Claude Code vs Windsurf: Why I Switched AI Coding Tools Vibing with AI Code, Episode 3: Brendan & Baird Brendan: today I think our primary topic is gonna be around Claude Code and our experience working with it. Um, you know, we've talked about it a few times in the past, but I think we've just got more things to talk about now. So we'll about that. And before we get into this, the topics, um, you know, I just wanna remind everyone that we are brought to you by Fireside fm. If you're looking for a podcast host, um, I think they're audio only right now, but they've been around since 2016. That's where our audio version of this is hosted. Um, there'll be a. You know, link in the show notes, check it out. And uh, yeah, let's go ahead and dive in. I know Baird, you've been working, uh, I mean you work. Day to day on code more than I do. And so your experience with Claude Code is, is, uh, I guess you, you've learned more or you just have a lot more experience recently than I do. and I, I know we talked probably last week, I think it was about, know, you're seeing some pretty cool stuff. So why don't we go ahead and just kind of dive in there and you can start with whatever you think is, uh, the thing you wanna share first. Baird: Yeah, sounds good. Yeah, it is one of those things where, you know, I use Windsurf pretty exclusively. Like, you know, I do some work in Claude directly and stuff like that, but. For a bit there. Windsurf actually had a whole issue with, uh, anthropic because of the potential merger with Open ai. So Claude actually got pulled out of Windsurf, and then when they finally brought it back, when they ended up getting bought by cognition, they actually are charging double for the usage for Claude than like every other model. And like Claude is, in my opinion, like better at coding than pretty much everything else I've used, like I used, uh, cheche or. Oh, uh, GPT-4 0.1 for quite a while. 'cause that was like like 0.25% of a credit and you know, it did okay, but it just, it wasn't quite as intuitive as my experience with Claude. And so over the last couple of weeks I just decided, you know what, I'm gonna pull clot into windsurf. So I still have the, you know, auto complete. Because the auto complete actually works really well in an ID like this. Like when you take control, it can still help speed you up without losing that. And I think you might be able to do it for free, but I'm not completely sure. I also have like the. Founders like Price for Windsurf. So I'm paying like 10 bucks a month, so I'm like, I don't wanna cancel it yet, just to kind of see how this works. But yeah, so I brought, you know, Claude Code in, it has the extension or plugin for Windsurf. So I'm running that and I'm essentially using it the same way that I used, you know, windsurf or how it people use Cursors agent in it. And like, I pull the terminal out, put it up kind of on the side just like I do with Windsurf and then just kind of run it in the same way. And, you know, the, the mechanics. Aren't radically different and I know a lot of people are using it, so I'm probably just catching up on it. But it's been really interesting. You know, the biggest thing I've noticed, and I haven't used cursors, so I don't know if it's exactly the same, but I don't think so from the research I've done is just like, it uses a to-do list like religiously, which is really interesting to me because it actually, where I run into like those endless loops with other agents, it tends to, because it has that to-do list. It does a really good job of like keeping on task and you can like point it back to it and it remembers what it was doing with better accuracy than I've had with other bots at this point. And you know, when you couple that with the quality of Claude, it's just been a really interesting experience to see how much more I'm getting done right now. And it's actually like. I mean, there's still the, Hey, you're doing it wrong. We have to start over and you know, do all that type of stuff. But that's less frequent than it used to be with the other bots that I've used. Brendan: Mm-hmm. Yeah, that's cool. And it sounds like what you just described there is what I mean, I've been doing a little bit with Cursor as well, Baird: Sure, sure. Brendan: the Claude Code plugin Baird: Yep. Brendan: you know, for those of you who don't know, I'd be surprised if you're watching this show if you don't. But just in case, you know, cloud Co, or sorry, cursor and Windsurf both. They're, they're, uh, built on top of VS code. And so like vs code, they've got the terminal. So the way you can run Claude Code is you can, know, there, there's the extension or, or plugin like you said, Baird: Yeah. Brendan: but even without that, like you can just run it right in the terminal. Baird: Yeah, Brendan: at least in cursor's case, I noticed that it's like, oh, well we've got an extension or plugin Baird: yeah. Yep. Brendan: so it, it recognizes that. But honestly, like I don't, I don't know what that's even doing. I just know that it's a thing. Baird: Yeah, what, what that does, and it's kind of clunky right now, like that is one of the pieces that isn't as refined as everything else that I've seen is that it gives context. So like the plugin actually gets fed context from your windsurf or cursor terminal. So like it knows what file you're in and it can actually get passed. Like you can highlight data or like, you know. Brendan: