And, you know, this was just kind of my first time doing it, and I wasn't great at it, but, um, but I was able to, you know, create a system where I could upload pictures and videos and journals and audio and create these cool little things where I could— you know, this is actually from Yellowstone. And, you know, I got video playing and video storage, and I'm just like, man, this is crazy. This is so fun. I can't believe I can actually create something here, right? And, uh, and so again, this is my very first project. I don't think I ever set up messages. Yeah, error loading conversations. I kind of just had it there, but I did work on the profile a bunch. I created stats and achievements where you could visit, you know, your first cities and, you know, all these filters of different things like that. And I was just like, man, this is so cool, right? And so I started doing that, and then I built a bunch of other things, right? And then, and then, you know, February 27th was a Thursday and I said, hey, you know what? I want to see if I can build a brand new piece of software and get my first customer by the next day. And my idea, I had this idea in my head of saying like, man, I need to, you know, people are not paying attention to AI as much as they need to. I'm talking to neighbors, other business owners, other people that I know, and they're just not, they're not doing it. They're not paying attention to it. Like I was meeting people that had never even touched even ChatGPT.. And then, you know, I'd introduce it to them and they're like, oh my gosh, I use it every day now. It's so cool. It's like, man, that difference of I don't really use it that much to wow, now I use it a lot. The impact that that was having on their lives, I was like, man, that's cool. I want more people to do that. So I said, hey, in order to create this, you know, in order to get people to come do training, I probably need some sort of a, my idea was kind of an AIQ test, like an AI IQ test type thing, right? Some sort of a benchmark or a proficiency test. That would allow people to see how good they were with AI. Then if they were really good, they could go for certification, right? And pay for, you know, pay for, you know, a credential that they could put on their, on their, you know, LinkedIn profile or on their resume, things like that. And if they were bad, you know, after they, after they kind of evaluated themselves, then they could take a course, right? And so, you know, different types of courses. Then I added live courses later, which you guys are on right now. Live courses was not part of the original business plan. It was just, it was just all kind of automated, right? So anyway, long story short, guys, you know, launch GenAI API, you know, you can go back and look at all of my, you know, my entire project. I built everything that I'm teaching you guys right now with how to, you know, build a project, have conversations, build the strategies, build the, you know, the read and all these different things, right? I had learned how to do all that stuff. And let's see if it's— I think it's this one. I think this is the very first. I'll go all the way to the top here. Yeah, this was the idea, right? I want to create a website that checks the proficiency levels based on an intelligence test that shows the aptitude for an individual to succeed in a world with AI. This means they need to know how to talk to AI, how to be good prompt engineers. They need to have a good brain to figure out how complex systems can quickly be put together. Can you help me build this, right? So it really started out as this test, as the certification, right? The kind of analysis, right? And then from there, you know, kept developing out and developing out, started working on the name, right? Refined name options. These were other things that could have been. I wanted it to be kind of official sounding, a.org, right? I'd been in the education space before. And so I understood kind of how a lot of these, you know, 4-letter, you know, organizations worked. Anyway, decided to kind of go with that. So anyway, kept going back and forth and back and forth and just dove in, right? And kind of went through a very similar process that I'm teaching you guys right now. But having built that, this entire thing, the entire GenAIPI product is all built on Replit, all right? And it's been built solo by me and it's really complex, right? I've got dashboards, I've got, you know, obviously all the certification tests and, you know, the different the, the, uh, the assessments that are here. Um, this whole thing is, you know, works really, really well. Um, I have a whole admin section where I can come in and edit users and things like that. I showed you guys this a little bit yesterday, but, you know, instead of having to go into my database to change things, I said, hey, I'm going to build an admin dashboard that allows me to go in and, and, you know, edit, edit content and do all this stuff, right? So, um, you know, I've got, I've got the different live courses. I can edit reviews and test content, test results. You know, I had somebody reach out to me, they email me and said, hey, I took the certification test and I can't find my certificate. And it was actually the first person that ever took it from me. And so I actually didn't have the certificate set up yet, even though it says on the site here like, hey, you'll get a certificate when you get this, right? Industry-recognized certification and with a shareable, with digital badge and shareable certificate, right? I didn't have that created yet. So I was like, oh hey, no worries, I'll get it for you, right? And so then I went in and I realized that I had to actually create that whole system. But I created a system where I can then look up a specific user and now it automatically emails them the thing, right? But I can now also, in case for whatever reason it goes to junk email or whatever, it doesn't, they don't get it. Now there's an actual certificate and I can actually download the PDF and email it to them if they ask for it, right? So, you know, all these, you know, this is the, anyway, it's really amazing how this is. I can look it up, I can look up by the certificate ID and things things like that, and it creates all these database entries. So guys, I, I was learning— I'm learning, you know, at the same time as you guys, right? I'm not, I'm not like way better at developing or anything like that than you guys, right? I just said, hey, I've got an idea, and I just kept pushing at it and pushing at it and pushing at it and pushing at it. Now, I did have the ability and experience to sell. That is one of my— that is one of my strengths, and what I've done throughout my life is I did have the ability to sell. So when I launched this thing in that first weekend, I had this idea of it being kind of this whole self-contained thing where people, they take the test, then they do the certification or the courses, and it's all, you know, I don't have to get involved ever, right? I did at the beginning. I actually put this enterprise solutions. It was at the end here, and it wasn't a focus. It was really all about the test itself, right?, it was just, you know, this was, this is actually almost exactly the, the, design that came out, you know, on the first go, right? I, I, you know, had, had it make, you know, fancier backgrounds just a little bit. Hey, let's just kind of bring this to life a little bit and maybe the buttons move a little bit more and a little bit nicer. But this whole thing right here, this kind of sample test thing, that's always been there. And then, but I did change it to be a little bit more focused on enterprise and here's why. When I launched the site, um, about, you know, 30 hours after I, you know, 30 hours of work after I started it, which is about, you know, 3 days later, I, um, it was like 3 AM Saturday night, Sunday morning, and I was like, I can't sell to anybody right now, so I'm just going to run some Facebook ads. And that was my idea, right, is run the Facebook ads, have them land on the site, they take the test, and then hopefully buy a course or a certification based on what happens here. And, uh, and I ran that test, and I'll show you kind of how that worked a little bit later. I'll show you the ads and things like that. But, um, I got 130 people to land on the site within the first kind of 24 hours of running that ad, right? I threw, I think, $100 