Subscribe
Share
Share
Embed
Founder Reality with George Pu. Real talk from a technical founder building AI-powered businesses in the trenches. No highlight reel, no startup theater – just honest insights from someone who codes, ships, and scales.
Every week, George breaks down the messy, unfiltered decisions behind building a bootstrap software company. From saying yes to projects you don't know how to build, to navigating AI hype vs. reality, to the mental models that actually matter for technical founders.
Whether you're a developer thinking about starting a company, a founder scaling your first product, or a technical leader building AI features, this show gives you the frameworks and hard-won lessons you won't find in the startup content circus.
George Pu is a software engineer turned founder building multiple AI-powered businesses. He's bootstrapped companies, shipped products that matter, and learned the hard way what works and what's just noise.
Follow along as he builds in public and shares what's really happening behind the scenes.
New episodes every Monday, Wednesday, and Friday.
George Pu (00:00.408)
So there's one topic that's been on my mind that I couldn't get out of recently has been vibe coding. I think that there's this guy who had mentioned the term vibe coding, which basically means that, know, coding should come from a vibe, right? You don't have to understand anything about coding. You just let your AI co-pilot, whether it's like cursor or windsurf or clock code to take over and be able to be the main driver for you. I think that's what a term vibe coding means.
And Vibe coding has since I think from the beginning of this year became a really huge phenomenon. And Vibe coding is basically when you use AI to code based on vibes, right? Rather than a clear understanding. So you know what you want, AI knows how to build it, but neither of you knows how it works.
George Pu (00:56.846)
So that's what I think basically vibe coding means. And from my personal experience, I've got into like, I've gotten into love with vibe coding, or love of vibe coding. And that was basically the first couple of months of this year. And ever since last month, my relationship with vibe coding has collapsed. So I just wanted to share that experience with you. I've also shared earlier this week on my Twitter about as well. So basically when the, when the vibe coding just came out, it was an aha moment for me. Right? So when.
When cursor first came out, my first reaction was, okay, this is really going to change industry. This is really going to change the world. And I'm sure many of you listeners are feeling the same way. When you first get your hands on cursor and coming as a, I will say myself as like a semi-technical person. I don't code every day, right? But I read code and I work with developers who code. So basically these few months ago, I got my hands on clock code.
And I couldn't get my head out of it. And all of a sudden, you know, obviously I jumped on the train of paying clod basically $200 a month. I'm jump on the max plan was 20, which basically means I have 20 more usage compared to the normal clock clock users. And I personally have found it to be really useful, right? I, I use it to rebuild a website, the entire marketing website of simple direct, which was rebuilt by basic clock code.
in less than a few hours, right? And actually the most amazingly, I learned the coding patterns and learned the design patterns. I learned Astro, which is a fairly new design language, right? And it was able to use front-end language and it was better to use it quite well, which is shocking. And I also use it for some of my backend operations. So I'll share a few projects I've worked on. First of them is the Simple Direct Financing mobile app. So we had this mobile app.
And it has been on the backlog for a while. We have this mobile app, we barely see our users using the mobile app. Most of them are using the web portal to make requests. So when we first go the mobile app, was 2021. And then we made a small improvement in 2022. And then ever since, has just been sitting there dormant. So it hasn't really been used by us. And then basically what happened after,
George Pu (03:13.519)
was that I asked our designer to give entirely new design for the app and then I think we just let it slide there. Like we didn't do anything for a year and a half, right? The design was just sitting there, it's not doing anything, it's not getting new features added and our customers are still using the old designs, right? And a year and a half forward, I was like, okay, now I have cloud code, why don't I just give it a try and see how it works out? So.
Very elegantly, I was able to get Cloud Code basically just to feed it a Figma designs. It's still a little bit bad at that. So I basically just gave it a Figma designs. I give it the dev mode. I copy the code. I paste that, take screenshots, I feed it. I'm sure that's not the best way to do it, but that's how I communicated with Cloud Code, by giving it my Figma file. I know now there's like MCP servers or whatever. So if you know how to do it better, for sure on the show.
or get some of your tweet about how you do it and I'll basically share with the community as well. So it was able to iterate quite fast. I think I spent about two days maximum, right? And I just gave him a front end tasks. I'll basically look at, hey, this is the app. Can you take the Figma design, right? Use it and basically make this app look exactly like how it's looking in Figma.
