Vibe Coding: The Future of Building Companies

What the Heck Vibe Coding is and What Does it Hold?

At this point, you may have already heard about Vibe Coding. If you haven't, it's a new kind of programming technique that is rapidly going viral on the internet—from Bloomberg to the New York Times to the Guardian to Forbes have already covered the concept.

But here’s the thing: When a new idea or concept first comes to the market, it’s really hard to predict how it’ll change human behaviors—or whether the idea itself is good or bad. But as time goes by, things start to become clearer.

And this is exactly what also happened with Vibe Coding. When it was first introduced, people thought, “Oh, wow! This is the future.” But then, a few weeks in, people started saying, “Vibe Coding is garbage, it doesn’t work.”

The concept of Vibe Coding is not too old, but old enough that now we can see the good, the bad, and the ugly parts of the idea. There are good things about the concept, but also, there are bad things about the concept—as every coin has two sides, right?

It's the perfect time to explore Vibe Coding.

Many interesting things are happening in the tech industry right now—every day, every week, and every month, there are new ideas, concepts, or products coming to the market and totally disrupting the existing model. And Vibe Coding is no different. So this deep dive is an attempt to help you understand what Vibe Coding actually is, how it's done, and whether it has the potential to change the tech industry or not.

Get your popcorn ready, and let’s dive in!

What is Vibe Coding

Vibe Coding is a new approach to programming, where instead of writing every single line of code by hand, the programmer describes the product/tool they want to build in plain, natural language to an LLM (Large Language Model). This shifts the programmer’s role from manual coding to guiding, testing, and refining the AI-generated source code.

The concept was given on February 03, 2025, by Andrej Karpathy, a computer scientist, co-founder of OpenAI, and former AI leader at Tesla. Here’s how he describes Vibe Coding:

There's a new kind of coding I call "vibe coding", where you fully give in to the vibes, embrace exponentials, and forget that the code even exists. It's possible because the LLMs (e.g. Cursor Composer w Sonnet) are getting too good. Also I just talk to Composer with SuperWhisper so I barely even touch the keyboard. I ask for the dumbest things like "decrease the padding on the sidebar by half" because I'm too lazy to find it. I "Accept All" always, I don't read the diffs anymore. When I get error messages I just copy paste them in with no comment, usually that fixes it. The code grows beyond my usual comprehension, I'd have to really read through it for a while. Sometimes the LLMs can't fix a bug so I just work around it or ask for random changes until it goes away. It's not too bad for throwaway weekend projects, but still quite amusing. I'm building a project or webapp, but it's not really coding - I just see stuff, say stuff, run stuff, and copy paste stuff, and it mostly works.

There are a few phrases I want to touch on:

Forget that the code even exists: It’s Vibe Coding because you’re actually not coding, you’re just giving instructions to an LLM, and they are doing the work on your behalf. You are not the actual coder, the LLM is.

Barely even touch the keyboard: The traditional way of programming is to keep stroking the keyboard until your fingers hurt. But when you Vibe Code, you don’t always have to touch the keyboard, because you can also instruct the LLM via voice—the lazy coder, if you will.

Just see stuff, say stuff, run stuff, copy paste stuff, and it mostly works: Did he mention “Code” anywhere? Nope. When you use an LLM to Vibe Code, you use natural language (not a programming language) to give instructions. You don’t even bother checking whether the code is accurate or not—if things work, nothing to worry about.

What’s even more interesting is, in that Andrej tweet about Vibe Coding, someone asked if he has built things doing this. Guess what? Andrej showed a new custom LLM reader app he built for himself. To prove the point, this thing works.

Link to Tweet

What’s even funnier is, Elon commented “Hmm” in that tweet lol (idea validated):

Link to Comment

Now you may still wonder whether Vibe Coding is real or not—or whether you can actually get stuff done using this approach, right? Nothing to hide, Andrej, the person who gave the concept is an example of this. He has already built many tools using this approach, and one of them was a calorie tracker app he built:

Link to Tweet

The screenshot above is a Thread, so if you’re curious enough, you can click on the above link and read all the details he has shared about how he was able to build the product in ~1 hour.

