Transcript · November 15, 2026 · 20.6K words

Ep. 91: AI Gave Creatives a Crash Course in Engineering

Captions uploaded with the episode. Expect the odd misheard word. Tap any timestamp to play the episode from there, or go back to the episode for the summary, receipts and chapters.

0:00:00Drew and Rory return for a two-host episode

Play from 0:00:11 What's up, everybody? Fast Hours Podcast, episode 91. Drew Brucker, Rory Flynn, and no special guest today. So we're rocking solo, we're excited to be here because we we haven't had time to ourselves here, Rory Flynn, which means that, Yeah, I get excited. you know, there's been a lot of great questions and topics we've gone into, but it feels like we haven't done this in forever now. Because we recorded like those one the, you know, those two episodes previously with Jack and Anthony before that. And we shot both of those like before before Halloween. And so what are we?

Play from 0:00:48 For how in. We're we're we're almost halfway through November here, man. And so a I don't know what's going on. lot has changed, bro. A lot has changed. And I don't we were just talking about before the episodes, like, what has changed? Like we we don't even remember what has changed because so many things have happened in the last couple of weeks. And we've just kind of been rolling with them, but feels good to be back. Does feel good. It's it's been it's weird. It's like you go through a week.

Play from 0:01:13 Like you I I know you were like isolated and you know, on a client visit for a week and it's like if you're so deep into that then you tune out of what goes on here, then you're you come back and you're like, what did I miss? And then you just end up probably like don't do any nothing that is I know, I know. relevant based on that. It's just like, Well what what did I miss? okay, whatever. Just go on too. It it feels exactly like that. Yeah. It feels exactly. And it's like, you know, I we've been through enough of these cycles too now where it feels like you, you know, you know how these things are gonna go.

0:01:36Staying tool agnostic as AI keeps changing

Play from 0:01:44 There's a lot of hype when you first hear about like a new model, coming to fruition. I mean, Fable 5.2 since the last time we did this. you know, new capabilities that are happening, new news, right? But it it it like it's got like a short sort of like half-life to it. And then, you know, it's it's just kind of like reassessing where you're at. And I was just posting about that yesterday. It's just like, dude, one of the best things you can do is like, you know, just be tool agnostic in a large, in a large way, because then you're not so worried about every little thing that happens.

Play from 0:02:16 and you know, frankly speaking, it's like, dude, you know how these things go? There's the hype, then there's the reality of what sets in. And, you know, also like when you get out of these conversations with. with people like in the AI circle, you realize just how far like the space Yeah. is between people that are doing this literally every day and then the average person sitting inside of a business, you know, that's trying to learn AI and use it in their work.

Play from 0:02:42 So And it it's you have to take yourself back sometimes because I I I do that often where I just like assume things that I'm like, Yeah, we're at this point where, you know, again, like are you using MCPs and sort of like, you know, headless browsers and all this other stuff and using technical terms like, Hey, just run it through a cloud pipeline and like it does this, you know, connect connects to Magnific. It also brings it back into Figma, Figma design and they're like, How do we prompt images? Like, All right, Yeah, yeah, yes, yes. let's start from the beginning.

Play from 0:03:09 and you know, you you you forget that that's sort of like I'm like, we haven't been doing this for the last couple of years and you've you know, you also now it's like I have my friends in sort of like PE and also, you know, in in more the finance space and they're like, What's going on with this you know, PHP critical vulnerability? I'm like, I I don't really know. And then I spit out an answer and I'm like, I actually maybe I do know because I've been running, you know, audits on my software to make sure that I'm not having I'm not like giving away things.

0:03:12Creatives picking up cybersecurity work

Play from 0:03:38 But it's Like I d do would I have expected four years ago for me to understand and care about anything that has to do with a critical vulnerability? Like, no. Like Yeah, yeah. It these are these are just kind of the things that like happen to come your way when you're when you're in this stuff every day, right? Like Yeah. I d I don't think any of us signed up for that. You know, but it but at the same time it's like it's also kind of this interesting like variable that exists in terms of like the information we hold on to versus the the information we're trying to basically find and then do something with.

Play from 0:04:14 So it's like not necessarily the retaining function. So for example, like when you bring up the PHP thing, it's like You probably don't remember a bunch about PHP. You just knew you had to do something about it or look into Yeah. it. Right. And so, you know, and so it's like that. And there's a lot of that. And then there's also the part where you're actually trying to like tap into something that sort of like feeds into your day-to-day or or workflows or whatever it may be. And so there's like kind of like those two different types of information that present themselves.

Play from 0:04:45 Yeah. And it's like, you there's there's things that I'm like I have a I have such a surface level knowledge of and then there's stuff that I don't understand where I'm like I you know, I'm asking my buddy who's in yeah, he's in more like computer science and and IT and things like that. And I'm like, Like, why why is this a problem with Python? He's like, Well, it's older versions of Python. I'm like, they how often they release new versions of a coding language? He goes, All the time. I'm like he's I'm like explain He goes, Well, it's like a dictionary, right? It's like every year they add new words to the dictionary, so a new update has to get Sort of made. And I'm like, okay, that's an interesting analogy and way to put it.

Play from 0:05:18 But I'm also like, like, how often do these things update? And then I go back and look at my code base and things. I'm like, you know, God, what what, you know, version of Python am I running? And it's like, you're running like like eight updates ago. I'm like, okay, let's update everything then. And then it's like, well, you know, if you update it this, then like everything that's tied to it, I'm like, my God. And then then you can see though, d yes, yes, yes. you can see how companies that are like, You know, like Microsoft just got hit yesterday and it's like how many how much code do they have?

Play from 0:05:47 I'm sure they have to like maintenance on that all the time. Either way, is that s I Like I think that's still down. I think Outlook and Teams are still down. Is it still down? I mean, it it was as of like early this morning. I haven't checked within like the last few hours here, but It's like how do you keep up with that? Like it the only AI is now the only way. How how can you possibly keep up with like a software company of Microsoft size, that size code base, infiltrating a, you know, some portion of that that might be in like, I don't know, a software that was released five years ago that's mildly used by people, right?

Play from 0:06:22 And again, like so it's you know, again, where we're We're not trying to become this podcast, guys, but it affects us now because why am I spending two, three hours a week looking at this stuff? Because I have to now, because I don't want my clients IP or API keys or whatever else I have in my code to get leaked to the internet and then it's my fault. Right? It's like now I'm now I'm playing, you know, defense or software defense.

Play from 0:06:48 What the hell? I don't even know the term. What is it? It's cybersecurity. That's it. Now I'm doing cybersecurity. W when did I see that coming? I yeah, I know. That's what it that's what it feels like sometimes. Just wanna just wanna create some images with Midjourney. And now I'm now I'm worrying now I got a scan for critical vulnerabilities happening every day. Rory, when's the when's the last time you prompted something in Midjourney? I am curious. probably like four days ago. I I just needed like yeah? a I just need like an outlet for something.

0:07:03Midjourney and making things for fun

Play from 0:07:14 I was doing I was making little characters for my drock bots. I was like, I want to do some avatars while I'm trying to like, you know, regain function of my brain, which I haven't had for a couple of weeks, and I was like, Let me find some like fun characters to do. And then I found, you know, let me play some styles, let me, you know, use some style y before you know it, it's it's been an hour and I've I've done nothing productive for society. And then I'm like, well, at have characters for my grock bots.

Play from 0:07:40 Okay. Yeah. There it is. Amazing. At least that. At least that. But it's it's like Midjourney comes for me in waves now where it's like I get really hype about something and I'm like it's like crack, like a dopamine addiction, where it's two hours of just like my face doesn't leave the screen. And then I leave it for three days and I don't come back. You know, it's it's it's very much so like a feels like a heroine sort of shot.

Play from 0:08:06 Just like let me get my fix and then I like all right now I'm gonna go until I until I relapse. Okay. Yeah, man. Yeah. Yeah. Yeah. Yeah. What about you? Are you do are you sort of like that now? Are you still on it every day? That's Man, I I was just thinking about the other day. I was like, I haven't been I haven't generated an image in Midjourney in a month. Yeah, at least a month. Yeah. It's the longest, Ooh. longest I've ever gone. I think it's just it's a combination of what I've been working on.

Play from 0:08:34 you know, a lot of the projects that I have or had just finished weren't necessarily creative projects. And then also just like creatively, a lot of the stuff that I've been doing, just for personal stuff, right? Like thinking about, I don't know, the game, the you know, that squirrel game we talked about, or, my website or any of like the personal stuff I've been working on. It's just like I've been using the MCPs, right? That have that have been available. Yeah. And it's like, I've got, I don't know, fucking eight hundred thousand credits on runway.

Play from 0:09:05 And it's like, I need to use them. Yeah. Like What you know, like it I have a I have a use for him. And then it's like I'm not even going into runway, right? So I mean it's like I'm not getting overwhelmed with the UI or anything, like trying to figure out where every everything is. It's just like here's what I want, go do it. Staying in kind of like that one place to get it done. So it's just kind of been weird weird time, but yeah, man, I miss I miss it. I miss it. So I I I wrote I wrote to myself the other day.

0:09:29Client work crowding out creative play

Play from 0:09:34 How about this? I I wrote Mm. to myself because I was getting a little bit frustrated and a little bit creatively tapped. And I I know probably if anyone's been listening closely enough to this podcast, you've heard me say, you know, a little breadcrumbs about this, probably over the months where I'm like not having fun anymore. Like this feels more like work now than it did like fun, because it is work. I mean, this is what our job is, right? But like the the fun of creating like a you know I don't have the f I don't have the actual like free time now to just go play as much as I used to.

Play from 0:10:05 So I'm not finding things, I'm not challenging myself. I'm just like get I'm like ticking boxes, you know, going back to our conversation with Jack, box ticker, one of those people, Yeah, but you know what? Like also just just to like get things done. the I think then the I I would say you're doing more than box check. I think I think we're just working on things that are removed from that, right? Like largely. You know, and we're Yeah. working on somebody else's problems. And most of the time that doesn't involve our own creative.

Play from 0:10:31 thing. And I think for people like you and me Yes. too, it hurts us on an underlying level. we we need to have some sort of like creative expression that goes outward to some degree, Like we have to. otherwise we feel it. And we feel like we feel like that distance removed from it. Dude, it was so bad. Like I was sitting there, I think it was before I did the the big chunk Romy Rome video. Probably like last Monday or something like that. I sat there for I had a Let's pull that up. Let me see if I g is that on X?

Play from 0:10:59 that's on X. Yeah, I had I had maybe an hour and I was like, Let me let me see like what I can create or dream up. And I I swear to God for an hour it was just blank. I had nothing. First time in a long time. Like I don't have anything I wanna make. I don't have any direction I'm trying to go. And like, you know. I'm like, we we took this one picture of him over the weekend during Halloween and they had him with this jack-o'-lantern hat on and this milk this a sort of like milk jumpsuit, which he's wearing here. And I, you know, I I sent a sent an image to my family and I'm like, you know, I they were like all cracking up.

