887: Vibe Coding Is a Problem

Is Vibe Coding the future of software development or just a fun way to experiment? Scott and Wes talk about coding with AI, the risks of ignoring the details, and the cool (and sometimes cursed) projects they’ve Vibe Coded. Show Notes 00:00 Welcome to Syntax! 00:31 “Personal Software”. 01:30 What exactly is Vibe Coding. 02:46 The challenges of coding with AI. Levelsio Airplane Game. 04:49 Brought to you by Sentry.io. 05:39 Vibe coding quality concerns. 07:25 Vibe Coding as a learning tool. 10:11 Things we’ve Vibe Coded. Scott’s Piano UI on X. Episode 876: MIDI & Music In the Browser. Wes’ Roomba on X. Wes’ Roomba on YouTube. Hit us up on Socials! Syntax: X Instagram Tiktok LinkedIn Threads Wes: X Instagram Tiktok LinkedIn Threads Scott: X Instagram Tiktok LinkedIn Threads Randy: X Instagram YouTube Threads

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Wes Bos?

  Welcome to Syntax. Today, we're talking about vibe coding. Now, before you shut it off because you're sick of the obnoxiousness of vibe coding, let me give it to you straight about like what we're trying to do. We're going to be talking about like just building random stuff with AI that might be kind of good, but you don't necessarily care. We're going to talk about what it is and whether or not we should be doing this. And some examples of stuff we have vibe coded. I don't know. People, the other words, people have been using this as like, well, like personal software kind of fits into this.

• 0:38Wes Bos?

  You know, or throwaway software I've been calling it is I've built quite a few things recently where I just need it very quickly. Once it served its purpose, it gets thrown away.

• 0:51Scott Tolinski?

  Yeah, so it's funny that you say that personally. And I do think personal software is connected for a lot of people. And I think personal software might be more connected for, I would say, more responsible developers than it is for other people. And if that's the best way to say that, because what I found with vibe coding is that specifically, it's a lot of inexperienced people doing it because they are becoming productive with it. But they cannot be that productive without it because they don't have the knowledge base. That's an assumption, but that's my experience, what I'm seeing on socials. So vibe coding, if you've never heard of this term, you're not on social media.

• 1:35Scott Tolinski?

  You haven't seen people talk about it. What the heck is it? It's basically coding entirely using prompts, things like Cursor or Claude Code, these types of systems that are agentic, that do a lot for you, that are adding multiple files, connecting things, running projects. And the idea with vibe coding mostly is that you, as the developer, are not getting really deep into the code. You're not really paying attention to the close details. You're mostly just running it, seeing if it works, seeing how it works, prompting again, tweaking, prompting again, prompting again, just on repeat until the thing does what you want it to do. It's a form of AI slop. And right now, you can actually make stuff with it.

• 2:19Scott Tolinski?

  You know, believe it or not, people might, if you're not getting into the stuff, you can be surprisingly effective vibe coding, but. Again, it ends up being a lot of slop and I think there's certain instances and projects and things like that that make more sense for exploring that type of way. And sadly, I think the inexperienced developers amongst us might not understand when and where that that line is.

• 2:46Wes Bos?

  Totally. I find myself when I want to get into it and I try to correct a problem, I find sometimes I'm two or three prompts in and I just have to scrap it and sort of start it, start again from that prompt because it just starts making it worse and worse.

• 3:04Scott Tolinski?

  Totally.

• 3:05Wes Bos?

  And it starts adding a whole bunch of stuff. But the, the whole idea with five coding is that you don't necessarily care. And as long as it works, it works. And I'm kind of somewhere in between because it's surprisingly, some of the stuff is surprisingly good. And it's kind of fun, the stuff that you can make with it where you might not have been able to possibly, the people are making games, Levels IO is, is building this like whole airplane thing. They're just building whatever they possibly would want. And certainly that will make people who are good at this stuff mad. But that always happens in every single industry. When something comes along and makes it easier, it doesn't do away with

• 3:49Wes Bos?

  the need for those people. But it certainly makes the intro level sloppy version of something very, very attainable. We saw it with, with drag and drop website builders, right?