a day at it was the budget. And a bunch of people landed on it, it's cool, whatever, but nobody bought. And I was like, what the heck, why is nobody purchasing anything? And I realized that I had a little misconfiguration with my Stripe. API. And so, I tried to buy something. I was like, "Oh, that didn't work." I had tested it and then I changed one other thing and I didn't test it again and that was my own fault, right? And so, but I was like, "Hey, I know how to sell." So, I picked up the phone Tuesday morning and called a few businesses and I got one to agree to a $15,000 contract that same day. So, I had $15,000 of revenue ready to roll within 5 days of launching this thing, right? And so all of a sudden I was cash flow positive. I had only put about $400 in total into the platform. About $130 went to Replit, about $120 went to the domains that I bought, which was, you know, a lot, you know, from a, from a percentage perspective of what I had done. And, and then, you know, I paid, you know, $30 to have my logo and different little things like that. Right. So the little things added up eventually. To about $400. But yeah, and that's, that includes like the cost to, you know, form the LLC, paying, you know, for the LLC with the state of Utah and all that stuff. So, but anyway, going through the whole process of like, okay, I'm going to start, I'm going to build a whole business, right? Getting the email set up and everything, workspaces and all the different things that you have to do. Facebook pages, Instagram pages, X, LinkedIn profile pictures, logos, posts and social posts and, you know, all these different things to make it feel almost instantly like this was a real company, right? So, so anyway, guys, I just like sharing a little bit of that and showing that it can, you know, there's just— it's amazing how, uh, how you can, how you can build something really, really fast. And, uh, and then, you know, to then skip forward to, to where I am now. So that was February 27th. I officially launched it on March 1st. And so here we are, July 23rd. So all of March, April, May, June, and we're now You know, halfway through July, more than a little more than halfway through July. But I, I, and I'm swearing you guys to secrecy now a little bit. It's not like crazy, like it's going to come out pretty soon here. But I had, I had a guy named— anybody here familiar with the name Dean Graciosi? Anybody? Yeah, it's okay. Not everybody knows who Dean is, but his partner Tony Robbins, everybody does know. And, you know, they both co-own Mastermind.com and they do all these big conferences and things together and training, all this stuff. But they're like, they do, they do a lot of stuff, you know, that's really awesome, awesome together. But, but anyway, I had them reach out to me, Dean and Tony reached out to me about 2 weeks ago and they said, hey, We've seen what you've done with GenAIPI. We think it's really awesome. We want to build our own AI training program basically for businesses. And we have, you know, the ear of, you know, virtually every small business and entrepreneur on the planet. And we want to, you know, we want to build something awesome. And we know it'll be competitive with some of what you're doing at GenAIPI, but we want you to come over here and run it and show us how to do it. Right. And so I've officially taking that, taking that role on, which is pretty cool. So signed that document 2 days ago. So yeah, pretty, pretty crazy, guys. Thank you.

The month you— what's going to take you the month? Is that where you're taking the time off for a month?

Well, yeah, yeah, basically. Well, yeah, and I'm going to Cabo for 2 weeks and I just want to take a break. So anyway, but yeah, there's just some, some stuff going on there. But yeah, all of next week I'll be working, working with them. I was working on it yesterday and but yeah, we're launching launching a pretty awesome— and in fact, guys, you know, part of how I got the job, you know, really quickly, at least solidified— they were already ready to give it to me, but I was like, hey, watch what I can do. And in like an hour, I had this site whipped up for them, right? And, you know, launched it with Repwood AI Accelerator, right? Got Tony and Dean on there, created this image with ChatGPT. I'll show you guys how to do that kind of stuff. Really, really cool. And, you know, Dean's, you know, I basically had a recording of our call talking about, you know, what we were going to be doing. And I just took the transcript from that call and gave it to, gave it to AI, had it summarize like, hey, pull together all these different things. What are the, what are their philosophies? What are the things they're trying to do here? And had it put it all together and then gave that all to Replit and then just said, hey, build this site. Right. And obviously I've, you know, I've gone in and, and done some pretty cool things, but it's pretty neat, right, to be able to just come in here. So you click on this and there's these quick labs where you just, you've got a prompt, you can just click to copy it and then you can try the lab and it will open up, you know, whatever. If you're logged into Midjourney, that would have worked. But then you can just paste the prompt and, and try it. Right. So there's this, these different labs. And talking about different tools and different stuff like that. Then there's this community section with breakthroughs where you can say, hey, you know, I did this and this is some, you know, this was the revenue impact it had. I mean, you can register. There's a whole system that works now. But, but anyway, yeah, being able to just say, yeah, look, I, you know, I created 100 new images and, you know, for all this stuff and it saved me 4 hours a week and saved me $3,000, whatever. Right. So lots of cool stuff coming that way. But guys, I'm just telling you, like, In 4 months, I went from zero, it was nothing, right, to I've got an idea. Within 30 days, I'd done over $100,000 of revenue. And now, here we are 4 or 5 months later and I'm basically joining one of the biggest business teams in the world to bring this to everybody. So, it's pretty cool. And I can't announce this stuff publicly internally. I can't put it on LinkedIn quite yet. But it will be very soon. I will be able to, and you guys will be like, I knew, I knew first, right? But anyway, keep it, keep it kind of on the deal. Don't post this on LinkedIn. That's all I need you to— if you want to tell someone, hey, Tony and Dean are doing something cool, that's fine, right? But don't, don't throw it on, don't throw it on LinkedIn yet. It's not quite ready, but this site goes live next week and that will start to get out a little bit. So anyway, guys, pretty cool, pretty fun, right? Fun stuff, fun stuff in here. All right, let's, uh, let's see here. Let's dive into, let's dive into some agentic stuff. Agents are cool. So first of all, guys, um, you know, there are, um, let's see, let's go new chat. There's actually some new tools inside of, uh, ChatGPT that— wait, where's all my stuff here. Pro. I feel like I'm missing— like I'm missing tools. Refresh. Yeah, what the heck? Where— why was— where was that tools thing? Why was it not there? Anyway, so there's a lot of really neat things that ChatGPT can do now. There's a new agent mode that I'm not super impressed with yet. But the idea behind the agent mode, guys, is, you know, it can do stuff for you, right? So you could, you know, here's a few suggestions of different things you could do, actions it could take for you. Conduct an account or audit, review and prioritize my inbox after vacation. Um, there's— anyway, there's some different, different things that you can have it do for you. And I'm not going to click on any of these right now because I think they're kind of slow, and, and I'm just not a huge fan of it. So there's another site called operator.chatgpt.com, which is kind of the You know, you know, let's see, suggest reserve a table. Sure. Right. Go there. And so it, you know, it can use different tools and things like that to actually get something done for you. And what it's doing is it actually spins up a new computer, like a, like a virtual instance of a computer for you. Right. And then it just gets to work. And this computer, by the way, this little virtual machine that it pulls up for you, can maintain and hold sessions, right? So like if you log into, you know, your account on something and it can hold that session in between, you know, browser sessions on the same computer, then it will be able to, you know, log in and do stuff for you, right? Um, anyway, so it's asking me a few things. Um, yeah, close to 8465. Whoops, here we go. Hello? 