Play from 0:11:35 So then I put it That's the best right there. on a, you parental advisory sticker. I put the parental advisory sticker on it, made it black and white, made him big chonk. And then I'm like, you what? I'm just gonna make a music video with him. I know I can do it. I don't want to do it in this sort of ridiculous style that we've highlighted on here before. But like Yeah. make it grainy, you know, have him, you know, have him rapping in like this sort of like green screen look where it's like he's in his he's in his walker. But you know, that that's Amazing. that's what brought the creativity back.

Play from 0:12:00 I'm like, let's, you know, what what can we make him do? yeah. Yeah. He's in a, you know, let's go, let's go make him go to a pawn shop and try to get some, you know, some grills. But I'm grills isn't that's too That's too adult. What about a binky? Well, all right, cool. Diamond encrusted binky. I'm like, that's not ridiculous enough. Let's go with grill Yeah. let's go with a binky that looks like grills. Like, that's where I'm in. Like, it's the most ridiculous thing I can think of with my child. putting him in his his car seat and sending him down a zip line into a party. You know, and like that's that's the so yeah.

Play from 0:12:31 The zipline, that kills me. Yeah, it's great, dude. I had I had this one, the the pit bull chariot. I was like, you know. I was just coming up with these I was just writing down ideas of what's the most ridiculous thing I could think of. That that shot kills me. So how long before how long before you show him this? What age do you show him this where he will appreciate it at in its truest form? This is we've we've had this discussion and I'm like, he better he b if he doesn't find this funny at age like fifteen, then I'm like, I've I've I've failed as a as a father in you know, giving sense of humor to my child.

Play from 0:13:08 Cause this is funny. You know, like how he can't even be mad at being on a pit bull cherry. Like that's you know, I could do a lot worse than than give him diamond encrusted grills binky and This is great. you know, have him have him partying with you know, with a with a paper bag wrap bottle. But you know, the the unintended sort of biggest fan of this is [name].

Play from 0:13:33 She she loves it. It's just funny yeah? There we go. hearing hearing your daughter say, Dad, can we watch big chunk Rome Rome? I'm like on the TV. Like, all right. Dude, she's loving it. I don't know how this I don't know how this this'll land psychologically, watching, you know, I don't think AI has been a part of this, That that's the that that's the thing, but dude. I I'm like thinking about that a lot. I'm like, dude, like we didn't even, you know, like my dad had like the the you know the camcorder stuff Yeah.

Play from 0:14:02 going on. And like there was a degree, I think, of where we were able to like maybe watch ourselves like later on. But even Yeah. yeah, it's just different, right? Now you've got all these photos and videos on your phone. We've got AI. It's like You know, I'm showing pictures and videos of my girls all the time of themselves, you know, and I'm just think, Yeah. wow, that's a really interesting thing. No other generation has had that sort of like ability to kind of like look into things that they've already done and come almost like re experience them like that in that way.

Play from 0:14:33 and it's yeah, it's just I'm like I'm constantly thinking about that, man. It's it's a very, very interesting time and generation for these kids. Yeah, and that's that's why I've been that's why I you know, we went over this maybe like ten episodes ago, twelve episodes ago was that that rare vision app that I have on my phone that records like VHS. Yeah, yeah. So I I did like phase whole summer and Halloween and everything in the VHS. So like I can edit it all together and it can like here's phase twenty twenty five, six, whatever year it is.

Play from 0:15:01 And you can look at that as like feels something almost like normal, not not like 30 second screen grabs of her doing something. But like here's like a like we used to watch it. We'd put it on the TV. He used to plug in the VCR or the VHS cam to the TV. Hope it hope it worked. Yeah, yeah, yeah. Yep. Yep. you know, or you go to the store and have it get made on the cassette tape, and that was and dude, it was I forgot to tell you this, but I don't know if I showed you or not.

Play from 0:15:26 But my wife's father has a they have a closet in the When everyone says, you know, are you in Montana or something? And when I've when I film at my in-law's house for the podcast, they have like a closet. Dude, Wood panels, baby. it's like a it's like a nostalgia bomb. You open it up and it's like they have every medium in there. They got everything from records, A-tracks, cassettes, VHS tapes. They have older, you know, they have their older sound systems.

Play from 0:15:52 They have you know, some things that it's just like it's like a time capsule of media in there. It's just cool to look at. And I'm I was Yeah. going through it the other day. I was just like, why have I never just spent time in here? This like it's awesome. They're like, What are you doing? Like, this is sick. You know, like the Sick. only people who like would cu would curate this, like they get paid money to curate a look like this. And, you know, it's got the the warm gold light in the in the closet. It's not like the the clinical sort of hospital lighting. So it just it just looks cool with the wood around it and everything. Just something that feels like it was made intentionally, but it's not. It's just the way that they organize.

Play from 0:16:30 Don't, don't let your significant other convince you to get rid of stuff that you want to hold on to. All right. There Yes. is a fine line between hoarding and holding on to things. Ha ha ha. And I, you know, like those are the types of moments, though, that it's like, dude, I'd rather hold on to it. You know, just, I mean, like, stuff like that's really, really cool. It's just, you know, go back in time. But I am curious. tie this up in a knot though, I did I do I did want to just go back to the real quick to the because I think maybe some people are feeling the little bit of creative depression is what I was calling it.

0:16:54Keeping up with AI's constant learning curve

Play from 0:17:04 yeah, yeah. maybe like a little bit of a little bit of creative sort of setback. I kind of I kind of like summed it up to myself when I wrote to myself in like three parts, and I was like, it's the pace, number one, where it's like the shiny object syndrome, it's something new all the time, but also like the the like the not the the the near mastery. of things where you're you really can't master anything before something new comes out and it changes the way you work. So it's like it's like learning a new version of Photoshop every week. Right. It's like okay now there's something different.

Play from 0:17:34 Now there's something different. We used to Yeah. have to do prompts. Now we have to build context. Now some of these need reference images. And it's like you have to like constantly learn and it's all trial and error and there's no like real playbook for it. That was one thing. Number two was like the consolidation Where it's like now everything seems to happen in text form, right? Like where I'm everything's in like the LLM. The MCP's come out. I like working that way and having the consolidated space. But it's like I don't need to go in different interfaces.

Play from 0:18:04 I don't need to learn the tools. But like again, I kind of liked working in those. But at the same time, it's it's just more efficient to work out of the LLM. But everything is I'm typing, I'm talking, I'm typing, I'm talking. It's giving me this long answer. I'm reading. It's like Yeah. there's like I'm not touching things. Like I like to touch Yeah. things. Yep. You know, so that was that was something I think I was frustrated about. And then like this this melding of like engineering and creative, that was like my third point where it's like, I'm building creative stuff, but I'm doing like everything is like code.

0:18:21GitHub and code enter the creative workflow

Play from 0:18:35 So I'm figuring like figure out what GitHub is and like the what's a work tree, like what happens when Yeah. I push to push to main and and merge a branch. I'm like w why like again, like I'm I'm not supposed to in if five years ago I'm not supposed to know what any of this means. Right. I'm just supposed to do my work in Figma and design or Fib or Photoshop and design. And now it's like I'm trying to figure out all these different angles, which is cool. I know a lot more now.

Play from 0:19:01 But I'm I just feel like I spend what used to be spent 100% on creating things and strategizing is now like part learning the two the new tools every week, Yeah, yep. Yep. part, you know, strategizing, part code, part, you know, how do I just talk, how do I verbalize all of this instead of just touching it? And it it's made me it made me a little bit crazy and it laid me like I feel like my bandwidth is spread across too like so Yeah.

Play from 0:19:27 many things instead of actually just creating. I think that's gotta be felt by anybody that's really in here every single day. because that's that's so true, right? Like you're you're not only focused on the things that you're trying to achieve, but because other things and new things are possible and accessible, it creates you run into these things that you didn't have to run into before, Yeah. and which you now need to at least have a surface level understanding of what it is or what it does or how it's affecting what you have going on.

Play from 0:19:59 Which requires, right? Like, yeah, space to to learn those things. And then you're kind of unlearning things at the same time because a lot of things have changed and you don't need to hold on to things that used to be the case anymore. And so it's like you're you're you're like really there's a very transactional process of like new information going in. You're you're probably unlearning a lot of things that used to be the case. And I I just think it's fascinating too, where it's like You know, we spend a lot of time with clients and so forth.

Play from 0:20:28 And it's one thing to go deep on something for a client. And then Yeah. then you have a day or two or or something that feels a little bit lighter. And guess what? Like you kind of need to, you know, clean up your own workspace and your own ecosystem. And is there other things that you could be doing? Or is there something that now is sort of broken or stale or you know, could be reconfigured, you know, it's like you could spend an entire week just doing that. You know what I mean?

Play from 0:20:57 Like, and then Yeah. even when you do that and you think you've got everything tightened up, like something changes and, you know. So I I'm I'm always thinking about like, well, shit is like, you know, I I I have a lot of things set up. Is everything set up in an optimized format? Like is everything really firing the way it should be? Right. Then you can go down a whole rabbit hole just on that. You know, it's like There's so much that it just feels unattainable to grasp it all at the same time. and and it can drive you crazy a little bit. That said, I think the advantage of playing both sides of that is you really get a good sense of a lot of the use cases and applications and thought processes and so forth that other people run into. So when you're doing stuff with client work,

Play from 0:21:42 chances are whatever they're asking, you've probably run into it before. And maybe it's not that exact situation, but there's some sort Mm-hmm. of framework or template or common thread that exists that you're able to tap into and understand of like, here's how I'd I'd approach that. Because even though I haven't maybe experienced that exact thing one for one or a hundred percent, I've experienced something that's like that probably like 80%.

Play from 0:22:07 Like it's got basically the same bones to it. So it should right be a similar way that I might approach this. And I think like when you unlock that capability. And you see things through that lens, like it's just it's really, really helpful when you're having conversations outside of yourself because you kind of have, you know, this sort of like expertise or you know, thing to to call upon and and recall in your own way as well. So it's like between that and then having the ability to tap into AI and troubleshoot through AI, what's not possible? You know what I mean?

Play from 0:22:38 And and I think like because there's so much white space or green space to go in. And people like us, we want to do so many different things, that's where we get in trouble. Right? It's like there's just not enough time. Not enough time in the day, not enough time in the week. Right. So And I I would also like to say that we're still we're still smarter than the machines. I had a I had a client project this week and there was something that I was I was sort of running trials and simulations on on this workflow build and I was just having sort of Fable orchestrate it. And I was like, you know, you go you go orchestrate and delegate to to Opus and Opus does this and whatever. And it came back and there was something wrong like every time and it was like min it was minor.