And surprisingly, it did that really, really well. think I did obviously have to tell it to change certain things. But within two days, like I said, it was polished. It was new, right? The app is like reborn. It's a complete new look. So I'm really happy about the project. Still today, I think it solves the problem, right? It solves the problem that this project or this feature has been in our backlog for a year and half. And without Cloud Code or these like, you know, vibe coding tools,
It basically probably wouldn't even see a light of day for a long time to come. So it's able to move a Dorman project into prominence simply by just giving it life. So I really appreciate that. And I think if you're building a demo, if you're building a prototype, if you're something that you want to release really quickly, it's definitely the way to do it. However, I think if you're building production level applications, or if you're looking to make money and make your app commercially available,
George Pu (05:27.201)
It's definitely not a good idea to just rely on vibe coding. Right. And then I think recently we had this like T app that's been completely exposed, right? The T app is basically like an app that I think the women uses, like young women or women uses to basically spill the beans about, you know, certain toxic males. And I think just like two weeks ago, there was a huge hack, right? Because how this app, this whole entire popular app, which is like a number one on app store in the U S
was basically using a very simple database on Firebase, right? And hackers were able to basically retrieve that without even having to hack the application. And that's how simple it is to hack it. And I can, based on my experience, I think probably 99.9 % of chances that this app is basically vibe coded and the vulnerabilities here, like people's driver sizes were stolen, people's addresses were stolen. And it's just been very bad. I'm not sure about what happened after, but...
we can all expect a lawsuit or it gets escalated to FTC. So nobody wants that as a developer, as a builder, you don't want this type of distraction. And that's the type of fear I have with my own application. keep thinking about, so basically after the success of the mobile app, I moved on to another project which has also been sitting dormant for a while. We're basically trying to build this like private AI chat tool for teams. And we're trying to reborn this app called Simple Direct Chat.
that we'll be using internally. It has also been sitting dormant since 2022. So it's been sitting dormant for three years. And I really want to take this project up again and just to do it again. So I basically ignored the advice of my developer team. And I basically just like copy this project. I forked it. I opened a new branch and I basically vibe coded away, right? So that was my second attempt at vibe coding. And needless to say, I think it didn't really go very well. Well, it goes well in the sense of like,
the application features are not built. So I was able to add like, I think like eight to nine, 10 features. I asked my friends, okay, like this app, like what kind of features do we need? And they will give me like the wish lists and then I will go home at night and I'll start vibe coding and I'll just have two terminals on clock code on the left, on the front end and the clock code back and on my right. And then basically just feeding it prompts to let it basically do the vibe coding for me. And
George Pu (07:48.481)
It was able to do it pretty well, think fairly well. think I was able to cross off all the major features that I'm trying to build. And I was able to build it so fast, I think one night I was able to build two, three features. very, very quickly that list is empty. I was able to build all the features that my friends have requested. And I think in and of itself it's awesome. However, the problem came after. I told my team, my developer team to take a look at the repo, I just study it and to see if there's any issues.
And I haven't heard back from them for a little while. And then now I realized why they didn't want to look at it. It's probably because the repo is so terrible and the application is so terribly written that it's basically barely understandable, probably, because I was able to pump, I think, of, I think maybe even hundreds of commits into this app. And...
Basically when they try and look back on this app, they're like, okay, we don't know what's going on there. These are, these are code. These are, there's like markdown files everywhere. There's like files, there's environment variables splash over the screen. There's like all these different things that you're committing and pushing right without checking. And I personally use GitHub cold pod to check it again. So basically I'm using AI to check as work. Um, and I think at this point I personally, I don't even understand the architecture anymore. There's been so many changes, so many branches.
