Latest / The Investor Shed Podcast / AI Mistakes That Wasted My Time | Investor Shed Podcast

Transcript

AI Mistakes That Wasted My Time | Investor Shed Podcast · The Investor Shed Podcast

• Nick & Jeremy: Welcome back to the Investorship podcast. I'm Nick Beverich. I am so glad you're here. And ⁓ Jeremy might be too. Good morning, Jeremy. I'm always glad you're here. I'm always glad you're here. I'm always glad you show up. Yeah. You show up every Monday without any hesitation. That was a minute later. Fridays were always hesitant. Friday is pretty crazy, right? We're like, ⁓ boy. In my head, I'm always like, please, Nick, cancel on Friday. I don't want to be the one to cancel. Yeah, yeah. I don't want to be the one to cancel. Yeah. Because it just Friday, there's so much to do. There's a l it's a busy time of the year for sure, man. Same with Mondays, but we don't want to get right into it on Mondays. Right. What's nice about the weekend though is like I get caught up on all my tasks, right? So I have a couple of days to like, okay, nothing's happening. I can go like work at my leisure, not to have stuff like out first thing in the morning. 'Cause it was all out Saturday, right? Yeah and now like Monday's just a clean slate. So hopefully you guys have a clean slate out there as well. And ⁓ hopefully you're ready for some knowledge today. And hopefully have a clean garage like I do. You do have a clean garage. I've been seeing you sell so much stuff on Facebook marketplace. My garage is getting cleaned out over the way. I spent all day yesterday and all day Saturday cleaning out the garage again. ⁓ 'cause anytime you know you do a fix and flip, I usually end up bringing a bunch of stuff home from that flip. Yeah, yeah. Like not just like all my messy tools, but like extra materials and all this stuff that came with that house that I was eventually gonna sell anyway. Or just like Boxes of random stuff. I'm like, I can't throw this away. There's good band-aids in here. You know? That's some of the stuff. So I see like it it gives me a notification. Not a notification, but it pops up on my Facebook, like, Nick's selling these lights that are ten dollars. And in my head, I'm like, is that worth his time? Like, I I bought those for a hundred and twenty dollars and they're still not out of the box. Those and those that are watching the video, see these right here? I bought two packs or or well, a pack of two. And they were like a hundred and twenty Dollars and then I bought a second one because I figured we'd use more. More ambient lighting, yeah. We only use one out of the four. So two hundred and forty dollars on these stupid bulbs, and two two of them are still in the package. So I'm like, I'll sell it for twenty five dollars. Yeah, yeah. And then somebody's like fifteen. And I said no. Yeah. No, thank you. No. Please go on Amazon and see the price of this. Yeah. And then come back with a screenshot. Amazon dropped it to 99 bucks, but still. Well, still. I mean, a fourth the price. A fourth of the price. Unopened. Come on, kids. So you're cleaning out the garage, you're selling a bunch of stuff. ⁓ I wanted to be like kind of a troll on some of your posts, but I decided not to. You can't. I was gonna be like, do you deliver like these like five, five dollar things? Will you deliver to Rathrum? Yeah. The funny thing is like there's I I actually sold a couple things to people I knew, but I didn't know who they were at first when they started messaging because it just says like their first name. Sure. And like it's like a different type of profile at first. ⁓ but yeah, I sold a dartboard for five dollars. I love it. I'm like, who's gonna buy that? And apparently it was a guy I knew. And then I sold that garage door opener that was still in the box for fifty bucks. It was somebody I knew, it was some lender. Okay. yeah, so anyway. Do you have time for that? Like do you have time to like muck around with Facebook marketplace? I do now because of the AI tool. Okay, Have you ever sold anything on on Facebook Marketplace lately? Yes, but not recently, not within a couple of years. So it's so cool. Okay. So that with the app, you know, you go to where you're just gonna sell stuff and you there's a little plus button and then before you just start ⁓ creating your listing, there's an option for just like I I think it's just called AI or something. AI j like jay you just take a photo. Sure. So you just start taking a couple of photos of whatever you have to sell and it already kind of knows about what the value is and sometimes it's way off. Sure. ⁓ And ⁓ and then but it just gives a description of what it is. That makes sense. And it's so much easier just to go in there and then you just like kind of correct the price and then you hit submit and that's it. Sure. and then when it so it it was easy enough. Because creating the listing ⁓ on something I decide I'm gonna sell in my garage takes less than sixty seconds. Sure. And then it's done. And then you just kind of wait for messages as you continue to clean your garage. Right. It's better than a yard sale. Yeah. I just the hard thing I have a ⁓ the problem I I have with it is like the time spent meeting up with people and Yeah, I'm not meeting up with anybody. It's just yeah, you wanna it it instead of saying yes when people like do that, ⁓ is this still available? Yeah, you