Vibe coding a 3D multiplayer game in 15 minutes—with no game dev experience | Cody De Arkland

We're talking about games and we're talking about building games here. But the same thing translates really well to like when you're building actual applications too. A lot of times you're starting with this blank framework and you're giving it like a broad idea of the thing you want to make and then you're diving into these individual features. What are your sources for figuring out how to scaffold with existing technologies? A lot of times I'll ask the AI if I wanted to build a game and I wanted it to run inside of a browser which technologies make the most sense. but then picking those things out and starting to do deep dives on it like traditional Google and feeding that back into LLM. So it almost becomes like a conversation with another developer where you're like, "Hey, I learned this thing from the internet. Can you implement this in the game?" I just think that's a really interesting process that nets out net positive. We're in this time period where everyone can go and do this. My kids have sat down and started playing with building games and things like that. So we're going to speedrun vibe coding this game. We are going to speedrun. I want to build a flight simulator. I want turning to bank the plane and arrow keys to control pitch. And there's our game. Hold your horses. You wrote like 27 words into this prompt and now you have a video game. Welcome to How I AI. I'm Claire, product leader and AI obsessive here on a mission to help you build better with these new tools. Today we have a very fun conversation with Cody Dearkland, senior director of developer experience at Century. Cody is one of the most prolific Vibe coders I know. Doing everything from building personal to-do apps for his family to automating just about everything you could automate at work. But today, we're doing something extra fun. Cody's going to speedrun building a 3D multiplayer game live on the show. Let's get to it. This episode is brought to you by Interprint. Enterprit is a customer intelligence platform used by leading CX and product orgs like Canva, Notion, Straa, Hinge, and Linear to leverage the voice of the customer and build bestinass products. Enterpris unifies all customer conversations in real time from gong recordings to Zenesk tickets to Twitter threads and makes it available for your team for analysis. What makes Interprint unique is its ability to build and update a customerspecific knowledge graph that provides the most granular and accurate categorization of all customer feedback and connects that feedback to critical metrics like revenue and sees. If modernizing your voice of the customer program to a generational upgrade is a 2025 priority like customercentric industry leaders Canva, Notion, and Linear, reach out to the team at interpret.com/howiia. That's e n t e r p r e t.com/howi ai. Okay, Cody, I hate to admit when people vibe code harder than I do, but I believe that you are one of a very few set of people who do vibe code more than I do. So tell me the the truth. What is up on your screen right now? What are you working on? Oh god. So, I tend to look at the different tools as like little junior developers who are helping me work on different things. So, I tend to keep a lot of tools up at any given time because they're all working on different parts. And so, I'm going to go and just share my screen and we're we're going to take a tour of what's running on on Cody's desktop. Share. Maybe I'll keep a running tab if I'm running the same things. That That's fun. All right. So, I was in I was working on a little a little bit of a page about like web performance. That's a little bit of a work test. So I I work at Sentry and so I was working on uh some things for real work there. Behind the scenes though, I've also got uh Cursor as well as Wind Surfera because I really can't decide which one that I like more at any given moment. So I just use them both and I use them both often. So I've got both of those. Uh I was doing some work inside of MCP and so I have clawed up and was doing doing some things inside of inside of there. Inside of Windsurf, I was playing around with the the game I built, which I think we're going to talk a little bit about a little bit about later on. Cursor is dived into a little application that I was building for, you know, testing testing sentry things out. And then again, tons tons of stuff in Bolt, both things like that are workrelated, but then also diving into some just like personal applications that are just fun for for home productivity. So, what is that? Windsurf, cursor, claude. Uh I I have Claude Cod open inside of inside of my terminal. Yeah, we're all over we're all over the place. We are truly unhinged in AI today. Yes, I think you you have me beat at least at least right now. I'm not currently coding while doing doing this podcast. So, you are right. I actually want to talk about something a little different today because I've seen a lot of your work product generated by AI. I've seen some of your personal productivity tools and what we call meme apps. These like little micro apps that you've built with Bolt, but the most impressive AI thing I've seen you build is your space game. So, I would love for you to show us the end product of what you built with AI, and we're going to work back into how you learned how to build that. Sure. So, we'll go we'll hop right back into sharing. And so I I built this fun thing and you know I've always I've always enjoyed kind of like flight simulators and there was this moment in time in in social recently where there's people popping up building all of these like plane simulator