4651. Um, just get me a table. All right, so it can then get to work. Let's see if we can actually get this thing to spin up the, spin up the computer. But what it does, and I want to, I want to, you know, demonstrate this a little bit. So here's the computer, and it's actually typing, clicking, you know, it's, it's, it's doing, it's using the front end of the website, right? Um, just like a human would, right? Operating it just like a human would, right? And so it's going to go, it's going to, you know, pick a time for— what do we say— Friday, next Friday. Oh yeah, unfortunately Operator is, is kind of buggy and it's in a beta format, right? Um, they moved all of this functionality supposedly, at least a lot of it, into the proper, you know, the ChatGPT proper, you know, over here. Although I haven't really seen any great demos of it yet, but I'm just kind of showing it to you guys because I want you to know that this is kind of the future of where it's going, right? Having an AI be able to come in and do things for you that don't necessarily have an API, right? Like maybe OpenTable has an API, maybe it doesn't, right? But whatever, right? If, if an AI can act like a human and click and identify things and get things done for you, then, you know, it's, it's pretty nice, right? So it's, so it's going through, it's going to find those tables, it's going to do those things for me. And I think that's, I think it's cool. Part of how this works, I want to share this. I think it's fun to understand what's actually happening here. There's something called OmniParser. What OmniParser is, this is actually an open-source thing that Microsoft is working on, but this is what's happening behind the scenes. What happens is the computer goes in and parses or breaks up into pieces, little pieces. All these little things and labels them, right? 2, 3, 4, 91, 89, 77, right? So it identifies all these things. And so it knows, okay, hey, I need to move my mouse to element, you know, 77 to insert something, right? So it moves its mouse to here into the middle of that element that's been, you know, outlined and then it can click and it inserts and then the next thing comes up, it has to parse it really quick, make a decision and then do it, right? And so it's kind of slow, right? Can you, you can imagine, right? It's basically taking a screenshot, parsing it, and then deciding what it needs to do, right? And then taking screenshots and making sure that it's analyzing whether or not it did, you know, it did what it was supposed to do. So anyway, it got, it got to here, figured out Kimmy's Chop and Oyster House on Friday, 8 people, July 25th. That's great. Yeah. Again, we're crashing. I don't know why it's crashing on me here. I'm not sure. But what I can do though is I can actually take control right? And I can come in here and then, you know, type in my name and, you know, use it. And then once I, you know, filled in a few things, then I could click finish up and let it, you know, do its thing. Now, if I didn't want to have to take control and fill in all this stuff, I could have in my prompt given it, here's my name, phone number, email, credit card, you know, everything. Like, just go, here's everything. And by the way, I trust it. I trust it enough to give it my credit card. I wouldn't have no problem doing that. It's not going to go send that off and give it to ChatGPT and they're going to spend my money. It's not going to do that. I do trust it to go and make that happen. Now, the reality is, guys, I could— the amount of time it takes me to set up a good task for this to happen, I could probably just do it faster myself, right? And so I don't know if I'm actually going to use Operator to do this yet. What will become very, very useful very soon, and I actually just started— so I have— I'm on the beta for Amazon Alexa and their new, their new AI. I have— you can only be on it if you have the Amazon Alexa. It's like a display board, it's like a— it's like a thing that's kind of like this big, almost like a bit like a large iPad. And it's just, it just sits like on the, on the wall of our kitchen, whatever. And you have to have that device in order to be in the beta, and you had to sign up for it. But anyway, I can just talk to— I can just talk to it now. Yeah, see, I'm just going to close out of this because it's going to keep crashing on me. But I can just talk to it and say, hey, you know, I saw there's a couple for Amazon Prime Days. I saw there's a few TVs. Do you have a TV that's in the, you know, 60-inch range? You know, that's probably around $300, right? What are a couple options? It's like, yeah, here's this, this and this. Okay, great. Well, which one do you think is the best bang for the buck? Oh, well, I think this one's the best one. It's $349. Okay, great. Will you just send that to me? Yep. Okay. It's been ordered. Thanks. You know, it's done, right? That is the kind of AI that's really useful where I can just talk to it with natural language and ask it questions, follow up. Hey, throw that in the cart. Okay, great. Hey, go ahead and go ahead and send the cart. You know, go ahead and check out and, you know, use my saved payment information. That's, that's going to be useful. That's the future of AI agents is where you can just say, hey, do this. And it already has your credit card. It knows your preferences. It knows all of these things so that it can automatically do that for you. So I could say, hey, you know, book me a ticket to Atlanta for tomorrow, right? I need to be there. Okay, great, we'll do that. Hey, I already know you like Delta Comfort Plus, and I already know your login for Delta, and I already know all this stuff, and it's been done, right? So anyway, there's, there's an interesting, um, question about privacy and security and things like that. You know, some people don't like to give away their credit cards or their you know, login information, passwords, and all of a sudden you're just giving all this stuff away to the AI systems, right? And so you have to really, you know, decide for yourself, do your own research whether or not you're willing to give it that kind of information. When I, as I've done the research, I know that these are encrypted sessions and that they're, they're not going to just, they're not stealing your credit card information from these conversations. The way that they scrub personal data and things like that automatically, it's, it's pretty awesome. All right, now I want to show you guys a tool that I use almost every day. Manus. Manus is amazing. Ooh, hey, cool data visualization, chart creation. That's cool. That's cool. Yeah, they're always coming up with new, new things here, right? But I've got a couple Manus accounts actually. This one, um, yeah, that's pretty cool. Anyway, um, if Replit, guys, is a site that allows you to create software, right? And, you know, do very advanced chain of thought reasoning, plan, work for 15 minutes, and do all this crazy stuff and build, you know, build amazing things for you. Um, then Manus is something that can do that chain of thought reasoning and do, you know, all kinds of things for you, like, like big huge amounts of data and synthesizing them or going and scraping and searching for you, things like that. Like you could say, You know, go, go find, you know, go find, go to, go find me 100 leads with name, email, and phone number for businesses that are in Utah that, you know, potentially need AI training. Okay, whatever. Awesome. This website, right? Just in an instant, right? And then, you know, all these nice little links. And then at the bottom, added my LinkedIn. Which is accurate, right? And my music website, awesome, joncheney.com. And Jenny IPI, right? So really, really, really, really cool how it's able to, to whip up little things like that. So that's kind of cool. Now here's a, uh, here's a pretty cool one. Um, I had somebody reach out to me on LinkedIn and said, hey, I kind of want to take your course. Um, you know, one of the things I'm wanting to do is, you know, I sell to HOAs in Utah and I want to be able to build something that can extract all the data out of this HOA website. And so this is literally