Vibe AI Agents are Universal Problem Solvers (Take Advantage of This!)

Parker Rex. I led technology and product for a startup that sold for 23 million bucks. And on this channel, I tell you how I'm going to make $100,000 in profit a month doing AI services. I go through a couple different things on this channel of what's working in the field. I made it this little prompt thing so I don't forget what to talk about. Then I'm supposed to do all these things. But cool. So I do these things every day anyways. So it's really nice because it's just like a stream of conscious for me. and it was my birthday a little while ago. So, shout out to birthdays, whoever's birthday is today. Happy birthday to you. We're gonna go through Q&A. We're going to go through some AI news only if it's relevant to making money or helping you with your job. We're going to go through some of the strategy that I'm working on internally and then a road map of stuff that I'm building. So, start with Q&A. We're starting to get questions, which is awesome. Definitely ask questions. I love them. There's no bad question or commentary, you know. RIP on me. Let me know what's wrong with me. Let's see where the heck is the questions. Let's go into here and let's go into here. Wow, this is taking a while. Sorry, fellas. There might be a girl watching. I doubt it though. Um, okay. Let's see. Community and Okay, first one here before you go viral. But seriously, I'm going I'm going asking this around hoping I could find an answer. I consider myself a good programmer. I know how to do great products that actually work and look good. My biggest issue is I'm not a sales person. I struggle asking for money. When you create a product, how do you know the right place with the right prices and get your first initial customers? What a great question. I've dealt with this the wrong way and the right way. So, you're not good at sales. That's one point. How do you do price and then how do you do customers? Okay, awesome. I love I love this question. So, the first one is sales. The second one is price. The third one is what was it? Just like first 100 or like initial customer base. And it's cool because I can use this AI community that I'm building this AI brand really. So on the sales side, you only get good at something by doing it. So it sets and reps. This is everything in life. You're not going to be good at it the first time. If you did, you're lucky. That's awesome. I feel this. So I totally feel it. But at the end of the day, when it comes to sales, let's say you made this awesome thing. And I'm actually curious what it is that you made. So let's find the tab studio. Cool. I wonder if you have it linked. Just Alex 337. Oh, it's Elite. Do you have any links in here? Let's do uh one of these. Um GitHub just Alex. Maybe he has something open. That is not what I was looking for. Well, point is, if you want to get your stuff out into the world, like you worked so hard, but then it's pointless if no one knows about it. It's literally a waste of time. And that's hard to hear, but it's just reality. And what I would do is just start posting about it anywhere. Like I have an example where I made these GBTs when I knew that the GBT store was coming out and it didn't really require much programming which was phenomenal but I wanted to learn how to write swift code. So I made these things that basically pieced together a couple different technologies. I had a scraper that would make it LLM friendly. I would point said product at Apple Swift UI kit. I knew how to do basic web stuff like the Nex.js stack, but I didn't know how to write any Swift code. So in this idea, what I did is I went to the docs and I made a scraper that could go through each one of these. So in all technologies, there's a bunch of these. And so I just took the most popular ones that people use. And I figured that out by just digging around in the Swift channel and I found out that there was an area for improvement for Swift developers, iOS developers that could be served. Now, I didn't think about the commercialization or actually making money with this, which was stupid, but it was a learning project. And as soon as I made one of them, I just shared it. So, I made this thing and it was free so no one could hate and I just said I train custom GPTs for the following. What am I missing? I need testers. 14,000 views, 49,000 views. Um, other examples of stuff that I built. I made this crazy health stats thing. Another like just research project frankly because I didn't know how to use shad CNN tables and get like live data in here. And so it involves some calculations, pulling in data, making it all friendly, reading from sheets, and then posting it back. But when I made that, I was like, "Oh, maybe this will be helpful for some people." And then this linked back to MAP, which is the SAS product I'm building. So that said, like whenever you have something that you want to get in front of people and you say you're bad at sales, well, you just need to yap about it. You need the uh it's like the build to yap ratio. Yours is off. So, you're doing a lot of building and you're at like 100% here, but you need to increase the yapping. And I say yap because it's just a funny way of saying it, but it's you communicating the value that your product brings to a market. And you say it over and over and over again. And the best salespeople like that that's not even required. Like that's not even sales. like you can just if you hate sales then you're just marketing it because marketing becomes or it it precedes sales. So if you're really good at marketing and you're pushing this message out if you're doing the Swift thing then it's you know slashr slift. I didn't show you all the posts, but there