Чи може AI насправді побудувати компанію без людей з Paperclip?

What if I told you that there's a free tool that lets you build an entire company with a CEO, CTO, engineers, but zero humans? You'd probably think that I'm full of crap, but in today's video, I'll prove you wrong. Because over the past 4 weeks, Paperclip has completely broken the internet. And today I'm testing if all that excitement is justified or if it's just another hype train in the AI gold rush. But first, let's talk about what Paperclip actually is because it's unlike any other AI tool that I've seen so far. So right now, people are using AI tools kind of one at a time. You'll be using ChatGPT or Claude or Perplexity, and each one of these tools is doing a different specific task. One might be doing research, the other one might be writing code. And so the AI tools are basically working as individual employees, which is fine if you're doing it one at a time, but as soon as you have multiple instances, it starts becoming absolute chaos. In fact, the creator of Paperclip is a guy who goes by the name of Dota. a pseudonym and he's incredibly mysterious. But he was literally living this problem. He was running an automated hedge fund at the time and had pretty much 20 instances of Claude code running. And he had all these tabs open. There was no sharing of context between these different agents, and it was super hard to track how much they were spending, what they were specifically doing. Not to mention the switching cost of seeing the result of one and you've ran 10 at the same time, and now you're kind of all mixed up. And that's exactly what Paperclip is built to solve. So if platforms like Openclaw and Claude code are kind of like the employees, then you would think of this as the company they're actually working for. There's an organizational chart. Each one of these individual employees has a role, CEO, CTO, CMO, and thousands of other roles that just start with C's and sound super fancy. So Paperclip gives you a dashboard that allows you to manage and oversee these different employees working with each other. The employees can actually create their own tasks and accomplish those tasks, and they also have shared context and skills as well. So if we come to their website, and yes, I'll be showing you how to set this up as well step-by-step in just a second, you're going to see that they call themselves an orchestration platform for zero human companies. Super awesome, but ultimately, if you scroll a bit further down, it kind of shows you what that looks like. It's basically a dashboard where you organize these different agents, and there's also an organization system where you can see how much these people are spending, and as you can see all of the to-do lists. So it's basically like a project management tool similar to Trello that allows you to actually manage that your business goals by organizing these AI agents. Now, the platform itself is actually completely free. It's simply just a repository from GitHub that you can actually install, and you bring your own bot. So if you work with Claude code, awesome. If you're working with Openclaw, that's cool. And any of the others that are mentioned here, including Codex, Cursor, Bash, HTTP. And so yeah, this seems like a natural next step. And the managing agents actually just pass responsibilities down the organizational charts and then create specific tasks for the employees to fulfill. And as you can see here, you can also track the cost, which is super valuable. Okay, Christian, but does this actually work or is it just a fancy demo? Well, to be honest, it's only been out in the market for about 4 weeks, so in this video, we're also going to discover that. But let's talk about a couple of use cases that the founder Dota has mentioned people are using this for. What actually blew my mind is that this is something that works really well for existing companies. That's the first thing that he kind of put a little bit of emphasis on. He talked about a security review company running automated audits with it. He mentioned a roofing company that was using it for lead generation via satellite data. And he also mentioned a marketing firm setting up multiple agents that were able to basically interact with each other. And so it seems that this is actually something that's amazing if you already have an existing process of making money. And so at this point, you're probably wondering, well, how do you set this up? It's open source. I don't code. Well, let me show you exactly how you can do that in just a couple of clicks leveraging a platform called Hostinger. Hostinger used to just be a hosting platform for the internet, but now they actually have lots of things, including website builders, hosting, domains. They even have a vibe coding tool called Horizons. What I'm going to show you right now is their VPS, virtual private servers. Basically a place that you can rent a server and run code on top of it. And that's important to know for the setup because basically, Paperclip is simply just a repo that lives on GitHub that you can pretty much clone and host on your own server. But even better, Hostinger actually created a one-click setup for Paperclip. So all you got to do is come to this link and basically click on deploy right here. And then it's going to run you through the setup steps to be able to run this on your own private server. So that's going to be your first step. You go to Hostinger, you create a new account, and then you come to this page, then you come here and you select your server. I would recommend the KVM2 for this one. We're going to go ahead and click deploy. It's going to bring you to the checkout. It costs $8.99 a month. Also, in the description below, you're going to see a coupon code that you'll be able to apply right here. You can just put it right inside of here and click on continue. Now, if you go through the payment gateway, that's going to show you exactly how to do it that way. And because I'm already logged