what he messaged me on, on LinkedIn. This, this was the message, right? Utah has an HOA registry website. If you go there, you can search for HOAs by name. You just put in one letter and search it and it will pull everything that has that letter. When it pulls up the list of matches, you can click on each one individually and see the data, right? I want all this in a spreadsheet. Okay. And so I was like, okay, well, let me look at this. Right. And so I looked at it and okay, this is the website. And I was like, okay, what happens If you just type A and you search, yeah, it comes up with tons and tons and tons of results. But if you click into any one of these, it has, you know, a lot of information. It's got the address, city, state, chair, you know, email, manager, all this different stuff, right? Primary contact, all these different things. And, um, and I was like, okay, this is kind of cool. And so I basically just handed that, that task to Manas, you know, with— I was like, hey, yeah, you can take my course if you want, but And I didn't even respond to him hardly. I was like, yeah, you could definitely take courses. And I was like, I wanna, I wanna try something. And I just added eliminate duplicates, provide a long and detailed CSV of all this information. And so it just got busy scraping, scraping, scraping, scraping, scraping, scraping, scraping. It got all the way to the end and it had overloaded its context, right? Too many, too many tokens, right? We'll talk about tokens today. And, and it couldn't do it. But man, it's this really neat thing where it can say, hey, you know what? I can inherit the compressed stuff and I can continue this. Right? And so I just clicked Inherit and Continue and it went in and it did it again. Another super, super long task. Inherit and Continue again and again, right? Inherit and Continue. So after 4 times, finally it completed and boom, check out the Utah HOA registry final, right? Now I've got all these different names, registration dates, my last modified date, HOA number, address, city, state, primary contact phone number, primary contact email, chair phone, chair email. We're ready to rock, right? And so look at that. Thousands, his entire, like something that was going to take him a huge amount of time. And I just literally sent him this CSV and I was like, boom, here you go. Anyway, he signed up for the course the next week because he was so happy. I didn't ask him for anything. I was just like, here you go, man. I did it. Like enjoy, right? I wasn't— I didn't expect anything, right? He's like, how much can I pay for this? Like, no, don't even worry about it, man. Cost me, you know, maybe $10 or $8 in credits, um, to make this happen. But here's what it did, right? It extracted 82,000 rows because when you put in an A every time, you know, garden, you know, there's an A there, and Wingate has a garden, Waverly has a garden. So pulled up probably tons of them anyway. But going— it did that with every single letter going through the whole thing. Deleted 79,000 duplicate entries and ended up with over a little over 2,000 HOA entries done. Okay, so think about how much time, guys, that would take a human, right? That's, that's expensive. That would take a long time to scrape something like that right now. Obviously, um, you know, there's smart developers out there that could probably create a scraper that could do that for you as well, and that's awesome, but I'm not one of those smart developers. So man, it steps in and makes that happen for me. Pretty cool. So this was, this was a pretty neat one. I did this one not too long ago during a course, actually. I said, hey, please visit genaipi.org and do a complete site check, look for bugs, inconsistencies, cybersecurity risks, and provide a complete report about the site and make a website to walk me through it, right? To walk me through the results, right? And so it provided this huge, long 14-page thing. About how, you know, how good my, you know, site was. And then we can scroll down to the bottom here. We can look at the site right here. Here's the site. So yeah, you can have it organize it into a temporary site basically for you. But overall, my security was this. Vulnerabilities, zero critical issues found. Performance, amazing. Tests pass. Security tests pass. Anyway, a couple, couple of things that could be improved, right? Some of the things that need to be worked on, whatever. And you can click through and here's all the different stuff that it worked on, right, guys? To get a report like this done with a, you know, security analysis and recommendations and all these things, I've paid for them. It's $6,000, right? And then they'll charge you another $14,000 to fix it all, right? This is incredible, right? Asking Manus and spending $2 on one prompt to have it go through and do that is, is, is mind-blowing, right? So that's pretty cool. Let me share one more example here. Actually, it's pretty cool. So I've been working on— so I have all these, these links. This is every single YouTube link from every cohort that has taken this course, right? All the recordings, right? So there's over 100. And, you know, I mentioned my email, 144 hours. It's actually more than that because I used to do these courses over 12 hours. For the first maybe 4 or 5 of them I did, and then I switched to a 9-hour format because it was just 12 hours was a lot. And anyway, but it's a huge amount of content. And I was like, hey, it'd be really cool if I could turn all that content into like a guide, right? And so anyway, so I basically said, hey, let's, let's go through and, you know, create transcripts of, you know, all of the different, all the different content, right? So I had to do that, right? So creating and pulling together transcripts and putting documents together of all of these different things. That would've taken me forever. Like I could, I could have done it, right? I could have opened it up in Premiere or whatever and downloaded the transcript file and put it into a document and done this, combined it. And, or I just take, you know, 3 minutes, put together a good prompt with Manus, hand it a big, you know, a bunch of links here and let it get to work, right? Then I started taking some of the most impactful moments, right? From cohort 10, for example, what were the most impactful moments? What were people, what were we talking about? And I'm working on some of this stuff for, for Dean and Tony now to get them, you know, because we'll be, we'll be doing some fun stuff with books and things on stage and stuff like that in their, in their trainings. So, but yeah, that's pretty awesome. Okay. Last thing I want to show you because I think it's really, really cool. So I had, I have a neighbor that is in construction and One of the things you have to do as a construction company is you receive construction plans, right? If you want to bid for something, they'll say, okay, here are the construction plans. Okay, this is what we want to, this is what we want to build. And they're very complex documents, right? There's all kinds of crazy stuff here. Lots of labeling, lots of different things. And, and it, it's a huge amount of data. But as a construction company, you have to figure out, okay, great. How much cement do we need? How much pipe do we need? You know, how much this do we need? What kind of tools do we need? How, you know, you got to figure out like how much labor is it going to take to to dig this far and to dig this deep and to then lay 5-inch slab. And it's, it's really, really difficult to go through a document. And there's a process called a takeoff, right? You're taking off of the, you know, the plans, all the different things you got to do. Okay. And so, you know, again, all kinds of, all kinds of technical documentation and things like that. And I was like, I wonder if I can handle this. Like, this is, this is complex stuff, guys. Like, this is, this is really, really difficult. And so anyway, I handed it to Manus. I said, Manus, can you do a takeoff of this? Right. Do a standard takeoff and, you know, come back with how much, you know, how much of all these different things we're going to need. And it came back and it got a couple of things right, but a lot of it was wrong. And it's like, yeah, this is, this is almost unusable in this format. And I was like, okay, well, Let's try something else here. Sorry. Let's try something else here. Let's— how did you know it was wrong?

