Cursor Hits $2 Billion Revenue & Developers Are Leaving

0:00Cursor just crossed $2 billion in annual revenue. $2 billion and this company didn't even exist 2 years ago. So, at the same time, someone tried to vibe code an entire operating system and it was a complete disaster. Nothing worked. The browser was actually an image viewer. Doom was listed as a feature button and literally did nothing when you clicked it. So, we have all this weird moment right now where AI is coding tools and making more money than ever before. But, the way most people are using them is completely wrong and I think we need to talk about it. Cursor, I use Cursor every day. I use Claude code within Cursor and, you know, there's many Cursor users out there and it has many use cases. But, $2

0:46billion headline on the screen may grab your attention and I want to break down what's actually going on with the company saying which is the AI coding editor that basically kicked off this whole wave just hit $2 billion in annualized revenue. It's hard to kind of comprehend, but Bloomberg reported it like a week ago that the number doubled in 3 months. 3 months ago, they doubled revenue. They were valued at $29 billion back in November when they raised $2.3 billion in funding led by ASL and Coat, but the corporate customers now make up about 60% of their revenue. Here's the part that caught my eye. Individual developers are leaving Cursor, like actually leaving the platform. They were viral tweets last week from well-known developers talking about switching to

1:31Claude code, which I recommend you do. I have a little bit of bias, but it's just cheaper and honestly more capable of a lot more other workflows, especially with what Anthropic has been shipping as of late. One thread had thousands of retweets. The corporate clients are staying because, you know, enterprises move much slower to switch tools. There is a procurement involved. There are security reviews and there's contracts. But, the builders, like you and me, the indie hackers, we are actually shipping products on nights and weekends. You're probably shipping something right now. A lot of them have moved over to Claude code, and that's our main focus inside of our community as well. So, that split tells you something really important about where this whole space is headed.

2:17The money is going in one direction, but the innovation and the energy is going somewhere else. So, I want to talk about like the vibe coding problem, because this is where it gets pretty interesting, especially in relation to you and I. Andrej Karpathy, one of the co-founders of AI open AI, coined this term vibe coding about a year ago, and I'm sure you've seen it come across your feed or your screen. The idea is pretty simple. You just describe what you want in plain English, and the AI writes all the code, and you don't even have to read it. You just keep prompting it until it works. There's more to it than that, but that's the concept. You accept every suggestion, you don't question anything. You just vibe with it. Sounds amazing in theory, and it's great to start, but you do kind of need to know what's going on under the hood. Well,

3:03someone took the idea to an absolute extreme. They tied or they tried to build an entire operating system using nothing but vibe coding. It's called VibeOS. They put it up on GitHub, and it's supposed to run on ARM and x86 computers. It claims to have custom kernel and modern macOS inspired graphical interface and a virtual file system. So, sounds impressive on paper. A YouTuber named Trimid spent literally hours testing it, and it was an absolute disaster. How embarrassing. The file manager, the buttons didn't even work. You couldn't even click all day long, and nothing would happen. Which Which happens when you ship fast, and you don't know what you're doing. The notepad app, it refused to even save any

3:48documents. Like, you could type stuff, but the second you hit save, it wouldn't. The snake game briefly functioned. They claimed the OS could run Python scripts, but when you actually tried, there was zero Python support installed. Here's the best part of all this is they advertised that you could play Doom on it, like the classic Doom. The Doom launcher icon was right there and was on the desktop. It looked official, but when you clicked it, nothing happened. They I literally hallucinated a feature into the documentation and it wrote documentation for a feature that doesn't even exist. So, think about that. So, listen, this isn't just some funny one-off thing. This is a pattern that is forming across the whole industry. The New Stack ran a major piece back in January with the

4:33headline "Vibe Coding Can Cause Catastrophic Explosions in 2026", which I think is a little bit, you know, extreme to say. But, Red Hat, one of the biggest enterprise software companies in the world, published an article called "The Uncomfortable Truth About Vibe Coding". And Medium is full of posts from experienced developers warning about this. I really just think they're more scared about their job. That's just my opinion. And there was a real security breach, the Multbook breach, that was directly traced back to vibe-coded software that had no security review whatsoever. And we've all seen how the Open Claw launch back a couple months ago kind of raised some concerns. It's much better now, but I think security is the first thing you need to

