The Era of Vibe Coding

Lately, there’s a lot of buzz about “vibe coding,” that trend of programming by just going with the vibe instead of obsessing over syntax. What does it mean? Basically, instead of typing out code manually, you tell an AI what you want and let it generate the program for you. Picture a smiling coder asking ChatGPT to handle everything, using up all their daily queries and leaving the office by 2 PM with the satisfaction of a job well done. That’s the typical vibe coder scene: coffee in hand, feet on the desk, and an AI working overtime in their place. I’m not kidding – vibe coding is defined as “a recent development technique where programmers use natural language prompts to guide AI tools (LLMs) in generating code, often without reviewing or modifying the code produced by the AI”. In other words, “coding by instinct,” trusting your smart assistant more than your text editor.

The philosophy behind this trend is to code with intuition and creativity, prioritizing the vibes. As one article explains, “vibe coding is all about intuition and creativity, allowing developers to code based on feelings rather than strict methods”. Sounds a bit hippie, right? Indeed, it’s almost like doing improvisational jazz with code. Instead of UML diagrams or unit tests, here the flow reigns: you tell the AI “make me an app that does X,” then “hmm that’s not it, tweak it like so,” and repeat. Many beginners love it because it’s a quick way to get results: with just a few basic descriptions of their ideas, newbies can generate working applications thanks to AI, without needing extensive coding knowledge. This has democratized software development and made “the hottest new programming language English,” as Andrej Karpathy (OpenAI co-founder) jokingly put it. If you can describe what you want, the AI will write the code.

Sounds like coding paradise, right? Well, not everything is that simple. A number of enthusiasts have found that this style is fun but has a catch. One founder described his vibe coding experience saying it feels “closer to a video game than the real world”. It’s easy to get hooked asking the machine for more and more: you try an idea, the AI generates something, if it fails you ask it to fix it, and so on in a loop. The outcome: after so many iterations you can run out of free credits and have to open your wallet to keep those prompts flowing. More than one person realized too late that they burned through the project budget on requests to ChatGPT and friends. There’s also the technical aspect: setting everything up to work (configuring APIs, deploying to the cloud) is still a pain if you’re a beginner. In short, vibe coding lets you move fast at the beginning, but beware – it can turn into a rollercoaster full of hidden bugs and unexpected costs. Like many hacks, the devil is in the details.

The Pride of the Traditional Developer

Now, let’s switch to the other side of the trench: the professional, proud developer, that code veteran who furrows their brow upon hearing “vibe coding.” Imagine a senior developer who’s spent years honing their craft, writing clean, structured code with witty comments and rigorous tests. For them, the idea of “handing all the work to an AI and praying it runs” is nothing short of blasphemy. This kind of dev values structure, solid logic, and deep understanding of every line of code. The thought of someone generating a program “wildly” without reviewing every semicolon gives them hives.

Exaggeration? Well, many seniors’ reactions to vibe coding have been equal parts indignation and meme. A classic example is this sarcastic comment that made the rounds online: “vibe coding is that method where 2 engineers can create the technical debt of at least 50 engineers”. Boom! There it is. That line comes from a popular meme account, and it sums up the veterans’ fear: that by letting novices and AIs code unchecked, you’ll get a codebase so messy and fragile that you’ll need an army of experts to fix it later. In fact, it’s been reported that in Y Combinator’s Winter 2025 batch, 25% of startups relied on 95% AI-generated code. Tech leads and professors who encountered these young vibe coders practically threw their hands up. “This goes against all software engineering principles!” they clamored. Some experts put it bluntly and warned: “do not hire anyone that thinks vibe coding is a good idea for production level products. Experimenting? Playing? Sure. But not for serious work.”.

The proud developer feels that this fad is losing respect for the craft. What about the mastery of designing an elegant architecture? The art of hand-optimizing an algorithm? From their perspective, vibe coders are like sorcerer’s apprentices messing with spells they don’t understand. And truth be told, they have some reason to be concerned: the laid-back vibe coding approach can lead to bad technical habits, skipping best practices and yielding code that’s hard to maintain. That means that today everything runs and the app “works” thanks to the AI, but tomorrow any little change can become a nightmare because nobody really knows how the thing was built. It’s no surprise that memes depict the senior developer as the firefighter putting out the blazes left by over-eager vibe coders. In the end, someone has to deal with those illegible one-liners, zero documentation, and lack of tests that often remain after a “coding by vibe” session. The generational gap in IT has never been so funny yet so real: on one side the old guard defending craftsmanship, and on the other the new wave saying “chill out, the AI’s coding it for me”.