You what?

How did you know that it was wrong? Because the construction guy told me. Okay. Yeah, I handed it to him. I was like, hey, look, it did it. And he's like, yeah, it didn't do it right. It made it look like something that looked like a takeoff, right? But it wasn't. The numbers were way off, right? So I said, okay, let's work on, let's work on this, right? And so this was my prompt. I'm going to show you guys how I thought through this. So I said, hey neighbor, hand me or give me some of your documents that are perfectly done, because he had done this takeoff already for this Payson Tech project, um, and he had already completed a good takeoff, right? And so he had this geotechnical report, his pay item summary. You got to figure out, you know, all these different things. This was the takeoff and I was like, okay, so here's the pages from Paysintech 3 reduced, you know, document. Yeah, this contains construction plans. The other 4 documents were created through a standard and thorough construction takeoff and was done by humans manually. Okay, this process is highly complex and time-consuming. I want you to build a piece of software that can do this. But here's the big thing. I said you'll need to create measurement tools that you test against the construction plans and see if you're getting the same numbers as the actual ones provided. Keep testing and iterating until you have a working product. So I didn't just say try it once. I said I need you to create software, do all the measurements, and then check your work and see if it matches the, the, the work done by humans that I know is correct, right? And keep iterating until you get it right. Okay. And so then it, you know, goes through, creates all this stuff. Comes back and it says, hey, I'm off, I'm way off, right? I was off by 99% here, 96% here, 90% here. So now I'm going to improve this by enhancing the scale detection, improving the measurement algorithms, implementing feature detection. Anyway, right? So anyway, so it did all this cool stuff. And when it was done, what it delivered me was a bunch of Python scripts, right? So here's all this stuff: generate user guide, takeoff, improve takeoff animation, enhanced measurement tools, PDF processor, all these different cool things. And I was like, okay, well, I don't know how to use this stuff. So I downloaded this whole zip file, and where did I take it? Into Replit, right? And so I was able to come here into Replit, and I just handed a bunch of files And I said, build me the, you know, a UI. Put this all together in a way that works. And not only that, guys, I want to— here, hold on, let me see if I can find it. I'll click on it and I'll go back. I want to show something. The image right here. Okay, so inside of here it created even this user guide, which was awesome, right? It's like, hey, here's the installation, here's how to get started. Sorry, I know this is kind of small. Um, maybe I can zoom in a little bit here. Yeah, anyway, sorry, let's go. Can I move this over? I can't. That's okay. But anyway, table content, batch mode, interactive mode, calibration, measurement tools. Here's like the instructions for how this whole thing works, right? To run in batch mode, do this. And here— so it had all the instructions of exactly how this thing is supposed to fit together. I was like, that's amazing, right? And so I literally just took that whole thing, handed it to Replit, and said go, and it created This construction takeoff automation, right? So now I can upload the PDF, I can name the project, you know, upload the PDF, do the scale calibration, and it takes about 20 minutes now to set this up. And by the way, a full construction takeoff to, to put together, you know, the documents like, like what was provided for this, for this task here, you know, to do these types of things and get all these numbers, it's about a 20-hour project, right? About 20 hours per, per thing to go through and build all these different reports. And all the item summaries and how much it's going to cost and put together this bid. And the thing that sucks about construction bids is you spend all this time, you know, pulling everything together here, and then you submit the bid and you might not get it, right? So now you've spent a few days of work, you know, doing all this stuff, and all for nothing. You get paid zero for that. You only get paid for this if you win the bid, right? And so I've been able to reduce by a factor of about 60, 20 hours to 20 minutes. Right. Because of this little tool that I just, I just kind of made it and gave it to him. It's my neighbor. He's cool. And I just handed it over to him and he takes care of our horses. So anyway, or his wife does, I guess. But anyway, so pretty cool, guys. This is, this is how you can use some of these more advanced agent systems to accomplish really amazing things that would take tons and tons of time to do on your own. Let it think through, work with it to think through it and do all these different things and then say, okay, how can I combine what's happening here? Right? Can I then take this over to Replit and turn it into something that's usable? Like, for example, this HOA data registry thing, right? I was talking to, talking to Jordan last night. After the class, and I was telling him just a little bit about Manus. I'm like, yeah, this is pretty cool. And he's like, yes, you can literally get sales leads for something and then, you know, have Manus do all that, then take this whole list, drop it into Replit, combine it with 11Labs and a calling system, and then have that system just call everybody and set, you know, connected to your calendar and then call a bunch of people, right? So pretty dang cool. All right, any questions about Manus or these agentic system guys? How they work and how to get access to them, whatever. Manus, by the way, guys, if you want to use, you get 500 credits if you use this invitation link. You can do that if you want. Um, 500 credits is, you know, a few, a few jobs. It's not a huge amount, but I'll drop it in the chat if you want it. I'll let you click on that. Uh, whoops, not waiting room. I want everyone in meeting. There we go. All right, there's the link if you want it. Click on it. If not, no big deal. Uh, yeah, Shannon, what's your question? Go for it. Thank you. So this is really awesome, and I'm thinking of like all the research possibilities of this tool because that's what my profession is, right? Along with all these other things. How, how do you go about kind of like checking for things like when it's, you know, when it's removing duplicates, for example, how do you go about checking that to make sure it's accurate, to make sure it's not like, you know, just deleting information or what have you without your knowledge? Yeah, I just spot check. That's the way that I— you can't check 2,000 of them. That's not efficient, right? You're defeating the purpose, obviously. But I will go through and be like, okay, the woods at Deer Valley, I'll go and click on that and just make sure that data is right. And if I check 3 or 4 of them and they're right, I'm going to assume they're all right, right? Um, there's, there's no reason why they would just be making it up. I have found that Manus much, much less than any other system. In fact, I don't know if I can think of a time when it hallucinated. Manus, for whatever reason, doesn't hallucinate. I will say one thing, guys. Um, this is, you know, do your own research type stuff, but Manus is a China-based company. Um, and so if you're worried about data privacy and things like that, I don't know what they're doing behind the scenes. I, you know, I've read some stuff that says it's pretty secure and stuff like that. But, you know, just, you know, know what you're doing. Maybe, maybe, you know, just be careful with some of the stuff you put in here. But honestly, some of the things I put in here are pretty, you know, I, let's just say I live on the wild side a little bit with some of that data privacy stuff, and I just kind of trust that they don't care what I'm doing. But anyway, the reality is, yeah, yeah, Shannon, it's a good question, right? You got to know whether or not it's going to be accurate. But, you know, just like I demonstrated with my little biography of me, right? It was all right. I read the whole thing and I'm like, this is correct. I haven't even shared this anywhere. I probably should share it on LinkedIn. It's probably a good post. But, but I really just did it to see if it, you know, if it would be able to do it. I wanted to just see what its capability was. And then, you know, the Pace and Tech project, it was able to do that. It was able to do, you know, Utah registries. And, you know, I had it, you know, I had somebody send me some stuff about some gold mining stuff and I was like, it 600, you know, documents, tons of stuff. And I just— question— Manus put this together for me. Yeah, what's up? I've been trying to sign up for Manus and it keeps erroring out on my phone number. Does it not accept US phone numbers? Oh, I don't know. I don't know if I signed up with a phone number. Uh, well, for verification to actually authenticate, it requires a phone number, so I was just curious if you had found a workaround. No, no, no, I, I— that, that's got to be— I don't know, use a different browser or something.