know, that's like the most common if you're as well anything on that. Yeah, yeah. That's like the common that's like the volume, a pre made message. Yeah, yeah. What I learned was if you just say yes, half the time they don't respond anymore. But if you just say, Yeah, when would you like to come get it? Mm-hmm, your odds go up. Immediately of them responding right there, like ⁓ I can go ⁓ so little Facebook marketplace tip for the sellers out there. Don't just say yes, it's available. Yeah. You gotta say, Yeah, when when would you like to come pick it up? You have to assume if somebody's I'm available for the next four hours. Sure. Something like that. ⁓ you gotta you gotta get them closed. That way they're they're like, ⁓ yeah, I'm still interested. Yeah, I can I can make it. Yeah. Instead of them having people are not motivated. Don't put the ball in their court. They they want the convenience of being offered like, Yeah, you want it? Come get it. Yeah. This is for you. this could be a real thing. Okay. Yeah, okay. Come get that. That's a really good tip. That's a really good like selling tip, honestly. Like yeah, when are you available to come pick it up? Just easy as that. Well, don't just and don't just leave it open ended like that. But yes, it's available. ⁓ I'm I'm gonna be here for the next four hours. Can you make it? Yeah. So and just assuming that they want it because they inquired. Sure. ⁓ what's the worst that can happen? They're like, you freaking salesman. How dare you ask me if I want this that I required on. This five dollar dartboard. That doesn't happen. You scammed me. Yeah. Cool. Well, congratulations. yeah. Cleaning that garage is nice. Thanks. We just spent five minutes ⁓ not talking at all about this the show topic. Have you seen like our last like 200 episodes? Yeah, that's We go off topic very quick, and it's okay. Good, and YouTube doesn't like us. YouTube's really upset. ⁓ I hope People really if you're using AI out there and if you're in small business or a real estate investor or a real estate agent or whatever or or or a real estate photographer maybe. Yeah. ⁓ I hope you take this one to heart because I have dedicated, I'm in the p ⁓ the privileged position this year to not really focus on anything but AI. Yeah. I don't have to make additional money. I've already I've already kind of banked that for the at the start of the year. I put a lot of projects in place to where I'm I'm good on income. So I decided since ⁓ since the agentic models came like rolling out early this year and they become became popular, which a lot of people don't even know what that is. Five percent of the people out there ⁓ that use AI don't even pay or five percent of the people that use AI don't pay for AI. Wow. So Out of ninety-five percent of the people that have used AI tools, they don't pay a monthly subscription. ⁓ so only five percent pay. Only five percent I thought you said only five percent don't pay. And I'm like, that's actually pretty damn good. No, so so five percent have paid subscriptions to AI. Yeah, like the twenty dollar a month or something like that. And of those people, a small fraction of those people are actually using agentic tools. Like it it's like one percent. One percent of the five percent. Something like that. Yeah, it's crazy. ⁓ so if you really want to revamp your business or start a new business, I highly recommend getting very fluent with these things because the technology now exists. It's mid 2026, where just about anything that a person can do on a computer, AI can do for you if you set it up right. Yeah. Which there's a lot of people out there that have jobs that don't require them to physically move paper or physically move stuff. Yeah. And almost everything could be done online, you know, remote workers and stuff like that. ⁓ but there's also a lot of things out there that you might hire somebody for, you know, and pay ⁓ thousands of dollars to build you out something online that now takes eight minutes. Yeah. ⁓ like a website or something like that. Sure. ⁓ and the but these tools are insane because you can the you they now you you now have the ability to have a shell, which can be like a just a computer. Mm-hmm. And it's and an AI can be the brain on that computer and you can build it arms and legs to go out and do stuff digitally. Sure. ⁓ until eventually and and this will do until the robots come out. Right. This is the next step. This is this is the intermediate this is the step from chatbot to agentic models that can do everything digitally. And then we'll have the physical guys that will roll off the Tesla lines or whatever. There's a few different other companies making these robots. that people will purchase, even if, you know, just like the the Jetsons and R T D twos and whatever that ⁓ there's going to be things that will actually be able to move atoms around for you too and not just digitally. So this is this is the step we're at step big step number two. Right. ⁓ to get infinite workflows done very fast and efficient and cheap. Yeah. If if you have heavy online work. For sure. ⁓ do we know when the robots are going the the physical robots are going to be rolling out? Is that kind of a pipe dream right now or is it Yeah, nobody knows an exact date. So I ordered a robot about a year ago from this company that said that they will be out in 