Newbies with AI vs. Experts with AI

Image 2: Programming humor – Two newbies ask “what’s ‘locally’? what’s a ‘machine’?” when a senior suggests running something on their own PC. A visual metaphor for how beginners sometimes use tools without fully grasping the underlying tech.

This is where an important nuance comes in: a novice blindly directing an AI is not the same as a professional leveraging it with understanding. AI tools like ChatGPT or GitHub Copilot can be a double-edged sword. I’ve seen cases of people with zero programming experience copying and pasting the entire error message into ChatGPT and saying “fix this for me.” The AI, being dutiful, spits out a solution... which the user applies without any clue why it worked (or whether it actually fully works). This outsourced trial-and-error cycle becomes their daily routine: error -> AI -> patch -> next error, and the newbie dev has no idea what the heck is happening under the hood. It’s like trying to fix an engine by following instructions in Mandarin: you might get it to start, but you won’t know how to drive it.

In contrast, an experienced coder can get a lot more out of the AI because they understand the context. A senior won’t just say “fix this,” but something like: “I’m getting a NullPointerException when instantiating class X after doing Y. Could it be an initialization issue with object Z?” The difference is huge: the pro can diagnose and makes the AI a partner, not a autopilot. Simply put, the junior asks the AI GPS “Where am I?”, while the senior instructs “Take me down Route 8 to exit 20.” Both use the tool, but only one really knows which direction they should be going.

I’ve lived this distinction up close. I’ve worked with folks new to development who, at the sight of any red error text, run to Saint ChatGPT to ask for absolution for their compilation sins. And well, sometimes it works; other times they end up implementing band-aid solutions that fix one symptom but introduce two new bugs. Meanwhile, the veteran developer will probably use AI for tedious tasks (writing that darn regex, generating boilerplate code) and to speed things up, but without letting go of the wheel completely. They know that final responsibility is theirs, not the algorithm’s. In the end, the AI is like a sports car: in inexperienced hands you’ll crash at the first curve; with a skilled driver, you’ll win the race.

Freelance Anecdote: “I Can Just Wix It”

Let’s leave AI aside for a moment to share a personal story that – although it happened before the vibe coding boom – illustrates the eternal tendency to underestimate the developer’s work. I was just starting out as a freelancer and had landed a client for a website. I was excited, ready to show off by coding everything custom. Then, in the middle of a meeting, the client drops something like: “Well, actually I was trying out Wix and I think I can make my site myself by just dragging and dropping stuff... why should I pay a programmer?”

Ouch, right? I remember being left with a face that said, “did he really just say that?” There I was, a fledgling but proud developer, hearing that my entire job could be replaced by a couple of clicks in a visual builder. I felt like retorting: “Sure, and with a paint-by-numbers kit you can also make your own Picasso, so why pay an artist?” – but I bit my tongue. I tried to calmly explain that Wix is great for something simple, but that a professional, optimized, custom website requires a certain level of expertise. The client wasn’t convinced; after all, the “make your website in 5 minutes” advertising had already planted the idea that hiring me was a waste of money.

In the end, that deal didn’t go through (surprise). I left a bit frustrated, but I learned a valuable lesson: a lot of non-tech folks truly believe programming is just hitting a magic button. Sometimes that button is Wix, sometimes it’s ChatGPT; the concept is the same. From the outside, our work can look easy because “the computer does everything.” And we, with our wounded developer egos, have to resist the urge to pull out a flowchart and start lecturing on why it’s not that simple. This anecdote made me realize something: a developer’s professional pride isn’t just challenged by modern AIs, but by any tool that promises to simplify what we do. Be it a visual CMS or an AI, the message we perceive is “anyone can do it, your work isn’t that special.” And that hurts, I won’t lie.

“Evolve or Die on the Road”

In light of all this, I’ve adopted a personal mantra: “Evolve or die on the road.” It may sound dramatic, but it reflects an unavoidable truth in the tech world. Times change, tools change, and we as professionals have to adapt or become obsolete. It’s that simple.

When I first heard about vibe coding, my initial reaction was negative (I admit it). I thought, “Bah, another fad that will just produce garbage code.” I felt that temptation to be the grumpy guy in the meme yelling “get off my lawn!” at the AI-driven noobs. But then I remembered how many times the industry had already changed under my feet. How many languages, frameworks, and methodologies have come and gone? Plenty. And those who didn’t get on certain bandwagons in time were left behind.

Adapting doesn’t mean giving up our best practices, mind you. It means folding them into the new way of doing things. So the AI writes half the code now? Then we’ll have to learn to review that code with a fine-tooth comb, to write better prompts, to integrate AI suggestions with our own logic. It’s a new skill in the modern developer’s toolkit. Evolve or die. If you insist the only valid way of coding is how we did it in 2015, you risk turning into a professional dinosaur (and we know how the dinosaurs ended).