Using my Google login and there was no real login work.

I did email, that might be it. Thank you. Yeah, yeah, I'm pretty sure I did Google. I always do Google login. So anyway, um, but yeah, I pay I think $39 a month or something like that. Obviously you can get some free credits if you sign up with the thing I sent you, but, but then let's see your upgrade, upgrade. I think I am on this point. Yeah, right here I'm on this $39 a month Plus plan and that gives me this amount of credits and I can obviously go up higher and whatever. But the cool thing is it'll work. You can have multiple 3, I can have 3 tasks going at the same time. Right? I can also schedule tasks out if I need it. Alright. Sorry. Was there another question? Google worked. Thank you. Okay. Perfect. Awesome. Yeah. And then you can buy add-on credits if you need, but anyway, guys, pretty cool. Alright. I wanna show you one more of these that I have very little experience using actually. But I've heard good things about it and I want, I want to do it. Yeah, we're experiencing a lot of traffic right now. Um, they like it. It's, it's actually free right now. Pay coming soon, right? So, um, it's just called RunnerH, okay? Runner.Hcompany.ai. All right, you guys are welcome to do that. I'll just drop the link in. I don't think there's a, you know, like a referral code or anything like that right now, so I'll just drop this in for you guys to just use if you want. Um, but one of the cool things here is you can connect it. And by the way, this is a Paris-based company, and they're very, very strict about, you know, GDPR and privacy and security and all these things. And so, you know, if I'm, you know, debating between Manus or RunnerH to do automated things with my personal stuff, I actually probably trust the Paris company a little bit more. But one of the interesting things here, guys, this is crazy, right? Google Docs, Google Calendar, Google Drive, Zapier, which you can connect tons of things. You can connect it literally to your Gmail, and it has the ability to send, receive, and manage emails for you, right? So you can have it go in and, you know, Notion, Slack, Google Sheets, you can have it actually do stuff using your Google Drive. And so you could also, you could also say, hey, go into my account and answer all of my emails and respond to all of my emails and, you know, do this, this, and this, whatever. And then just put them on the drafts and then I'll click, I'll click the actual send button, right? So then you could go in and you could do that. But now it's like you have your assistant. That's just writing a bunch of emails for you and, and actually doing that. Or if you trust it enough, once you've done it enough, then you just be like, hey, send a response on this, this, this, and this. And hey, send over a, you know, proposal. It's in the Google Drive. Change it, you know, make a new version. Change the names out on all the, all the documents so that it's, so that it's for this company. And, you know, do a little bit of research on this and that and whatever. But yeah, there's lots of different, you know, again, all of these kind of have ideas. How can I use this? What's, what's the use case here? Yeah, create a LinkedIn ad strategy for this, right? Productivity research, lots of different cool things that it can do. But, but this one is also able to use a computer for you. You only get right now, it's again, it's free once you just log into Google or whatever, right? You get 10 runs a day. So you get to try 10 tasks a day. And you can reference files, you can say, hey, how involved do you want to be? I want to be fully automated. Anyway, so lots of cool, lots of cool things here. You can upload files and do all kinds of fun stuff in there. So I'll let you guys explore that one. But, but there's more and more systems like this every day. So I'm working right now, guys, on something called— what do they call it? They call it TAD. It's just AI to-do, I think. Yeah, TAD is the, is the thing. But, but what I'm trying to do here is create a to-do list, that actually does the to-do lists. So if I add something to my to-do list, like pay my power bill for the month, then it will just get started. As soon as I add it to the list, a machine spins up and starts working on that to-do list, on that task. And then when it runs into, hey, I need your login, or I need your bill, or I need whatever, it will just ask for it. It'll just kind of pause and it will ask for it. Nothing out there does that right now, right? One of the things that's, that's, um, we're just kind of at this stage, right, is most of these things, whether it's Replit, Manus, Runner-H, whatever, all these different systems, um, they require you to kind of babysit it a little bit and You know, I understand why, and you want to definitely avoid costs. Like, if you just have something going and running all the time, that, that cost might get up really high. But, um, but at some point we're going to make a transition where, where it really is just like working on stuff for me all the time, and I, I don't even have to be around for it, right? It can just, it can just do it. Um, and heck, I mean, that's kind of what I'm building with the app that I've been working on this week with you guys, right? Or this, you know, the last couple days with this social suite, right? To automatically generate all of these posts and then get them, you know, get them over to a system that will post them for me without me ever having to touch it, right? So anyway, Runner H, Manus, Operator, you know, ChatGPT's internal agent tool, agent mode. Right, is, is, uh, you know, I don't know, it's coming along, I guess. It's, it's supposed to— here, let's go library. There we go. New chat tools. Let's just try— I want to just try one agent mode really quick. We'll just have it— we'll just click on something. Um, it's not connected to any of my things here. Let's have it just make a presentation. Let's click on create a market entry. Luxury hospitality brand. This is a copy of a conversation between ChatGPT and Anonymous. Okay, that's cool. So it goes through, it's searching the web. I wonder if I can just skip through here. Builds all this stuff. Lots of stuff. Yeah, see, it's got this terminal that just pulls up. And let's see the presentation. Let's see what happens here. So a couple things I'm noticing here, guys. I think this is fine. The numbers are not centered on this circle here. I think this design of this slide is pretty crappy. I'm just not very impressed with it right now. Right, like low, local, high. Like, what is happening here? This is definitely off, bad formatting here. I, I'm just not a fan of what, of what ChatGPT has available at this moment. I've seen a few posts about some of what it can do. Manus right now blows this out of the water. It really just does. It can create very nice presentations that look good. They're designed well. All kinds of stuff like that. And apparently now they can do really nice graphs too. That, that was the little, little notice that was happening there. Um, anyway, so keep playing with this stuff, guys. That's how you learn. That's how you learn and you figure out what works. Um, know what's available and then just keep coming back to it and check on it. Now that you're signed up for, you know, a couple people are signed up for Manus, you're signed up for different things. Now you'll start to get an email here every, you know, couple weeks or whatever. It's like, hey, new features, just hit this and, you know, whatever. So, um, that'll keep you, keep you in the loop. All right. Those— that's agents. That's agents. Any questions about agents before we move on from this topic, guys?