0:22:53Human judgment fixes a failing AI workflow

Play from 0:23:18 But I'm like, you know what? I think one day it was maybe like Wednesday or something like I'm going in here. for an hour and I'm gonna just give f like figure out the problem and go through like every single little piece. And I just rebuilt it from scratch and got a better result in one hour. And I was just like it's just like the the the the knowledge the like the failure knowledge that you can't offload to the machines. Like, I know that's not gonna work because I've tried that on other projects a million times. It didn't work. you know for ex Yeah. Yep.

Play from 0:23:46 for example and I'll pull this I'll pull this up so we can see some stuff here. It was like very speci would it's a for For context, it's a like a PDP generator. So something that's just product shots, but the product shots seems easy. We're just gonna put the product in a studio and take these different angles and shots that are necessary. But like if the if the output specifications are very tight and you need things positioned appropriately and the shadow needs to be consistent and the lighting needs to be exactly consistent, every single time on every single image across thousands of products, then it's like the you need to build the engine to be that way. And Everything was coming out with like subtle variances.

0:23:52Building a consistent AI product-shot system

Play from 0:24:23 It was like off by five or ten percent, which doesn't work, because then you just end up doing more in post. And that's that's like what they're trying to avoid. So I just had to build it. I had to like find I went through like seven different iterations of input. For example, like what am I putting in to get out? I know everything that from a workflow standpoint is basically it comes down to the input and the context and then what I'm writing in the system prompt or what context I'm providing in those, you know, sort of directions to get the output that I want.

Play from 0:24:50 And I went through You know, do I make this like a style card? Do I make it just a background? Do I just make it an image with a PNG and no background? Do I do it with a white background as the input image? Do I do a 3D version of it? Do I do a you know a light sphere, whatever? So instead of that, I just built this tool to simulate like the exact process of a photo shoot that we'd be looking at. Now this looks simple, right? But it was basically like, let me drop in the project the product, have 3JS render it, right?

Play from 0:25:19 Something like this, the shelf line needs to be exactly the same every time. So I put a guide here to make sure that it's at the same place every time I toggle it on and off. But like again, I can play with within here, I can play with the camera distance. So I can go further back. I can do the field of view, make it wider or you know, tighter. The I can change sort of the I can change where the marble is. You can hear Rome in the background voicing his displeasure. you know, I can change the mat.

Play from 0:25:48 the matte version of it, I can make it lighter or darker. But also like, you know, I can change the ambient occult to make this stuff sort of like darker on the shadowing or not. I can change the product scale. Is it bigger? You know, so I can basically create the perfect image. I can change the light, which is great. Like this will give me like the shadow reference that I need, or I can change like, you know, the height of the light or the direction of the light or how powerful it is. Right.

Play from 0:26:14 So I can basically create the the exact image that I want. now with three D and then like, you know, giving it the isometric view so I can see everything where it is. Again, if we're doing like that's just one shot that's necessary. If we're doing a different shot that requires like this heavy flash and it's on a reflective surface, right? I can play with I can play with this stuff. I can pull it further away from the wall. I can put it closer to the wall. You know, right? Like I can play again with the light bounce and where the key light is and the reflectors.

Play from 0:26:43 I can the acetate ripples on on the ground. We can make it Softer or more s you know, have more strength. The ripple spacing can be a little, you know, we can basically play with everything that needs to be here so I can download this and then use this as a reference image. So once once this was built, everything got better. And then it was, you know, I have I have the matte version, the silhouette version, the surface normal version, so I can really like map the texture and the surface curves and everything to give it more dimension.

Play from 0:27:15 but like This is a problem that took for I'm like, I let me try this a million different ways. One of the other problems that that probably would find interesting. This is like a this is like a totally technical problem that I was I was some some people might be might be you know dealing with. So like the reference images I got were of the product, but the product was already in studio. So there's like a direction of the lighting. Here's like it's like a stupid little problem that probably cost me three days, right?

0:27:45Fixing silver-foil text in AI product photos

Play from 0:27:46 One of the products has the the the text is silver foil, right? So think about that with the reflection of the flash wherever it's coming from. Yeah, it's already a big problem. So half the text, half the text looks silver, Yeah. Yep. Yep. half the text looks white. So the the model doesn't know which one it is, and it'll just default to one of the two. Sure. boy. I'm like like racking my brain on how to do this. So again, anyone who's problem solving here, right? Like we're thinking about like going through this problem.

Play from 0:28:11 What do I isolate? So I was trying to, I was like, let me. Extract the text, make it like a flat neutral color, put that back in, didn't work. Right. Sizing was never right. Like, let me try burning it onto the product in the in the 3D version and just make it like this just a slightly darker shade of matte. Didn't work. Like so it would actually just be on the bottle, like the actual product label. And then the other thing that I finally came to after like 50 iterations of how to get this was let me take the initial image and then remove the light source from it.

Play from 0:28:41 So it took the initial image and flattened it out, made it sort of like this matte version of it instead of instead of the my god, instead of like the the light version. So it stripped all the light and just made it like this matte neutral version and then put that in and out of the light and it worked. And I was like the tenth iteration of how to get this input image correct, because I'm like, it can be done. I've done this before. Yeah, yeah, yeah, yeah. Yeah. It's just this one little stupid silver foil, Text piece that's ruining it all.

Play from 0:29:12 I know. But like that's the level of iteration like I like that problem And that's the left, dude. solving. Those and and I think that's the thing, right? Like maybe the commonality with that, and then going back to your earlier point is just the fact that like we because people that are probably listening to this and us, we we're kind of we're so early in that we're we're experiencing a lot of the the problems that exist.

Play from 0:29:38 we're almost like we're we're basically the QA. Like we we are the ones that Are finding all of these things and working around them. And I think, dude, I mean, I I I would just imagine that I don't know, whether training data is on or not, or what they can access inside of our individual accounts. But it's like there there's gotta be a feeder for things that are prioritized based on how often they're run into, how they were solved, right?

Play from 0:30:10 Like I would just think that's a proactive. sort of net that's set up. I would I would imagine like if I ran the company, I'd be I'd be very interested in that because it it it allows you to get customer feedback and then also solve those things like in a much faster way because you have somebody that's using it and tapping into maybe different ways of running into that same wall. Well how are you gonna get a walk around the wall? You gonna climb over it, you're gonna knock it down, you gonna find a way around it, like dig under it. Like what is the answer?

Play from 0:30:37 And I I just think it it's just funny because like that that reminds me of Kind of what we were doing on the Midjourney stuff, right? Like I always bring up that example of you doing the multi character consistency thing and like how many loops had to be run around just to get there, you know? And it's just like this, it's kind of the same thing on the on the LLM side. I just feel like we we run into some of those things. Like, you know, like I don't know if this is like the best example, but it did make me think of this, which is like, you know, we we talk about taste a lot.

0:30:42Teaching AI taste without boxing it in

Play from 0:31:06 Can taste be codified? Can you, you know, is it actually a moat? You know, blah, blah, blah, blah, blah. Well, mean, like, you know, like we do I do, I've been doing like a lot of like artifact stuff, website work, you know, the squirrel game, you know, like all these different things. And a lot of just voice prompting, references here and there, right? but like when you get back the visuals, the mock ups or whatever they may be, right? Like iterating on those.

Play from 0:31:35 And those do take time and you have to kind of explain, right? Like, no, this isn't gonna work. And here's why it's not gonna work. And you know, it starts to pick up on those patterns. So I was just thinking, like from a taste standpoint, do we really want to feed great examples of our taste? And I think to some degree the answer is yes, but also does that box us in, you know, in terms of the things that we potentially like? Because you know, AI likes to hold on to examples and sometimes too Mm-hmm.

Play from 0:32:04 closely, right? Like examples are great, but also it can they can kind of work against you in the sense of maybe it stays really close to those things. Right. And kind of using them as loose examples, you know, like you just kind of gotta really be very almost semanti thinking semantically about it. and that that part's kind of weird, but like I think about like the the the net new side of taste and expanding on that, does that box you in versus going in the opposite direction?

Play from 0:32:35 And having it know exactly the things that you reject and you don't like. So I've been kind of like thinking on that opposite side, right? So like I got, I got to thinking. I was like, you know, especially with doing some of these deliverables for clients, I was creating like these, these little artifacts, and I can show some of that. But it's like, no, I want to design this way, but I was running into things that I generally don't like in design, right? And so I just thought, like, why not set up like basically like you know, taste you know, rules and guidelines around the things that I don't like. And so I went back and had it record all of the things I've ever done where I where I rejected it or, you know, made a tweak on something or, you know, asked for like a visual fix on the work, right? So it could record like the exact things I said when

Play from 0:33:21 I rejected it as well. And now that bakes that in. So I shouldn't hypothetically run into those same problems again when I'm designing new stuff. You know, so I'm just thinking like I don't there's there are just so many of these little intersections, right? Like when you're in Yeah. this, where it's just so fascinating to think through. It's like a so you know, one hand, it's like, yeah, I do want it to know my taste. And but does the value of my creative expression go down if it knows my taste so well and it boxes me in?

Play from 0:33:50 Doesn't really quite feel the same. It doesn't hit the same. Versus, It doesn't hit the second. versus like in and even if it doesn't box in, it it sort of shortcuts the creative process too much. Even if like even let's just say it's damn good, that doesn't really feel like if if I were in a client project and time was of the essence, sure, that would feel great. If it's for my own personal stuff, I it feels too manufactured, feels too automated, too hands-off, right? And then versus like this other side of it where it's like you're almost setting up the things that you just don't like and you're kind of leaving over open the the space.

Play from 0:34:28 sort of like for creative expression and net new ideas that come into the fold for your taste as well. I don't know. Just like an interesting thing that I ran into last week that I've just been thinking a lot about. It's like, dude, there's so many of these little dilemmas, right? And like fine, you know, fine hairs that you could sort of split and and think about differently with how might the LLM might interpret them and how you might actually build out leverage for yourself. And sometimes it holds on to too much memory, right? Like we always complain about that.

Play from 0:34:54 Also true, right? Also true. But it's like, you know, I'll I'll say something like I want to create stuff. It's like, well, based on the stuff we've done with Midjourney I'm like, no, no, this is different. It's different than that. Yes, right, right. Don't don't don't bring that up. Like it's it's totally different. We're not using that that mindset for this one. But I see what you're saying though. I kinda like that I kinda like that thought experiment of like the the taste not being too codified to the point where it doesn't leave any room for new stuff.

Play from 0:35:21 Right. Which I th which I think is you know, which I think is important because i the the one thing I find that's always interesting is I I I end up like whenever I see these these documentaries that come on and it's like, you know, like the eighties or the nineties and you see like the radically different style, but it all seems tied together from like where society is, what's going on in the arts, and then like what that essentially outputs to visually, right? Or like what technology is available.