are other posts that I made and I need to pin those. But they got a couple hundred thousand views and then it drove a lot of traffic and I still get emails to this day from people asking about it. I made friends through it. Shout out Rich at Apple and he's a great person. Like it it expanded my network. Um point is is like you need to if you're scared of the sales thing that's fine. just start talking about what it is in public. It's like what I'm doing here. And you get money out of it. You get friendships out of it. You get people that are more similar to you and that they want to support you. You're going to meet a lot of different like indie hackers and there's communities for that too. Like if you go to X and you go into my communities, of course it's not going to load, but you can just search up any hackers. So, I have a bunch of lists that I create. Why is no literally nothing's loading? Oh my gosh. Well, you can pull up different lists and then you can make lists that are private and you can add people that are bootstrapping companies that you want to learn from, people that are building web products. I have one that's basically I have like dozens of these and they're all categorized by different types of founders and stuff. So, I would do that. And then after you do that, the next question was the price. Price is funny. So I'll give you an example with in the past I led product and did very little coding. It was mostly UX design at a company called Delivery Dudes and that's the thing that I always lead with because there's a framework to starting these videos to get people to stay. It's like, oh, you know, I did this and that and it sold for whatever. But at Delivery Dudes, we started by saying, hey, it's going to be $5 if we deliver you your food and you pay in cash. That's our fee on top. It's going to be seven if we do credit card. Because back in the day, we had these clunky credit card machines that we bring in and the processing fees on it sucked. But then the technology changed and we just said this. We looked at the numbers and if you're doing SAS, your profit margin should be very large like you shouldn't have much expense beyond your actual hard costs like servers and stuff and maybe it's usage base for using AI but you can count around that and then as technology changes we went from the five to seven to okay it's just five because we got better terms because as your company scales just so you know as your company scales let's say we got to 73 million RP peak we sold at 47. But our cogs for payment processing in the early days were well actually I guess I should say that this should be the x-axis but you get leverage as you go on is essentially what I was saying and it plays into the conditions of how you treat price. So, in the early days, we had a really bad fee and then it just comes down and down and down and that comes from literally flying the Stripe account manager in and having tough talks with them. But you don't have to worry about any of that, you know, cuz then other things, just so you know, in this case, can happen where some competitor comes in and in our case it was Door Dash and they've raised how much money did they raise? How much money did Door Dash raise in venture capital prior to going public? Uh, I shouldn't have done four or five, but I think it's like what? 320 million. 2 Okay, JK. Lol. When you have $2.5 billion to go to work, then you have to kind of figure out how they're pricing things. So, you set up scanners and all this stuff and uh it's just going to be dynamic. So, what I would do is just throw an arbitrary thing at it. Um, arbitrary but researched. So, I guess that's maybe a paradox, but like at the end of the day, it's just something that you slap a label on. You can figure it out by looking at the the comps. Don't use that as a gospel. Like, if someone else has something that's nine a month, don't just be like, "Oh, it's eight." Cuz then it's a race to the bottom. you actually will benefit from being more expensive if you deliver more value in the software world. So that can be one like thing that you're looking at in terms of the price. And then also just like your internal goals to like do this napkin exercise where each one of these is a you can do financial period or month. It's probably the easiest one. You have like Jan, you have Feb, March, April. Look at me with my month. May. And then you can have a couple of inputs. So this is just basic 101 on doing your maths for it, but always make the inputs blue or some sort of variable. So you'd say like uh cost and then you'd have you can do like cost per user to figure out what that actually costs you internally and then your fee. And all these end up lading down to your MR and then you'd have one for churn. You need to have assumptions. These are all the like inputs and assumptions. You have your turn. You have arr and then you can bake in other things. That's what I do. And then for the initial customers, it goes back to the first thing is just start yapping. You want to do this manually. In my case right now, what I did is I always do things that are performance-based. So I market the thing as it's out but I limit it to a small subset of channels. Behind the scenes I call it alpha then I call it beta and then I call it you know like GA. So in the earliest days I'm going and alpha for me like no one knows these terms externally. They don't care. They just know that this thing's early and you can say that you want your first users. Don't sound desperate either. But my alpha for my uh brand that I'm building called Vibe with AI, I just launched it on a YouTube channel that had a very small following. So I was like, "Okay, I'm going to do that." But behind the scenes, I have other things. I