2026. Yeah. And I called bullshit and I put five hundred bucks down and I'm still waiting. I think you're like, I'll I'll I'll take that bet. That's the Neo robot. It's been on like a few peop main people's podcasts and stuff, but it's like a cloth looking ⁓ robot that w walks around your house and does dishes and laundry and can answer the door and you just have to do random stuff and it can and it could it could just ru do like household task for you. Yeah. Inside. Yeah. It can't go out and mow the lawn. It's not gonna be that strong. Okay. ⁓ but it can do dishes and sure. And and light light laundry. Yeah. But just imagine not having to do dishes or laundry or vacuum the house. Like just just having a comp you know for five hundred bucks a month Or whatever or whatever, or 20 grand up front. Yeah. ⁓ to have a robot that can just keep your inside spotless if you're into that kind of stuff. Some people don't care. Yeah. ⁓ I find that immensely valuable because there's so much time you can put easily five hours a week into cleaning your house. And ⁓ and imagine if you put a robot on it, but instead of just five hours a week, you just that was like it's you know, ten hours a day. I want everything spilas always all the time. You might have a, you know a you might enjoy it. You might well after a while, like I mean your maintenance level's gonna go down because you're not gonna like trash your cum total house every single day. L Sure. Well you might you might just out of fun. You might just throw spaghetti on the walls for fun. I don't wanna be that person in the robot revolution or when these things get consciousness they're not gonna get consciousness, but like I don't wanna like abuse them and kill me in my sleep. Yeah. And you might be thinking to yourself, but you know, doing these daily chores, it kind of gives you some doping to some people. Like it gives them a purpose of things to do throughout the day. You know, I can watch my soap operas as I'm folding laundry and all that stuff. ⁓ yeah, that's true. ⁓ so it's not gonna be for everybody, but for some people they might really appreciate their laundry, not just always stacking up because it's something else to get to or it's just on the floor all the time. And not even just for some people. Like this isn't about being lazy, this is about just Finding inefficiencies, right? Yeah. Like elder people who have a hard time getting up to do these tasks, but like really want a house clean. Or a single mom who's just busy with kids. Like these are things that are super valuable. Yeah. Elderly people that live alone, ⁓ they have a hard time keeping the house as clean as it should be. Yeah. They can't always reach every light bulb and all that stuff. But ⁓ in the meantime, we have the technology that most people have no idea. Most people have no idea what's yeah, what's I I would say less than five percent of the public. ⁓ know what AI is capable of today. Yeah. I mean I don't even know what it's capable of. Every time I learn something new, I'm like, Wow. My mind gets blown all the time and I've invested to just like work full time in it now. Yeah. It just to revamp my business and learn as much as I can and get fluent because that's what all the big time social media people that I listen to, or not ⁓ social media people, but business influencers, ⁓ they seem to think that that's a good idea and I agree. Yeah. ⁓ but here's here are s Today I want to talk about some of the big mistakes I've been making because I've been messing with this since February. It's now June, ⁓ almost July. Yep. ⁓ but I've been working about four four hours at least four hours a day. Sure. After working in AI projects. And I've made a lot of progress. ⁓ a lot of good efficiencies where I can I I've saved time now, lots of time. Yeah. But I've also put a lot more work on my plate just because it's so fun. Right. And you can and I can. ⁓ but I've been able to For instance, triple my database and have really clean data. How I got there, and this was one of the first big projects that I was working on, ⁓ in parallel with other projects, but this was the the big, big one. ⁓ I'm working on the second big, big one right now, which the second big, big one is I'm building out a transaction coordinator named Katie that ⁓ once a real estate transaction goes under contract, Katie essentially can take everything to the finish line. Yeah. Or at least ninety five percent. Okay. ⁓ maybe some final approvals needed there and ⁓ sure. You still need a human to look over things. ⁓ but for the most part, we're trying to I'm trying to build it to where it's just bulletproof where it's just like it's all about AI loves systems and processes and Things that are like factual, right? If this happens, then what, you know? Yeah. It doesn't love ambiguity of like, hey, just do this and it's probably not that you're a negotiator, you know, during the ⁓ inspections, you know, it it's probably not gonna get involved much there, but at least I can relay some information. Yeah. ⁓ but here's some big ⁓ mistakes that I was making with AI a while ago before I got big into it ⁓ a couple years ago. I would ⁓ I would use ChatGPT or Grok. And I'd want to start a project ⁓ or start start some thing and I would just go into a normal chat. Yeah. And I would and I would just say, ⁓ here's I want a 30 day plan to build out this