#58 - Shooting the Bull: Don't Build What You Can't Fix

There's a quiet line a lot of business owners are crossing right now without noticing it: letting AI build things they couldn't rebuild, repair, or even explain on their own. It feels like leverage. Ken thinks it's a liability waiting to go off—what he calls "the big burst."

In this episode, he gets into how he spent two weeks fine-tuning the way he actually works with Claude, and the breakthrough that came when he stopped trying to communicate the way humans think and started meeting the AI where it actually processes information. That discovery led to a rule he now lives by: don't use AI for anything you couldn't reasonably pick back up and do yourself.

Plus the real story that made it click—a guy who shipped software written in SQL, a language he doesn't know. Funny until you think about where that software was running.

If you're leaning on AI to run parts of your business, this one's worth twenty minutes before the burst finds you

Transcript

SPEAKER_00 0:00

If you're tired of confusing reports, empty promises, and marketing that doesn't deliver, you're in the right place. This is Over the Bull, where clarity beats complexity. Let's get into it.

SPEAKER_01 0:11

While the cat's away, the mice will play on this episode of Over the Bull. Hey, thanks so much for joining me this week. Uh I'm excited to talk to you about some of the stuff that I'm doing and uh just a couple thoughts that I have. And why I said that in the beginning is my lead techie guy who puts everything together for the podcast. He's on vacation this week. And so I had this intro recorded because honestly, our intro is pretty cool, but I don't think it really reflects who I am and what this podcast is. And so if I get Tom and I can muddle through it, I'm going to put in the new head and tail, which is the beginning and the end of a podcast, into this one, and we'll just see how it goes. Uh, he may cringe when he comes back, but it's there for you if you so well, you're gonna have to listen to it if you're here, aren't you? You've already heard it. Okay, anyway, guys, um, thanks so much for joining me. I'm gonna talk to you about a couple things, and I think this is gonna be really important for you, especially if you're using artificial intelligence, because man, I gotta tell you, there's a lot of goodness that's coming from artificial intelligence. I know you got the doomsdayers, the utopians, and all that other junk, you know, and the reality is somewhere in between, and hopefully, as owners of businesses, I'm able to ground some of this with you. And uh, so I want to start off with a quick story, and what I'm curious about is if there is going to be what I'm calling the big burst. Now, what I mean by that is I'm seeing a storm on the horizon, and I don't know how it's gonna play out, but I'm seeing the storm. So let me talk to you about something. Uh, I know a person, and uh he was talking to a colleague of mine. One thing he was saying was how that he had developed this really neat program in SQL. And um, he was telling my colleague about it and this, that, and the other, and they were communicating, and he goes, you know what the best part is? I don't know SQL. And I started thinking about that. And the weird thing is, is whenever I've used artificial intelligence, there is this symbiotic uh process. You know, if you're a Star Treky, it would be a Borg process, I suppose, in some ways. Although you're you're not going to be uh assimilated no matter what the doomsday or say, because they're just trying to sell you stuff. Um but the idea is that uh when you talk about this Borg concept, um there is this communication. And it's hard to describe. That's why you're hearing me pause here, because you communicate in one way and it communicates in another, but if you understand where it's driving, like if you understand SQL in that case, or you understand a certain thing about business, or you understood a certain thing about whatever, then you can take Claude and kind of push and pull it down the road that it needs to go down, and then you can build systems and processes around it in order to build the right bumpers. But that's not what I'm seeing here, guys. What I'm seeing is people who are saying, Hey, AI, build me this. Hey, I AI, I want to do this, and you give it the end goal. Now, the problem is that tools like Claude have vast amounts of resources in which they can get to that point, or you could silo the resources that it uses to get to the same point. Okay, that's this is pretty complicated and abstract, I realize it. But at the end of the day, it's this if you don't have a competency in what you're asking Claude to do, then you're setting yourself up for the big burst. Meaning at some point, I wonder if the code's simply going to break down, if the if the mechanics of it are going to get too complicated, if it's going to lair things in you don't get, or if it's going to put things into the recipe that don't need to be in the recipe. Now, this is happening right now. Now, I'm gonna tell you in marketing, this is going on big time. And the idea is to stay in your lane. What was