have an email list with a couple thousand people on it. I have a Twitter with a couple thousand people on it. I have an Instagram with a couple thousand. You know, I have an internal network in my phone, so I would just be able to literally send texts to everybody in my contacts, but I don't want to touch any of these until I've had this first cohort because first impressions matter a lot. Let's say you had 100 people on your YouTube and you brought them in. 10% of those convert. Talk to that 10% every single day. I was up late last night talking to someone in this uh brand that I'm building. Now, as soon as I say the word school, it's insta cringe. And I agree with that. I think there's a lot of challenges that they have. But I wanted to launch this as this thing called troublefree. And I made a satellite site for it that's running uh like this SEO like automation on it. Has like login for me so I can go and do the CMS. But this is the first version of it. And then I launched it and I was like, well actually this troublefree AI thing doesn't really make a whole lot of sense. I think I want to lean into this term that's going viral around vibe AI and vibe with AI and it plays into the cultural thing that's happening where everyone's typically scared of AI outside of our little little bubble but I launched with this and it took pretty much no time to put together and then you get the feedback from the real people. So there's people in here and when I say something there if it's not a good thing they disagree with it then they'll let me know. So, I put out a question that's like, "Hey, this is what I'm planning on doing." And then it's like, "Yo, like this sounds all over the place, blah blah blah blah blah." So, then I go and I work around that. I'm going to reach out to him individually, go talk to him, figure it out. I was up late last night, like I said, talking to one of the members in here to be like, "Yo, here's the thing." And he's like, "Well, this is what I think. Do you want to jump on a call?" We talked for an hour and a half. So, it's, you know, the classic hologram talk to your customers thing. and you get such better feedback than just sitting there and pontificating oxygen. Uh, sorry, feedback equals oxygen. So, think about that. Next up, let's just reply. Hey, thanks for the comment. Appreciate you. And I went in depth on this for maybe too long, but I'm really passionate about this topic and I think that you're going to do great. Also, share your project. I want to know. Follow me on Twitter and DM me. Let's connect. So, next up, watching a PM try to explain programming through what they've picked up from Vibe Coding is pretty wild. I don't mean that in a negative way towards you. You're learning a building. That's great. This just wasn't on my bingo card. Boom. By the way, you're sort of on the right track and thinking of 3D for vector DBs, except there are way way more dimensions than three. Yeah, if there were just three in the array, then yeah. Hard to conceptualize in our 3D world, but yes, there's a sense of directionality in the higher dimension embedded space. I love this. This is awesome. Yeah. So, I looked up Adam because I just wanted to see like what a G this guy is. Uh, let's see. Adam, what's his username? Adamfield 7 something. Adam Icefield. Uh, GitHub. So, he's a Swift developer. Probably more than that. He probably does really awesome stuff, but he's got some really cool repos. I just want to like show you guys. You should follow him. Um, users. No, Adam Icefield. What is going on? Adam Icefield. I had him pulled up earlier. What the heck? Well, that's too bad. Let me go to my history and let's search Adam Icefield. Dang it. I don't know. Well, I pulled him up and he had some interesting stuff. And um yeah, I think when I when I make these videos, I've what I've learned is I want to there's a skill in communicating things in a really simple way. So, some of it might get missed and also when I record at night, frankly, I'm pretty tired and I get lazy. But anyways, some of the best communicators in the world that have the largest audience speak in really simple language. So Trump, whether you hate him or not, he speaks in third grade language. Mr. Beast, fifth grade. Obama, sixth grade. So I'm really trying to do that. that has probably nothing to do with what you were talking about because I didn't nail anything that you were talking about with the explanation of vectors, but I know that they're helpful and I know that it puts them in a multi-dimensional space so you can do similarity search and a bunch of other cool things. I know that I wasted a lot of time honestly trying to vectorize things that didn't need to be vectorized but I do code um professionally so probably not a programmer probably a coder but like you can just see like I push a lot of code um really didn't start taking it seriously till here before that try spell some stuff here and there um but like yeah going way back it's you know I've been doing stuff for for a bit definitely not as good as you and um it might be cool actually if you're willing to come on and just kind of explain vectorization to us. So, I'm going to reach out and um Hey Siri, remind me in 30 minutes to reach out to Adam about vectors. Doctors really? Yeah, you'd think they'd make that better. But thank you for the comment. And um yeah, I guess I could just real quick, I wanted to sneak in this demo where I use vectors because they're really helpful when you use them correctly. So something I was working on with map was taking all of your data across the five different apps that are in there. So map is a platform and it had it started with I wanted to accomplish more stuff. And so this demo is going to be three parts. In