today. Baird: Yeah, so that, that's the biggest thing. The rest of it, I mean, yeah, you could just spin up a terminal in it. You could spin up a terminal out of it. You'll still get all the benefits of like the coding, but you do get that context like. By default if you're using it with the plugin. So I, I have noticed though, that it disconnects quite frequently. So it's like it'll be working and then it'll stop working and I still be out of habit. And just to make sure I give it like the paths to the components or the pages or whatever that I want it to work on. Just anyway to make sure that it actually knows what I'm talking about. But. Brendan: Which is funny 'cause that's something that I was doing, like even when I was using Cursors Agent and not Claude, like we Baird: Yeah. Brendan: There's a way, and I assume I, I've used Windsurf a little bit, but not a lot. And I, I forget how similar they are in these res regards, but like in Cursor at least you could, you know, mention a file or you could like. Baird: Yep. Brendan: Right click on the file or add it in to your, Baird: Yeah. Brendan: chat. Baird: Yep. Brendan: still found myself constantly having to, to be like, no, I mean this file. Like, and just like put in the path in the text. Baird: Yep. Brendan: and, and I did notice, so I, I just happened to be working on something this morning. Where I was using this whole setup that we talked about, back to the extension, I, I noticed at the bottom of my, my terminal window that it said something about the file I was on and I was like, oh, that's Baird: Yeah. Brendan: But it didn't even cross my mind like how it knew that, but yeah. There you go. That's the extension or the, the plugin. I don't know why I keep saying extension. Baird: Uh, I, I think, are they called extensions? That, that's actually a a good point. Semantics. Yeah, Brendan: way. Yeah, exactly. Baird: it's the same thing. Brendan: what it means, Baird: I would hope so. Brendan: yeah. Um, there's another point I had there, but I, I, I think one question I have for you, just because you are Baird: I. Brendan: code day to day, where I don't do that still, um, like how much of a, a production gain have you seen from, you know, going to this Claude Code in Windsurf versus just having windsurf. Baird: Yeah, I mean, what I've noticed is like when I would try to have it, you know. And one of the things I've been doing over the last couple weeks is doing a lot of like tool building for internal work stuff that I'm building and it's more like prototypes than it is. I'm like fully building production stuff, but it's kind of a, you know, at work we're trying to do a bunch of new business stuff and we're trying to do a bunch of like integration with AI in different places and being able to manage all this data and act on this data. And so. Outta curiosity, I was like, let's just start generating this stuff. And so like I've really tried to get into the habit of like, 'cause are you familiar with, I, I know you know about Claude MD files, but are you familiar with like how you can nest them? So like you can put them in the folders where you're working and give more like specialized context? Brendan: Sort of, but yeah, for the benefit of our, our listeners and viewers, let's go ahead and talk through it. Baird: Yeah, so essentially like as Claude Code works on your files, it will look in your route directory, find a Claude md, and that's usually a good place to put in like the general, like this is the technology you're using, these are the rules that you have in place, like kind of the overarching context for your project. But if you're like building a new feature and you create like a directory called. You know, source slash feature. You can then put a Claude Code in there and describe like what you want that feature to do. And like, you don't want to put the overarching context in there because it gets picked up as Claude Code goes, it goes recursively through your folders that you're working on to pick it all up. But what you can do is give it a to-do list in there, like your own to-do list and say like, I wanna build this, this, you know, like give it all the detail. You can give it references where you want it to go. Like research things if you need to. You give the overarching kind of description. And then you just kind of say, you know, look in this folder, grab the Claude MD and start working on it. And it'll go out, grab that. And it essentially picks up your to-dos, kind of changes them a little bit to kind of fit its mental model, but essentially like just maps that it'll ask you, like, do you wanna start with, you know, a certain one or do you wanna work through sequentially? And it's cool because you also then have the options of, you know, if you. Wanna make sure you're doing it right. You can just be like, okay, let's go. And every time it's making a change, like, you know, after it creates the plan, it'll start making the changes and you can like approve it each step. It'd be like, okay, yep, put that in, put that in. Or if you're like, I don't really care if this is like fully fine tuned right now, I just want it to build something for me to get started with. You can just say like, Brendan: or something. Yeah. Baird: exactly. Or just like something that's not super important or like in my case. Brendan: change it later. You're not too Baird: Yeah, and it's like it's internal facing, not going into