• 4:02Scott Tolinski?

  Yeah.

• 4:02Wes Bos?

  Us WordPress devs hated these Dreamweaver drag and drop or these Squarespaces, things like that. We hated that because it, the code that it outputted was garbage, you know? But the reality is that a lot of people don't necessarily care as long as it works and does what the thing that they're trying to do.

• 4:21Scott Tolinski?

  Yeah, I remember the first Adobe Web Builder that I had that had a drag and drop UI. And it's like, oh, I can get something looking really nice. And then it wouldn't work in various instances, or it's full of bugs, or all kinds of weird stuff. I think right now, in 2025, you know, March 2025, as it is right now, I think vibe coding is good for demos. It's good for personal software, which we'll touch on that.

• 4:46Wes Bos?

  Proof of concept.

• 4:48Scott Tolinski?

  Proof of concept. And it's also good for Sentry at Sentry.io, because your code is gonna suck sometimes. It's gonna be terrible. Uh, it's, and you might not know it's terrible, which is, man, that's like, that's Sentry's ideal landing spot is like, uh, you have some bugs in your code and you don't know about them. And like, if you're vibe coding.

• 5:07Wes Bos?

  AI slop running, yeah.

• 5:08Scott Tolinski?

  Yeah, you don't even know, it's not like you don't even know about your bugs. You don't even know about your code, let alone the bugs, which is gonna be full of them. So, uh, so check it out at Sentry.io, solve all your problems that way. And there is actually an AI tool within Sentry too. So maybe at some point Sentry part becomes part of the AI vibe coding flow. And you just click fix it for me. Oh, there's a bug, fix it for me, fix it for me. And then it just fixes itself. So check it out Sentry.io forward slash syntax. Sign up and get two months for free with the coupon code TASTY TREAT, all lowercase, all one word. Uh, this podcast is presented by Sentry, in case you could not tell that. But either way, I do think that it is a major concern for vibe coding specifically. If you are an advanced developer and you're using these agentic coding solutions

• 5:50Scott Tolinski?

  as an assistant, as like a junior level programmer that you're watching very carefully and you're approving all of the code line by line and you're reading it and you're really comprehending exactly what's going on for every single thing that's added to your code base. One, it's gonna be a lot better at the end of the day. Uh, but two, you might actually get very frustrated and wanna do some of it yourself because the, the AI stuff can do a whole bunch of changes at once. And you're like, that's not what I would have done. Uh, and that's a bad idea. Whatever you're doing, it's a bad idea. Let me do it myself, move, it's mine, you know? But other people, I, I think they just, they're just checking to see if that, that shit works and that's it. You know, so I mean.

• 6:28Wes Bos?

  It's, it's really funny because like when the, the AI images stuff first hit, I don't know, probably three years ago, everybody was posting the most hilarious images that popped up, right? And then when the chat stuff popped up, everybody was posting the hilarious. Chat back and forth and then that sort of died out and now that you can just like you literally make a game or build some software. We're having that in as well. People are making fun software. They're building little games and whatnot. And I think that will will die down as people go, yeah, of course you can, you can generate a dog eating a hamburger,

• 7:10Wes Bos?

  you know, like, oh, wow, you know, like, that was, that was the most hilarious, mind blowing thing a couple of years ago. And now you see, like, of course, like, anybody can do that. So I think we're sort of in that stage right now for building these types of things. But not to say that it's just like, it's just garbage as well, because like, I'm thinking right now, my kids want to have wanted to make a math game for a while, you know, they want to make like a flash card, but like, imagine we could build, I thought like, yeah, we could build that. Let's set across some time. And for me to make a little flash card game from scratch, like, that's, I don't know, I could probably do that in six hours or something like that.

• 7:51Scott Tolinski?

  Yes, totally.