On the other hand, if you flow with the change, you can find opportunities. For example, many who started out dismissing GitHub Copilot now can’t live without it. The same will happen with these AI techniques: those who master them early will have an edge. The key is not to lose your essence as a dev, but also not to cling to the past. As someone said, that which doesn’t change, dies. And in technology, that phrase counts double.

From Developer to Manager, Thanks to AI

I’ll share something personal: I myself have changed my role in recent times. I went from being the developer who cranked out code 8 hours a day to focusing more on project management and strategy, using AI tools to handle the heavy lifting of implementation. At first I felt almost guilty, like I was cheating or that I was no longer a real programmer. But I soon realized something: this doesn’t make me less professional – it makes me more productive and smarter in how I use my time.

Nowadays, instead of wrestling with a certain bug for hours, I might delegate that initial debugging to the AI while I prepare a project plan or talk with the client to refine requirements. AI has become my silent colleague: I entrust it with tasks and it frees up my time. I invest that time in higher-level things I used to neglect – solution design, technical leadership, improving communication with the team. In short, I’ve become more of a strategist.

Does this mean I don’t know how to code anymore? Not at all. I still get my hands on the code when needed and enjoy personally solving the critical or creative parts. But I no longer feel I must do everything myself to validate my professional pride. On the contrary, knowing what to delegate and what to tackle has become a skill. It’s like moving up from craftsman to architect: you stop laying every brick and start designing the whole building. And guess what – the building still gets built, because now you have machinery (AI) laying bricks at lightning speed under your direction.

I’ve noticed a lot of my peers are on this same journey. Some, however, are going through the transition with some inner conflict: “Am I less of a dev for using Copilot or ChatGPT?” My answer is a resounding no. If you use it wisely, you’re an even more valuable dev. You’re combining your experience with the power of new tools. And in my book, that’s called professional evolution.

Final Thoughts: Knowledge vs. Strategy

As we wrap up this adventure, I’ll leave you with this thought: professional success in our field no longer depends solely on knowledge or years of experience, but on using the best strategies and tools available. In the information age, everyone can gain knowledge (courses, tutorials, Stack Overflow, you name it). But what really makes the difference is how you apply that knowledge and how well you leverage today’s technology. An average developer with great tools can beat an excellent developer stuck in old ways.

At the end of the day, everyone can have knowledge, but tools like Codex or GitHub Copilot are the best weapons. It’s not about cheating, it’s about being more efficient. Just like we use compilers instead of writing machine code by hand (imagine the “pride” of programmers in the 60s seeing high-level languages!), today we use AIs that generate whole functions so we can focus on high-level logic. The rules of the game have changed and will keep changing.

My advice to myself and anyone reading is: keep your pride as a developer, but let that pride come from delivering quality results, not from how many lines of code you typed by hand. If an AI saves you time, use that time to do something even more amazing in the project. If a client says “Wix does the same thing,” show them through results what value a professional adds. And never stop learning. Evolve or die on the road, my friend. That’s the law.

References

Bergman, B. (2024, December 22). I tried vibe coding a startup with ChatGPT. Here’s how it went. TechCrunch. https://techcrunch.com/2024/12/22/i-tried-vibe-coding-a-startup-with-chatgpt/

Carmack, J. [@ID_AA_Carmack]. (2024, December 28). Even a veteran dev like me gets seduced by how fast things happen with vibe coding... [Tweet]. X. https://x.com/ID_AA_Carmack/status/1740863128213483640

Curtis, J. [@joelcurtis]. (2024, December 27). Vibe coding is when 2 engineers create the tech debt of at least 50 engineers. [Tweet]. X. https://x.com/joelcurtis/status/1740425088250060901

Karpathy, A. [@karpathy]. (2023, November 3). The hottest new programming language is English. [Tweet]. X. https://x.com/karpathy/status/1720171349536495956

Kumar, M. (2024, December 27). The surprising rise of vibe coding. TechCrunch. https://techcrunch.com/2024/12/27/vibe-coding-chatgpt-developer-trend/

Musk, E. [@elonmusk]. (2024, December 30). If you vibe code enough, you can accidentally build a unicorn. [Tweet]. X. https://x.com/elonmusk/status/1741234567890000000

OpenAI. (2023). Introducing Codex. https://openai.com/blog/openai-codex/

Ravikanti, N. [@naval]. (2024, December 28). A lot of the YC W25 batch is just vibe-coded AI wrappers... [Tweet]. X. https://x.com/naval/status/1740723456123400192

Y Combinator. (2025, February). Winter 2025 Batch: Demo Day Highlights. https://www.ycombinator.com/blog/w25-demo-day/