production for mass usage. You then can just be like, okay, just do it. You know, tell it to stop asking you for permission and then it'll go through and just build the whole thing. And kind of coming back around to what we were talking about was like where I've seen the most gains is the last couple of days. Especially I've had stuff that I just didn't have a time to like babysit going through it. And I just wanted to kinda see where I got to. It's built like two full like dashboards for me that are pretty well built that it did without me like touching anything. Like there's definitely still not like I would leave the code as is and walk away. Like I still have to go in and as I'm, you know, testing it and seeing stuff, I'm like, okay, we gotta tweak this or that. But yeah, like figuring out the whole context at the different levels and like giving it those to-do lists at those levels has. Made it so, and like again, you might be able to do it in Windsurf, but every time I've tried the Windsurf agent to like be autonomous and go that far, I've never had success, you know, and Claude Code, it was like I didn't have to stretch to get it to be doing things that were super productive super quickly. So it's been interesting. Brendan: Yeah, that's really cool. I mean, I forget if I talked about it on the show or not, or if that was just a separate conversation, but I had been doing for a while, this was even before I introduced Claude Code into my workflow with Just Cursor, is I would, you know, have a a to-do list, but it was just in the root of my, my Baird: Sure. Yep. Brendan: and I would kinda say, okay, we're gonna work on. This now, and I Baird: Yeah. Brendan: it was definitely not as smart as what this is, and that's really, it sounds like a really good way to do it. I'm curious so. I have thoughts on this, but I'm curious just for you, and just again, share with the, the listener, how do you go about getting those to-do list files to begin with? Before you start? you're, you're, you're, obviously, you're, you've got an ID in mind, but are you planning with Claude or you know, the LLM ahead of time and creating those. Baird: Yeah, I. You know, you can do both. Right now I've just been kind of testing out, like in my head, I just have been giving it relatively simple instructions of like the stuff it needs to do. But there definitely is, and I have actually worked through with Claude Code to say like, let's evaluate what we already have and then figure out like how we can proceed, or like, you can have a conversation directly in it. So that's the, the, the piece that isn't. Readily inherent about it is like you can actually ask the questions and have it go, give you answers on your code base based off of, you know, whatever you want it to do. And you can have it like, you know, give me a detailed list, like what are the priorities of how we can approach this and stuff like that. And it can then, and this is the interesting part about that, which again, like I haven't done a ton with, but I'm glad you brought that up, is. You know, if you go out to Claude, you go out to chat, GPT, the, you know, interfaces themselves. Yeah. Like you can do it and like, I've had a lot of success with Claude planning stuff there. But what's really cool about doing it in Claude Code is that it's in your editor so it can look at all of your files and can be like, yeah, I see we're doing it this way, so we'll continue to do it this way. And so I, I think there's definitely some potential there that I wanna push on to, like. Do more of those planning pieces and creating to-do lists with Claude's help versus just like me coming up with stuff off the top of my head. Brendan: Mm-hmm. Yep. Yeah, because I would imagine if you do it that way with Claude, then it would. Probably just no to, okay, we're talking about putting something in this directory like you were Baird: Yeah. Brendan: So let's put the to-do list file for that here. And then the other one's here and there versus, you know, what kind of what I was talking about with my old, uh, workflow, which was talk to the LLM Baird: Yeah. Brendan: in the web browser or in the, you know, desktop Baird: Yeah, Brendan: whatever, but not inside of Claude Code Baird: yeah. Brendan: even inside a cursor in that a lot of these cases. then come up with A-A-P-R-D or a Baird: Yep. Brendan: and then throw that in. But then, yeah, that, like you said a minute ago, like that didn't have the context of the app. Baird: Yep. Brendan: wasn't thinking about that stuff. It was just all the, as we've talked about before, it's kinda like the average planning, like Baird: Yeah, Brendan: is what the average thing has done Baird: exactly. Brendan: build this sort of thing out. so that, that's really cool to just think about. part of it that like you can plan with your app's, uh, scope in mind. Baird: Well, and and that's another, I think, huge thing too, Claude Code. And that this is versus like a lovable or a V zero or whatever is like if you already have a code base or if like you have an opinion at all and you don't have to start with your own code base. It can be like you go in, you start modifying things and then you have Claude Code come in and actually be like, evaluate how this has been written and create, you know, your clot md that actually like puts in those rules so you can keep consistency, whether it's code consistency, design consistency, like whatever