apps and I was like man I really want to fly in space you know I I I loved playing X-Wing back in the day. I loved playing No Man's Sky and I was like I bet we could make something like this. So I started playing around with h how could we do this? And so this is the game. This is little game called Spaceflight and anyone can go and play this. It's on spaceflight.gg. And so you can go in, you can pick your name, you can pick the ship that you want to fly around as, you enter into the game, and then you're in. And so this is all multiplayer, too. And so that wasn't it wasn't initially, but it it has become multiplayer now over time. And you know, it started off as just kind of like a random idea of I can fly around space. How could I make the scene? And then how could I how could I make this thing look like a ship? Well, what if I really want it to look like a ship instead? And so it it kind of grew uh sort of out of control, but there was a lot of a lot of fun that happened along the way to get there. Okay. I joined the game as well. So it proof is it is online and it is multiplayer. So I'm in it. Okay. So you built this game which when I was growing up all the only reason I got into computers is because I wanted to make video games. But tell me, have you done game development before? Like this is very impressive. No, honestly, you know, I think I I played games growing up a ton and I was always pretty intimidated by going and trying to trying to build one out and so I just I played them and so this was my first foray into trying to trying to get something together that I could actually go and go and play and and have other people play with me and see where it goes. Okay. So, how did you learn how to develop this space game, which still totally blows my mind? You know, I think we we we joke about the whole vibe coding thing, but it this was this was probably my first real experience in like truly vibe coding a thing, right? Like, and I jumped in and I was like, I want to I want a flight simulator in space. And so, it it ended up producing this thing that had like the starry background kind of like what you're seeing here. Uh, but ultimately had like a gray cube in the middle that was that was supposed to represent a spaceship. And and really it just ended up being this kind of this back and forth asking of like, I want to change this part a little bit. I want the stars to move when I when I go forward. I need to handle controls. And there was all these things that were like that I didn't really think of before I started the project out as far as like what is movement like in three dimensions in that case like how do you handle that? How do you handle which way you want to move at any given moment? I had this idea then like well there's tons of 3D models out like for example I was on Sketch Fab and I started looking up different different spaceships and I was like well these are all pretty cool. I imagine I could bring these in. And so I went in and asked asked cursor to to go in and and set these up. And I started reading more about what it was actually doing. And I think that's like maybe a little bit of a different approach that I take is like when these things when it starts going and building this stuff out and I see like different technologies being used, I'll go in and I'll start asking it about how does this work? So like as it started talking about glTF models and GLB models, like how do these work? How are they actually brought into the game? and start trying to learn a little bit about like what it's actually implementing so that I can use it use it also and so like it set up the scene in 3JS 3JS is a pretty well doumented out platform overall and so there's a lot of information out there it built out the scene I told it I wanted to replace that gray cube with the with one of these models instead and we went back and forth on that a bit because there's a lot of there was a lot of nuance in making that happen right for example like when you pull in these models the game doesn't know what that model actually looks like. It's just a skin over over a thing. And so figuring out how to like position it and how to give the game an idea of what forward actually is because some of these would load up sideways. And so I'd have to tell it, oh, that the front of the ship is actually 90° turn to the left horizontally because then I do 90° turn and it would just rotate at 90°. And so you it's a lot it's like an interesting lesson in being really specific with the models and how how you want them to behave and how you want to um have them like act or how how you want them to bring things out inside of the game. But then once you get it done once, you can say, "Hey, apply this same logic to my next ship." And like I have a local version of it that's running that has, you know, five different ships in it that I just haven't put into the into the real game yet. Uh because like it's really easy to prototype those out more and build more now. So you did you start this from scratch in cursor? Like walk us through how you really got from an idea to here. What tools did you use? Where were you doing that learning? You know what? Let's have some fun. Yeah. Let's start building one a little bit. Okay. Let's just do it. We're going to build a game now. We're going to we're going to actually build one. We talked about doing this. Now we're actually going to do it. Okay. So, if I wanted to start this from scratch, I'm gonna go mpm create and I'm gonna give it a different name. We'll just go instead of flight, we're going to go boop flight. So, you're starting a blank basically react. Yep. Empty React project. Great. And so, I'm going to come in here and I'm