Now the real question is, how does Vibe Coding actually work? Pretty simple: For one, to Vibe Code, they need an LLM platform. You can’t actually Vibe Code on VS Code, Xcode, and other traditional programming platforms. What you need to Vibe Code is the new LLMs platforms like Cursor, Lovable, v0, Bolt, Windsurf, and other similar platforms.

Precisely, any LLM platform that doesn’t require you to write every line of code manually to produce stuff, and instead the platform accepts the instructions in plain, natural language and helps you build the product/tool you want to build is a Vibe Coding platform.

The Paradigm Shift

Vibe Coding is truly changing the way programmers program. But it has only become possible because of the existing LLM platforms—Cursor, Lovable, Windsurf, Bolt, etc. If these platforms didn’t exist, Vibe Coding would never be a thing.

Before these platforms and Vibe Coding, one had to learn a programming language for years, refine their skills, learn from their mistakes—and hopefully by doing so for years, they’d be able to build a product. But the downside was, one small problem would take hours or days to solve because there were no AI chatbots like ChatGPT, Grok, or Perplexity.

Vibe Coding is a byproduct of something that has already happened. But it’s truly a paradigm shift—one does not need to learn a programming language to successfully build a product, and if you are stuck making progress, there are countless chatbots available online ready to help you with whatever problem you have—for free.

The barrier to entry now is also almost zero—anyone with an internet connection, a laptop or phone, and a few bucks in their bank accounts can start building the product they want. Forget about building a huge team, tens of thousands of dollars of upfront investment, and shipping the product in months or years. Vibe Coding is all about building what you want, shipping before the product is ready, and getting the same things done with 10 times fewer resources.

If you’re a passionate “Builder” who doesn’t have much programming experience, this moment? Right this moment is for you. All you need is just one weekend with 5 to 6 hours of focused work, and you have a working prototype, in some cases even an actual product.

Many people have also indirectly claimed that “People don’t need to learn to code anymore.” For example, the co-founder and CEO of Replit, Amjad Masad, has publicly said that one does not need to learn to code—all because of the advancement of AI and these existing LLM platforms.

Link to Tweet

Vibe Coding with these tools is real…Google’s CEO Sundar, has said that 25% of Google’s new code is AI-generated. Another example of this would be that a quarter of YCombinator’s current batch (W25) has 95% of their codebase written by AI.

Moreover, what if I told you that you won’t get a job at Shopify if you don’t know how to use AI in your work? Yup, you read it right. The CEO of Shopify, Tobi Lutke, recently made it abundantly clear that their minimum expectation is that employees know how to use AI for their work.

Link to Post

You see? When a multi-billion dollar company does something like this, know that things are real. Instead of increasing the headcount, Shopify is prioritizing AI and trying to get as much stuff done as possible with it.

Things are changing very fast—it won’t take too long when most of the codebase will be AI-generated—commanded by the programmers who don't even know the fundamentals and basics of programming. What social media and the internet did to content—AI tools and Vibe Coding will do to software.

The End of Software?

There is this famous essay named “The End of Software” written by Chris Paik, an investor and founder of Pace Capital. The essay is almost a year old, but the premise of the essay is how software will get commoditized the same way we did with content.

Here’s one of the paragraphs of the essay, which I think is true:

Software is expensive because developers are expensive. They are skilled translators–they translate human language into computer language and vice-versa. LLMs have proven themselves to be remarkably efficient at this and will drive the cost of creating software to zero. What happens when software no longer has to make money? We will experience a Cambrian explosion of software, the same way we did with content.

Of course, creating software was expensive until these magical LLMs. But now, because of these platforms, building software is no longer expensive—it will continue to go down as we progress. The software will get commoditized because anyone and everyone will be able to build it without spending tens of thousands of dollars.