it is like. You know of easier or lovable, you start from, the bot spits all this stuff out. Hopefully it's consistent. You never know. And I think it's getting a lot better, but then it's like you have no real control over that. You know? Like you don't have your opinion, you don't have your style, which is fine for somebody that doesn't code, that doesn't care, you know? But this is kinda like almost the, and this actually kinda leads to another thing I, I saw that I really liked about. The whole vibe coding thing is that like it's less vibe coding when you do it the way we're talking about here, because the, the way I've heard it was like, if you know what you're doing, you're not actually vibe coding and like, so it is really kind of like you're just using a tool versus like you're working off of vibes. You actually know what you're doing. You're actually like directing it in in the right way versus just being like. Build me an app, you know, and so it's an interesting kind of shift in it. And, and I find, you know, vibe coding works more for like people like us when we're prototyping, when we're like trying to push into things that we don't know as much about and like we're trying to figure out without having a fully formed idea. But once you, like, get it into a production day, you know, like day-to-day routine. It isn't vibes. You're not just like leading, like letting it lead. You're like giving it the requirements and then having it go do the work for you. More like we've talked about before, like an intern versus like it's actually doing the work instead of you because you don't know what you're doing. Brendan: Yeah, that, that's a good point. It it, it makes me think of a lot of things about just like it, I, I like the vibe coding its, know, what, what it's meant to be, or, you know, what, what that, that definition actually Baird: Yeah. Brendan: from a standpoint of I can build an, uh, an app. In five minutes or 10 minutes or whatever it is to get that greenfield, Baird: Yep. Brendan: out of the, the way I'm not blocked by that. And I can instantly see, not instantly, but I can quickly see, okay, does this make sense? Or, you know, it's like a planning mechanism. It's a Baird: Yeah, Brendan: also planning, like, does it even make sense? What would that even look like? Even at like Baird: for sure. Brendan: quick building sorts of thing. And then you can throw it away like almost instantly and go Baird: Yeah. Brendan: and, and that's something that I've seen several people on X Twitter talk about that. You know, they're, they're building in public and they talk about how they can just like something and then, then rebuild it the next day, Baird: Yeah. Yep. Brendan: like a, a cardinal sin in software development, Baird: For sure, Brendan: to rebuild things. And obviously once you get in production, you're also not gonna do that, Baird: for sure. Brendan: It's, you can iterate so quickly. It's, Baird: Yeah. Brendan: of nice to be able to do that. But yeah, to your point, like this, what we're talking about here is not the, the same kind of vibe coding and, and yeah, Baird: Yeah. Brendan: but we don't want it to be, we'd want to get the, the Baird: Yeah. Brendan: of using ai, but we know enough that it's, it's, it would be kinda silly to not use our expertise with, with software development to, to help us. Baird: Yeah. The, the other reference I've heard about Vibe coding, which in this context I think is absolutely true, is that Vibe Coding is essentially legacy code. It's code that gets put into your code base that you, Brendan: I, I, I marked a tweet to talk about, so maybe we Baird: yeah. No, that, that, that was hilarious. That was actually the same article where they kind of said like, if you know what you're doing, it's not actually vibe coding, but yeah, it, it is totally a, the more you have code generated for you that you don't understand and don't take the time to understand it is legacy code. 'cause it's the same thing. You come into an app, you don't know what's going on with it because you didn't write it and it just sits there and nobody wants to touch it. So it, it's an interesting concept. Brendan: Yeah. that reminds me too, I should look it up here quick. So I, but it's, it's related to it. Um, uh, let me just see if I can find it quick, but, so. DHH He's the creator of Rails. happened to see a tweet from him today about AI coding I, I, I also saw that they have a podcast episode out recently. Lemme see if I can find it quick. Otherwise, we don't need to, uh, spend the time. But I, I know that that 37 Signals is not, definitely not doing a whole lot with it. I don't know Baird: Sure. Brendan: I haven't listened to their podcast episode yet, but, And shoot. It looks like I, I didn't save it, so I might not be able to find it, but I, I wish I remember what it was. 