going to go into this Boop Flight project now. Oop, not Boot Flight. Boop Flight. And now I'm in this project. Now, one of my favorite tools. It's kind of the upandcomer right now. I know you love it, too. We're going to go right into Claude, but we're going to live dangerously, and we're going to skip all permissions. So, it's just going to yolo all of this into place, and we're going to see how it goes. Great. So, we're going to speedrun vibe coding this game. We are going to speedrun. I'm going to have it init. And what this is going to do is it's going to start bringing in that initial context of the project as part of like Claude's understanding. I think it's going to create a claw.md file that's going to make it understand what the project structure is, what application or what a framework it's using. So, it's just good to like start off uh the project with like good understanding, a good context around this. And like I really did start from just an empty an empty V project. And you know, I think like we're talking about games and we're talking about building games here, but the same thing translates really well to like when you're building actual applications too. A lot of times you're starting with this blank framework and you're giving it like a broad idea of the thing you want to make and then you're diving into these individual features. And so like I built a lot of that inside of like real like company level software. I took those same lessons into how I was building the game. Like I had this broad idea and I just started working on the individual features of that idea and building it out. And so we have this broader broader game here now or broader uh project overall. So, let's say uh we're going to do a flight simulator again, but we won't do space this time. What we'll do is uh I want to build a flight simulator game that uses 3JS. Let's use polygon style art for the planes. WSD for movement. I want turning to bank the plane and arrow keys to control pitch. Oh, look at you. Kid raised on flight simulators. Kid raised. Uh let's see. Let's um let's have it take off for my green field with trees and other objects around. We're going to be super vague. I have no idea if this is going to work out well, but we're going to see what happens. And real like this whole idea, I think people struggle a little bit with like the vibe coding thing because a lot of times they give it far too big and too specific. I try to stay pretty wide in like what's the broad strokes of what I want this thing to look like because I feel like that gives that gives the LLM enough room to actually build something that works and then I can tweak the individual parts as I go. And so like in my head I literally think through these chunks of the thing that I'm going to improve and I just want to see like what's the the vzero prototype of this thing that is the earliest version and how do I start making progress on it. Again, the same thing translates to like the productivity apps. I built this like tracking application for like tasks around the house that I could share with share with my spouse. And so like the same thing applied in there. I want an app that tracks everything I'm doing. I want to create subtasks off of that. Now I want to integrate, you know, superbase for authentication. Now I want to have a backend a back-end database. And like this gradual step of complexity over and over again. It has always been a much more approachable way for me to handle like building things out inside of these applications. But I see people pop in sometimes and they do these like very detailed project level plans that are like every single stage and they feed it all into the context in one shot and then they get frustrated when like the AI doesn't produce the perfect thing at the end of it and it's like well you gave it a mountain of tasks that are hyper specific. Let's try to break it down a little bit. So, let me ask you a question because one thing I will say is you may not be a game developer, but you're clearly a developer and that you know what libraries to ask it to install. You know, some of these components and I'm just curious, how are you learning for anyone vibe coding project? What are the libraries that people are using? Like what are your sources for figuring out how to scaffold with existing technologies? I stay fairly connected in like developer communities overall. But something else that I've started doing a lot more lately is like going and a lot of times I'll ask the AI for like the game one is a great example. If I wanted to build a game I I remember actually doing this. I went and I said if I wanted to build a game and I wanted it to run inside of a browser which technologies make the most sense and like it'll go through and tell you like in this case it was like 3JS as an option when I did the multiplayer. That was another one that I did this on was I want to implement multiplayer inside of the game. what's the best route to do this? I think websockets might be the right answer, but are there alternatives? And it and it said, you know, websockets are are an easy one, but and allow for great customization, but also have the highest uh the highest amount of work associated with them. Here's a couple of other open source libraries people use. since like that's one way to learn, but then picking those things out and starting to do deep dives on it like even outside of the LLMs, just going in, you know, traditional Google of like building scenes in 3JS, how do people approach this? And and taking that and feeding that back into the LLM. So it almost becomes like a conversation with another developer where you're like, "Hey, I learned this thing from from the