in and already have my own VPS, I'm going to go ahead and set it up on that VPS. So I'm logged into my account now. I can go down to VPS right here, and I can see the VPS that it wants me to now set up. I click on setup right here. Going to basically ask me a couple of questions, and I'm going to say United States. Then it's going to ask me what I actually want to install, and I'm going to put Paperclip, which is pretty much the same thing as going and clicking on that page that I showed you earlier. Then it's going to ask me for my API key. In my case, I use Claude, so I would be grabbing my Anthropic API key and putting it right inside of here. So in my case, I'm just going to go to console.anthropic.com, and I'm going to find my API key. But if you're using another platform, you're going to want to go and ask ChatGPT or whatever you use how to actually get an API key. You're going to put it right inside of here. So if you use ChatGPT, it's going to be OpenAI. I use Claude, so it's going to be here inside of Anthropic. And then I'm simply going to click on deploy. Now, the project inside of here is actually deploying, and the cool thing is that I actually have Docker Manager, which you won't have to know anything about. Just literally follow the instructions that I'm giving you here on how to set it up, and then you're going to be able to run your own instance of it. We're going to want to give it a couple seconds here. It is deploying, as we can see here. And now this project has been created, and live, it is running. Now, once you see that this is actually running, you want to come into manage, and then you're going to want to click on deploy. Now, as this is being deployed, we're going to give it a little second here. And now we can go ahead and open, and we reach a page where we want to log in. You're going to want to use the credentials that you used when you created the account. So remember when we gave it the API key? That's when we actually determined the email and password. I'm going to go ahead and log in. And now we can create our company name. Mission: Build a zero-person company. It's going to bring us through some setup steps. The first thing we're going to want to do is set up the actual CEO. So we're just going to keep it on default mode for this one just so you can see how it works. I'm going to go ahead and click on that. And we can see that there are companies, agents, tasks. And once you've set it up, you're going to want to give it its first task. So hire your first engineer and create hiring plan. In this case, I'm just going to use the default. I'm going to go ahead and click on next, and then it's going to ultimately say, "We're ready to launch. Let's start." We're going to click on that, and now we reach the actual page where you're going to be able to see everything that goes on inside of here. On the dashboard here, you can see that it's hiring first engineer and creating a hiring plan. So you can actually see that it's going through the process of starting to run. It ran into an error, but it's completing certain tasks. Boom, now that has successfully been run. So think of the issues as basically being a task management tool. Routines are things that it would do on a recurring basis. Just like you, you wake up in the morning, then maybe 9:00 a.m. you're already planning out your day. So you can have agents that do that. And then the goals here, and we already set up the major goal is to build a zero-person company here. Obviously, you want to be more specific with these things, but that's kind of the overall gist. We have now one CEO, and that CEO we can go and check out. They're already running their first task, and we can pretty much see everything that's taking place here. We've spent zero dollars, and here we can assign tasks to this person here. So I can just be like, "Okay, let's hire a CMO." And then I can describe that. So I'm basically I'm I'm just kind of describing what I'm looking for here. I'm going to go ahead and create this issue, and now it's going to work on that. So it's first it's hiring that engineer, and then it's going to have a to-do, which is basically to go out and find a chief marketing officer. Down here, I can basically see the organization. So as soon as it hires successfully that first agent, it's going to come as a direct report probably to the CEO. Here, we can basically see their skills. There are pre-built skills. We can talk all about this in a little bit. There's the costs that your employees are costing you, and then there's kind of the overall activity. So here I can track the activity. So think of this kind of really like a project management tool. CEO request requested approval. Let me go ahead and check this. Okay, hiring engineer, chip code, implement feature. So it's found an engineer. If I click on approve here, it basically has approved this. And now down here, I can see my engineer, and that engineer has specific set of instructions as skills as configurations. And all these things, by the way, are going to be able to be completely adapted. And now if I check out the issues here, we can see that there's basically some issues that are going to take place. There's some issues that are already done, like this one. And so if I come into issues, I can actually see it. But most people are used to this view, right? It's like the to-do, in progress, and then either blocked or because it's waiting for my approval or something else. And so you can really think of this as a task manager that organizes everything. But there's also an organizational chart where now we see that the CEO has his direct report, which is the engineer. And eventually, we'll also have additional things, right? Now, that was pretty much how to set it up step-by-step with Hostinger. You saw it took zero technical skills. But now the big question starts becoming Boom, as we were speaking, the CMO just came in here. So now we have a CMO. That task was done. But your next question is probably, how do you actually train