'cause the way he said it, what he was basically covering, I remember, was he was talking about the fact that when you are, are using AI to try and vibe code essentially, Baird: Yep. Brendan: just keeps going in that loop that we've talked about. I think Baird: Yep. Brendan: first episode around, you know. It's not, you can't just rely on it to get everything done. Baird: Yeah, Brendan: know? He was just, I wish I could remember the way he worded it because it, it made a lot of sense. And it's like a, like a black hole essentially, is what he was Baird: sure. Brendan: Like, you, you could have spent that 30 minutes doing something productive and actually learning something versus trying to over and over ask the AI to fix it. Baird: Yep. Yeah. Brendan: know, I, I do less and less because I have realized that it's not gonna fix it. Baird: Yeah, Brendan: the times when I would run into that was mostly because. was curious to see, Baird: for sure. Yeah, Brendan: like if I was doing that day to day, I would've probably learned way faster. Like yeah, Baird: yeah, Brendan: gonna happen. And there's certain boundaries, which, which we know now, Baird: yeah. Yeah. Brendan: but it was a good reminder of that sort of thing and, and kind of goes hand in hand with what you just talked about, the legacy code. Baird: Yeah, no, and, and I agree completely with that. I find I let it try like once or twice with an error because like, if it does it right away and it does it well, it's something that you don't have to like search what's going on. You don't have to dig it. It does save you some time. But if after like one or two tries it's just not getting it, I'm like, okay, I'll take over and run from there. And, and so yeah, it, it's definitely a, you know, and, and this is. True for all of it right now is like, what is the right balance? Because it's also like, I don't want it to do all of the work for me. You know? Like I tend to let it do stuff. I'm just not that interested in like I know it needs to get done, but it's not super complicated and it can be a lot quicker if I just let it like run and do its thing. But when it comes to things like actually creating a decent looking app, like most people say AI generates like. Nice looking UIs, but they look like Tailwind UI or whatever you call it. Like it, it is just stock tailwind. There's no differentiating between anything that it creates. And that's actually another thing with Claude Code that I wanna play with, and I've created it, but I haven't pushed on it at all, is um, you're familiar with subagents. Brendan: Oh yeah. Yep. Baird: So I'm pushing on the idea of. Both like through brand guidelines, design files, stuff like that, creating kind of rules and like a design system that, um. Bot can understand and having a subagent that actually is like in charge of the design. So it's like, you know, go through, check everything you've done as a designer and fix things based off of these rules, and like really trying to play with that. And, and like I've, again, I keep forgetting about subagent, so I haven't pushed on it too hard, but like yesterday I fixed a bug for a site and I was like, oh, you know what I should do? I should just have. Claude Code, go in and write a regression test for it just so that it, it can, you know, make sure that if this happens again, I know about it and. I actually had forgot I created a subagent about a week ago. That is essentially my test writer. And the point of it is like as I create stuff, it should actually run periodically and kind of go in and like create those tests almost automatically as I go to just fill in those blanks because I've never been great at writing tests, but I need to get a lot better and like AI is a place where I think that can happen pretty nicely. But again, I've heard some horror stories where like. AI will actually like mess with your code to make sure that the test passed, even if it's not right and Brendan: was just gonna say that I, I have, I would agree in general, yes. Like that's a good place to even start with AI coding is Baird: yeah. Brendan: tests, but you need to be very explicit when you say write a test and say, and. Something along the lines of, don't change anything but the task code. Baird: Yeah. Yep. Brendan: still have to be very specific about it. 'cause Baird: Yeah. Brendan: I've, I've heard that same problem or thing and it's just like, oh my God. Baird: Yeah, so Brendan: also comes back to you've gotta be reviewing this stuff no matter Baird: for sure. Brendan: because you would catch that obviously in a review. But Baird: Yeah. Brendan: vibing it and you're not paying any attention, or you don't know any better, then you're gonna be introducing lots of problems. Baird: Uh, and that that's such, you know, it's such a, a hard thing because you see, or maybe you see people out there like, oh, I'm writing 10,000 lines of code every day now because of ai. And it's like, that's scary. Like how are you reviewing 10,000 lines of code. every day? Like, is that your job now? Like, are you just a code reviewer and you're going through every line, or are you just like, yeah, I generated it, I'm good to go, like I'm done. And that just. Because again, it, it, everything with AI that I've experienced is like, even when it's wrong, it is so damn confident about being wrong, but it's just like, this is perfect and it's not, you know, like I just, even today with Claude Code as good as it is, there's still times where it's like, well, no, I take that back. It was yesterday when it was doing the test writer for me. There was some issues with my test setup. And it kept being like, okay, I fixed all the issues, we're good to go. And then it would just fail. And I'm like, did you though? And so I'd go back in and be like, oh yeah, no, I, I missed this one. And so you really like, yeah, it makes a lot of things easier, but it also makes things harder. And you know, I've, I've heard there's the perception that AI makes you like 20% faster is like what everybody