internet. How do you can you implement this in the game?" And like having it tell you if it's a good idea or not. A lot of times that's another thing I'll I'll do is I'll ask the game or I'll ask like the LLM is this a good approach like I've implemented websockets in this way because I wrote like the first part of websockets myself just because I've done a lot with websockets in the past and when um when I got done with it I went back and said hey can you look over this implementation and tell me if this is right and it caught a few things where it's like oh you're you're having a race condition here you're not optimized here you're not disconnecting sessions appropriately here and so I really do look at this as like another developer that I'm handing off tasks to to see how they how they end up going. This episode is brought to you by work OS. AI has already changed how we work. Tools are helping teams write better code, analyze customer data, and even handle support tickets automatically. But there's a catch. These tools only work well when they have deep access to company systems. Your co-pilot needs to see your entire codebase. Your chatbot needs to search across internal docs. And for enterprise buyers, that raises serious security concerns. That's why these apps face intense IT scrutiny from day one. To pass, they need secure authentication, access controls, audit logs, the whole suite of enterprise features. Building all that from scratch, it's a massive lift. That's where work OS comes in. Work OS gives you drop-in APIs for enterprise features so your app can become enterprise ready and scale up market faster. Think of it like Stripe for enterprise features. OpenAI, Perplexity, and Cursor are already using work OS to move faster and meet enterprise demands. Join them and hundreds of other industry leaders at workos.com. Start building today. Okay, so where are we with this game? It looks like it wants to run. All right, so it looks like it finished up. It finished building. Everything's good. So now I just need to go and run it locally. We'll do that. and we'll launch it. And there's our game now. Shut up. Hold on. Hold your horses. You wrote like 27 words into this prompt and now you have a video game. Now, there's a couple things though. Like, it looks great. This is fun. I feel much joy. But what we I see on my side is if you look, I'm banking to the right. So, I'm I'm pressing D and it's turning opposite. Also, the tail is on the back end. So, the camera is actually looking at the fly at the at the plane. So, instead of being behind it, I'm on the front of it. Do you see what I mean? And so, this is a good example of going back in. We'll go back into here and we'll tell it looks like my camera is facing the front of the model. Now, this might not work right because like the front is a is is more of a of a human term in this case. We say the nose of the plane. This is my liberal arts coming into play. Yeah. Yeah. The nose of the plane. We should be looking should be fixed to the tail of the plane. Can we I don't want to zoom out yet. I also feel like it's a little too close to be We need a little bit more We need a little bit more range here. Look, it even gave you a little um guide to the flight controls. You didn't ask it to do that. So, it's making some changes. I love I love that we can watch it actually happen here. What is actually updating? Cloud code is the ultimate like vibe coding tool. I swear I definitely can tell that it's reversed because when I go forward, the plane is flying at my screen as opposed to away from my screen. So, it's definitely that the camera is totally reversed. What I love about Claude Code is you can see your tokens go really use up those uh token fees. I really need I need this game to take off so I can make like just $100 a month for people so I can pay for these tokens and not get yelled at for spending my household funds on AI every single month. So, looks like it's rotating the camera here. So, we're getting a different look direction. it's making these updates. And so like what I think is really interesting about this is this is all running. Now something else that I think is like a not so often used hack is I'll open up a new tab sometimes. And so in this case I'm I'm still in Boopflight. I can make a directory for server. I can switch into server and I could say npm anit. I can install express in here. And so like I could easily go back up a level now, open up a second claude instance and I can start in instrumenting like the multiplayer stuff that we talked about. And so because I know just from the previous project that the websockets works really well, I can say I want to start implementing multiplayer for this game in the server directory. handle player joins and give me a chat interface on the top right that shows when people join the game. So you have dueling clawed code going right now working on the front end front end uh just setting up the framework the visual framework and interactivity of the game and then the back end um setting up multiplayer. Okay. Oh, we got a tail now. But now we're now see but this is goes back to what I was talking about like that things don't always work the way you expect it to. So now instead of being slightly above I'm slightly below. I'm I'm still reversed. So when I go forward I'm flying backwards. So visually we look right but all of our controls are reversed now. And so in reality what was going on is like the camera was not mispositioned. The camera was right. It was that everything else was positioned wrong. So now we go back in and we tell it all of my flight controls are reversed. Now we should fix that. I also want the camera to be slightly above