the first one, we're going to cover the original prototype where the thinking began and then the next one we'll get into level two and then the current one. So in this first one, the toy from about 6 weeks ago and this is just an original like heads up display where you kind of get an idea of where you're at. You can set goals. This is what the first time user experience would be. The right side, this is just a console showing you that it works. And so in this case I was setting a goal that so this is multi-agentic in that like it has a a set of different really uh defined tools. So I wanted to be able to speak to the best version of myself. That's the thing like I know if I do and follow the actions involved every day towards these goals that I will become the best version of myself and maximize potential. So in here it covered a couple different things. It covered the goal setting piece and then it was trained on how to set goals using different frameworks that I like. And then there was calendar. So the calendar would then go and plan the events for said goal around your events throughout your day. So I just you can see I have notion calendar pulled up but I built that in. I don't know where it is. This is kind of how the multi-aggentic thing works, but the vectorization comes into play with the is it the activity journals? This is the health dashboard. So, you'd have all the health information pulled off of your whoop. This is more just multi-genic stuff. Um, let me see. This is calendar sync. So, no. Um, that's the calendar. Where where were vectors? We did vectorization on the tasks, the notes, and the calendar. So, yeah, basically just making a bunch of these things and then you get that array of numbers. And he's totally right where it's like there's a lot of dimensions that like we really can't think of in a 3D world where then you'd selectively use those against finding relevant information. Um, it makes it faster to pull it. But yeah, please, please, please enlight me on how to better understand it. Um, because I just literally just set up the thing because I know it's helpful for the use cases that I have. Cool. So, last question. Hey, great video. Just wondering what you're using for transcription, specifically what you're used for answering comments. Yeah, just whisper. And so the way that I did this, and it's important to note, I made this a um background on my computer so I remember. But when it comes to automating stuff, hang on, I'm getting a phone call. Sorry about that. We're just quick phone call. Uh but I made this thing on my background just so I always remember that when you go to automate something, you should do that last. So when it comes to pipelining these automated comments for growth agent, I'm just wondering like well what is actually required and then deleting the stuff and then simplifying the stuff and then speeding it up and then automating it. So that's Elon's uh fivestep algo and it really does apply because if you automate stuff prematurely then it's kind of stupid. So, I never start with automating now at least as you can see by literally putting it in the background on my computer. But yeah, I started it with just spinning up. Let's see. Um, let's I don't know why I still have this open, but let's see. These workflows that you can use are really nice where if I go to workflows in warp and you can see I was doing it manually where it's video to text or YT video that's for downloading but yeah it was literally like okay I'm going to download this video in this case it's the auto reply comments but then vid to text just using whisper you can use the new one as well um but yeah whisper Did I say Whisper? Uh, hey Tron, thanks for reaching out. I just use Whisper. It's really simple. However, they have a new one that you can use. 4 o mini transcribe or mini transcript, not sure. Check it out. It's the same price, but far more accurate, at least in other languages, but slightly better in English. It's Whisper by OpenAI. Cool. And if you're wondering what I'm using to yap, that is uh what's it called? Whisper flow. Hey, Adam, great work on the physics engine that you made. Objective C is a language I never touch because I'm a scrub. I just did Swift for about six months. I appreciate you letting me know the explanation on vectors sucked and thanks for circling back on this. Would you be interested in giving a short master class on vectors so that we can all benefit from it? If not, totally cool. Either way, cheers. Hope you have a good day. I appreciate you commenting. Okay. Uh would you be interested? Appreciate you commenting. Cool. Rubber duck. Yeah. So, if you say rubber duck, then I send you um the asset that was in the vid. So, DM and then I'll go and comment or uh reach out to them afterwards. Cool. So, it looks like that's all the questions for today, I believe. And let's get into some of the news. So yeah, the GPT40 latest came out yesterday, I believe. And the news on it is pretty positive, the vibe on it. So, oh, I was going to talk about this next, but go follow this guy. This guy's got prompts and he is like pushing like it's very interesting to see how he can take literally just a transcript and then make an Instagram carousel out of it. He is doing some cool stuff. if you can extract the thumbnail profile JSON from existing thumbnails. I covered this yesterday. We can get a JSON output and I'm imagining they're just making that on the fly using probably not image trace nets, but I just want to know how they make it. I'm really excited for that API to come out. You can extract brand profiles from existing ads and then use those against the style of your ad. This guy's he's got the alpha. So, he gets a hard follow. Um, but on the 40 side, I just pull