And this project basically is live and ready and you know, people are using it. I'm using it. My team is using it, but I think I'm going to scrap it just because I don't do not see it being a commercially viable product. And I do not see it being viable for me to release this product to commercial customers, right? Right now just internal use. So similar to a T app, think many of those like vibe code app became basically they became a nightmare once you cross a feature threshold.
And I think the app I built, Simple Direct Chat, has basically became that product. And it's like basically impossible for me to present this product to my customers with confidence because I don't know what's going to happen tomorrow. Someone might come in and basically say, OK, this app has a lot of huge issues. You guys can just solve it. Or maybe my customers' credentials, locking credentials, will be stolen. And that will cause them, you know, and that will basically bring lawsuits upon us. And I became weary.
George Pu (10:08.033)
about this and I think that was when my first wording sign with vibe coding has started. And then basically this new project, which I'm sure all of you could understand. Well, that was last week. Actually. So the third project, which I built vibe coding was a simple drag block and it's using this really different language called handlebar, which I think the ghost block or two in a very simple terms are simple drag block is hosted on this platform called ghost and it's open source project that's using
a theme, a language called handlebar. And I'm sure Claude or Chachapiti doesn't really know much about handlebar, right? And basically I tried to change the designs of Figma, which basically I did it very successfully with my first project to this project. What happened? It broke. It broke the entire site and it made the site almost unrecognizable, right? Whereas Claude Coe is telling me, okay, this is perfectly done. And it lists me a list of things that I thought it was successfully done.
But in reality, when I load the app, when I load the site, it's broken. The fonts are like 72 everywhere on mobile site. The apps are like, it looks broken. The text have like, doesn't even have distance between each other. The margins are broken. There's a huge chunk of like white spaces at the top. And I try fixing it for hours. I spent my entire evening, right? I spent my time on the subway doing this because I was like, okay, I really need to fix this. And after many hours, I just gave up.
I give up because I do not see it being viable and do not see our customers will find this acceptable either. So that's when I caught it. Okay. I'm gonna, I'm gonna call it. This is it. This is me with the vibe coding. I'm not vibe coding anymore. Right. And so my personal lesson, I think from this is it's basically that you need to really understand what you're coding and it gets really tempting to just code without proper planning. And we have seen, and first of all, I,
appreciate how easy it is to code nowadays, right? It has made me a very dormant coder into someone who actively codes now, Through, through clock code and cursor. And there are many command line prompts that I do not understand or I was not aware. cursor is able to, you know, push this, push this, these very elegantly to me and accept them, which is another problem because you don't know what you're accepting. so that that can
George Pu (12:36.079)
destroy your laptop or your environment very easily as well. Although it has not happened to me, I think it has happened to people on the internet. So I appreciate how much value is creating to people who hasn't doesn't have a chance of learning into coding before, right? Because I think I think I was reading something a few years ago, only 1 % of people can code. There's a really small segment of the population, maybe even less. And if you think and how many percentage of that percent can code really well,
probably another 10%, 20%. So you really diminish just the amount of like work that you could do as a solo entrepreneur or a small business or like, you know, like, a small team, right? It's basically impossible. But what it's doing now is is enabling developers with some knowledge to build really elegant applications, to build really cool applications. And I think that in and of itself, it's like a truly amazing thing that
We're creating value or giving powers to people. We're democratizing access to coding. That is amazing. However, I like to make this episode to create a warning for all of you guys for listening that if you think vibe coding is the future of commercial applications, it is not. It is not. When everyone gets to be able to do something, right? What happens? Right? So basically think about this way. Everyone is able to pick up their phone and record, right? So camera industry has changed significantly.