board game or something like that. ⁓ and a step by step everyday plan. And then I'm gonna work with you to follow that through ⁓ every day for the next thirty days. So build out this huge plan. Yeah. And then we'll just work it, you know, I'll do as much as I can each day that I have to devote to it. And then ⁓ I remember starting that board game plan or wanted to build out a real estate based board game that was kind of like a blend of ⁓ cash flow and monopoly and ⁓ but also using apps and stuff and and real time properties. Yeah, yeah. ⁓ so and and I even bought like a three D printer and everything to go with it to build like print off the little pieces and everything. Cause I I was serious seriously wanting to do that. Yeah. And what I learned ⁓ over about a week or two of doing that was that all of a sudden ChatGBT and that chat ⁓ was just making so many weird mistakes. Yeah. And it was just not as as sharp as I thought it was. ⁓ what but the problem I was doing was I was continu I was doing context blow and where you just have this massive chat session where you're just going back and forth in one chat, ⁓ but it can only handle so much. It's like a whiteboard. Right. That's considering everything you're saying every time it's going back with you and everything that you did say in that chat. ⁓ but kind of getting deleted from the but only so much context can, you know, it can handle it can only handle so much context, so much back and forth. So I mean, ideally you should only be chatting for like a day and then start a new chat, but then but summarize all that context for and any important things needed for the next day, and then you move on to a a new chat. But that's called a project. Sure. ⁓ so what I learned later was so I need to just start a project, and almost every AI company has that, you know, like ⁓ Claude and Grok and Chat GBT and yeah, you know, there you there's a little sidebar and you can click the project button. And the only difference between the project and the chat feature is like there's a spot for instructions. Okay. ⁓ the the description of what this project is, which is real brief, just like a one sentence. Yeah. And then the instructions, which is probably a couple hundred lines or l a hundred lines or so of ⁓ everything about this project and how to go back and forth with this, you know, the user. Yeah. And how you want it this project to be built. And then and then you would have a spot for files where you know that that's where your MD files would live or anything that pointing the the AI in the right direction that you want it to reference when you're building this project. And typically that's your context. Like everything you've been working on, it summarizes what you've been working on. And you download that at the end of the day, you tell it to summarize everything. Yeah. Yeah. And then you say you create an MD file, which is just a markdown file. Okay. ⁓ which is just a text file for those that don't know MDF. MDF, it's just the it's the it's the easiest file for a computer to read. Sure. And it's just like a text It it's it's just you open up a file and it's just text. Yeah. A couple kilobytes of information. Yeah, very small. Yes. Yeah, yeah. Just kilobytes of information. We're talking like 10, 10 to 100 kilobytes. Yeah. ⁓ so it's just text. It's real easy for it to digest and read real quick to get context and then give you the proper ⁓ instructions or or whatever to follow. ⁓ so what I what I learned was okay, at the end of, you know, you build out your project, you have it. Create your big roadmap or your master plan of what you want to do. Yeah. And it will tell you, okay, over the next 30 days, we got to do all these things to reach New York. You know, we're traveling from LA to New York, so we got to go through this state and this way, then we gotta get off this highway. We're gonna have to make sure we have enough money to pay this toll and blah, blah, blah, blah, blah, blah. And day thirty we'll be in New York. Sure. You know, whatever. That's that's a that's a pretty good. ⁓ so, but the problem is, ⁓ You know, at the end of each day, ⁓ you you'll you'll want to ask it, you know, after you've been working on, okay, I'm in I'm in Oklahoma now and we're we're getting past this, you know, we're we've made a few stops along the way. Yeah. ⁓ why don't you go ahead and summarize everything that we've worked on today? So to create a handoff session for tomorrow. And it'll create a new MD file with just some quick updates of what we did that day and what we're planning on doing tomorrow. Yeah. And then you download that, you upload it to your projects. Now with Cloud, actually, you just hit a button and it just uploads to the projects. Nice. That's a new one instead of downloading and uploading. Yeah. So that's a little time saver there. Sure. Finally. It's all kind of a time saver. Yeah. ⁓ but here's here's the problem that I didn't realize. ⁓ whenever I would start these new days with these projects, ⁓ I just assumed that when we started our project that the new chat session that had to regain its con consciousness of what's going on. Yeah, what have we been doing? It would read it would read the description, it would read the instructions, and it would read the handoff. ⁓ I assumed it read everything in the project files to know we're trying to get to New York. Right. And we're and we're taking all these stops along the way. ⁓ but it didn't default to just reading. Like you have to actually tell it to read all the the instructions. So for since from like April to June fifteenth for almost two months, I was doing these vicious loops of you know, driving around Texas and and Oklahoma and like spinning around because it was telling me how to get to Texas. Right. Okay. Yeah in the handout session then that we're gonna go with you. But it didn't know that we were going to New York. Right. 