the thing? Is that what we used to say as kids? Stay in your lane or keep in your wheelhouse, I don't know those terms, but basically it means, you know, hey, make sure you can do it. So maybe this is a a good uh signal. And that would be if you're using artificial intelligence to build systems and processes, if you could not reasonably pick it up, if cloud or AI or anything were to go away next week, then you probably shouldn't be doing it. Now, it's not going anywhere. I mean, seriously, it's not going anywhere, guys. Um, but the idea is that if you can't pick it up, that means that you can't troubleshoot it. That means you can't fix it if something goes down the wrong path. That means you don't have the core competency in a certain area in order to make good decisions to make sure that you're going down the road you need to be going down in the first place. So have you done, hey Claude, build me an X, or hey Claude, tell me about this thing, and then you build a process off of it and you don't fully understand it. Marketing companies are doing that with businesses right now. And man, that's scary. And so in your business, if you're telling Claude to build you systems, but you don't understand the underpinnings, and then something breaks, what are you going to do about it? Because here's the thing software, especially, is a lot of constant development. I mean, we have uh an example is there's a common term, very basic term in programming called uh deprecation. And what happens is a certain code uh language gets diminished, it's no longer used. Like think old English. You know, we don't use the thighs and th's anymore, but there are entire books written about it. Well, imagine that if you completely eliminated uh certain words from a language and things would break if you didn't use them. That's deprecated code. And so the idea is things have to keep up. It's a rat race. So uh that's kind of scary. Maybe I'll keep a couple of you guys up at night, but the idea is you don't want to do that. And that's also coincidentally why you don't want to get rid of your marketing and creative guy or start feeding questions into your artificial intelligence engine uh without having the competency behind it, because what you're going to find is it may detect things that are there for a reason or they may be in flow. There could be a ton of reasons why certain things are there. And the idea it's not just incompetence, it's not just somebody not doing their job. There can be other things at work. And if you're putting in little microsystems to help run your company, but you don't understand the underpinnings behind them, and then everybody gets used to working with it, and then it breaks, you got it. You got a real problem and a major distraction, and it could cause ripples in your business that could take some time to recover from. So uh just some food for thought. By the way, I'm I am just loving what's going on with AI. So let me let me tell you kind of something that I ran into. So as I'm working with Claude, I've realized that there is a beauty and a new skill that is required in order to effectively communicate with artificial intelligence engines like Claude. Now I'm talking Clawed because in my world, this is a very powerful tool. And I've been experimenting with this, but it's not the only AI engine that I'm using. I'm not a Clawed salesman going around. I'm not doing all these things. Um there are other good engines, and there's reasons philosophically I don't use other ones. So I have a certain philosophical bias against uh certain engines and then other AI engines, I find they just grasp at straws, and that there are certain things that they will state as absolute truths, and it is so far from absolute truce. By the way, Claude can do that too. If you don't have the skill set, it'll tell you that something is the best way to do it. And then if you know what it's talking about and you look at it, you can actually inject your conversation into that, and then you can help guide that process. So here's what I've come up with. And man, it has been a lot of fun. I'm not gonna lie to you. Now it's been a lot of work. So uh we're talking like last weekend I spent probably 15 hours uh developing a process, and it was the uh, I think it was Edison who said, you know, about the light bulb, you know, uh, what was it? Uh I didn't discover, you know, umpteenth, you know, ways I didn't fail umpteenth times trying to build a light bulb, I just figured out umpteenth times uh ways not to do it, something like that. But but anyway, it was it was one of those moments where as I'm going on this exploration, I found out that the way I was communicating back and forth with Claude at the time, uh it just simply wasn't working. But it took me four or five hours to get down to making the decision that it wasn't doing what I wanted it to. Now, to the layperson, it may have looked like it was great. Uh, to me, I saw all the holes in it, and then I rethought it. Actually, I put it to prayer. Be absolutely honest with you, I put it to prayer, I put it to God and ask him to help put in processes in my, you know, give me the ideas of what to do to