these agents so that they become incredibly good? Well, that's where this becomes a lot more nitty-gritty. Meaning that if these are not trained properly, then it's probably not going to give you the results that you're looking for. And for Bond, we're actually in the process of pushing this to the limits by running our full engineering through this platform. We've only been using it for about 3 days. We're going to be able to give you specific results of what we got from this later, but for the moment, it's very promising. But, now let me give you my critical analysis of the current state of this tool and what and how you should be thinking about it. Okay, so now you've seen what Paperclip actually is. We've set it up. Now, it's time to really break down what I actually think about it honestly. Okay, so first of all, let's talk about what I'm impressed with. Number one, this is a fundamental shift that I think is inevitable. It's pretty much obvious that if we're creating AI employees, we're going to have to have AI management systems that humans can interact with and choosing how much or how little they're going to want to interfere with those systems. So, yeah, I'm completely bought into the mission and I think the execution has been super solid so far from what we're testing, but I think time will really tell. Number two, I think that the government's over cost control is actually super smart. Someone actually said this, I think on Twitter. It was like it's structurally impossible to spend too much because it basically asks you to set a max and when it's 80% to the max, it basically warns you and at 100% it stops. So, for anyone who's ever heard those horror stories of people spending thousands of dollars because one of their agents went completely rogue, that's not a problem here. Now, another thing that really blew me away is the fact that you have a lot of control over the skills, over training these individual employees. I love how the skills are sort of on a company basis, so you can actually give those skills to multiple people in the company. And the idea of shared context, which is just super important when you have one ultimate goal because so far these agents are kind of doing their own thing and helping towards a big goal, but they don't really all understand or share information around what that big goal is. So, I think that is super, super interesting as well. Now, another exciting thing that they've just launched is companies.sh, which is actually importing full companies with all the trained employees and implementing those as templates. You want to run a marketing agency, you'd basically just download the template and run it. But, this again is completely unproven, incredibly exciting, but still relatively unproven, so don't put all of your eggs in that basket for the moment. Test it out. See how it works. See if it works for you. See if you're actually generating revenue before you go all in. And finally, I think probably the most important for me is that this is something that is already being deployed in actual real businesses because it shows that this technology in and itself is great, but it's actually just a bunch of fancy-pantsy, you know, nerds nerding out on stuff. And I love the idea of having real hard businesses actually being able to take advantage of this. But, now let's talk a little bit about the downside because no, not everything is sunshine and rainbows all the time. Number one, Paperclip is only 4 weeks old and Dota himself is actually very transparent about this fact. And they have so much exciting stuff coming out, but also this is the early stage of a product. So, we have to take everything with a bit of a grain of salt and understand that they're working towards the complete end mission. This ties us directly into the next thing, which is that Paperclip itself actually doesn't generate revenue because it's completely free. It's literally just a repository. And so, we're still waiting on some real case studies with money-generating companies. Some of the ones I mentioned will probably be in the first to generate money with this or to grow. So, it still comes down to are we solving a real problem with the business? And then if so, how well can this thing help? Now, another point on the negative side is that when you run multiple agent workflows, you're bound to see some errors. And even though it has a 95% success rate, well, those errors can actually easily multiply. So, it's really important to have some supervision, a little bit of intervention, quality control, and more importantly, to make sure that your agents are properly speaking to each other. This is something that we're still at the very infancy of learning about and figuring out the best methods to do this. So, my honest breakdown is that this is not today going to run a zero-person business that makes a hundred million dollars. But, it's absolutely what the future looks like when it comes to orchestrating multiple agents. And I'm a huge believer. We're going pretty much all in on this and I recommend that everyone go into this platform and set up an instance, but not just play around. Actually try to see how this thing can solve one or two of your actual problems. Start small, replace real workflows that you're currently doing inside your company or in your daily work and then expand out from there. And I think that there was an article written by Dealroom that pretty much sums it perfectly. 2025 was building AI employees and 2026 is all about building AI companies. And that's where I think Paperclip really shines. But, Paperclip and everything we just talked about is actually completely useless unless you have a certain level of product-market fit. So, if you don't currently have a company that's generating revenue, but you want to start one, then I highly recommend you check out this video cuz I'm going to show you exactly how to build one from scratch with zero technical skills. Hope you enjoyed this video and I'll see you in the next one. Let's go!