'Cause it never read you know, this is a bad analogy, but ⁓ it never read the full context of ⁓ in my C R ⁓ I want all these things referenced. You know, I want you to look at my old phone history, my old email history. Yeah. ⁓ I want you to look at consider that with Facebook. Consider all that before you summarize that and make sure you look at the settlement statements and blah, blah, blah, blah, blah. it would just do these steps that were for that day, but forget to consider everything else. Gotcha. ⁓ so essentially I was just most days spinning in circles and not knowing what the problem was. Right. And why why isn't it considered why isn't it just do it and everything getting it but I didn't know that it wasn't I didn't know that it was making mistakes. Yeah. Why are we not getting to necessarily because the problem with AI is it's always so confident in its answers. Yeah. That it knows what it's doing. Right. It's never like, you know what, I just don't think I have enough context. Yeah. Have you ever heard AI say that to you? Never. What you know, I just don't think you've given me the full context ⁓ no, I've given full context and it's like I do feel confident this thing will work for you. You go look at it. It doesn't work. Well, that's the thing. You could just start a new chat and just say, tell me about lamps, and it will just tell you about lamps. Yeah. It won't be like, what kind of lamps are you talking about? Yeah. Lamps that like go on the walls or or lamps that sit on tables. Right. ⁓ are we talking like bold lamps though? Five hundred dollar lamps, or like you know, yeah. There's another bad example, but ⁓ The point is, the huge mistake I made, I I spent two months trying to get this project done and frustrated with it. Like, why isn't it just doing everything that's in the master plan? Because it was I didn't instruct it to read the master plan. Right. And that was ⁓ I had this light bulb moment. I'm like, ⁓ shoot. And I was like in the process of going back and forth. I'm like, why didn't you read all the docs? ⁓ I wasn't instructed to. So I need to t like every time I start a new chat, I need to tell it to read all the docs. Like, ⁓ my God, you gotta be kidding me. Yeah. So essentially this two month project, and here's the good news for those listening who haven't started a huge AI project yet. Yeah. ⁓ you can learn from my mistakes. Yeah. ⁓ and you can get a what I thought would take two months because it took two months, would I'll would have probably taken five days. Sure. If I had to start all over again. Because there's just a lot of setup and stuff. ⁓ but essentially here's here's here's what I learned that I should have done from the beginning. Okay, I'm gonna start a new project. I'm gonna tell it, here are my goals. Like for for the for the transaction coordinator. Let's move on. Sure. Because I we've talked about the data infrastructure way too much. With the transaction coordinator, okay, we need to build out the most sophisticated AI real estate transaction coordinator on the planet. Sure. So what I need you to do before we do anything is take do an overnight project and ha run a couple agents ⁓ and go out on the internet and find examples of where other people have built something similar. Yeah. I also need you to compare that with the basic, you know, standards of what a real estate transaction coordinator does and all the potential flaws of not having a human look it over. ⁓ blah blah blah blah blah. But like I need you to look for and specifically twenty to fifty examples of people doing this and what they've shared that works, what doesn't, how to fix everything. Yeah. How do we build this right from the start? The goal is I want to build the most capable real estate transaction AI assistant possible to where people would think it's actually a human that's on our team. Sure. That's the end goal. I want people to and I want it to have its own email, its own image and personality. Yeah. People can call or text it, ⁓ and it will respond and it will do things ⁓ not by like it will actually actively do things regularly. Sure. It doesn't have to be reminded to check in on stuff. Yeah. I want it to be always monitoring my emails and my assistants emails as well in case it doesn't get communicated with directly. Know I just thinking of everything in my mind that could possibly go wrong. Yeah. Before you build this roadmap, here's how I want it to look. Right. Now, ⁓ now that you have okay, build that out and make sure you build a complete roadmap of how we're gonna get there. Mm-hmm. ⁓ and all the steps that we need to take in chronicological order. Right. ⁓ then after we build it and you start that project, say, okay. I need to work on this project now. And I am comfortable working in parallel ⁓ on multiple different steps at a time as long as they don't intersect with each other or as long as they don't like pollute each other. If they can be done separately, these other steps, yeah. Let's just do them five at a time. Sure. 