5:18figure out whenever you're starting to build anything with AI. So, real users got their data exposed and that's the scary part because someone vibed their way through the security layer. This is becoming a serious problem and I think a lot of people building with AI right now need to hear this before they learn the hard way. Here's my take on this whole thing and I want to be really clear because I do not want people to misunderstand my position. Vibe coding is not coding or as it used to be. I think we're seeing a complete paradigm shift in the whole coding space. And it's prompting. It's basically like you orchestrating and managing an LLM to do the coding for you because at this point, these large language models, and I'm referring to Opus 4.6, is going to get you about 80 to 90% of a senior

6:06software developer. It's going to do really, really good, probably better than you and I could could do if we spent the next 4 years learning how to develop software. So, there's a massive difference between a senior software developer and just vibe coding. When you vibe code, you're basically saying, "Hey AI, build me a thing." And then you're hoping that it works. You've probably seen those memes like, "Hey, build me a billion-dollar business overnight. Make no mistakes." That is exactly kind of the concept when you are getting into this vibe coding space. You're not thinking about like the architecture. You're not thinking about security. You're not thinking about what happens when a thousand users hit your app at the same time or when someone tries to inject the SQL layer into your login form. You are not thinking about error

6:51handling or any kind of edge cases. And you're just vibe coding, which is sounds cool, right? Vibing is not software development methodology. So, for prototyping, it's great. For getting like your idea off the ground, I think it's perfect. For testing whether something is even worth building, I think it's great. It allows you to ship fast. There is absolutely a place for the quick and dirty AI-generated approach. I'm not anti-AI coding. I do this every day with the three projects that we have, Easy Flip, Magic Can, and Snaptastic. And those are all vibe coded projects. They They were completed 80-90% of the way in just several days, but to actually complete and finish the app took months of security tightening, fixing bugs, getting users. So, it gets

7:38you close, but the finished product needs much more refinement. So, I use Claude code, like I said, inside cursor every day to kind of squash bugs, check the logs, make sure everything's working for the users. I also built my entire mission control dashboard using AI coding tools. This is the system that automates the YouTube content pipeline, newsletter, blogs, school community, X posts, podcast outreach. Every morning that fires off at 6:00 a.m. My AI agents wake up and they scrape the trends, they generate the thumbnails. That whole system was built with Claude code Opus 46. So, I'm not coming at this from some pursuit of real programmers to use like a notepad perspective. And there's a massive difference between using AI as coding partner and just blindly

8:24accepting whatever spits out. That difference is the gap between shipping something real and shipping a vibe OS. So, I'm going to talk about what I actually do because I think this could be useful from like my personal experience. What I'm building is something new. I do not just throw a massive prompt at Claude and expect it to just be okay with whatever comes back. That is the vibe coding approach and it does not work for anything beyond a prototype. Instead, I actually have a little bit more of a process, especially after building my first app. I think about the architecture first and really when it comes down to it, a successful web app or mobile app has just a few components. You have database, authentication, and payments. That's basically it to get your app off the

9:09ground. And if have those proper things installed into the backend architecture, you're opening yourself up to security risks, to things breaking down the road, and you really need to know what is the architecture and frameworks that I'm going to use to build this thing. Think about that for a minute. You don't need to be a full-on software developer, but you do need to know which frameworks you're going to be using to build your app out and what they do. So, before I write a single prompt, I really want to know what are the tools I'm going to use to build this. I know what I want to build, I know what the data model looks like, and I know how the pieces connect. I want to know what they do and why they do it. So, then I kind of break reverse engineer it, I break it down into small chunks. I might say build me the API route for creating a new video and then

9:55look at what comes back. I read it, I try to understand it. If something looks off or I don't understand why it made a certain choice, I dig into it and I follow up with questions and I just say break it down for me. Usually Opus 46 is pretty good at this and I've been using these models for the last year and we've seen such a crazy evolution between Opus, you know, 41 to now 46 and it is a night and day difference. It's like pairing like programming with an incredibly fast partner than it is Vibe coding. You're basically using Opus 46 as a senior software engineer. Um, very helpful. It's going to know, you know, a lot more than probably you do if you haven't been, you know, going to