everyone's able to pick up the phone and record, right? But who is the really better photographers or who is the better content creator when everyone can still create content, right? There's still this like threshold of like the top content creators and the bottom 99%. Right. And I predict this is going to be the same with vibe coding. When everyone gets to understand how to code, the industry would change significantly, just like what has changed for the camera industry with iPhones and phone cameras. This
software industry is going to be changed forever, which the topic we'll talk about another day. So basically everything we're looking at today will be different. However, that doesn't mean if you know how to to vibe code, you're going to win in the new market, right? Whatever collapses, you're just gonna, you know, use vibe coding to rise from the ashes and be able to take everything home. Unfortunately, that's not how it works. And my prediction is that only one to 5 % of the best companies or the best projects
George Pu (14:57.359)
I'd be able to control 90 to 95 % of the total output or revenue from the software industry going forward. So we're going to see even more concentrations. And it's hard to imagine this even for me. However, think about it this way, right? So when everything is so readily available, the market would naturally choose the best option for them. When there are hundreds and hundreds, for example, of customer support apps, there's only going to be one that people know and going to use.
going forward, given that there are probably similar prices. And I think that's going to happen to every quarter of the software industry. And in which cases we will have a few mega names, which I don't know, I don't think they exist right now or they're in the process. We will know the names now, but in a few years we will know and they will be household names and will be default and be using their products. And how will they do it? They will be probably having the best engineers in the world to be able to use AI to make them building more better applications.
right? AI agents might take over and the software application interface might disappear, right? And that's another thing that's going to wipe the average developers away. Because in order to imagine the future, you have to be a mastery of what you do, right? You have to be a masterful developer. have to be the top of the world developer to see what's going to happen after the application interface is over. For just one example. And there are many, many other examples. So I personally think that's going to happen in the future. And this
firm belief is changing how I see my current projects and how I see my current technology companies, especially as simple direct, and we're doing a huge pivot and change away as we're recording this, right? We're making the change because we see what's coming and we're making change to avoid being slaughtered in the upcoming, I will say software slaughtery that's going that that's already playing out in some sense. And I think be humble be
really able to learn new things, new technologies, new AI tools. And if you're a company, if you're a builder, they're really competent developers to be keeping an eye on your application. Make them empowered to use AI tools because what you are doing is you're giving responsible adults access to toys in which they understand. So basically every new line of code they understand and they're able to take it further to the next level, right? And they're able to make sure that there's no
George Pu (17:22.209)
application errors, there's no limelines in your code that you're going to regret a few months or years later. So that's what I'm doing now. have personally quit the vibe coding community myself. I barely use this cloud code or cursor now as of the past few weeks. And I'm focusing more on designs and enablement and helping my developers do better every day. And so I personally tried and failed at vibe coding.
And I'm sure there might be some better stories from people who have really made it. So if that's you, feel free to write to me and tell me why I'm wrong. Right. So obviously, I'm always open to understanding that, okay, I'm wrong. And I'm very glad to be wrong. But in this case, I think I see the future in this one. And I think the future is basically what I described. So be ready, be ready, get ready, learn, be humble. And, you know, that's what's going to happen in the future. Right. And I think last words,
for folks who are still in school or who are studying computer engineering, computer science or something related to steam and who is just coming out of college or who are worried about being replaced by vibe coders, be confident that you're not if you're really good at your craft. Right, my experience, my personal story has told you that companies will always be hiring developers from this point forward and you will not be replaced easily.
If you do your craft, if you practice every day, you do really well in the field that you do, if you're a system design engineer, do that really, really well, right? In which case people will still need you and you will be enabled 5x, 10x by AI. And that's how I see the world moving forward. And that's my advice to you. So I think that will be this episode. It's really a topic on top of my mind. I made this mistake myself on live coding. I just want to share it with you.
with you all. If you love the content, you know, follow us at founder reality show at founder reality.com. You can also find me George Poo, the host on Twitter as the George Poo. We also have a newsletter that's coming out every Monday, Wednesday and Friday. So sign up for that whenever you can. I'm also going to be releasing more of the templates tools and you know, things that I use as a startup founder and all those will be given for free on founder reality.com. So stay tuned.
George Pu (19:41.241)
Stay tuned for this podcast. Let me know what you think. Let me know what you think. Email me at george at founder reality.com for anything that you think I'm wrong or if not something you want me to discuss or just tweet at me. And as always, this is George. I'm really grateful to be your host at Founder Reality and see you next time.