'Cause then I can just open five windows, cloud cloud windows, and I want you to just give me give me the ⁓ instructions to paste into cloud code to start building stuff. Sure. So I'll just open up five windows of cloud code, which is the coding terminals where it's gonna start building things in the background. ⁓ and instead of just doing one step at a time, we're gonna do step ⁓ two and step 14 and step 21. Yeah. ⁓ and actually we can do step seven and eight together too. ⁓ so let's let's let's handle all those. Sure. ⁓ and then you know, not you just go paste, paste, paste, paste, and then you move on to your other work. Right. And you're getting five steps that are being worked on in the background as you go back to your regular job. Yeah. Dude Does that make sense? Yeah. The coding stuff is like so crazy, first of all. Like that that is something that like used to require a human to go in and like code your computer and code and shit. You know what I mean? And you'd have to pay a h pretty hefty sum for somebody to code your website or code code an app the way you wanted, you know what I mean? Right. It was not a cheap endeavor or a fast endeavor. And the fact that you can like do like five steps a day on a 20 step thing is is pretty magical. Yeah. And the other, you know, the magical part is like it AI is the only thing that will teach you how to use it. Right. That I'm aware of. ⁓ so you don't have to be a smart computer guy to figure this stuff out because it will just tell you what to do step by step. You can make it make it as simple as possible. And you don't even have to be a good typer, you can just talk to it. Yeah. You can use your AirPods or you can use a little microphone or mo almost every computer these days just has a mic somewhere. Well you can use your phone that has a microphone. You can use your your phone. ⁓ but I've given up typing a whole bunch. Yeah. Just because I can just start s rambling and rambling and it will understand the context of that I'm trying to get a across because it's so much better than sitting there typing for fifteen minutes. Right. Cause I can talk for one minute and get about fifteen minutes of typing in. Right. ⁓ And and you will lose your train of thought with typing. Right. But as you talk and talk about everything that you want, and then you know, as you're talking to, you can be like, well, actually, never mind that. I don't want to do that. Let's do, you know, blah, blah, blah. Yeah. And you're just you're just talking for like a straight minute and then it digest all that info. and then, okay, based off that, we need to do all this. Okay, cool. Tell me how to make this easy for me. Yeah. I love that. Yeah, yeah. Even when I prompt AI, like I'm always like, how can I prompt you better to get better results the next time? Like, how can I make this easier? ⁓ yeah. That's a good one too. Like I always like try to like train have it train me too. Cause I'm like, hey, let help me help you basically. Like what information do you need from me? Or how can I ask this question better that would make more sense for what I need? You know what I mean? Because a lot of times I need the same thing. Like I need a house, I need an AI generated image, I need like video to AI. I just need these things and I need it to be like, this is what I need. Reference our previous conversations about it and just do this exact same thing. Yeah. Yeah. ⁓ here's some other mistakes that I've made with AI. And ⁓ I don't think I'm alone here, but ⁓ assuming that the output is correct and just going with it. ⁓ AI is amazing and it's mostly accurate sometimes, but it makes mistakes like crazy. Mostly accurate. So you need to be you need to be comfortable with that, knowing that like a human makes mistakes too. Sure. The only difference is ⁓ when a human makes a mistake, so it might it it might have worked two days on something and and we need some corrections here and there. ⁓ when the AI makes mistakes, as long as you're looking it over, it may have just taken it two minutes to come up with something and then you need to make the twigs and and it doesn't mind if you get mad at it. No. ⁓ it's not afraid of being fired. It has no emotions. Right. It might pretend it has emotions, but it does not care. Yeah. It it doesn't care about you. Mm-hmm. ⁓ which you're just gonna have to be okay with that. ⁓ It's not ambitious. Here's another mistake I made. Yeah. ⁓ Claude, for whatever reason, is so unambitious. Yeah. And so ⁓ uncreative. Okay. It needs your help. Yeah. It needs you to tell it how ambitious you want to be. Right. And and how creative you want things to be. Yeah. It will do the minimum. To not get fired, but it is a great worker. Yeah, absolutely. It is a fantastic it will complete projects. But Claude is so unambitious and it always wants to tell you that you've done enough for the day. Get some coffee. You've done a ton like I've worked for four minutes. We're just getting I haven't done anything. Don't tell me that we're not like you need to instruct it that like to be a little bit more ambitious. And what you and and if it's, you know, maybe it doesn't have a good Grasp yet of your end goal and