up OpenAI devs. Highly recommend following this or adding it to a list. Um, I do lists, but yeah, they basically say it's better update API, but stay tuned. Better at following detailed instructions. Cool. And then here's the nerd chart. So, it's apparently really good at coding. Moved up quite a ways. I was using it. Well, I use it every day, but that came out. It's so hard to keep up. It's so crazy. Now, let's talk about no code versus code. And I'm going to take this sweater off cuz I'm literally sweating. You guys are making me so nervous. Code versus no code. There's a thing going on from Replet Founder. You're familiar with it. It's the like no code get your hands off the keyboard thing. I think his name's Oman or uh Replet founder. It's kind of blowing up because this guy says, "Hey, you should not learn how to code." And I disagree strongly. I started taking coding seriously way before the Vibe Code trend um like three years ago and you can still use the AI as a tutor. So that's why I made the GPTs for dev thing from earlier. But you just always have to ask what's the incentive of the person saying the thing. So it's somewhere in here, but kind of blew up where he's like, "Hey, you shouldn't learn how to code." Yeah, there we go. 5.8 million. So issue with that is the muscles that you will gain from learning how to actually code, understanding the underlying concepts. Those muscles are great because you'll be able to see around corners. And I talked about this in a previous video, but if you have a spectrum of amount learned, so learned and then you had your method. So I think with AI, you don't learn much. You see enough of it, you might understand like you might do some pattern recognition on things that you're seeing. Then if you do video tutorials, midts, so done with you, you're up here, you're definitely going to learn more stuff because you're going to recognize those patterns and you're doing the code along. Then the DIY method is just the best, right? Because you're going to take longer. There's another dimension to this than it would be like linked to product. But it will give you the ability to learn systems thinking. Learn what's going on underneath the hood. Learn how all these different packages, technologies, libraries, how they work, how databases work, how authorization works. There's a meme around don't roll your own. I think you should roll your own off because you learn how it works. And it just takes away all these abstractions. And then what won't change is the models will get better. But the skill of knowing how to architect, knowing how to speak to the computer, that's an evergreen skill. Like you're going to get a higher multiple on your leverage if you have done it yourself. So I hard disagree with that. Get your muscles up on Cursor. A new version came out. Again, I've been using Cursor since it came out and they're an awesome team. They listen, which is the best part. A lot of the features that I've recommended show up. Probably not me, probably a lot of people saying it. So little things like going to beta and flipping on early access well worth it. Just select the model that you know works well. 37 runs rampant. I don't use 37. I use 35. And um when you do that, you'll start seeing these different features up here. So one of the nice ones is a little notification bell where I still read all the outputs and I typically air on the side of using chat and then just bringing in context. But having a nice little notification where if I like walk away for a second and it goes ding. That's nice. Um and then custom modes as well. I haven't used them that much. I just made one in here. Yeah, memory bank, which follows the client best practices for using this memory bank feature. If you're not familiar, it puts in a series of files. So it loads things into memory fresh and um it's got nice structure in it. So has a project product context the active context of what you're working on some client rules around that like oh we use Pixie for client side off the progress of things that you've done so your checklist a project brief which should not look like this it should be much more detailed but this is your working file for things you want to get done I think of it as a prompt spec a technical PRD whatever you want to call it all the things that you want to do with the relevant files and relevant snippets relevant libraries and all those things should go in here um system patterns and tech contexts So that mode is nice because I it just it's preloaded with the thing that it needs to do. So as this gets better, this UI sucks, but it's conceptually interesting. You'd imagine that the different types of roles on your team will exist in here. Like this could be a problem identification or an incident response engineer who's responsible for going and finding the issue and outputs a thorough root cause analysis. Another one I'm doing these manually with hotkeys and it works super well but it is nice that they put it there. The next one would be the incident uh expert solution architect. So again like it's setting you up for success when these are separated and well defined. So then it takes in the RCA and then it returns an atomic step of list atomic meaning like each one is a next action and it is very well defined and then you'd have um the incident remediation engineer which goes in and takes all those things as inputs and does it same thing on the feature side. So net new stuff you'd have those three steps. So this is why I talk about nodes a lot on this channel because I think having these in a guey where you can just drag and drop and try different variations, it's just faster. So to all the people that are like, "Oh my gosh, you're using NAN." I only use it because it's