Play from 0:35:51 at the time you look at like the 'cause like you go to the nineties and it's like, yes, the technology that was available at the time, you know, you had the VHS, you had very early computers, you had C D ROMs, you had cassette tapes. And then you had like this sort of like in America at least, you had like this very it's like felt like an like an angry movement during a period of of like very minimal conflict for us. Like there was nothing really wrong, but we were angry for some reason. Like our music is angry. Right?

Play from 0:36:17 Like our like Wait, what's the first thing you think of when you say that? Because I have something and I'm curious. Like like limp biscuit and corn and like then like you know yes. Yeah. Yeah. Rage again rage against the machine. and then the hip hop of that era was was way more Yeah, Yeah. Yeah. Yeah. like we're we're angry, but like what are we really angry about? Like we weren't like we were doing really well. Like like US was we're not in, you know, any sort of conflicts or yeah. So I think it's just all relative, right? Like people now it feels like that we wouldn't be, but then it was like this big deal 'cause nothing else was bigger than that, you know, kind of kind of feel, Yeah. And then you look at like you know? But boy bands, you know, in like the two thousand.

Play from 0:36:54 Like we're in conflict constantly in the two thousands, like boy bands and you know, emo music and things like that, and you're like it's like this doesn't seem as it's doesn't Dude, yeah, dude, that's so true. as angry. Yeah. MTV, MTV at the time, what was it? my God. What's the name of the show? The T R L TRL. Total Request Live with Carson Daly. Yeah. So like every dude, there was a period of time there, like a pretty long period of time, where when you got into like they do the top ten, it's like in that top five, it was always between boy bands and limp biscuit corn.

Play from 0:37:33 And like, you know, it was, Slipknot. Yeah. dude. It was. You know, it really was. It was like it was a weird time, man. It was like, yeah, it's just like these are the two genres that are dominating at the time. But it was it was weird because it like again, you go back to like taste, right? Remember how we've had that conversation where like taste, yes, it's all it's like being able to visually represent something that is attractive to the masses, but it also has to be the right time, you know, wrong taste, right time.

Play from 0:38:01 I think that was the quote that we used at that point. There you go. or right taste, wrong time, whatever I said. That it it goes back and like again, like it was a little bit grungier back then, like stuff sounded better. you know, there was on on like more of an analog style. You're probably looking at you're listening to you know you're listening to to music still on probably things that were produced in the eighties, wooden sound system, things like that, had a little bit more like grit and like texture and tone to it. So CDs and cassettes work.

Play from 0:38:30 Then we get phones, right? And then we start getting TVs and the TVs are four K. And then everyone wants things in four K. So everything's crystal clear. And I think you see like Yeah. this is what this is my theory on everyone could be Rory, you're totally nuts. But like I think we've all had so much like perfect airbrushed, crystal clear, 4K, 8K liquid cli crystal display that like you see everything's moving back to like this VHS look because it's like one a little bit of like I don't want to like see every pore on my face. Like I like the scan lines, I like the saturation. I know my skin isn't red.

Play from 0:39:04 It looks cool though. You know, it's like this it's like this ability to sort of like like every memory isn't crystal clear. You know, like when you remember something, like right, it's not crystal clear. Yeah, that's right. It's like kinda we wanted to remember it being like fun and warm and not I don't know. That's just how I feel about what why you see this, you know, especially a lot of the Gen Z stuff. They're into the nineties, the two thousands. They like the way that stuff looks. They like the fashion. You know, there's like there's something about it that's like kind of this counterculture of like high tech and like super everything is so like polished and sort of curated, right? And it's like more sort of like that's why TikTok works.

Play from 0:39:39 It's more like Authentic and real. You get just people talking to a phone. It's not like perfectly branded visuals. things like that. So, you know, that's why UGC is the Hundred percent. thing, but it's it's interesting. It's a it's a good concept. I like I like that your head's there, by the way. It's I d I also don't want to codify it. I like don't want to, if that makes sense. Like 'cause my taste changes. That's what I'm saying. I I think I think there's like there's a there's a cutoff there. Also, we should talk about just Claude in general.

0:40:00Fable 5.2 vs. Opus on real work

Play from 0:40:07 A lot of a lot of things that have been announced, also released. So I mean you had Fable 5.2. Would love to hear your take on that. Okay. my take on that is Opus clears, in my opinion. Better. that said, I have not Played with 5.2 Fable, with strictly like creative application from like a 3D motion standpoint, right?

Play from 0:40:39 Like I was with Opus V. So don't have that baseline. But any other tasks that I've been running on it, I dude, it's making mistakes, bro. I mean, like mistakes that I wasn't getting with with Opus. It's like it's I don't know, like kind of underwhelmed with that so far. Again, still gotta explore on the creative side. I've been seeing really cool creative examples, and I've like got some bookmarked, I'm sure, here. But like, I don't know. That's that's sort of been my particular like experience with it. But I mean, a lot of people have had great things to say.

Play from 0:41:15 The other thing I was gonna mention though is Claude dropped a couple different things that haven't hit us yet. But I think they are gonna be really interesting to see how they work. So this Claude meetings thing, if you guys haven't heard, so basically you can add Claude to your team's video call now, and it's gonna be able to build docs like whiteboards, other artifacts while you're still talking. Yo, I'm interested in that, but that's not available to us yet.

Play from 0:41:44 So it's starting to roll out in beta on team and enterprise plans. But but that's interesting, right? Yeah, you're on you're on mute, but I know you want to say something. Go. Screwed that up. I just said I I thought I d I thought I dropped that perfectly. I was like, do they listen to podcasts? Can they can they be in here live animating what we're talking about Dude, that would be that would be wild. and burning billions of tokens? That'd be wild. But like, you know, just thinking about like stuff like that. So interesting because like, okay, let's remove ourselves.

0:42:00IT approval holds up connected AI workflows

Play from 0:42:14 Let's go into, let's go to in-house, let's go to client experience, right? IT is is a blocker. It's a blocker, guys. one of the clients that I was just on site with, again, Microsoft Shop, they have to use Copilot. So copilot is existing for their email and their calendar, but they do have a Claude Teams account, but they can't connect Claude to Microsoft. They don't have approval to do it. And I I think this is a common thing that a lot of companies they're running into. And I would have I would imagine this is very common if you were to talk to other people going in-house these teams. And that's just one example, right? Let alone maybe other connectors and so forth that come into play.

Play from 0:42:54 And that's where that's honestly where you get a lot of these teams that just when you remove yourself from these conversations like we're having here, they're still operating a very basic level, right? They don't really know how to create skills. They don't know the value of the connectors truly, right? Like they know you could connect it, you could probably get information, but they don't know all the use cases, applications from stuff like that.

Play from 0:43:21 I mean, that's that's where a lot of these companies are. And so, like the cloud meetings thing is interesting, right? Cause it's like, okay, well, well, could we give it access to that? Because like even that sort of setup, right? It's like you could record the calls, but you got to get approval from like Zoom. Can we can set up the Zoom connector? Right. Or can we have granola or Fathom? And like some of those things are restricted. Then you've got like the email comms sitting over here and like co-pilot, right? So it's like none of this stuff, like none of the like the biggest stuff that could be automated and provide the most value, but are kind of like locked in a way.

Play from 0:43:59 So you kind of have to like find the loopholes around those things. Just makes it harder, Yes. right? It's just like, so it's interesting to see Claude come for this because like a lot of people like us, I feel like, that are listening to this, like we've set up ways for we've got granola or fathom or fireflies or whatever, ingesting that information, ingesting calls, emails, coming back to LLM for contacts. We have all this information about our clients or us or whatever, you know, every call that we're on. The comms layer, and like we know how valuable that is.

Play from 0:44:26 It's a second that is truly a second brain in itself. And a lot of Yes. these companies just don't have that, man. I mean, it's like so it's gonna be interesting to see how Claude sort of comes for this. And Roy, I've been kind of going on a little tangent here. So I'm gonna shut up here in a second. Also, they did, you know, Claude announced this whiteboards thing, right? Like this is now part of this this meetings piece, which The whiteboard, and then they also announced the, I think, this like sub agent, like this managed agent thing.

Play from 0:44:56 Self-hosted Claude Managed Agent, private beta. So you can host the agent harness sandboxes session history in your own cloud or on premises and choose where Claude requests to Claude are sent. Huh. Interesting. They do so many things that I'm like, I don't even know what that means. You know, like when you say manage agents, But I I agree. So, so so here's how I read that though. I'm like, explain. It's like, okay, we're we're talking about IT being this barrier.

Play from 0:45:22 Yeah. And like, so, so I get the benefits of Claude, but it lives on my premise. So all those security questions that I have in place, or a lot of them from an IT perspective, could start to dissipate. It's not training on my data. That's what they're all concerned about, right? And even if you sign up for Claude Enterprise. There's no way to fully get out of that. They may not train on it, but they can still ingest it. So it's like, so how's that actually working?

Play from 0:45:49 You know, we don't know. So Yeah. We don't know. It's it's very weird to me. And also go back to your point about Fable five point two. Like I was I think I, you know, we get excited about anything that's labeled Fable because it was such a jump. But like I've I've had a lot of issues. Same thing. And it's it's a it's a token monster. It just got it's like Pac Man, you know, just ripping through tokens one by one and I'm very This one is fast though, I will give it out. It is fast at least. fast.

0:46:14Usage limits and running a solo AI business

Play from 0:46:15 Very fast. But I'm finding it like I look every time I send a message and it's like 1%, 1%, 1%, 1%. And I'm like, dude, look, what are we doing? I'd like that's not Chippin', chipping away. even and then I look at my overall usage and I'm because now I'm at the point where I'm like, yeah, you know, like you got me to pay $200 a month for Claude. Would I pay $500? Yes. Because like it's so ingrained, I don't want to worry about the I just don't want to worry about my usage limit because I don't know if anyone's gone over like your usage limit lately. But if you like if you go over and you have to pay API prices, which scares me for what's to come post IPO, is that like it's it's insane.

Play from 0:46:51 Like I just it's like every run is like fifty dollars. And I'm like, okay, this isn't sustainable long term for the independent. you know, we we use it as a business expense, but it's it's getting crazy. And I think it's like, you know, I'd rather pay a higher fee than have to pay API price. because it does it does do the work of I don't know, dozens of people for us. Right? Like We wouldn't be able to scale like I personally, I would never be able to scale a business to what w you and I are able to do by hiring people because I would not be able to still do the work I would be managing, which then would not be able Yeah, a hundred percent.

Play from 0:47:27 to scale a business, right? So it's like this really interesting, like delicate maneuver of like how much would I pay for a whole staff that's run basically by, you know, synthetic humans or like actually build a business. that with with real humans and you know, really have to like not grow at the same rate or scale product at the same rate or scale my services the same rate because I can't actually do that and think and without managing Yeah. Yeah. Imagine imagine all of the other things. So it's just hiring somebody right now and they're s they're starting tomorrow. You know, Yeah.