• 7:52Wes Bos?

  But I could probably make something with the 3D world where you run around and answer questions. I probably make that in like an hour or so. And now, and that would be so exciting for them. And I think like, that would be a net win because they're excited about computers, they're excited about coding, and they're going to be learning math.

• 8:12Scott Tolinski?

  I know. I was thinking about this a lot with like, you know, my son and I are getting into game development and it takes so long to see any major real change in it, especially because I'm not great at it, that like, he's losing attention. And I'm doing my best to be like, look, we got a guy, he's walking around now. He's not pointing in the right direction. He's not doing the right animations, but he moves. And my son's is like, okay, great. When are we going to get to like, build Mario?

• 8:39Wes Bos?

  Yeah.

• 8:39Scott Tolinski?

  Because he'll play Mario Maker where you're just dragging and dropping everything onto the screen and it's so nice and fast, you get instant results or whatever. And that's what he really wants, you know, out of the experience. And yeah, I do think that instant feedback or visual, whatever, it's great for being inspired. There's all kinds of things. I think a lot of it. Falls apart when you try to build longstanding production code bases. Not that you can't use these tools, but you need to treat them more like a, an assistant that you're, you're really watching closely, rather than vibe coding it. I think about this all the time. I have friends that are like, oh, I made this app that does this whole thing. And I just did it with prompts. It was super easy. Like, I'm going to turn this into a product.

• 9:23Scott Tolinski?

  And I'm thinking like, all right, well, do you know anything about authentication? Do you know anything about databases, security? Do you know anything about anything? And the answer is no to all of those things, right? And without those answers and what the, you today probably can't confidently ship something that is good without those, those tools and knowledge and, and you know, there, there's still that gap there. So yeah, vibe coding your production, for now at least, right? Vibe coding your production code is only going to be, I'm going to say it's not good at all. Because again, vibe coding kind of implies that it's, you're not even looking at the code really, or you're not getting deep into the code. But yeah, for throwaway, personal software, projects, demos, inspiration.

• 10:07Scott Tolinski?

  I think it's, it's totally valid and, and works just fine for that stuff.

• 10:11Wes Bos?

  Totally. I, I think I'm going back and thinking about the stuff that I've built in the last little while. And quite honestly, I don't even remember half the stuff that I've, I've coded because I just, I build it, it does this thing and I throw it out. But yeah, we'll, we'll throw an image up on the screen right now is I'm building a, a parametric grid that I'm 3D printing and I'm, I'm going to be hanging this grid in our gym. And then you can put all these hooks into this grid and hang up all the different attachments for the gym. And normally this stuff is you just print off a sheet of it. You print off as many as you want and they all click together and you're, you're up and running. However, because I want to put a nice border around it, I realized it's much more complicated

• 10:53Wes Bos?

  because I need a top right, bottom left, like it's basically the sliding panels of CSS in real life. I need the, I need the corner pieces. I need the tops, the bottoms, the lefts, the rights, and the internal pieces, nine different puzzle pieces. And I was like, man, I'm having a hard time visualizing this. So I started like writing it, scribbling it down on a piece of paper to try to visualize what I needed. It's a four by six grid. And I was like, you know what? I'm just gonna start prompting my way into this. So I built a little thing that I said I have a four by six grid.

• 11:30Wes Bos?

  And I need to define what each of the edges are called, right? Top left, top right, edge right, edge left, et cetera, internal. And then I just, I just kept saying, good. Now visualize it. Good. Now make me a cut sheet of how many need to be printed for each piece. And like, that took me like 15 minutes. And I was just like, okay, I, I never would have, not never, but that would have taken me much longer. And I don't have a nearly as nice of a visualization. If I were to build this myself, I probably would not have built it myself. I probably would have made a spreadsheet or, or a piece of paper or something like that.

• 12:08Scott Tolinski?

  Yeah. And to be clear to, to people listening, it's not because you cannot do that. You are very capable of making this interface.

• 12:16Wes Bos?