faster to put tests into place. If I'm using a MIMO notebook, which is more of a pain in the butt to set that up. So yeah, that's uh a big deal for cursor. If you're having trouble getting on the latest version of it, use Homebrew. Homebrew is helpful for that. Now, I wanted to talk about this visualize any codebase thing because I think I'm going to bring this into my file structure and then edit the client memory bank for that. So, let's see. This guy made this cool thing and we're going to check it out if it opens. Whoa, it's playing in there. That's kind of cool. So, yeah, if you've seen this pattern before, it allows you to put something ahead of the sub route and then that's the way that they built it. So, it it's a subdomain of their actual site, but it makes a nice I think it's mermaid diagram, which I have a prompt that I use for that helps with figuring stuff out. So, just to make it larger. Yeah, it just gives you the IO and um it's nice. So, what I think is I'd take this because it's open source. Then I just have that be like the plus up version. I want to comp it against this mermaid agent that I have. When I say agent, it's it's like a tool. So it's not really an agent. That's just a prompt. Agent would be tool plus loop plus conditionals. But yeah, so that is pretty helpful. And this will likely go into the quiver. And then the shaden dashboard. Uh shout out. So he made a new thing and when makes things you listen. So this is pretty cool. This is using the V4 tailwind stuff which I haven't used at all, but um yeah, it it's just updated for that and it looks really nice. I like the mono stuff. It's got everything in there. I like the look of it. But yeah, you can check that out and then it's obviously open source so worth a look see. On the strategy side, what I'm thinking a lot about is how do you give the people the most amount of impact when you're building a brand around educating and reskilling folks? Um, this thing's happening faster than we thought by 2030. It's stated by Goldman Sachs that 300 million million 300 million jobs will be displaced. So, I want to be on the side of history that's helping people out. I think it's like literally the agrarian revolution happening and I'm not alone with that where everyone that worked in farming I think it was like 85% of GDP in the US was generated by farmers. It's kind of like we're the farmers. So how do we leverage this stuff? How do we get ahead? And then how do we teach people? And in my head it's building a brand around that. Right now it's just a community but it'll be events as well. It'll be probably info products, coaching, seminars, all that. But at the end of the day, it's a mixture of a media and an education company. And what won't change is yes, the models will not get worse. So I think of never replace human beings, but you need to take some action. You need to learn how to use these tools because just like the this is a video sales letter that I made. It's not public cuz I think it's a little cringe, but farmers way back in the day that thought they were going to be replaced by big machinery. Well, they learned new tasks and they became way, way better. So that's you. That's where you're right now. You're wondering how do I do this? How do I learn how to use AI to leverage all these things that I don't want to do throughout my day? I want to take you from where we take the all the world's information and make it accessible and useful to everyone. People started to learn how to find information on the internet. We're in that same place. Jeff Bezos says there's an even more important question that I almost never get asked and that is what's not going to change over the next 10 years. We all know that AI. So yeah, I think that's the question that everyone should be asking themselves. The models aren't going to get worse. You're going to get better, but won't change is how to like you're going to want to learn how to talk to the computer better and you're going to want to learn how to problem solve with it because any problem that you have can be solved with more knowledge. I'm putting together a deck on this that like kind of like hammers this point home. Idea being that all problems are solvable with enough knowledge. Before AI, it's you plus computer plus internet. Now you have this diff which is like cool. So you get these models. So this is like the average Joe using AI and they copy pasta things around. And then the third iteration of that is okay now you've trained each one of these individual models. They're connected to one another. They're dependent on one another for certain tasks. They have hit or human in the loop stop gaps. So if you reach a point where you don't want this thing to do something typically around like payments or uh something that's like a a customer communication, you want to actually see the output of it before you go live with it. But it'll make that first draft. It moves that first version forward. And then once you've doing that done that workflow dozens and dozens if not hundreds and hundreds of times manually, then you go and you build the system out and then you start the HITL process. So I just think that that's a big thing. And then I'm trying to just like think of how do you have evergreen framework such that you could teach people how to problem solve. And um that's what I want to like go through is it's like how do you iteratively problem solve with AI for any problem so that you're not sitting there trying to educate on every single use case. I've tried that making like a massive list of like by job type by category is it content creation are you a recruiter are you salesperson but at the end of the day it does just flatten the learning curve. So it's it's more of like a meta