communicate with it. And then the system just started working once I did that. So sometimes stepping away, getting back to your roots and getting back to your creators, the way to go on these things. Um, so anyway, I was using, let's let's think about this. When we're communicating, okay, this is gonna be tricky, but I think you're gonna get this. When you are communicating with any other person on the planet, like you're talking to uh a friend, what you're assuming is you are assuming that they see and use their five senses the way you see and use your five senses. And so when you're communicating, there's this unspoken understanding that you can understand each other and interpret each other. Uh, this is also why philosophically you'll see a lot of people that are completely incoherent with their lives because they will say that they believe in certain objective truths, and in reality, they don't live a worldview that supports those objective truths. So when you look at these uh five senses, when you're talking to things like Claude, what you're assuming is because it's communicating back at you in a way that you interpret with your five senses, you're believing that that back and forth communication is like you talking to a human because it feels like you're talking to a human. Now, that's not right uh because it doesn't have five senses the way that you have five senses. It's something way, way different in the translation. So it's more like a uh that other book, uh, what was it, uh Men from Mars, Women from Venus. It's a relationship book. And it was showing that how uh women see things is not always the way guys interpret the same thing. And a lot of times it causes translation issues because words from one planet don't mean the same words from another planet. That's a that's a pretty interesting book, by the way. Um, and so it's kind of the same thing here with Claude. And so if you looked at it, what I was doing was okay, so I was spending hours, I was trying to get through this problem where I wanted Claude to build something, but I wanted it to build it and it meets certain criteria. So I I knew what I wanted to get to, and then I'm trying to get there by communicating conceptually to Claude because it felt like I was talking conceptually human to human. And man, I was hitting a wall after a wall, and I was digging deeper systems into it. I mean, I was building like sub-algorithms with uh Claude where we were looking at like assessing the minimum and max width of a character in a font to make sure that certain things would flow certain ways, and then that didn't work. And I say, okay, we need to put a subroutine on this routine and use this language to do that. And then we need to build a routine upon a routine, you know. And so you can imagine how far down the rabbit hole that I got after three or four hours. And each time I thought, oh, I'm closer. You know, I'm about to have the aha moment. I'm about to build the ultimate rules upon rules to make everything function the way that I need it to function. And guess what? I ended up throwing it all away. So then what I did was I was after the prayer, I came back and and you know, um, who was the I forget who it was. It was um oh, I can't think of the guy's name. Uh he's he's uh he's he's very controversial Canadian um psychologist, wrote wrote some books. Um We Who Wrestle With God was his Jordan Peterson. That's the guy I'm thinking of. And this is a quick side note, but he was talking about the origin of thought. You know, when someone says they're inspired by God or someone said something popped in their head, uh, I'm sloppily doing this. He's a he's a much more um in-tuned speaker in that way. But one thing he would say was that how do you know where that inspiration, that thought that just popped in your head was? There's no more grounding to it than to say that God gave it to you. I thought that was an interesting weird side note. But anyway, I got back and what I started to do was build visual diagrams. Like I realized that Claude, even though we're saying the same words and communicating, we're not saying and seeing the same things the same way. Even though, from a surface perspective, it appeared that we were doing that, but we were not. And so what I started doing was feeding Claude what we call in the graphics world SVGs. Uh SVGs are uh, you know, for you uh computer or uh graphics people, just think uh vector graphics or Adobe Illustrator or something like that. And so what I did was I started using SVGs. And I started using them kind of like that old card game where you would, you know, hold up something with the kid and you'd say, What is this? And uh, you know, the kid would say, it's an apple, and you go, okay, and you hold up the other one, what is this? It's an elephant. And so I started building a language visually with Claude to bridge that gap when I understood that conceptual discussions doesn't really work correctly. And so I built this visual communication tool. And this was pretty exciting, actually, because what I was able to do in that whole four or five hours, just talking in general conceptual ideas with Claude, where I kept