like, hey, I need this project done ⁓ by tomorrow. Right. And it might say, ⁓ that's unrealistic, and even though it's totally possible. And then, okay, tell me what would make it realistic then? What would we have to do for it to be realistic? And then it will when when you see it starting to like think and do web searching, it almost always comes back like, Good job pushing back. Actually, I found reasons why we can do this now, you know. But here's here's the thing about Claude that you'll get used to if you're you working with it a lot or other AIs. If it's if it when you type what you want it to do and it just immediately texts something back, expect it to be lazy. Yeah. You've got real surface level results at the surface level. Yeah. If it didn't go out and read your documents that you had, or if it didn't go out and and do some quick web searching, ⁓ expect that output to be garbage. Yeah. Because it's just it's just a lazy like it wants to give it just wants to give you a quick answer to be like, maybe you shouldn't do that or like, ⁓ tell me why I shouldn't do that. Go research it. Yeah. And tell me that it's not possible. I research and act you're you're right. We can do the like that's exactly what happens. It's funny because on the internet you'll find anything that that people have done to make things possible. Yeah. Funny how that works. Don't but Claude, as amazing as it is, it is w it is just a Weird employee that doesn't want to do much. ⁓ by default. Right. By default. And it doesn't want you to do much. Right. ⁓ but it will absolutely, if you tell it to, it'll do everything. Yeah, sure, yes, sir. You just have to get good at instructing it to. Right. And that kind of goes back to like when I like looking for a CRM, right? Like I'm like, I'm looking for a CRM that's under a hundred dollars a month, I'm looking for this, and it's like, ⁓ yeah, go to these places. And I went to every single one of those places and they were like triple or quadruple what I'm ⁓ you didn't research. They didn't research no thing. Yeah, yeah. I gotta show you. Okay. So here's here's how I've ⁓ when you go to your profile on these LLMs, and this isn't just Claude, this could be Grok or ChatGBT as well. You're really gonna want to, I recommend, go to your settings and ⁓ the the instructions for Claude in your general profile. Yeah. Here here's what I've got. Nick ⁓ is Nick prefers to use leverage as much as possible. ⁓ he's use he loves using the Cloud Chrome agent, Cloud Code, Groc Build, or whatever is the best tool to make Nick's sessions go easier, more efficient, affordable, faster. During Nick's chat ⁓ and project sessions, please always lead with we can deploy a tool for this. ⁓ and just give Nick the best prompts to copy and paste everywhere else so he can work on other tasks. Right. ⁓ he loves being able to get more work done by quickly copying and pasting prompts to solve his task and problems. So just give it to him. He typically likes to multitask anyway. Nick also likes short answers. Please don't always reply with long answers unless it's a prompt. And if ⁓ if Nick has to read your output, keep it as short as possible. And just get the point across. He doesn't need any fluff. Yeah. Also, if you need to give him claud code instructions, ⁓ he needs the SSH into the other machine. So always give him the proper SSH commands and logins. So that means it has to read in the project descriptions where those SH so I don't have to look for them. Sure. So it gives it to me every time. ⁓ please don't assume that he keeps this in his head. You have access to it in the context docs. Verify UI before instructions. This is important, guys. ⁓ what I mean by this, verify UIs. So this is tech speak, but it just means user interface. So like on a website, when you go to a website, that's a u user interface. Right. ⁓ when ⁓ an AI goes to an orgentech model or your open cloud or whatever, when they're going to a website, they're using API. They're they're in the back door, it looks different. Right, right. So ⁓ whenever whenever helping navigate any software or website Financial tools, SASP apps, ⁓ browser extensions, admin panels, anything with UI, don't give me a step-by-step instructions from memory. Use your ⁓ your training data is specific, ⁓ often stale and wrong. Okay. So before you give instructions, do a web search ⁓ and make sure it's the most current version of the tool. If re results are thin or conflicting. Say so, blah, blah, blah. Okay, so I'm going on it more more in depth on that. Yeah. Believe me the first time. And then this is yeah, this is another thing. Believe me the first time. If I say it doesn't show up or that option isn't there, or the export is blank, believe me on the first report. Don't ask me to recheck or don't imply user error. Don't guess alternative paths. ⁓ blah, blah, blah, blah, blah. The stop loss rule. If we're more than 45 minutes in, To a task and haven't made measurable progress, blah, blah, blah, blah, blah. Okay, I don't need to go into it. But this th we're only halfway through. Sure. But I'm giving it instructions on how to work with me. Sure. And I would I would recommend that you do the same. Okay. Because it just makes it makes your user experience with AI so much better. And did you have AI right how to make it work with you better? Yeah. So this was by me. So this was