thing. It's a framework where it's like okay I always think of this like in both my entrepreneurial journey but like in general like this is how it usually works where the thing that you want whether it's like making more money or uh getting a girlfriend or anything it just takes shots on goal and it's the thing that you want on the other side of those attempts or shots on goal. But if you are levered up then it just reduces the time to that. So, I'm like trying to make it look like that. And um yeah, this is obviously like the thing for the brand that I'm building. But I think in that case, how do you take iterative problem solving and bucket it into frameworks that are really piffy and understandable so people could just read it and get it? And uh for whatever reason, I unbound this stupid key in my mouse, so it keeps doing that. But I thought about this with a member of the brand and we basically said there's a couple different buckets. One is you don't know where to begin. One is you have some context on where to begin. And then I guess another one I need to think this through. But I'd love your guys feedback on this because iterative problem solving with AI will make everybody better, right? Because then it's agnostic to the problem, agnostic to whatever it is that you're doing. So it's like typically you'd be like okay well do I have any idea on how to solve this problem on how to solve then if no then you need to do another step. If you do have an idea then that's when you do context gathering. So that's uh let's say like some text maybe examples of the thing. If you have no idea on how to solve said problem then you need to enter the research phase which just is before this. So then you'll end up going to that. So then now we're in this context gathering and then now and you know the method the method would just be something like this. you know, you just open it up, but you probably have some sort of glossery on like which model to use when, if you want to use deep search or not, whatever. But then once you have the context gathering, there's probably some sort of like rubric that we can put together for here. So it's like, yeah, text would really be like could be a PDF, but any sort of sources, some source material, maybe there's a best practice for that. So like an exact summary, a talk a YouTube video pod and if it is these things there's probably a conditional here where it's like okay go grab notebook LM with this prompt. So, I'm trying to, as you can see, I haven't thought about this a ton just from like a high level. And this is super ugly, but there's definitely a system that allows people to get to the root of what it is that they're trying to solve and to get the answer. And that whole like flow, this cycle, this loop if you will. It's just going to help that thing that we spoke about earlier of like getting the attempts, the shots on goal. Then if you can shorten those feedback loops, shorten the problem solving, then you get to the thing that you want quicker. Example, this was yesterday. I was setting up a multi- docker container home server. I call it the agent team. Has a bunch of different agents on there, database instances, nextjs apps on there. And it was the first time I done that. First time I use a couple of these different pieces of technology to like string it all together. And um like I just had a bunch of questions at like the system level for my computer and I'm like I have no idea what is going on. But luckily, you know, if you have this then you can just do command I. you ask much questions. So, it's like strapping yourself with the right tools which can then teach you the thing and then you need to know how to ask the right question. So, a lot of this is obviously uninformed just chiseling away at the idea. But if you guys have any insights on this, please drop a comment and let me know how you guys problem solve. And uh the best one will get featured in the next video and I want to walk through how do you problem solve? That's right. Wake up everybody. How do you guys problem solve with AI? Let me know below about that. That was cool. And then I was going to do this example. So, we might as well I was going to do this example where like I'm want to run through this process just raw. I'm going to raw dog it. And uh I really like the look of this thing. So, I want to like snag a screenshot of it. And I wonder because there's all these plugins that allow you to take a flat design and then make the actual Figma stuff out of it. So, Figma works with nodes. So when I draw a shape that is doing something behind the scenes, it says, "Hey, it's this gray thing. One second. I got to take my food out of the oven." Behind the scenes, this is drawing all this stuff. And then it ends up in some sort of data format. You know, they've got a billion of them. But let's see. What I would do is I just go into a chat instance. And so I'd have the screenshot of it. Then I want to see how it works. So I'm going to just see if it can do this. Um, let's do a deep research on it. I'm building a Chrome extension that allows you to take a flat PNG or JPEG static image and have it rebuilt with all the necessary information in Figma. I know this is possible because there are existing plugins, but I need to do this for myself. Can you explore how this is done? So now this is a really bad prompt. So let's make it better. You are an expert at reverse engineering technologies based on existing technologies and then giving step-by-step instructions on how to recreate said technologies. So okay, that's like a decent role. Um, okay. Context. I'm building this thing. Um, and then we're going to provide examples provide five examples of the most popular Figma plugins that allow you to take a