Okay.

Yeah. Is there a way to make your site especially agent-friendly? Yeah, probably not. I think a lot of the standard stuff it's doing, because it's able to parse it, it's going to do that parse thing, right?

I, I just wondering if you have a, a site or app or tool or source or, or service or something that can, can be favored by, by these agents, then they'd probably in the future use it more frequently. If they can complete an order on OpenTable versus, you know, Seth's Table, if you can get it done in, in 2 seconds and OpenTable takes, you know, another, I don't know, minute and a half or something like that, it makes a big difference.

So I would, I would, this is how I would do it, Seth. I would literally in my, in my Replit thing here or whatever, you could do some ChatGPT, you could have this conversation there too. But I would say, um, sites like, uh, ChatGPT's operator slash agent, Manas, Runner, etc., all are agentic systems that can can do things for users on their own. Is there anything that we can do to this site to make it more accessible by these type of agent sites? Don't write any code, just tell me. Okay, that's how I'm gonna, that's how I'm gonna answer that question. So I don't know, I haven't ever even read anything on that topic. I think it's a great question and probably something that people should be thinking about. But let's see, let's see, you know, so it has access to Claude, it is Claude, it can search the internet really quick, it can do a quick— it understands what Manus is, Runner H, ChatGPT's operator agent, whatever. And so let's see if there's anything at a technical level that can be done here.

It should answer this pretty dang quickly. Yeah, I just think we're at the beginning of Major changes, just like when people started adding the sitemap.xml and all those things that helped with SEO friendliness and being scraped is, this is gonna be one of those features that the speed that you can compute all this and deliver value back to either the agents slash back to the user, I think that you'll gain high preference.

Yep. So this is pretty cool. Basically, you know, we don't have to read through the whole thing, but But here's, you know, here's a lot, right? Like, here's a few things that could make it more accessible. You want structured API endpoints. So that's cool. Add comprehensive documentation, right? Agent-friendly authentication. Implement API key authentication alongside session auth. Scope permissions. API rate limiting, right? Content management for agents. Programmatic content operations. Bulk content creation endpoints. All right, template-based content generation, campaign automation. So you're basically setting up really awesome APIs. That's what it would be doing, right? And so you could then— so here's the integration capabilities, data exchange, all these different things. We don't need to read through all of this, but when you get down to here, so here's the implementation priority, right? An API key authentication system. So it's basically you're making— just like you have API keys on these other systems You could, you could have an API key system for your, for your thing, right? Set up a whole API, whatever. Basic REST API documentation, bulk content creation endpoints. Then phase 2, do this webhook support, right? And so then I could say, great, make a plan, make a plan for phase 1. Don't write any code, just tell me. Boom, right? And and it will, it will then think through, okay, now I understand phase 1. Here's what I need to work on and here's how we're going to make this happen. Here's how we're going to set up an API structure on 60 posts per day. Pretty dang cool. And then it's going to set up all these documents and all these cool things. New endpoints, GET, POST, PUT, POST, GET, POST, right? List and filter campaigns, create new campaigns programmatically, schedule content. List available content templates. So I would tell it to just get started on this and work on this cuz this is something I do want, but I'm not going to yet cuz I know it's not doing some of the things I want correctly yet. I wanna get it working correctly and then I'm gonna automate it, right? That would be the way that I would set this up, guys. Don't start with, you know, let me leave this timeline estimate 7 to 8 weeks. No, we're gonna get this all done in 7 to 8 minutes. Yeah, maybe 17 to 18 minutes. But then, you know, you're going to do some pretty extensive testing, right? Can you, can you get your APIs to work right? And so you have to literally start another Replit project and say, hey, I have an API for you from this new 60 posts per day project and here's what we're doing. This is what we're going to make happen. Yeah, it could be pretty crazy. You'll also have to figure out what your cost structure is related to the API, right? Because if it's coming in and creating a bunch of posts, well, it's creating posts using your APIs and your AI credits and things like that. Okay, so let's look at, let's talk a little bit about tokens and costs around APIs, guys. I think that's a really useful thing to do. So let's look at a, let's see, OpenAI API pricing. Okay, so, so guys, if you, if you integrate AI into your project, your product, whatever you're doing, and let's say you use OpenAI to do it., then you're going to use one of these models, right? GPT-4.1 or 4.1 Nano, or, you know, there's detailed pricing here. Yeah, I don't know why I'm yawning so much. I wasn't up that late. Anyway, you've got all these different models and inputs and outputs and all these crazy numbers. Okay, so what it's talking about here, guys, is price per 1 million tokens. Now, a token, guys, is about equal to a word, right? It's really more like 0.8 words. But anyway, on average, you just call it a word, right? And so if you tell it 10 words and then it says 10 words back to you, right, well, that's only about 20 tokens. So obviously a million tokens is a lot. API pricing is extremely, extremely low. You obviously want to pick a model— you don't want to pick GPT-4.5 preview unless, you know, you want to be spending $75 for $150. Because you need like really, really, really, really advanced reasoning. I can't even think of a reason why anybody on this call would want to do that. You probably want to use something like GPT-4o Mini, right? Or GPT-4.1 Mini maybe, right? And your API costs are going to be, you know, fractions of a penny for the most part, right? Going all the way down the line, you're going to be able to do a huge amount with just, you know, a couple dollars of whatever. So to set this up, I want to show you guys how to, how to do this really easily here. So we're going to do this, um, create a ChatGPT clone. I have an OpenAI key when you're ready for it. Okay, go. So guys, one of the cool things that people might not understand is that ChatGPT itself is a— actually, we'll open it in this other one over here. ChatGPT itself is a platform. It's a piece of software