Play from 0:48:04 how much time are you gonna have to spend with that person till they till they really have a solid foundation to operate on their own on your behalf? And do all the admin that goes with it. I mean, you're gonna have to show them so much shit. And and even then, like how how AI pilled are they? Are they gonna kind of know how to go about that process? Like, I would just assume, like, I don't know, I could self-serve and and probably rely on AI to pull things for me, but I don't know that those are the right things for you, right? Like there would have to be I would that would be an interesting experience to go through.

Play from 0:48:37 I'm curious if anybody else Yeah. has had to do this recently. That is maybe operating a small business or solo business, and you've hired someone and what that process has maybe unlocked for you, or what you've realized as a result of that and how you went about it, that would be very interesting. I'm not even sure what I would do. I mean, I haven't thought about it, but offhand, I'm not sure what the most efficient way to do that would be. I mean, I think like even the fact that you and I, right, We work on things together, but we don't have, we don't share a claw account.

Play from 0:49:14 No. It's too hard. And so our claws can't really talk to each other, right? We got to export something or, you know, send something, copy and paste something, attach something that you can then take and then feed into your system, right? Like even that's a mess. That's messy right now. That's gotta be fixed. We'd love to see that fixed. Maybe we could do it in Slack. when that when that happens, bro, when that happens. Yes, you're right. Yeah. Maybe, yeah. I I there probably is a way around that.

Play from 0:49:44 but that I mean, that should be something that is so natively built in. And I think that would be a game-changing moment in this space for sure. Like that feels like something crazy. I think so. Here's that whiteboards thing too, right? It's like claw whiteboards, clawed sheets, clawed data. Think a plan through on a board, build a tracker and a sheet. Tell a story with your data. And that's what we've been trying to do.

Play from 0:50:11 Right? Like that feels like everything's sort of coming to like this head now. It's just like everything can be turned into a visual very quickly. Like it's very good. I mean, you see it w if you any of you guys watch our clips, like that's what happens. Right? Like Claude just takes our raw voices and narrative and transcript and then just turns it into visuals. Right? So that's that's what's happening. But I like the I do I do like its own it's Yeah. It's like a consolidated way. It's coming for Google. I mean, like what else can you say about it?

Play from 0:50:38 I mean, like it's it's this is Yeah. It feels like this is like consolidating a bunch of steps, right? Like that Again. you would manually do. Like it's not like these things weren't happening or able to to to do before and a lot of us were already doing these things and but they were disconnected, fragmented parts of the process. Whereas now there it seems like yeah, it's the compression. And that's i it's very it's very odd again. It's like I feel like I'm I'm always the fable itself, again, going back to this, it's so verbose and it just writes so much.

0:50:57The mental load of reading AI all day

Play from 0:51:07 And I'm just like, dude, just list the problems of what we just ran and that's it. Like I don't need the whole explanation. Like I don't need like the like that was the fable, the initial fable thing too. It was like I'm reading like three pages 'cause it's so smart, it's telling me everything that happened. Like I don't need everything to happen. I need like the I just to isolate, You need a you need you need the you need the isolate. ASD STE one hundred, brother. You need that. Yeah. Dude. And well, the other thing that that's been crazy is like I used to be able to work around chaos, right?

Play from 0:51:35 Like I could I could go to a coffee shop and like be around the chaos or like just like I could Yeah. be in my own zone because I'm I I'm not reading like for every word. And now I'm reading for every word all the time. I have to like think about it. Then I have to think about the context, think about how I'm gonna respond. Yeah. Yeah. Yes. If I say the wrong thing, it's gonna go the wrong direction. Right. So it's like I I can't I can't be around as much stimulus as I used to be. I need to be like isolated, which I don't like. And I'm like, I needed to be to like think I can't have, you know, I can't have Luna licking my arm while I'm reading a clot output, which is just like you know, it's just like she's licking me.

Play from 0:52:08 Yeah. I'm like, whatever. I'm like, you know, I'm drawing stuff on my canvas. I'm like, I can, you know, I can do that. Now it's now it's just so much focus. I I feel like I'm a lawyer. You know, like every word matters in a transcript. Every word I say. i is gonna either like make or break the next thing I do. So it's you know, again, this is like this consolidation thing. I don't know if I like it or not. I'm still like I I like it for what it's able to do for me, but like as a whole, I don't know if I love it.

Play from 0:52:38 you know, everything coming into this one place and it's all text based. Like I did I've never liked working just text. Yeah. Yeah, yeah. Me either. Right? Like Me either. I I think that's where we're gonna get though. I think we're gonna get there. Yeah. I and I you also brought up something too, which was what did you how did you say it? I'm gonna I'm gonna say it in this way where I'm gonna have to pause. I'm gonna have to do a purposeful pause after working with your hands, something something along that line, Yeah. Touching things. Yes.

Play from 0:53:08 right? Touching things. There you go. But like I I feel like that's that's a that's an experience we're gonna get at some point. And when that happens, that's gonna be the the best time for I think creatives because right now it's like I I I think words are the death of most creatives. We we don't wanna read shit.

Play from 0:53:33 Yes. I don't want to write either all the time. We want things visually communicated, we want things we can skim, we want things that convey a message. Right. So going back to like last week with Jack Butcher, like convey a message simply, right? It could be a complex thing. It's simple. I get it. I can move on. Doesn't take up a cognitive load in my head. I can keep going. And and to your point, right? Like these long blocks of text. And again, multiple of those going on in a day at the same time.

Play from 0:54:02 And you're doing that the whole day. Like it's a grind. And then to your point, having to worry on the back end, I better sure I'm I'm locked in enough. Where I don't go in down the wrong direction. Otherwise, I'm just wasting time. Right. So Yeah. it's like this really odd thing that I don't really know that has really existed before. at least for me, I I haven't really experienced that. But it's a it's a different thing altogether. It's it's building the language of knowing how to talk to it, number one. And then it's like at the same time having like really selectively choosing the right words and and sort of like.

Play from 0:54:38 Explicitly saying and implicitly saying things, right, that it could account for. And it's it could be tricky. It could be tricky, man. Here it's a perfect encapsulation of that. Right? Like I I this just reminded me. I did this I did this this morning in like five minutes and then forgot I did it because I was getting frustrated with something. This is I'll I'll share this. This is here, let me share my screen.

0:54:46Visualizing GitHub branches in 3D

Play from 0:55:04 I did this because I was getting frustrated, like trying to track my my work trees and like like the GitHub repos. So I built like this little three D visualizer to like track what I did. Cause you know it's like, you did it on this branch, but you didn't merge it. Or you did this here and like, in inversion, whatever. So I'm like, I need to actually see it. Like I can't just read what the repo is. Yeah. And like, you went on this, you went on this direction here.

Play from 0:55:32 And then it like went to it went to something else. And I'm like, okay. Like now I can start to visualize what the work tree looks like, what things have been pushed to the main branch, where I like screwed a things up or what I need to go back to. Right. So it's like This is this is five seconds. This isn't like beautiful Yeah. by any means. But like I I I just need I needed a visualization of a work tree Yeah. and not just a like a what did I do in version five of this?

Play from 0:55:57 And like then it tells me in like eight hours, That's that's where this that's what it's a long paragraph. that is exactly like what's what we need now, right? We need the visualization Yeah. of it, but like not just like one place. It just needs to like be intuitively built in in almost Yeah. all of these formats. Like there was a point in time where you know I was working with like this one skill I had, and I just wanted every time it works, I wanted it to provide a visualization of what was happening as well.

Play from 0:56:28 Right. But it's just like it's it That's that's cool. and it's cool, it's cool, but it's like I w I wish it would just do that automatically, you know, and then I Yeah. would still get the speed, whereas it really would kind of like slow things down. And sometimes it wasn't useful, right? And you know, it's just like one of those things. But I just feel like there there needs to be, and I think there will be, right? The visual component that is baked in to a much larger degree than it is right now.

Play from 0:56:54 Also, I think like we eventually get to that touch feel. Hold, you know, craft sort of experience in the next 10 years, you know, that's just really going to be wild at that point because I think that's really where it's going to feel like creators are are still creating. You know, like could you just imagine like seeing something that's on a screen, but you have it in sort of like a hologram format and you can kind of twist it around, move it around and, you know, make a change, you know, right here by touching it.

Play from 0:57:21 Yeah. And, you know, just it's a, it's an entirely different experience altogether. And so, you know. Yeah, dude, the cognitive load right now with the reading text is is really an issue. Well, it did it did it for me. Th that's the one thing Fable I've I have noticed is that it has been good. Is that like the it it does anticipate needs sometimes and does find solutions, like maybe not on per my execution, but like I didn't ask it to do this for like the Yeah.

Play from 0:57:48 going back to the pro the project I was talking about with the three D visualizer, right? Like I wanted to I wanted to compare and contrast outputs against like other versions and then like find the small little pieces. I'm like, like the the shelf lip. Like I wanted to see the shelf lip and where the product sat on there. If it was like above it by a pixel or two, if it was sitting below, if it made sense like Yeah. physically. So I was I was telling it all these problems. And it just like, okay, I built you a an HTML artifact where you can like like swap through all the test the testing version.

Play from 0:58:16 And I was like, I didn't ask you to do that. But that that makes sense. Now I can see it. Right. Instead of like I think artif I think artifacts are just generally underused right now. Like they're Yeah. they're really awesome. And I love too, like when you get into let me just show this, like the the comment kind of component to where now you can just kind of like click this, right? And it's like I wanna change that. You know? Or I don't like this button. Yes, that is that is great. Or just like highlight it. Because like before you had to like be like, you like.

0:58:19Artifacts and visual AI handoffs for clients

Play from 0:58:42 Okay, in the first section, you know, where it says like one Yeah. week in numbers, you know, there's like this toggle right underneath there, not the left side, but the right side, you know, it's just like, bro, that was just, you know, exhausting, right? but like maybe going back to the visual point for a second, like this was something I did on site for a client. Basically, like going on site there. I've been out there before, basically doing like a an AI assessment and roadmap.

Play from 0:59:08 And then this time I went out there to kind of do individual builds with. I don't know how many people I ended up doing it with, maybe 15 different people. I white labeled sort of the brand name here. and I took the employee names out and their their pictures. So this isn't quite as visual as like the actual deliverable that I gave them, but this is these are deliverables. And I think about this in terms of like, hey, let's, you know, let's think for a second, right? Like going back to that text, like it's like if I'm doing this, I could say a lot of things that we did.

Play from 0:59:40 This is huh. a chaotic company. They're busy. You think they're going to read through like blocks and blocks of text? Hell no. Shit, they may not even really fully dive into this. But it's like, I got to visualize this in a way that's going to communicate these things pretty quickly. And so like I built out sort of like this piece of it, which is more of like everything. And then I built out something that is a little bit more consolidated into one page for like an executive type of view.