flat image and turn it into Figma layers. Let's use 40. Let's use the internet. Wrong one. Okay, it's off to the races. Kodia. So, let's grab this list. here are five examples of plugins that do this. I don't know if it's I guess it is plugin. Okay, grab that one. Grab this one. and that one. Well, let's just do three. Cool. Let's move. I'm building a Figma plugin to allow users to paste a flat PNG or JPEG static image on the Figma canvas. and automatically convert it to the editable Figma design by analyzing the image and generating the corresponding layers. That's decent. So, let's take that. We're going to kick that off and then we're going to open a new one. And we're going to do that with Grock. We're going to do that again. We're going to go to deeper research. Cool. That's on. And we're going to do Gemini. And we're going to do deep research. And it doesn't have the image, so whatever. It doesn't need it. We're just learning. Okay. Do you want the plugin to support just basic layouts, complex illustrations? Basic layouts is fine. Use whatever you think is best. For number two, we're prioritizing Figma native layers. And yes, looking into their APIs would probably make sense. Cool. I'm going to pause this and come back. While it's going, I want to use this meta prompt that I found. So, let's do I have a bunch of them in here, but I want to try this new one. So, let's paste in that as a system message. And let's do how about a three mini? No, let's use a thinking model. Yeah, let's use let's use a one and let's take the original one that we had. Let's plus this up. Let's take the one that was templatized. Let's put this in here. It's going to plus that up with 01. That should take about a minute. Then we're going to take this. Um, oh, and I just am curious like what this would look like. So, let's do that. Let's see what the difference would be between the prompt that I put and then the prompt that it's going to give us. Oh, okay. That was quick. So, Figma uh Grock came back in how many minutes? Two and a half minutes. Gemini is still ripping. What's so great about Gemini is its ability to go on to YouTube. Well, that's going. Oh, here we go. So, okay. You're an expert below is a reversed one. You're reverse engineering expert and Figma plug-in developer. That's better. Using simple terms, explain this. Break your explanation into five parts. Image analysis, object detection, and layer extraction. Yeah, this is sick. Oh, okay. But as you can see, like this process, I'm not going to go through the whole thing because this video is going to go forever, but this is a nice starting point. So, you probably should just do the research regardless. Guess it' be dependent on the problem size and the urgency, but this is so helpful, right? It's like, okay, I now know that this is the Jason that gets like put into use, which is dope. You can use OCR. This is sick. Yeah. I mean, imagine if I put this meta prompt in there. Wow. So it may be that like let's just to kind of wrap here. It may be that this process is really like always doing the research. So d y r um it would be something like this where for every problem you're trying to solve you do deep research. Well really that'd be three. So one is draft your prompt. Two is plus up your prompt. Three is deep research. After you have said deep research, you really should know what to do then. like that almost then it's almost like you could do some sort of like architect the solution and it's going to look different in every industry but at the end of the day this is the thing you want and then execute on problem go get it so yeah this could be something interesting I'm going to keep kind of like thinking on this because it needs a name and um acronyms are are really memorable. But yeah, I'm excited about that. And then lastly, I vibe with AI.com. So, I'm a deep believer that these trends are going to persist. It may not have like the staying power forever, but right now I think it's an interesting thing to kind of like grasp that cultural acceptance of like tying a fun word like vibe to an industry. And so I'm not a vibe coder, but I think it is beyond that. It's like I'm trying to like lean into the fact that a lot of people that are learning this stuff need to learn the hard way. um in that like they need to learn the fundamentals, but it's not that hard because you can leverage AI. So, what I'll be doing is like tearing this site down, which had a bunch of like the original like workflows that I use and then the models and the steps and all that stuff. And I'll be tearing this down and then replacing it with that URL that I purchased. And it's going to be a fun website that's like branded like that deck that I showed you earlier. It should be really really uh helpful. And then I just want to make sure that I did everything I said I was going to do in this video, which was yeah, I think that's that's pretty much it. If you guys want to be a part of the brand that we're building, it's a great time to join. I'm sweating. It's a great time to join because we're in alpha right now and we're going into beta next week. And uh yeah, eventually this is going to be its own platform. We have bunch of different things that I'm excited about working on with different pathways. It's like choose your own uh your own route on which stuff that you want to learn. And I want it to be industry agnostic. It's just teaching people how to talk to the computer. So if that sounds interesting to you, then go ahead and join. It's going to be cheaper. Now obviously like everything with these communities is you like anchor it low and then you rise it up and then find like what makes sense. But make sure you like the video, subscribe. I'm gonna go eat a pizza. See you tomorrow.