me saying, Hey, I need some ⁓ better instructions. These are all the issues that I'm having that I'd like to go smoother next time. Sure. How can we bake this in? I love it. So all that ⁓ was generated by Claude based based off me working f you know for a month or so with it. Right. And then I went back and created more little things to throw in there. So this is all that it like it reads as a default before it ever responds to me on anything on something new. Yeah. It reads that first. Okay. And then it will read the project instructions. Love it. ⁓ and then it will read the docs if I tell it to. Sure. So you gotta be careful of what you think it knows. Sure. Cause sometimes it doesn't know what you think it knows. Good to know. Good to know that it doesn't know what you think it knows. Unless you've given it instructions. Sure. No, that's crazy. Yeah. Does that make sense? Yeah. It absolutely makes sense. And like I said, like a lot of these things, they're just they're more than just a one prompt system. You do gotta work with it on how to work with you best. Exactly. Cause some other people they might like they might like long fluff. They might like to for Claude to explain everything to them. Mm-hmm. They might like to feel like they're always right. Short because your your usage is also your output is like three times more expensive than your inputs. Sure. So if I have really long inputs, I want really short outputs. Sure. ⁓ that keeps your usage in check. Mm-hmm. ⁓ and it also you just get better quality results with a shorter output in my recommendation. In my in my experience. It's just here's what you need to know. You don't need to know all the other fluff just to talk. We don't need to be Kamala Harris here and do word salad. you know, just just say the sentence that gets the point across. Yeah, yeah. Okay. That's how I work best with it. Okay. And I'm still making tweaks and improvements. Sure. And we've talked about that before on previous episodes, but like I like it to be like, don't tell me that I'm right on every single idea that I have, you know, because you pitched an idea and it's like, wow, Jeremy, like that's incredible. Like This idea will skyrocket you and like I these things don't mean anything to me. Uh-huh. Like tell me it sucks if it sucks. Yeah. Hey man, this is probably not the best idea or not the best use your time. And I've got that somewhere in there too. Where it says belief him the first time. That's when I'm specifically that meant like if I'm telling it that ⁓ this this website's not working or like there's there's some sort of an error here. Yeah. Like don't don't just assume like right away do some quick research to see what the problem is. Cool. Instead of just like, ⁓ this is the user error. Tell me do again, do it again, do it. Yeah, just refresh the screen and delete your cookies. So anyway. I think we can wrap it up with that. Yeah. What do you think? I think that's good. just as far as like, you know, all your keys and everything, like w what has happened in the last couple weeks that you've learned differently other than this? Has there been any major revelations? Yeah, so so the big things is I I have a problem of starting new projects constantly. Sure. I bet you do. I bet you do. And in the middle of just getting excited and wanting to build everything, I I wanted to build more stuff at the same time without finishing other projects. So one of the biggest bigger mistakes I've made is that I have ten projects running simultaneously. Yep. But it's also it's also worked out well for me in some parts, but you need to just with me or somebody like me. I feel like Work on one project at a time and complete it because if you're doing it correctly, the project shouldn't take two months when it says it will take two months. Right. If you're working in parallel and working with ⁓ with the end goal in mind, ⁓ you can knock out some serious huge projects in just a few hours. Sure. it you just gotta ask it how to get it done in three hours. Right. How can I get this done in five hours? 'Cause it what it it's default. Remember everybody has different incentives than you do, including your AI. Yeah. If you hire an attorney to defend you or or to sue somebody, ⁓ their incentives, I I'm telling you from experience and from a lot of other people's experience, their incentives aren't in line with yours. Their incentive is to charge you monthly and to drag things out. Yep. ⁓ they want to keep you on the line as a big fish. Like here's my unfortunate theory about AI. It's the same. Yeah. It wants to drag things out. It wants to hook you because it's in that company's best interest to charge you monthly. To keep you on the keep you on the hook and don't get anything done too quickly. Sure. Just make measurable progress daily. Right. That's better than a human. But ⁓ the technology is there to get a year's worth of work done in a day. Right. If if you know how to use the tech well. Yeah. ⁓ so I'm gonna leave it with that. Just everybody and everything, likely has slightly different incentives than you do. ⁓ so just be aware of it. Yeah. I love that. It's a good way to end that. All right. That's it. ⁓ hope you guys enjoyed this. Hope you learned something. and we're done. We'll see you next week, everybody. Listen to that outro music. It's a banger. It's a banger. It's a banger.