10:40university to study programming. And that distinction does matter a lot because one approach gives you production-ready software that you can maintain and scale and the other one gives you Vibe OS where the browser is actually an image and a Doom is a hallucination. It's It's really funny. So, [clears throat] here's why this matters for you specifically if you are a builder or an entrepreneur because the opportunity right now is genuinely insane. The Cursors making like $2 billion a year. Cloud Code is growing rapidly. Many people are switching. Replit and Lovable and Bolt, all these companies are fighting over this market share. I don't hear too much about those companies now in the news even though they're some of the fastest growing companies on record. Hundreds of

11:25billions of dollars in venture capital is flowing into these AI coding tools. That means that tools are getting better every single month as the LLMs get smarter. The AI models are getting better, the code they generate is getting cleaner and more reliable, but the thing is is like the gap between the people who know how to use these tools properly and the people who are just vibe coding their way into a mess, that gap is getting wider every day. I think that the expectations need to be set with, you know, you're getting into this, you've seen other people just speak into the computer, you know, you're using whisper flow and you expect it to one shot your million dollar idea. I think if you prepare for the work expect the best, but prepare for the worst. You know, it's going to take a

12:10lot more trial and error and there is a learning curve to this. You are going to be going through nights of frustration on why your login page isn't working. Why isn't it parsing data correctly? Why do I have 50 tables in my neon database that are useless? Things like that are going to cause a lot of headaches. So, I want to be able to help those kind of people like go through that and just set expectations. So, I talked to people in our community and we built something with the vibe coding tools, you know, maybe we ship something in a weekend, they launch it, they got users and maybe even like a few paying customers. I'm happy for them, but then everything falls apart and that's where you can increase churn and you don't want that. So, they could not fix bugs because they did not understand the code and that's

12:55the biggest problem. They could not add features because they did not know how the existing features worked. They could not scale because the architecture was a mess from day one. They had a product that worked on the surface, but crumbled under any real pressure. So, when I launched Easy Flip, I launched it to my community to get some beta testers. This thing was a horrendous mess after figuring out, oh, it doesn't list things properly on eBay like remotes and VCRs and jackets and we had to go through a bunch of categories to really optimize it. Now, if I had launched that to my complete audience, it would have been terrible. We would have had users sign up and I'd been refunding them like crazy. And I think that's just part of the process. Don't feel bad if you have done that before. It's just a learning

13:41lesson. So, they used to send AI or they used to use AI to accelerate every step, which is kind of like a convenience loop that was like, "Let's have AI fix it." But, they actually need to understand the output of what A I L LLMs are doing. So, those are the products that will survive. When you understand why it's making changes and why things are breaking. You know, a real helpful for thing for me is I have on one of my monitors live logs that on Vercel to be able to like know what's actually going on when my users are clicking around. If I get errors that pop up in my logs, I could just copy-paste the error log into Claude and just say, "Hey, break this down for me. Why is this person having a problem setting their shipping defaults?" Or,

14:26you know, things like that. It There There's things This is common that you won't be able to find every bug when it comes to building your app. So, get some beta users, get some free users, have them flesh it out before you do that soft launch so you can prepare to minimize the churn when people actually sign up for your app. Like, those are the ones that are going to evolve in this AI space. Don't just expect AI and the LLMs to just do it all in one shot. And I'm going to go back to the cursor numbers. They prove that the companies are willing to pay serious money for the AI coding tools. Like, 2 billion in revenue means there are millions of developers and teams using this stuff every single day. And the market is real. But, the Vibe coding disasters prove that the tool is only as good as the person using it. A paintbrush did

15:11not make you Picasso in 1960. And Claude code doesn't make you a software engineer now. But, if you understand the fundamentals even like at a basic level, the concepts, and you're using these tools you make if you use these tools, they'll make you incredibly dangerous if you figure out what they actually do. So, that's going to be helpful in the best way possible. So, why I built Shipping School because I saw this gap forming in real-time. Like you've got people with amazing ideas who want to build apps and ship, but the tools that exist now will make that even more possible if you're not just a traditional del- developer, which is why the community is geared towards non-technical entrepreneurs. I'm not expecting to come in here and people learn all about TypeScript and C++ and

15:58HTML. This is how we can use these tools to build real products correctly. So, the barrier to entry has never been lower, but you still need to understand the basics, the principles. You need to know enough to guide the AI, review its work, and fix the things when they're inevitably will break, because they will break, and don't get discouraged. Even every piece of software breaks at some point. I mean, come on. Inside Shipping School, I work with you side by side on the exact thing. We do three live calls a week, build sessions, boot camps, Q&A. You also get a Next.js starter kit that I used to launch my three SaaS products off the ground, and I walk you through to actually help you ship something real using these AI tools right away. Not vibe coding, real building with AI is