And maybe we’ll even reach a point where, just like today, creating content is free, building and creating software will also be almost free. So when software becomes free to build or create, it doesn’t need to make money.

Software will act just like content.

So is this the end of software? Not really. For example, anyone can create content today, but do people consume mediocre content—essays, articles, videos, podcasts? No, they don’t. In the same way, there will be good software and there will be bad software. The one that will get users and attention is good software.

Having these LLM platforms and Vibe Coding in place doesn’t mean “All” software will die. It means that people will have a hard time finding the right and good software—the “Noise” will be way more than signal, but at the end of the day, Signal (good software) will win.

Don’t worry about it, just build good products.

The Problem with Vibe Coding

There are good things about Vibe Coding, but there are also bad things about Vibe Coding. Everything seems to work when things are going pretty well. But everything feels irritating when you get errors you don’t know how to fix, debug things you have no idea how long they’ll take, and deal with poorly written documentation that you can’t comprehend.

There are some serious problems with Vibe Coding, as I Noted:

Vibe Coding and LLM platforms go together—Vibe Coding is not possible without these platforms. So basically, whether you’re Vibe Coding or not, you’re always using these LLM platforms if you’re thinking that what you’re actually doing is Vibe Coding.

And here comes the problems:

#1: Lack of Fundamental Knowledge: Vibe Coding is good for programmers who have already learned the fundamentals of programming and know how things are done. But since the internet is permissionless and everyone wants to get “Fast” results, many non-technical people who have no idea how programming works, will Vibe Code. And because they don’t have any idea about the work they do, they’ll face some serious issues when things are not going as expected—need to fix errors, debug, correct the files and documentations, etc. Vibe Coding without learning the actual programming skill is like exploring the Amazon Jungle without tools.

#2: Security and Data Concerns: One of the major concerns about Vibe Coding and, in general, about these LLM platforms is that you have no idea how secure the product you build is and how encrypted the data you give to the machines building it. And it’s not just a security and trust concern for the “Builder” but also for the users who use the product.

#3: Digital Pollution: Programming is a skill—great programming isn’t just about writing code, it’s about communicating your ideas clearly, explaining them in a way that is understandable, and removing the unnecessary stuff the moment you catch it. But guess what? Vibe Coding isn’t the type of programming that helps you do this—you don’t write the code line by line, so you have no idea what the LLMs are actually writing. It’s a digital pollution that sometimes does more harm than benefits.

#4: Competition: This is going to be one of the major problems of Vibe Coding. Let me give you a hypothetical example: You’re on X, scrolling the timeline, and you see a founder just released a product about X, it’s good, but you whisper, “I could do better than this!” Guess what? You open Cursor (or any other LLMs) and start building a similar product, but better. It doesn’t take too long to ship the product. And since the idea is already validated, maybe if you’re smart enough, you’ll outcompete the founder who built the product in the first place. Anyone can build anything in no time, which means it’s all red ocean.

The Future of Programming

There is this paradigm shift I’m seeing where more and more engineering people are becoming “Product people” as most of the engineering tasks are handled by AI. The question isn’t what to build but “How” to build.

The success of a company would largely depend on its superiority. The superior the product, the more success it’s going to have—less engineering (programming, debugging, solving technical problems), and more thinking about the actual product (ideas, features, benefits, simplicity, etc).

This Vibe Coding and these LLM platforms (Cursor, Lovable, etc) are here to stay. Companies that don't use these platforms will get left behind, and the ones that will Vibe Code using these platforms will not only enhance productivity but also build things faster and cheaper.

Now I can also confidently say that it won’t take too long to see companies reaching $1 billion in valuation or $1B in annual revenue with just 2-3 employees, along with the founder. Cursor has already broken the record of reaching $100M in ARR in just 12 months. This growth timeline will continue to shrink as we progress.

Thanks for reading, catch you on the next one.