Play from 1:00:07 You Mm-hmm. know, and so, but every but everything is sort of the same, right? in terms of me focusing on the visual aspect of it, right? So, okay, here's what we need to do to create sort of like a team brain, right? We've got these things connected, these are the things that you can do with those things, right? You click into any one of these, kind of tells you, you know, kind of gives you some basic things that you can try, here's kind of what it can do for the company, here's where we started, Like vi everything's visual. Right. Like everything's visual, trying to eliminate as much text as I can.

Play from 1:00:35 Some things do require text, right? Like actual step-by-step things, but in large part, and these are just prompts, right? But in large part, everything is being communicated visually. So, like even, This is cool. you know, created like an upskilling sort of center for their team that they could use for people that, you know, really aren't using this or need a refresher, or if they have new employees, right? Now they have something that actually comes up. And then taking like actual.

Play from 1:01:01 Like there's screenshots that are pulled into this too, right? So I could say this, but somebody that is not in AI is not gonna know what the hell that I'm talking about, right? I could have created, Yeah. you know, 30 videos of myself walking through that. I chose not to do that. Instead, I wanted as little text as possible and screenshots. And so that's what I did. I actually didn't even do the screenshots. It did them for me. Did it for you. Yeah. All of them, right? Like think about how much time that would take.

Play from 1:01:28 Right. Okay. So here, my god. you know, here's what fits the job. Here's why. Right. Think of a high, you know, haiku is the bike. This is the car, opus, truck, right? Use this for this. I wanted them to operate on this as default because I want them to, you know, I don't want them to underpower what they're doing. You know, so like just thinking through all these little things, right? One-on-one sessions. Again, visual, these would be headshots of people. These are actual names of people.

Play from 1:01:53 Here's what we did inside their session. Right. So like going in here. Here are the things that we did. Here's the things we handed off. Here are the things that are happening. Here's how it runs. Right. So here this goes in here. This goes here. Here, here. These are the things that aren't live yet, but we talked about. And here's what you need to do to finish them. Right. Like all this is in a visual, you know, sort of like perspective. and so I don't know. Again, going back to visual. And I think, you know, it's a really cool story that you can tell in that in that type of way. Right. So here's like, You know, how many hours they got back live.

Play from 1:02:28 Here's what we could do if we filled it finished every build that we talked about or started, right? And how those numbers change. here's, where the time comes from, right? This is broken down by person. Here's every time it runs. So things that aren't necessarily by week, but every time they run them, here's how much time those things save. You know, just just lots of lots of information, but it's like I can't just throw blocks of text around. Yeah. You know, and so it's just like visualizing each of these sections. And right now it doesn't think that way.

Play from 1:02:59 So my my process was I went through this and it's like, okay, when I have things like this in the future, I'm gonna treat this as a template, right? Like, Mm-hmm. is it gonna look exactly like this? No, because like this is kind of in their brand style, but that can be swapped out, right? And there are gonna be certain things that we talk about here that aren't gonna be relevant for another client, but things that are gonna be the same are are thinking about like ROI elements, thinking about like individual builds, thinking about like, things that they need to do next, things of the calls that leadership team has to make, right? Like all those things could be codified.

Play from 1:03:35 So it's just like anything I'm trying to build that I could use again, going back to our rule of thumb that we always talk about, I i I gotta, I gotta turn it into a shortcut. You know, it's gotta give me a head start on the next one. It's gotta be front loaded work that has compounding leverage over time, you know. But It this is cool though. I I mean if I was if I was receiving this, I'd be pretty I'd be pretty amped about it. You know, if I was like, this is just like broken down all this crazy amount of information to something I can like look at and comprehend in about two seconds.

Play from 1:04:07 You know, like that's that's always sort of been my style. I mean, I I feel like Jack took that to the next level, which is like, how do I get the idea in, you know, point five seconds without having to explain it? And he does it very well. But this is also the same way. Yes. Yes. It's like that's what this all is. And and like honestly, it's like I this is this is probably like the first time I've done something this fully across an entire org, but like there's room here, right? To to go further and push it even further.

Play from 1:04:36 Like to your point, like I probably still provided too much information. You know, and so Yeah. so there's kind of like there's that balancing act there too. But like even some of those visuals, like they could be even more minimalistically, you know, communicated, probably. Yeah. and sort of like discerning what's there. But that's like one of those decisions too, where I think we got to make where it's like you you want to kind of give them everything of value, let them be a little bit overwhelmed because it kind of exacerbates the problem that they have, you know, it's like there is a lot here, and that's the truth. But Yeah. And it's almost like you have to like like like seed it out and tear it out like the like w release by release, almost like how we learned how to use this stuff.

Play from 1:05:19 Right. Like I still go back to that point. Always we were lucky with how early we got in, not because we got ahead of it, but because it was so simple back then. There was one tool, one thing you could do, and that was it. It was like Midjourney, text prompting, enter. Now it's like trying to go back and learn every single little feature and function. Again, it's the same it's the same analogy of like the iPhone. The iPhone used to be just like you opened it up and there was like one app for music and one app for the internet and one app for text. And now there's like a billion apps and it all functions a different way and you can change everything else. But if you like gave you we understand how to use every new iteration because we've had every single iteration of the past. And then I would say like just this whole visualizing situation.

Play from 1:06:04 Like it just needs to become more real, right? Like we're We're all like super visual, right? Like I now I need to now I need everything to be built in visual because I I just like can't read through the text. Like I had like, for example, like here's a here's a fun Grokbot experiment. I've been trying to build this second brain for you know, using the the Karpathy method, which everyone seems to love, but it actually works really well. So I'm not gonna knock it. That was just one of those things I thought got really overblown for a minute, but I guess it wasn't.

1:06:08Building an ad-creative second brain with Grok Bot

Play from 1:06:33 if you're not familiar with it, it's just a way to like structure data for LLMs to read effectively. So I've Yeah. been having Grockbot do this daily scan. And now I've now I've had to do this a couple different times for people, you know, like specific type of creators on X that are that are talking about static ad creative and like breaking ads down and like putting on, you know, creative strategy and things like that. So have it scan every day. And then when it scans and finds stuff that I like, I have it go, okay, now break this down into, you know, the the the method and add it to notion in the second brain. And then the second brain becomes like this whole little thing that I tap into. It's actually re it's actually works really well. I've only been doing it for like a week or two.

Play from 1:07:14 But it it it comes up with much better ideas working off that than working off its general memory, if that makes sense. But then I'm like, I w I wanna see this. Like I want to see what this looks like. And again, it's like everyone's got their own little second visual brain. I know things like Obsidian sort of exist. But there was a I I No one never really caught on for me. I went through the trouble of doing all the setup I didn't get it either. for that and it's just like this isn't a tool I've ever used, so it it it faded quick.

Play from 1:07:41 So there where is this let me see if I can pull up this artifact here. So yeah, it's like a it's a way for me to just look at the information in the second brain and sort of filter through it the the same way, right? Where I'm like, okay, this is these are like all the connections between everything. And like I could go and scroll through it, and then if I'm like, I want to find some of the experts that I've been following, right? Or can like get rid of all of these little pieces here.

Play from 1:08:08 And just b maybe stick to the experts and the insights. So like maybe it's the experts and where their insights go, or it's the experts and maybe some of the formats that they like, right? Like what's tied to each other, or their hooks and their angles, or their brands that they're working with. I I do like the insights and the experts the most, like these two sort of like connected to each other. But it's like if I click on one of these people I can see where all of the posts go and all the different things, right? If I click on another person, it's like, okay, what's where does this branch out to?

Play from 1:08:41 Or, you know, if I highlight everything, it sort of gives me like the full the full context of where this all lives within this within this brain. You know, no one no one no one comment down in the comment section what this looks like, but hey, you know this is this is It is cool though. It is cool. I think like for me it would be really interesting to visualize something like Cont like content clusters, you know, in the space Yeah.

Play from 1:09:10 and, you know, things like that where it's like what's being talked about, what isn't? You can do that. And that's, you know, again, that's where I think this this Grok bot thing is interesting. Cause then I've had so like here's another one. My n my new little favorite bot is that I have a you know, for for all of our Japanese listeners out there, if you're if you are there, I've have a for some reason, I have like a seven percent, eight percent follower count on on X from Japan.

1:09:15Finding fresh references beyond your feed

Play from 1:09:37 Look at you. And yeah, I don't know why. But I I I love you guys. I've always been a fan of of sort of like the Japanese craft and the meticulous nature of everything they do. I love I just sort of love that whole culture. It's just like we like the thing this is my own assumption about Same. Same. Yeah. Japan. I've never been there. I know a very few Japanese people, but I'm just like I look at stuff and be like, they don't put their name on it unless it's a banger. You know, like you you don't make you don't make any sort of thing in Japan. You make like the best thing you possibly can, which I I will, you know, to take Feels like they put a lot of pride in everything that they do and I like that. And so I've been using Grok to help me find Japanese creators that are doing really cool stuff.

Play from 1:10:18 And like now I get all this stuff on my X feed, it's like tailored my algorithm to like a lot of different avenues. But I but I've been finding these people that I would never find because it would never show me people from Japan. Mm. Yeah. But I'm like, I'm like, I want I like their style and it's really cool. It's so different than the American style. And there's like a lot of cool things that it's pulling up for me. So I love like I love that function of of the GrokBots. I know everyone's like running their business on it and stuff, but it's like things you already built that stuff. It's already like running.

Play from 1:10:46 I don't need I don't need another Yeah. ops tool. I have I have five of them. so you know and it's relatively cheap and it's probably the only one that can really scan X the way that it scans it without having to have an API. So I think it's great. And you could do like some of the custom filterings on the search and you know you can make it really like I I limit the followers. Like I like people that are Like within like three and twenty five thousand followers, you know, like it's still in that like trying to get it range instead of like, you know, you we now have three hundred thousand followers and I just do brand deals and shit all day.

Play from 1:11:20 So it's like and it's Japanese are very I I'm I'm I'm doing a lot of lot of, you know, You're glazing right now. glazing right now. They're Yeah. very tutorial based, like a lot of here's exactly what I did. But I'm like, this is this is cool. Like I like that people share this. It's not like a trade secret. They're like, here, did this. I tried the You know, I had a couple recently had a couple people reach out and be like, Can I use like what you did? Like even like the the Negroni thing I did.

Play from 1:11:48 They're like, Can I use this? And then like create a post for it with like the same process and like explain it. I'm like, Go ahead, thank you for asking. That's that's wonderful. Appreciate that. At least they're asking. No one else asked me here. Yeah. So I don't know, maybe I'm starting to show up in theirs a little bit more. They're showing up in mine a little bit more. I just been loving the fact that I can filter by location on there and be like, find me cool people from there.

Play from 1:12:15 And it's like, Yeah, all right. So, you know, I think it's I think it's probably good to expose So this reminds me of ourselves. this reminds me of like day one, you were telling me about this. Like you you were tapping into the Japanese X feed. And now it feels like we're at like week one. So you're day one, now it's week one. Yeah. We go to month one. You're gonna be wearing you're you're gonna be shirtless wearing sumo gear doing this podcast.