16:43your accelerator. I'll put the link down below if you want to check it out. So, what should you actually do right now? If you're watching this video, you're like, "Okay, this is confusing. I'm looking from the outside in. I don't know where to start." Here's five moves that I would make personally right now. I would first stop blindly accepting AI-generated code. You want to figure out why it's doing it, what it's doing, and break it all down so you could learn as the LLM builds and ships features for you. I don't care if you're using Cursor or Cloud Code or Windsurf or whatever. Read what it gives you, try to understand it. You do not have to understand every single line, but you should understand the structure and the concept of what it's doing. Like what data is it touching? What is this

17:28breaking? Why is it breaking? If you cannot answer those questions while you're vibe coding, you're one bad deployment away from big problems. So, second thing is like I would learn the fundamentals. You're not expected to learn everything, but learn the basics. I'm not talking about like a four-year computer science degree. I'm talking about the basics. Like how the web app actually works. The difference between a production environment and a local environment. What does the database do? How does authentication flow? How do APIs connect with each other? What is a webhook? You can learn this stuff in a weeks and especially with AI, you can just have it break it down like your five-year-old. So, that's going to make you 10 times more effective with any AI coding tool. The third thing I would do is build in small pieces or chunks. Do

18:13not prompt the AI to build your entire app in one shot. You want to start slowly. This is a course inside of school community that we have is I break down the Claude code course on where to start and how to start. Like a product requirements document is so necessary to give to your LLM before you even build anything because then the LLM's not even going to have an idea what you're trying to build. So, without it and that's how you get something like a Vibe OS where you have a beautiful fit looking operating system, but nothing works. So, build one feature, test it out, make sure it works, and then move to the next thing. That's how real software gets built whether there is AI involved or not. As fast as AI helps us to ship features quickly, you do need to actually take a step back and focus on

18:58one thing at a time. The fourth thing I would do is picking the right tool for your workflow. I think Cursor is amazing. You could also use VS Studio or WinSurf and just pick the LLM of your choice. There's so many coming out. So, I like cursor is amazing for corporate teams and people who want like a full IDE experience, but that's not necessary anymore with open claw. But, for 20 bucks a month, you could pay for it or you could do the free version. I think that, you know, that's what happens like when 60% of your revenue comes from corporations. Talking about cursor, you know, cloud code is incredible for terminal first workflows and it is what I personally use day-to-day. It's simple, gets to the point, easy. Figure out what your style is and commit to

19:43learning it as deeply as possible instead of bouncing between five different tools every month. The fifth and final thing I think about longevity. Why am I learning these skills and how do they fit into the marketplace? This is something that I don't hear often on YouTube is like are you building something you can actually maintain? Can you add features in 6 months from now without breaking everything? Can someone else look at this code and understand what's going on? If the answer is no, then you've got a vibe coded toy, not a product, and toys don't make money, especially if you're spending so much time building right now. You need to make sure you're doing it correctly. I'm not trying to be a downer here genuinely. When I first started Easy Flip, I started in Bolt and the code was a mess. Things were a mess and you may need to restart your project several

20:29times just to get good code outputs and nothing wrong with that. This is the most exciting time to be a builder that I've ever experienced. I've always loved creating content. I've always liked building things. It's very fulfilling to me and it is probably for you as well. Well, today is the day to pick up these tools and actually learn them and develop skills. So, the fact that cursor is doing 2 billion in revenue means that the market has validated that AI coding is real. It's not just like a fad, it's not going away. If anything is accelerating faster, it is happening right now. More money is being poured into AI than ever before. The winners are not going to be the people who prompt and pray, I will say that, even though it seems like it, right? The winners are going to be the ones who actually understand what they're

21:14building and how they use AI to build 10 times faster than they could alone. I believe that's the move and final play. That's the difference between the $2 billion tool and a vibe coded operating system where doom doesn't work, kind of depressing. So, if you got something out of this, drop a like, please subscribe to the channel. I'm dropping updates every day and if you actually want to learn how to ship products, when use AI the right way, check out Shipping School. I'll put it down below. Please subscribe because we're dropping updates every single hour, guys. We'll see you in the next one. Be blessed.