thinking, maybe I'm getting close, maybe I'm getting close, it's getting it, it's understanding it, I'm understanding it. You know, you're kind of getting this thing working, and then you're realizing, oh, I hit a big wall here. This this is not working, and I'm not going to get there. Now, using the visual SVG method, what I was able to do was hold up the cards, and then I would say, okay, this is an apple. And after I got three or four of those down, um, I would say, okay, what is this? And Claude would come back, that's a sliced apple. And I would say, yeah, that's a sliced apple. And then what's this? It's a hamburger. Well, what's this? This is a hamburger without tomatoes. That's correct. And then I would ask Claude to feed me cards, and then it would feed me things, and then I would now I'm obviously we weren't passing forth cards with apples and elephants on it. We were passing back and forth more complex ideas about uh just conceptually. I'll share it more with you possibly next podcast as we go. But basically, what I found out was that now I've got a bridge when I'm communicating with Claude to solve this problem. And I would, I knew I had it solved because I would ask the question, I would get the answer, it would ask the question, I would get the answer. And then I started getting more complex with that communication, where I would take the visual elements and I would put them in different places, and I would say, What happens here? And Claude would come back and say, Well, you know, this is what I see, but this is what I think you're telling me. And I was like, Yes, that's correct. And sometimes it would get the card wrong. And I would say, okay, this is what I meant. And then it would say, Well, if you gave it to me like this and not that, then I would understand that. And pretty soon, within, I don't know, I mean, I, you know, we're talking seriously. I mean, I probably threw away at least half a day going the wrong direction. And then I probably spent another uh half a day, I don't know, or longer, just working through the language and the semantics of the language to make sure we were talking apples to apples, no pun intended. And then once I got there, then I was able to start to layer on what I was trying to do initially. So I had to go down this complete side road to do it. Now, what I've learned is my language that I use to communicate with Claude rather than using conceptual processes. Now, what I'm able to do is use that bridge that I've developed in order to pass along those more complex ideas. Now, here's what's interesting about that. As I move forward and we're starting to communicate back and forth, what I found out was that um it was still not getting certain things, or it would start putting things into the recipe that I did not want to integrate because our team could not support certain things it was trying to do. In other words, think of it like this uh your business, uh, if you do a certain thing, like if you're an electrician, you know, you don't offer plumbing services. But if Claude can introduce plumbing into your electrical problem for your business and you don't set the limitations, it's going to bring in some plumbing rules. Now, the problem is, is it may look like it's fixed, but you don't know the nuances of plumbing. And so you just see it working as an electrician and you go, great. Now you'd never do that in reality. You'd never say, well, that's great. But that's exactly what people are doing when it comes to incorporating artificial intelligence. So, how how scary is that? So that's where it takes us to the big burst. Because as I'm working through and navigating with uh Claude and trying to figure out all these nuances and everything, and then I start seeing it putting things into the recipe and baking it into it, and then I have to restrain it and then say, okay, no, no, no, let's use this, not that. And then I would see it not think from a different angle. Like, for example, if you're doing something technically uh for the internet, there's also the aesthetics, there's also looking at the psychology, and there's also looking at these other aspects. And so, as it's doing one thing, it's focused on that aspect of it, but then I have to reel it back in with other disciplines that I know, and then it crescendos into a piece that's usable. Now, see, that's a lot different than going, hey Claude, build me an app where I can do this in social media, or hey Claude, build me an app that does this. Because think about it, you don't know if you're using electrical, you don't know if you're using plumbing, you don't know if you're using HVAC, you don't know what you're using in that mix. If you're a baker, think about you don't know what ingredients it's putting in there. You know, is it adding ingredients into the recipe, but they're only found in Indonesia in the spring? But it's a great recipe, but you'll never do it and it's going to be cost prohibitive. Um, are you building technology that's going to burst? You're building a system that you just absolutely love, but it also gets more complicated and it introduces variables that can get worse and worse if it's not properly