kinda says that they feel it is. But it turns out that. The only people that actually makes 25% faster are people that actually know what they're doing or that are experienced, but the people that are actually inexperienced actually are 20% slower. Like it actually is bogging you down and it makes perfect sense. Like I've seen it. It just from the loops it gets stuck in, or the confidence that it has done something, but it really hasn't done something. And so you just, you have to be wary of those things. Like it is not a reason to not use it for by any means, but you have to know about it. Otherwise, you're gonna get yourself in trouble. Brendan: Yeah, yeah, exactly. I was trying to look, I thought I'd linked to something else that was similar to what you were talking about there, but I wasn't seeing it, so no big deal. But, um, I think that that's a lot of good stuff and maybe that's a, a good place to stop. Baird: Yep. Brendan: know, just kind of on that part, but. You know, I just wanna check in if there's anything cool you wanted to share before we, uh, sign off here. Baird: Yeah, so, and there's several of them out there now, but, uh, I just came across a, a tool called Mirage, which is essentially like a game engine generator using ai. And I know, um, Google's created one and I think there's like several other that are out there. But I, I'm really curious to see how that develops because like, I've always wanted to build video games, but it's. Really freaking hard. So I've never dug into it that deeply. But it could be really interesting to be able to create, you know, things in that realm without actually having to know exactly what you like. We're, we're gonna do the vibe coding for video games without the experience. So we, we'll, we'll see how that goes. How about you? Brendan: We will be the, the perfect use case. Maybe Baird: mean, we at least Brendan: maybe understand some of the code, but we don't know Baird: I was gonna say, yeah. Brendan: development. Baird: Um, I, for, for the most part, I would say over, you know, the, how many years, 30 plus years of gaming. I, I've got an idea at least of what goes into it, but definitely not an expert. Brendan: Yeah, exactly. Well, the, you know, the thing that, that I have got to cover here is of related. It's not building an own your own engine, but, uh. It is AI assistant. It's basically like Claude Code is my takeaway for Unity. Baird: Oh, interesting. Brendan: watching, uh, I'm interested in game development as well. I, like, we've already established, I don't Baird: Yeah, Brendan: with it. Um, I've been playing around a little bit, like just looking at different things and following like people that are doing it, like indie developers that are, have like YouTube channels and I wanna get his name right. So I'm just gonna double check. I got it here, but, uh. Where is it? Thomas Brush, I think is his name. Um, of course I didn't link to it directly, but, uh, yeah, Thomas Brush. his YouTube channel, I, I think it's his name. I'll link to it just to be sure. Baird: Sure. Brendan: he was reviewing this thing called Bei, which is, I said, it's it from my interpretation and kind of watching him. So his video is cool. It kind of goes through like he. He's already a developer. Uh, you know, he's already making games. I think he's Baird: Sure. Brendan: couple maybe, he had like four different things that he needed to do in his game. I think one was like they were all kind of bug related or like Baird: Sure. Brendan: things he wanted to get done. And so he was using this BEI tool in, in the video to do it. And it was pretty cool. Like not knowing a lot about game development, but just knowing how. AI coding stuff Baird: Yeah. Brendan: It works a lot in the, the same manners. And what I took away, like the difference between say like, you know, using Claude Code with a React app or a Rails app and using it with something like, you know, a game is that this tool, at least from what I, my, I took away from it, like actually takes into account all the assets you have too. Baird: Hmm. Interesting. Brendan: just the code. Baird: Sure. Brendan: of the things that was really cool that he did was it was like, um, he wanted a, uh. Basically like an auto build tool Baird: Okay. Brendan: you know, building a space for his character to play around in. And it Baird: Sure. Brendan: about all the assets so it like, took Baird: Hmm. Brendan: and used them seemingly appropriately. Baird: Sure. Yeah, Brendan: cool. And, you know, not being a Unity developer at all, I was just kinda like, that would be fun to mess around with. Like, I'm Baird: yeah, Brendan: build anything worth Baird: sure. Brendan: really cool to kind of scratch that itch of can I build something? Baird: No, Brendan: anything about game development. Baird: that's actually super interesting. My, my son keeps pushing to teach him how to build video games, and I'm like, uh, uh. Brendan: I think be between you and I. Like any kid that has found out that I know how to program, they're like, build me a video game. Baird: Exactly. It's like, uh, no, no. I've gotta learn how to do it first. And then so like, yeah. That, that would be interesting to, to get AI involved and, and see where we could push that. Brendan: Yep. Yeah, definitely. Cool. All right, well, I feel like that's a good place to wrap it. Baird: Yeah. Brendan: All right. Thanks for joining. We'll talk to you soon. See.