A junior developer stares at their screen, fingers posed over the keyboard, and then does absolutely nothing. Instead, they type a prompt into AI build me a full stack SAS app that lets users write office coffee using sentiment analysis and a crypto wallet. Seconds later. Poof, like code appears. It's not perfect, it's not even going to work, but it runs, sort of. Welcome to the age of vibe coding We're writing. Software isn't about syntax or architecture anymore.
It's about the vibe, man. Think of it as coding by gut feeling. Except your gut is now a chat bot with ACPU and a questionable taste in logic. In this episode, we'll be diving deep into what vibe coding actually is, why it's gaining traction, You'll see it on LinkedIn every 5 minutes, where it shines, and more importantly, where it crashes hard. We'll explore use cases, real life case studies, and why this trend might just be the fastest route to technical debt. You can't refinance.
If you're a business analyst, a product owner, a developer, a business leader, or someone just trying to stay one buzzword ahead of your next meeting, this one's for you. The Better Business Analysis Institute presence, the Better Business analysis podcast with Kingsman Walsh. Let's get into it. Here is 10 things that you need to know about vibe coding before it's vibe puts you into technical debt. That's right. So number one, we really need to talk about what is this thing
called vibe coding? It sounds cool. Vibe coding is an AI driven programming approach where developers provide high level prompts to large language model models. So AI, think of ChatGPT as an example, which then generates the corresponding code. And with things like MCP that we've talked about recently, this could actually start building multi layer architecture. This was coined in February 2025. So it's really new and it shifts the developer's role from the manual Kona to guiding and
refining AI code. And while we're on this number one point, I do this, I do vibe coding all the time because I'm not a programmer anymore. I don't know the latest languages and I feed in prompts to AI, more complicated prompts and areas of code and it allows me to prototype quickly. So we'll get on to that in a minute. But you might be one of these developers out there today. But this really vibe coding is when developers are doing this.
This is instead of them having to writing lines of code, they can be, some of them can be notoriously lazy anyway. They're actually getting AI to do the work for them #2 there's an appeal about wide coding. It sounds cool. It accelerates development by automating routine coding tasks that you might have to do again, and it's boring. It's accessible. Even individuals with limited coding experience like myself, we can develop functional
software. And you see a lot of promotion about this online and innovation. It fosters rapid prototyping and experimentation. And that's a really good point. So #3 what are the practical applications here? Well, in start-ups, if you have small teams, you can achieve outputs that are on scale with larger teams by leveraging AI tools, especially for some of the manual work, but not all of the workflow steps.
And enterprise solutions companies like Amazon utilise AI assistance to enhance development. They have like a a code plug in, Visual Studio has it, the GitHub plug in for Visual Studio I've used and it helps you for looking up classes and what not and educational tools. So facilities can allow students to focus on problem solving rather than syntax. So they can kind of think about how they might program a Python script and get it about right.
And they're like, well, don't worry, you didn't get the the syntax right. We'll move on. And syntax, to be honest, is just the like grammar of programming. So I use AI to fix my grammar all the time. So programmers can use AI to check their syntax and their syntax checkers. So why wouldn't you use AI for that part of it #4 OK. And there's a company called RE Pets AI generation game.
So there's a, a developer uses Repat's AI to generate a racing game featuring actually Elon Musk and Sam Altman from Open AI. And despite initial bugs, the project actually demonstrated the potential of AI in a simplified app development. It, it worked and people were able to play it. And this gave people the idea of, well, why don't we utilise that model for our business? And that leads us to #5 which are the potential risks here? Number one, it's, it's technical get.
So the architects and people that are really hot on good effective code. I I I do not like this concept when applied to production environments. Rapid development without proper structure leads to maintenance challenges and bad architectural decisions. All the code is usually a couple of files. It's not spread out how do you maintain it and this makes it worse. The other issue is around security, and this opens up security holes.
The reason we have programmers and the reason we have people that know about security, security architects, is that AI driven code may overlook some of these practices if you haven't fully explained what they are. And then there's an over reliance on AI. So developers might become detached from the understanding of the code and then when something breaks, they don't know when to fix it. And so let's talk about mitigating that risk.
So number six is mitigating the technical debt problem, and this is an argument against the technical weights Det starts and that is that you can have code reviews to ensure that the code quality is there. You can ask AI to generate clean documentation and to provide it's thinking about what it's done and include that in the code itself as comments and
testing. You can make sure that your testing is way more robust now and it incorporates rigorous testing to identify and fix the issues early. So you can actually use this technique. There is a comparative study here. So #7 here is around Airbnb. They use this concept of modernization, right? And the challenge was that Airbnb had this monolithic architecture and it led to increased dependencies and slow releases.
And we talked about that in the last episode around architecture and the putting the bits and pieces together. And the solution was to transition to micro services, which is the other architecture we talked about using various, I won't go into technical ways they did that, but they did that to improve scalability and reduce technical debt.
And that is a kind of a better way of doing things is breaking things up. And then maybe if you had a small micro service, maybe you could apply AI to that micro service to make that better, but it's kind of not applying AI and vibe coding across the whole structure. So get your architecture right first. Some of the organizational considerations we need to think about, which is number 8 is making sure that you are
training your teams effectively. You need to equip your teams with the skills to effectively effectively the keyword here use AI tools and not replace themselves. There's governance. So what are the policies how we manage AI generated code responsibly, but allow AI to be the true copilot as opposed to the writer. And if you I can tell you that if you're using AOI to write code for you as a developer, then you are not going to be needed soon enough. And integration.
So ensure these AI tools align with your existing development workflows, checking in code, code reviews and what not. And any of your best practices need to be built into the vibe code ring experience, unless you're going to be using that for prototypes, which I'm going to finish off with #9 is strategic implementations. So if you are want to try vibe coding and I I suggest you do, I'm not saying it's all bad.
Start with a small start scale project or a micro service to assess the effectiveness of that in your organization and whether or not by allowing this option, you know all the rules go out the window. Have feedback loops, look at feedback, provide AI prompts and outputs and look at reasoning behind the AI prompts which are available now and look and see if that's correct reasoning. I would say that in my experience, the tools out there today are not that great at
programming, to be honest. I don't think they're full. They're good at creating good structured architecture and look at scalability. Can this AI tool that you've just create actually be scaled? And if it can't be scaled, then obviously it's just a peripheral concept and can be thrown away at the end of your vibe coding session.
Before I drop into the final thoughts, the best use case for vibe coding in my mind is you're an entrepreneur, you're working with a developer or you're working with people who know a bit about architecture and you want to rapidly produce a prototype to show the market to get funding to do it properly. OK, now that doesn't mean you don't use AI in the generation of production materials and documentation, but you do it in
a non vibe way. You do it in a very structured, methodical way and you use the vibe, the vibe to get the funding and to get people on board by, you know, showing them something cool fast. My final thoughts are vibe coding represents a significant shift in software development, offering both opportunities and challenging. And I think it will evolve into just code coding as opposed to vibe coding, apart from the use case I just mentioned which is
prototyping. And by embracing AI thoughtfully and implementing best practices, then organizations can harness it's benefits while mitigating the risk. I will see you next week and I hope you know what vibe coding is. See you later and Happy Easter.