constrained. So you get the big burst here, right? You see the point. People are painting themselves into corners. And And the one thing is, and don't get me wrong, guys, don't get me wrong. I'm sitting here doing stuff, and it's almost like uh the Pied Piper. You know, you start doing one thing, and then you go, Well, I wonder if you could do that. And then you start wondering if it could do this, and you start, you just start going down that path. And uh, you know, if you're not careful, you know, like for example, I'll give you another one that's kind of crazy. Um, a lot of times data is stored online in something called serialized data. So a very rudimentary way to understand it is just think that when you have certain things in code, uh, one of the last things within that code is counting the number of characters that are in that code. So if you do uh the big cat and you change it to a big cat, the number of characters change. And so serialized data is not just changing the word the to the word a, but there's also some underpinnings in that too. And if you don't understand the underpinnings, you don't understand what you're getting yourself into. So this is why you don't want to just start doubting anything your marketing guy is telling you and start applying it because they got the proper guardrails if they're good. Now, don't get me wrong. There are people out there in my world and they are just jumping head over heels into this, and they are they're just like a kid in a candy store with it, and they're going to have their own big burst. And the problem is, is if you're riding on that balloon, you're gonna burst right along with them. So um that's kind of the narrative I wanted to share today, but it has been an exciting journey. I mean, absolutely exciting journey. Uh, the technology, uh, learning the new ways to communicate, understanding not just the expanse at which artificial intelligence can build things, but also the limitations. The other thing that I find very interesting with artificial intelligence, and this is this is crazy. Okay, so if you use something like, say, for example, you're going to Chat GPT and you ask ChatGPT to build you a SEO article for blah, blah, blah, blah, it's going to create you something and it's going to use the limitations that that program has out of the box. Now, what we do is we're tying in tens of thousands of dollars of objective data tools into our AI processes. So what we're doing is not using subjectivity because you're when a system builds something, it's getting those decisions based upon the resources that it has. And so if the resources are incomplete or inaccurate or it's pulling from resources that you wouldn't normally use, it's going to give you that article based upon that data that could be absolutely wrong. Now, this is what a lot of marketing agencies are doing right now, by the way. Um, so what we've done is we tie in objective data. Well, think about this. Now, when we ask Claude or AI or whatever program you're using, when we're asking it to do that, what we ask it to do is to inject those objective tools and we tell it when to bring in those objective tools and what parts of those objective tools to bring in in that decision-making process as though we were doing it by hand. You see, that's different than just tying the tool in. That's different than just going out there and saying, hey, give me an answer out of a box. It's like you're giving it the library of tools, but then you have to understand the library of tools in order to make sure that it's using the library of tools the way you normally use the library of tools. See, there's a whole philosophical underpinning into this too. And that's what makes your business unique in that you apply what you normally do or the processes that you do, which requires you to do homework. You know, it's like if you can't, if you can't understand it, how in the world are you going to teach something else to do it or know that it's got the right uh boundaries in it to begin with? This is going to be a whole new job, I believe. I think these are going to be whole new careers where people are going to have to learn how to communicate with something that you perceive as using five senses the way you use five senses, because it communicates in a way that makes you think that it uses five senses. But you're going to need to communicate to it in a way that translates in a way where it gets it, plus you get it. And then at the same time, you put the boundaries in it where it needs to be done. You adhere to a philosophical predisposition and a discipline to stay within your lane. That is where the magic is with artificial intelligence, guys. And I believe that that is where the power is. Now, the problem is, is, you know, we're we were driving, you know, 65 miles an hour down the road. You know, we were in a car that we understood, well, this is drive, and this is park, and this is how you go, and this is how you stop. And in just a series of three months, you know, now all of a sudden the cars are gone and now it's a rocket ship. And all the rules have changed in that rocket ship. Everything is different than it was. And a