and behind the model. Okay. And so let's check in on your multiplayer now that we're uh making some some modifications. Okay. So it's it's setting up some it is setting up the game itself. Great. are setting up the multiplay or the multiplayer stuff. Oh, we just dropped a whole bunch of stuff in. So, we can see the web sockets, right? So, it's going in and saying like when someone joins a message, what does chat look like? When the socket detects someone move, what does that look like? What does it look like when someone disconnects? So, it's building out all of this stuff now. Uh, it's adding in the dependencies for socket. Our tokens just keep climbing. Things just keep going. So, we're still fixing things there. And so, like I really am having, look at that. Look at us. And we have better and we have our So our flight controls are a little wonky still. So we're still we're still we're banking right, but we're turning wrong. So that's a fix. But visually we're much better off than we were before. It looks amazing. I know you want your kids to play it. My kids are definitely definitely going going to play this. Okay, so let's I think this is blowing my mind in terms of quality of gameplay. I know that you have higher standards for your game controls, but we can let that rest for a little bit. Yes. So, now you're setting up multiplayer, which would be another aspect to this game, which I think would be very complicated to set up and prototype, just taking a step back. Totally. How long do you think just writing this code would have taken? You know, I I think that like writing it completely from scratch is probably at least several hours to get to a good like baseline of what this is and especially like without sitting down and drawing up like diagrams and starting to think through what all of that looks like. Just getting it to a point of of being able to connect, have multiple players be able to connect, be able to see how they act inside the game. Like there's a lot of architecture choices inside of there that you would make as you were building it. So, I'm going to say at least at least a few hours. And we have at least a Vzero prototype of it coming in coming in pretty quickly right here in just a few minutes. This is amazing. We now have a we have a flight strip that it went and implemented. I didn't even tell it to implement these mountains. What is What is all this? What is all this stuff in the distance growing up? This is how Skynetut starts. I'm telling you, this is very very cute. Okay, so before this turns into a Twitch stream of us just playing your very fun flight simulator later game, let's check in on multiplayer. Did you make this a real game? Well, we're going to find out. So, looking through. So, the run finished inside inside Cloud Code. So, it made all the changes that we asked for. What's really cool is like it gives us some nice like documentation along the way of of what's actually happening here. So, I I love that we've got it up and running behind the scenes. So, our game is our game is up. the multiplayer screens up. I'm seeing some joins which are like the browser I'm guessing the actual like cloud code browser session trying it out. So we got we have good signs. We have we suggesting that things look good. Let's jump in. It changed more. Look, there's mountains in the distance as it was correcting the other things. Okay, so you joined. You're you're there and then Oh, and I see I see multiple players. So now let's open up another tab maybe. Oh, there's three players in the game now. Claire's in the game. Now we I mean obviously we have some things that are broken here, right? So like this should be down here on the bottom should be the top of the screen. It's like I can I can mouse wheel scroll. This is a new feature. We're stuck in camera views. It's a It's like a cockpit view. So like these are some of the problems though, right? Because like I didn't ask it to implement this and so like it implemented that anyways. But we are seeing Claire we're seeing like the chat the chat window is a little broken. These are things that we have to fix, but ultimately like think about how long it would have taken to learn how to build this from total scratch, implement all of this, set up the multiplayer. Like this would have taken several hours if not days to go and do just for like a hacky fun prototype. And we did this in what 15 minutes total. Like we got long ways in a short period of time. This is this is amazing. Okay, if I know you, we're going to end this podcast and this is going to be your weekend project. if you'll send it to me and my kids can start playing, which they're already doing with this space game. So, I I understand that we could iterate and fix some things. It seems like, you know, part of why I love this demo is you've shown us some of the frustrating moments of working with these LLMs to code. And it's not always perfect. You know, you take two steps forward, one step back. You add multiplayer, but you break the window. You add zoom, but you accidentally zoom through the airplane. And so, um, I just think that's a really interesting process that nets out net positive. What What do you think about that? Yeah. You know, I think, uh, one of the the things that's like worth calling out here is like I talked earlier about the problem with people throwing too much plan at it, but I think that like a little bit of plan is a good thing and some rules and guidelines around it are a good thing. But I think like again the iteration is what really matters here and like how far we were able to get in such like a short period of time. like as far as like creating velocity and creating speed. I think that's like the