that uses OpenAI's API. OpenAI, meaning OpenAI and ChatGPT are two different products. OpenAI has these things called LLMs, or large language models. It's the things that power It's the things that make AI possible, right? It's the flux capacitor of AI, okay? And, um, ChatGPT is just a user interface, right, built on top of OpenAI's platform, their own platform. So they're a customer of their own, you know, their own stuff. Now OpenAI owns ChatGPT, right? But that's what's happening, right? And so that's why I'm able to literally make a ChatGPT clone with one prompt right here. Okay, and so it's going to go in and, you know, create essentially the equivalent of it, right? I've got this little free user plan. Welcome to ChatGPT clone. I can do a new chat. Here are my chats, new conversation. So again, it's building out the visual preview right now, but I'm gonna— we're gonna, we're gonna build this out. We're actually going to build out a ChatGPT clone right now. We won't sit here and wait because it will take a minute, so we'll do a couple other things while it works on this, but then we'll fully integrate this so that you can see how it all works. Um, is anybody confused by what I said there about how OpenAI— about how ChatGPT uses OpenAI's API to make their system work? Everybody get that? Okay. Is anybody confused? It's okay if you are. It's very— I'll show you how simple and easy it is once we get in there. Okay, um, but yeah, you can make your own, right? Maybe, maybe, uh, Jon Heats wants to create his own, uh, you know, GPT for his own business or his own, you know, whatever, his own company, and be like Ask Jon, right? Kind of like AskJeeves.com. Anybody know AskJeeves.com? Anybody old enough to know, to remember that one? Yeah, I was a Lycos fan myself. Yeah, so, uh, yeah, anyway, you could make ask, ask Josh, ask Jon, ask whatever. And, and it would literally just be like ChatGPT. And then you could get on Facebook and start selling or running ads to say, hey, you know, ask Jon, he's way smarter than ChatGPT and he's funnier too. Right? And so you could set that, you could set that up in a way, or you could make a ChatGPT that doesn't glaze you. Like, I'm mean, I will tell you if your idea is bad, you know, whatever. Anyway, so you can create your own prompts. So let's talk about that while this is building. We're going to talk about building your own GPT and what that means and how these prompts work. So if you are a subscriber to ChatGPT, okay, and you all are, so that's good. You have access to this little thing here called GPTs. Okay, this button right here. And a GPT is basically this, you know, it's a purpose-built AI. To do, you know, very specific things, right? So like this one makes images and this one analyzes data and this one helps you as a creative writing coach, right? And they're actually very simple. They're very simple to set up. And one of the cool thing is you can create your own. Okay, so let's build— let's build our own AI right now. And it's so easy. And you guys already saw me do this yesterday inside of ElevenLabs when I was setting up an agent. It works basically the same way. Okay, so let's have a GPT here that tells you— let's see here. Yeah, what am I trying to say here? Criticism. Criticism bot. Idea criticism bot. Okay. Tells you why your idea is bad. And then provides— actually, you know what? Yeah, and then provides constructive feedback. Okay, so what does it do? Well, tell, you know, okay, you— and actually what you would do is you'd actually say, you tell users, you tell users why Their ideas are bad. And then you provide constructive feedback to help them improve. Don't ever be kind, just tell it to them straight. If their idea is bad, make them feel a little bit bad. Okay, so this is my custom GPT I'm making. So the instructions here, guys, is the prompt. I'm actually kind of excited to try this one out. Okay, we can have it create a little photo for us automatically. You can obviously upload a photo if you want. You then can, can come in here, guys. Conversation starters: I have a business idea. Um, okay, um, I don't know, I can't think of another, another way to do this. Let's just see. I— let's see here. Now, whatever, it's okay. We'll just start with that. We'll just do one conversation. So for now, but anyway, knowledge, you can upload files here. So if you're building a GPT that needs to have access to a bunch of data, just like we were building in ElevenLabs yesterday, and I was showing you with Outpost X how it had, you know, access to all this information. Right? You can, you can upload that into the knowledge base and then, you know, recommend a model, whatever. Now what happens here, guys, you can give it some capabilities, different things like that. This GPT is, look at the image, this mean robot. You can only access this custom GPT if you have ChatGPT if you subscribe to ChatGPT. All right, create. Anyone with the link can use this one. All right, so let's try it out here. Let's view the GPT. Okay, I have a business idea. Great, lay it on me and prepare yourself because if it's dumb, I'm gonna tell you. All right, um, I want to sell ice cream at the top of Mount Timpanogos. There are no shops up there, and people are so tired when they get to the top. Wow, that's impressively bad. Let's break it down. Logistics nightmare. Mount Timpanogos is over 11,000 feet. You're not running a food truck. You're talking about hauling frozen goods up a mountain. Do you understand how hard it is to keep ice cream from melting while hiking uphill for 7 miles? Zero infrastructure. No power, no roads, no water. How are you freezing anything? Solar-powered mini freezer backpacks? Please. Limited market. Sure, hikers get tired, but you really think they're craving dairy at altitude when they're dehydrated and trying not to puke from exhaustion? Legal issues. That's national forest land. You can't just set up shop wherever you want. You'll get shut down the second you make your first cone. But hey, you're on to something with the unmet need. People at summit do want refreshment, just not your melty doomed dream. You want to help turning this into something viable, or are you still romanticizing your fantasy gelato stand in the sky? It's pretty good. That's pretty good, guys, for, you know, just a couple minutes of setting up our own custom GPT to actually get it to tell me something like that's a pretty useful response. Like, that's actually what I want to hear when I can't tell it a bad idea. Like, if I were— like, look, guys, I want to just try this really quick. What do you think of this idea? Bold, creative idea. It's got potential as a memorable novelty experience. But there's some serious logistical and legal challenges you need to address. Here's a breakdown. Viral potential, first mover advantage. No one else is doing this. Challenges you'll need to tackle. GPT is Look at the image, this mean robot. You can only access this custom GPT if you have