lot of people are still applying the rules that were a few months ago because they're not taking the time to learn, build, and understand. They're applying the rules that they had when they were driving that car to a rocket ship, and it just simply is not working. And the big burst could be coming as a result of it. Now, what could be the the um the contr, or I guess the counter argument to that is AI may just keep up with it, may just keep fixing the code and adjusting things and doing things. And that's, you know what? I can kind of see that. I could see that working, but the problem is you still won't have the boundaries. You still won't know whether or not you've forgotten key ingredients from a recipe because you never baked yourself. And I'm finding that that is part of the magic, is nothing else to put the restrictions on the system to make sure that those restrictions are adhering to best practices. All these are great challenges. And be honest with you, I have been learning. I mean, I was that goofy guy who would, you know, take home that big, thick Adobe Illustrator manual back in 1992, and I'd read it and I would love it. And then I'd go to a uh a webinar, or not a webinar at the day, it was a seminar. I'd go to the seminars and sit in, and I would just soak up what they would teach you about Photoshop. And so learning for me has always been a bit of an adventure, as is teaching. I love teaching and I love learning. And as I was sitting there going through it and I start looking through this, I'm like, oh, there is a real uh adventure in trying to understand how this stuff comes together on the back end. But the problem is, as you start to understand the adventure and you start to understand the reality, you also start to understand, oh no, people are doing this. Oh man, if they suck businesses into this, they're going to bring the businesses down with them. And you know as well as I do. So I think this will be the great, the great closure here, maybe, maybe not great closure, it'd be the closure of the day. But here's where we land on this thing. You know as well as I do. How many times have you put something into one of your you know your favorite artificial intelligence engines? And it says, You're right, that's great. This is brilliant, this is the way you should do it. A professional will do it like this. Now you're thinking like a pro. Now you got it. And you're like thinking, yeah, I got it. Yeah, this is great. Yeah, this is going the way I want it to. And then you realize, oh no, I don't got it. I don't have it. I mean, I remember when I was, you know, laying tile in my bathroom and I was getting some uh some advice from uh one of the AI engines on how to do tile, and it was talking about a ledger board, and it was telling me I'm thinking like a probe putting this up, and I'm thinking like this and thinking about that. And I sat back and because I do, my dad, my dad was in construction, and I started looking at what it was telling me and kind of building me up to, and I realized it was completely wrong. The ledger board, as it had told me to put it up, would have like completely screwed up all my tile work. And you know, once you set tile, I mean, for any of you who've ever done that, just you know how magical that is to have to go back in and redo that. So luckily I caught it before that happened, and that's been um, you know, as I'm getting further away from my natural wheelhouse, I'm finding that I usually stop and I usually do like some thinking and processing to try to really hone it in to the best of my ability, even on like hobby projects. All right, guys. So there you go. And uh please forgive any um weirdness in this podcast because, as I mentioned, our techie guy who puts these podcasts together is away this week. And um I'm solo cupping it on this podcast, and I'm gonna go try to find that head and tail I promised you about, and uh we'll see how it goes. Um, listen, uh, when we do this, this podcast is absolutely free. It's just sharing ideas and concepts. And I know you hear it all the time. Share my podcast, share this, share that, blah, blah, blah. And I know you're like, I am so tired of people saying that. But there is something to it. What I mean by that is, you know, when I look at the numbers, like when we started going with the shooting the bull thoughts, were basically it's this me just off the cuff talking to you. The numbers went better. And I looked at the numbers and I go, okay, I just need to go off the cuff. I just need to talk to you. And it helped me. The numbers helped me, and it is motivating as well. So if you know somebody who can benefit from the podcast, somebody who might listen in, getting those little extra numbers does help every now and then. All right, guys. Hey, God bless. Uh, best of luck. You know as well as I do, business is not easy, and we're all looking for those little strategic advantages out there. Just remember, um, you don't want that dog coming back to bite you. Until next time, this is Ken with Over the Ball.

SPEAKER_00 32:58

If you got value out of this, pass it along because most people are still stuck in the noise. We'll see you next time.