really cool thing here is like yeah, it wasn't perfect, but we got it doing the main things that we wanted it to pretty painlessly. Okay, so let's do a quick lightning round and then get you back to your game development. So question one, how do you strike a balance between sort of this like fun exploratory stuff, building games, learning totally new technologies, and figuring out how to apply AI in your kind of professional day-to-day? You know, I think um the thing about this is that it really isn't that much different from the workflow that I do as like day job stuff as far as like building building software, trying things out, trying new things to hack out. I build a lot of repros off of off of AIS like when I work with customers and things like that and so like the skin might look different's a and one is inter enterprise SAS software developer SAS software but uh ultimately like the workflows end up matching a lot and so I'm fortunate that the things that like I enjoy doing are very similar in in a way but I think like the answer is really like intentionality right like I started building the game because I wanted something fun to be able to show kids and show my kids show your kids to play with have will be able to go in and have fun with that was different from building just another another web app. And I have to like be really like conscious about choosing which one that I want to do. But the workflow matches up in a lot of the same ways. You know, the whole taking a broad concept and breaking it down into the things I want to build inside of it and iterating on it. So it feels very similar, but it's just choosing what am I going to what am I actually going to do? Am I building the fun thing or building the the work thing? And they overlap sometimes. And sometimes the fun thing is as we as we showed a little frustrating. So, I'll wrap with my favorite question to ask, which is when your vibe coding AI assistant is constantly breaking things, stuck in a loop, or adding features you did not ask it to add, what is your tactic for getting it to listen? I mean, I just scream at it. Why are you this way? No. No. Um, I tend to ask it a lot of times. Hey, can we start over? Can we start fresh? If we get like far off the path or far off of what I wanted to build, hey, we've been wrong for a while. can we take a fresh look at this problem? Here are the main requirements. And I'll like list them out as like here are the main requirements. Let's take a fresh look at implementing these. Sometimes I'll call out like complexity. Like, hey, I feel like we've layered on a lot or I'll literally say, we've layered on a lot of solutions. Can we simplify this in some way and start over? Um, I like to also like early on set up like a good cursor rules file or have like windsurf memories in place. Windsor has it own it own rules files too but like I find that the winds surf memories work really really well for this and like set big guidelines around really what I really want it to focus on but that doesn't necessarily help when it gets squirly to your point off off the off the beaten path. So a lot of times just being very clear with it and very clear about hey let's start over let's reset this part and also like taking these bigger problems and scoping them down smaller and so like if we get way off the path let's take that problem that's way off the path and let's drill it down and say like I want the menu to be here like the chat window in the other and what we were showing a moment ago was way off the screen. I'm going to go back in and I'm going to say hey the part of the upper part of the game is moved to the bottom the chat windows off. Can we fix those? and I won't I'll ignore everything else and just focus on that to keep the context window focused. So, just being intentional, talking to it, giving it clear directions and clear expectations. I I hate to tell you this, my friend, but it sounds like you're a product manager. Oh, god. How dare you? Okay, Cody, this was so fun. I think you're going to inspire a bunch of people to build their own 3D multiplayer game in 15 minutes. I think we speed ran it. So, where can we find you and how can we help you? Yeah, I'm I'm on X, so at Cody Darkland, I'm I'm I'm on there often. I live live digitally online, especially when I'm building building the games and stuff. As far as helping me goes, like I get a lot of inspiration out from seeing other people tell these stories. And I think like going in and learning how to build something new. We're in this time period where everyone can go and do this. Everyone can like my kids have sat down and started playing with building games and things like that. And so the thing that would help me the most is just being excited about going in and building and like sharing the software you're building, the fun games you're building, and sharing stories like that. There's enough things going wrong wrong in the world that like being excited for this new world we're in where anyone can come in and build and build cool things is just really inspirational. I draw a lot from that. So share it with me, tag me in it. I'll I'll hype it up. You'll hype it up. It'll be it'll be a lot of fun. Well, cheers to that. Thank you so much. This was really fun. All right. Thank you. Thanks so much for watching. If you enjoyed this show, please like and subscribe here on YouTube, or even better, leave us a comment with your thoughts. You can also find this podcast on Apple Podcasts, Spotify, or your favorite podcast app. Please consider leaving us a rating and review, which will help others find the show. You can see all our episodes and learn more about the show at howiaipod.com. See you next time.