Play from 1:12:44 And then second month like month two, you're literally gonna tell me you're you're moving to Japan. I mean I might as well, you know. I the buying buying all Japanese products. You'll you know, I'll I'll have my I'll buy my, you know, my my nineteen nineties Toyota or Toyota Supra or I'll get my Skyline or whatever. And I'll be I'll be very into it. I'll have dude I've I've been I've been wanting to go to Tokyo for a really long time. Anyone who's from Tokyo listening, I wanna go too. Yeah, I'd like to go.

Play from 1:13:13 invite us over for something. Yeah, let's go. Conference, something, man. Let's go. Something we We love we love it. We love everything about it. I love I just I I'm s so curious. And it's it's just this this other thing now is like it opens up the the the tasting, right? Like what's crazy is I I was going th we we talk about this at at one of the projects is that the design styles that convert in different regions is so radically different.

Play from 1:13:42 Like when I did the conference in Bangkok a couple of years ago and I was showing like the highly polished like US version of like how things are presented, like from an advertising standpoint, versus like what works in Bangkok. And it looks Yeah. like it looks like something out of like a, you know, not to generalize or use a term like this, it looks like a Chinese menu. There's just stuff everywhere. It's like very busy, Yeah. like very in your face. But that's the style that works there. It's like there's no, Huh.

Play from 1:14:08 there's no sense of spacing and kerning and like letting stuff breathe. It's just like jam it all in, make it so like Really? visually explosive that you can't not look at it. But it's like it's so crammed. And you're like, how different is that across, you know, aesthetics, right? Like you look at some of the Scandinavian stuff, right? A little bit more minimal, right? It's not as clustered. It's a like a lot more room to breathe. It makes sense with the environment. You look at something like Bangkok, it's everyone's on top of each other.

Play from 1:14:35 It's a big city. It's it's crazy. Like that makes sense visually. Yeah, the cultural sort of leakage, Right? You just see what I'm saying? Yeah. right? Like yeah, integration into it. Yeah. It's it's cool and like that's why I like that's I'm I'm probably gonna do more now of that filtering via set 'cause I I just like there's a very there's a very US sort of look to everything, you know? Like it's very there's like a same thing when you go to like the US movies versus like a French film or like a you know, like I always go back to City of God from you know, those filmed in Rio. Right?

Play from 1:15:06 yeah. Yeah. Yeah. Yeah, it's got such a it's got such a different vibe to it. It feels like it's very You know, the the music is on pace with like sort of the reality of the of the movie, like the the shots, the like sort of like the the feel, the analog nature of it. It feels like very in the local spirit of where that like w what time frame that takes place in, what location it takes place in. Like it's on brand to like the the environment, if that makes sense. So it's Yes.

Play from 1:15:34 like a so I'm so I'm I'm looking at this. I'm like, have I been ignoring like do I see do I ever see anything from Africa? There's probably some really cool I don't know, I've always loved like that. you know, the the African a lot of the different African countries have very specific design and, you know, different, you know, color palettes and the way they use patterns and things like that. And I'm like, I really like the way that looks. Some some different style styles in South America. You know, the Peruvian ha have like a very the gaucho sort of culture, things like that.

Play from 1:16:00 Goucho. There's a lot of cool stuff that like has a very specific vibe to it. But I don't get it on my feed. Right. So I'm like now I'm gonna go and look for it. And I feel like there's a good way to find it. So I hope Hope I didn't bastardize any I feel that. culture description there. Just you know, just Yeah. know that I'm I'm looking for it now and I'm I know I can find it now instead of hoping the algorithm shows me. Yep. No, I feel that man.

Play from 1:16:25 I feel that. I'm curious, It's cool. like what what what hot takes you you sitting on lately? Is there anything that's like coming to mind for you where you're just like, dude, people got this one? This is always a bunch of hot takes. I feel like I spit out every day, like either to myself or to to someone in like a in like a text. Like Fable five point two, too expensive. Too expensive for what it produces. It's not it's not Opus five point five when it comes to like the value per token. So that that's that's one, I think. The you know, recently I I'm I'm tired of I'm tired of how easy things are.

1:16:34Easy AI work and the search for a challenge

Play from 1:17:02 Like that's that's it. That's the one thing. It's like Yeah. I I now go away from projects that I know are too easy and that I can do because I if there's a problem, there's a solution. Right? Like it used to just be like you can't like like you said, with the consistent characters things use You do you mean you mean from an angle of like, and I'm curious, the reason I'm asking this because I think this is maybe the thought I have I connect to with that. But is it is that because you feel like once it feels easy in your mind, that means somebody else also can figure it out or has figured it out and can offer the same thing.

Play from 1:17:41 And then in a sense, it doesn't feel quite as valuable or something to hang your hat on. Yeah. I'd say that's part of it. kinda that's kind of part of it for me too, I'd say the n the right? Like it's like novelness of it is will wear off in about six seconds. that that th that's maybe the best way to verbalize it. It's just it just can't Mm-hmm. be it's just like yeah, we used to figure out things to do in Midjourney that a lot of people couldn't do or didn't have That's right. And then when you get the vision to do. right. And then now either because the t the models have changed or it's just common knowledge now, it doesn't quite feel as valuable.

Play from 1:18:12 And it's not as valuable, but to like a client perspective, it it still could be. But even then you're kind of a little bit turned off by it because it just doesn't feel quite like it's a true representation of something that's gonna get you excited. Yeah. Yeah, it's just it doesn't get me excited, like 'cause now I'm like, if I wanted to do this, someone could just throw it into Claude, say reverse engineer everything of this build the same thing and then That's right. That's right. do it again. Right? Like Yeah. like like that Negroni video, that would have been two years ago.

Play from 1:18:38 Dude, we were, but dude, isn't it, isn't it funny though, how like something like that that is true now and I think a lot of people are unlocking, like because we were part of that like really early wave of like the photo thing, and the idea of reverse engineering really came from that, right? Because like we would see really cool images, or at least like my experience, I'd see like cool images on another site. And it's like, I want that look, or I want to create something like that.

Play from 1:19:06 How do I? come up with a prompt for that. well, you know, I'm gonna analyze this photo and I know the prompt structure I like, right? And I'm gonna have it basically describe itself to somebody that it can't, you know, that doesn't have the ability to see. You know, and and then now I've got something that is probably producing like somewhere between like a 95 to 99%, you know, accurate depiction of. And it's like, dude, that same framework, that same ideology is is has given us life throughout this whole wave of the video.

Play from 1:19:42 And then you get the 3D. And then you get like, I don't know, website elements, or then you get like any any reference. Right. And like Midjourney was really, I feel like ahead at that point in terms of the way they were thinking about it because that unlocked a lot of like, I think connective tissue for us to that still, that still plays today. And and and regardless of whether we're in Midjourney or not, right? It's just like that same sort of common thread and approach, kind of going back to the beginning of the conversation, is still true.

Play from 1:20:11 A lot of those things are still true in how I operate. Like those principles that that we learned and uncovered back then still hold true in a lot of ways today still. Yeah. And I think like building context has become sort of like frustrating for me. If if that's the that's the way to put it. It's like every time I want to start something, I know what it takes to get a good output and I know what it takes to get to the result I want, which means I have to spend a significant amount of time building the right context so that I can one shot it.

1:20:20One-shot prompting and dictating the goal

Play from 1:20:38 Yeah. Yeah. Yeah. When in reality is one shotting it that important? Do you know what I mean? Like is is Yeah. one shotting it and that that's a hot take I have. Is one shotting it that important? Because if it if it is, like No, no. Because I mean, like anything complex, you can't one shot anyway. Right. Like Yeah. it like and it'll tell you too. Like sometimes I've gotten to a point where I'll get in some of these threads and I I know I've done more iteration than I than I should like could have.

Play from 1:21:04 But I'm like, how Mm-hmm. how could I have prompted this in a way to give me as few prompts and back and forths as possible? Yeah. It's Because it's now it's got all the context of what what went in and how we were thinking. And it's like most of the time it's like, well, you couldn't have done this in one. You know, we still would have had to know in this, this, and this. And so like one one of the kind of going back to how I do this with teams too, is like I think one of the most important things you can teach them is like the dictation.

Play from 1:21:31 So you're not typing. You just get used to just word vomiting everything out. You state the goal or the outcome, right? Doesn't you don't have to know the how to get there. Just Really have sort of like a clear understanding of what you want, right? Like what is the ideal thing? And then have it ask you clarifying questions so you eliminate all ambiguity and speculation. And yeah, you might have a couple iterations of those questions, but it's like, dude, that is still probably one of the fastest ways to get to the end goal is possible.

Play from 1:22:03 It's just like you want it to have you ask questions. Does it take more time? Sure, but it's not going down the wrong path when it's asking you a bunch of fucking questions. You know, and and Yeah, I think so like It's also decision fatigue at a lot of stuff where you get to these burn you get to these like forks in the road with the with the things that you're at. And it's like, well, I want to do both, but also I don't want to run both in parallel 'cause that'll branch into seventeen different new problems. And then just say Yeah. But even just saying that, right, it's like you kind of need the dictation on at all times.

Play from 1:22:33 Like, so so like when I when I get in some of these engagements too, on site, it's like I'm I'm recording the session for a lot of reasons. One, I don't have to worry about taking notes, but two, it it's got full context of the entire transcript, what we've talked about. But but when you get into like having them prompt too, like you can see they're trying to figure it out in their head. And They're answering questions, right? In betwe they're kind of talking it out in between talking and dictating into the the box.

Play from 1:23:02 It's like probably just leave that on. Like just leave that on right now. Because even right now, you're tossing around. I don't know whether I should do this or do this. And you know, like I know this. And it's like, dude, tell that. Like, record that, right? Like, because It's really about like that that does answer questions without answering the questions. You you being unsure about this or that, or even just your example, it's like, well, I kind of want to do both, but I don't want to do this.

Play from 1:23:28 You know, it's like there's value in that versus just having to pick one. You know what I mean? Because it might actually have a smart approach to doing both, right? So it's sort of this really mind fuck thing, And I find myself Yeah. you know, like you just And then find myself I find myself like just trying to see what it can do by itself instead of things just I should do by myself. Like this workflow thing I was telling you about. I'm like, I could do that. Like I I can do this. I know I can figure it out. Yeah. I just have to spend like an hour in there.

Play from 1:23:54 But I'm also like, if you can figure it out, go ahead. And then and then Yep, yep. it's not figuring it out. So then I ended up doing double the work. Right. So I ended up spending instead that hour that I was in there doing it myself, I spent an hour of trying to tell it what to do and it not figuring it out. So I ended up spending two hours instead of one. Right. Like there there is some level of this to that where it's again, like, yes, is it is it the most helpful thing we've ever had? Yes. Could I do it by myself?