  Yes.

• 12:16Scott Tolinski?

  A grid or any of this stuff. It, the, the benefit really comes from the speed and effort that it takes. That's really where, where you see the benefit. Like for instance, um, we had my whole MIDI controller thing. And like, yes, could I have absolutely coded a MIDI piano UI, which is the keys are all, you know, they're the way the piano keys are shifted. Or I can ask Cursor to create a piano for me aesthetically in CSS and HTML. But I can guide it. I can say use Grid or Flexbox for this. I could use this or that, give it this aesthetic, whatever. You can guide it with the technology. You can tell it the approach to use. You can give it hints and stuff like that, but I didn't vibe code that entire app.

• 13:00Scott Tolinski?

  But I vibe coded the keyboard interface because I just flat out didn't want to take the time to do it. And it's again, it's just a, a for fun project, right? So I do, I do really find that, that to be like a nice little sweet spot. It's something that you could do. You know how to do it. You know what it looks like when it's done right. And you can just ask it, save me some time here. Do me a solid.

• 13:23Wes Bos?

  I think another area for this sort of vibe coding is you're you're going into things that you're not necessarily comfortable with. So the the Roomba, I wrote a whole bunch of C++ I think I forgot what language it was, you know, and I think it was C++ and I was debugging serial connections and and sort of going through that and and I built the whole UI to control the the vacuum, the Roomba thing. I have a video on, you can go watch it. I built that whole whole thing. And that was, I guess you call that vibe coding because I, if you look at it, it's it's not ideal. There's a lot of like weird react hooks and stuff, but but it

• 14:07Wes Bos?

  worked. It worked really nice and it was able to get it up up and running really, really quickly. So both languages where you're not really comfortable with or areas of stuff you've never worked in before and then UIs and things that you need to just crank out as as quickly as possible. Because because you're just learning and it's not to say that I didn't learn anything. I think I came out of that project way smarter than than I went in, right? Like, I know so much more about microcontrollers and I feel so much more confident in C++ now, not that the robot can write it for me, but like, I feel like I understand a lot more about it. And I don't know that I would have gotten that understanding, reading a book, but controlling

• 14:51Wes Bos?

  a vacuum, I would have certainly got there.

• 14:55Scott Tolinski?

  Yeah, it is. It is a whole thing because like, I think there is some degree where vibe coding is stupid. If you don't know what you're doing and you don't know what the code is is doing, and you're not learning anything from it, and you're just pushing out more slop into the world. I don't like that. I think that's bad. But if you are using it as a learning opportunity, you're using it to excite yourself, you're learning it to make things be productive, all of those things, I can get behind that and I can see a benefit to that. So yeah, different.

• 15:24Wes Bos?

  I think I think of like, somebody like a kid right now, trying to build a game. And and they hit problems and they're trying to debug it and they're typing in and they're reading the replies. Ah, it might be because we have a race condition here and I go, oh, that's what a race condition in this is. And we solved that. And there are so many different things you can learn very quickly, just building these silly little throwaway applications. So as I know, we had a lot of angry people, a lot of angry replies about when we do AI stuff on this podcast.

• 15:57Scott Tolinski?

  Yeah, totally.

• 15:57Wes Bos?

  But I think I think it's a net win for when you're trying to learn something and you're trying to just you either trying to fix a problem that you have for your own personal stuff, or you're trying to learn something new. I think it's a net win with this stuff.

• 16:11Scott Tolinski?

  Yeah, totally.

• 16:12Wes Bos?

  All right. Anything else you want to say about vibe coding? Or do you just want to vibe off?

• 16:18Scott Tolinski?

  I want to vibe off, yes.

• 16:24Wes Bos?

  I don't even know what else to say about it, but that's it. Let us know what you have vibe coded. Love to hear it. It's it's great to great to see everybody's replies.

• 16:33Scott Tolinski?

  Yeah, totally. Love it, hate it. What do you think?

• 16:35Wes Bos?

  Peace.