Play from 1:24:19 Yes. Is it worth spinning the wheels on something complex if you're not gonna get it? But then again, it's also like you don't know if you if it can do it or not. So you have to try. And then so then no for the second time. Right. But like the other example that I that I would like to use on this one too is the is the clipping agent. Right. The first the first run of it that we we showed on here when it was like did our clips, had the motion graphics, whatever. Cause I went I ran it again last week and like I hated all of them.

1:24:28When AI turns feedback into permanent rules

Play from 1:24:46 I'm like, what what the hell happened here? Like it's nothing has changed since the first run. Like I've just kept it the same. It didn't it didn't drift or what? What was your what was your observation? It it just drifted. It was just it was just like with the with the logging of results and like my feedback, it started to like self-optimize in places where I didn't want it to self-optimize, which I said, don't do that. And then it was like, well, per our rules from this one conversation, I was like, I didn't make that a rule. That was just feedback. You know?

Play from 1:25:12 And then it was like showing me the side by side of like what Yeah. Get in the fucking lost in the sauce with that. the original was. And it it this is one of those one of those very human things to me when it was like the first it was like the first time you said for us to do this, you just said Go crazy, do whatever you want, make it look good. And like that's what it came up with. Then it was like once I built a bunch of rules and like had it confined in a box, then it started making wrong decisions. I'm like, this is this should be the opposite of what it is with computers and automation, right? Where it's like it does the same exact thing every time.

Play from 1:25:42 And like that's what's written in the code, but it doesn't. It it wants to it like it wants to do its own thing and come up with its own solutions. Like it doesn't want my prepackaged solutions that we already worked on. And Yeah. That's what I'm finding interesting with it is like I'm like, why are you doing this? And it was like, well, because of the rules from the feedback and now like we're constricted in this area. And normally when I would have gone here, I wouldn't I wasn't able to because of the rules we put in place. So I'm like, are skills really becoming like a detriment in some ways?

Play from 1:26:11 Are they becoming sort of like a you know, when you're a creative and you keep putting boxes, you know, you know, box around yourself and saying like you can't go, you can't go outside these walls and then you're like, I don't know what I should do. you know, I really want to just like let my brain go and then I'll come up with something cool. But when you have all these constrictions, it's like working at a brand, you're like, No, you can't do that. No, we can't use those colors. No, the person can't be facing to the side. And no, the text is too big, right? Like it like sort of depletes you and it makes Yeah. Yeah. You know what else, you like, this sucks. So I hope we're not doing the same thing.

Play from 1:26:43 dude? I was thinking last night, these, these you know, one of these shower thoughts, but it's like, I should create a five wise skill, you know? Like when you get to the root of something, you want to get to the root. You just ask, you keep asking why, huh. Yeah. right? It's like we do this podcast. Why? Well, because we like to talk about AI and it's fun. Why? because something we work on every day and we're passionate about. Why? You know, like you you eventually hit like a bedrock situation.

1:26:45Five Whys and reusable AI skills

Play from 1:27:10 I was like, dude, I should hit a s I I should just create a skill for that that I can run kind of with the outputs that it gives me sometimes, you know, I like that. and and like learn lessons from that. Like that could be interesting. You know, like I don't know. And then like I was also just thinking about I was like, dude, why also like I feel like I should be stacking skills more. Like I I've created so many skills in sure, like some agents and so forth have those attached, but I'm like I'm sure there are even more creative, maybe even obscure ways that I could be stacking skills in parallel. Right. And Yeah.

Play from 1:27:46 maybe I would encourage other people that have done that, right? Where you've probably got a lot of different skills. see how they might pair up, right? And kind of work in tandem. You know, they can run you know, in in sort of like a chronological order or as sub agents simultaneously, like whatever it may be. But like I feel like that's. That's like one interesting thing that I was thinking about just last night. I was like, I should be probably exploring that a bit more as well.

Play from 1:28:13 So that goes back a while to another conversation, like maybe last March when I was doing when I was sort of visualizing this for the presentation, when I was saying I take all my skill after I build something, I always have it mine for individual skills and strip and make them sort of make them sort of analogous. No, you or not analogous, that's not the right word, make them sort of abstracted so it's not like specific to any project.

Play from 1:28:39 Like when I did the comedy writing pipeline I broke it into and it's like the analogy engine and this and that. Yeah, yeah, yeah, yeah, yeah, And then just like yeah. dump it somewhere. So it's like here's all these random skills and it's like now you can tap into whatever you need to build, Yeah, you break, yeah, you break out of the specific yeah. here's those other ones. Like and and that's kind of cool too. Like I've done that with like client stuff too, where it's like white labeling skills where it's like Yeah. I have my own use case for it, but it's like, how could you get the benefits? How could I strip out, you know, those maybe dynamic parts versus the Mm-hmm.

Play from 1:29:06 static ones, kind of kind of deal. that's I don't know. Like just there's a lot here, dude, that that could be tapped into even just within your own ecosystem that's still A lot. Same with scheduled tasks. I feel like those could then be stacked, right? Like so scheduled task huh. one runs. Okay, now that you've got those results, here's something a s you know, maybe another task grabs that handoff, right? And then relays it here.

Play from 1:29:32 Then it maybe runs a skin, you know, like it's just like, there could be something could something interesting there that, It's a lot. It'll make you nuts. you know. It'll drive you nuts. I have more scheduled tasks now than I did a month ago and now I feel like every morning is a bigger I mine is ridiculous because routine. they all run on Monday too for some god awful reason. And then God, how many do I have live right? I had to turn a couple off, bro. It's like I'm not checking them anymore. Some of are vanity. I probably I think I have like twenty that are that are live, like still not paused. And I've got probably another twenty that are just like I had to turn off that it's like I'm not either not using it or I'm just not, you know, like maybe it's lost some value, but It's interesting.

1:30:15Comparing AI skills before adding more

Play from 1:30:15 Audit your skills. That's another one. That's a th that that goes back a while What's that? back. All this stuff audit your skills. Stuff's coming around full circle. Audit your skills. i I okay, Coming around full that I circle. that was one other point I was gonna make. I still see things all the time where it's like, You know, like taste skills, or you might have like a a design type of skill, right? But it's like I've already got some design skills loaded, whether it's my own or somebody else's, right? Or I've taken parts from it.

Play from 1:30:45 I always am running the experiment of I, you know, it's like a skill, my adopt where I'm like digging into it to see, do I already have something that exists? Does this overlap any parts? If so, I need to put them head to head. And I want to understand visually what those look like head to head, especially when we're talking about design stuff, right? Like, okay, like if I'm gonna adopt this skill, I need to see how that compares to what I currently have.

Play from 1:31:12 And you're not gonna see that unless you ask for the visualization, right? So otherwise you're just guessing. And so I I don't take anybody else's skill unless I've I've ran it through an assessment process because it might break something I already have set up that I like or do something different. And You know, I I also like to just kind of it it gives me basically a list of this part of, you know, adopt fully or adapt, meaning like take pieces of it or skip it all together. And here's why.

Play from 1:31:40 And like that saved me a ton of time because you always see this cool stuff and you're just like, maybe that could work, but I don't know exactly how that would fit in or if it does. And, you know, rather than just blindly accepting those things, you know, going back to your point of like knowing what you have and auditing what's there, have to do it. I think that I think we should probably should probably wrap up on that one. I think we we just d tie it tie a bow. Is that the is that the actual terminology? Tie it. Tie it up.

1:31:57The sign-off takes a detour into Bravo

Play from 1:32:06 I think that's pretty pretty well said to end this one here 'cause I think it Shout out to all those bow tires out there. That's a you know, shout out to those bow tires. The bo bow tires? Boat tires. Did I say boat? I think I did. That was well you do have the dock hands, you know, the or the deck hands too that do that at marinas. That's a that's another good one. Deck hands. Wow. Wow. Getting nautical. That's right. First mates. Well, we haven't touched we haven't touched the maritime professions yet.

Play from 1:32:31 Maybe we need to go that way. Get nautical with it. It's a very it's a very very cold nautical type type you know, type day up here in the northeast. You're below deck staff. Yeah. You're below that your first your first mate. Your you know, your deck hand, your your L Cappy Ten, whoever else might be on board there. I don't know all the positions. I don't know all the positions on on board there, but let's just cover all. Yeah. you know them. You know the terms. Okay. I see you. You are I was gonna say, are you watching Bravo? Okay, I got you. I got you.

Play from 1:33:02 Yeah. Yeah. I used to watch that show back in the day. yeah. That one's that one's if it's gonna be on, that that's not that's not the worst one to have on. You just look at these people and you're like, can't believe you you like pay for vacation to go on and be on a TV show. Like the people that are renting the boat. I'm like, I wouldn't want my I don't know, I don't want anyone filming me on vacation. Yeah, hell no. Camera crew there too? I was like, yeah, no. Dude, no. Catch me drunk saying dumb stuff on TV and then like getting blasted on the internet about it.

Play from 1:33:29 No, I think I'd rather pay twenty thousand dollars somewhere else. I wonder if they like give an extended stay on those trips where that's it's actually longer than they say and then like the the crew gets off. Like there there's probably like I wonder if there's like I would hope so. yeah, like they can actually enjoy it. my god. But yeah, man. Well going back to my days on Bravo, last quick story. My my days on Bravo when I was on Real Housewives. That was the producer's wood. Let's just call it that.

Play from 1:33:55 It's like boy. it's it's the it's the the the pre the pre scene you know, liquid courage sort of situation where everyone just you know free drinks, there's a couple pieces you a couple pieces of food. The whole goal is to get everyone a little bit saucy before they sit down for the dinner scene and then that's what you see like the I I told you this and then like you know, all that stuff comes out. And it's 'cause, you know, they've had two martinis and and nothing to eat and in a you know, high stress situation with a bunch of cameras and lights all them Yeah, loosen up a little bit.

Play from 1:34:26 and there you go. What loosen it up, it makes people more stressed. I think that's the thing. And the producers get in your ear like, Why don't you talk about this? Do you remember when she did this? And like that And then You know, before you know it, it's like yeah, hair pulling and and cursing and there's our next viral clip. it's engineered a lot of it, my god. Dude. manufactured. but those those little they're like Who's that character in in Game of Thrones? that stirred the pot with everything?

Play from 1:34:55 Was it Varys or and Littlefinger, Littlefinger. Littlefinger and Varys. that's the one. Yeah. Just like stir the pot, And Varys though. Yeah. Always. you know, get everyone going. Look at it from a look Always. at it from a ten thousand foot perspective. But yeah, that's Yeah. We'll get into my Bravo appearances another time. You guys can you guys can hang on for that one. Bravo had a run. I love Bravo had a nice little run there. Another episode. Yeah, man. well, dude, yeah, glad we got to do this. next week, we'll be back.

Play from 1:35:20 Thank you guys for joining us. And yeah, everybody have a good, good little break. We'll catch you guys next week. Peace. Stay sane. See ya.