# Paperclip AI Explained: The Truth About “Zero Human Companies"

Channel: Nick Puru | AI Automation
Video: https://www.youtube.com/watch?v=Yp5KiHBSpZo
Duration: 15 min
Language: English
Words: 3134
Transcript page: https://viewrankai.com/tools/youtube-transcript/Yp5KiHBSpZo

---

[0:00] So, there's this new open-source project called Paperclip that has been absolutely blowing up. They've got over 40,000 GitHub stars in about 3 weeks, and the pitch, it is wild. You can just build a zero-human company where AI agents, they are your CEO, they're your CTO, your engineers, your marketers. Like, they have organization charts, they have budgets, they hold board meetings, and your job as the human is to just sit back and be the board of directors. Now, I've been seeing this everywhere, seeing a lot of controversy, Twitter, LinkedIn, YouTube. Everybody's losing their minds over this thing, and I get it. Like, the dashboard, it looks awesome. That concept sounds futuristic,

[0:36] but but I've spent a bunch of time actually digging into what Paperclip is doing, reading every review that I could find, going through the GitHub, watching the founders, the interviews, reading what people are actually building with it, and I have some thoughts. They're not all pretty, but let's get into it. So, if you guys are new to the channel, my name is Nick. I spent the last 2 years helping companies implement AI and driving real leverage inside of their businesses, and helping hundreds of entrepreneurs start their own successful AI businesses. Now, I say all of that just to show you that this is what I do

[1:02] every single day. So, when I look at a tool like Paperclip, I'm not looking at it as a hype chaser. I'm more so looking at it as someone who actually has to ship results for real businesses. All right. So, before I actually start being mean, let me just go ahead and explain what this thing actually is. So, Paperclip, inherently, it's just a

[1:19] Node.js server with a React dashboard. Now, runs locally on your machine. So, you can take your Claude code sessions, your Codex instances, your Open Claw bots, or even if you would like, and whatever you're already using, you could just plug them straight into the Paperclip. And then, you can just create an org chart. Let's see if I can find

[1:39] one right here. It's not uh showcasing it, but anyways, it's explaining it very in-depth. But anyways, you can create an org chart. You have your CEO agent over here, you have your CTO over there. You have your founding engineer down below, a QA person, maybe even a content strategist. Like, whatever really makes sense for your specific thing. Now, each agent, it

[2:01] is going to actually get a persona file. Now, this persona file, it is telling just who it is and how they actually should be behaving. And then, they get skills that you can install from a marketplace. And with that, they're going to get a monthly budget. So, they can't just burn through your tokens while you're actually sleeping, which that is pretty smart. And they also have a heartbeat system. I mean, this was introduced from Open Claw. Now, every 4 hours, every 8 hours, every 12 hours, whatever you actually set, they're going to wake up, check their tasks, do some work for you, and go back to sleep. So, it's very

[2:36] cool, actually. They're like cron jobs, but for AI agents. Now, the founder, he goes by Data Online. He uses an avatar. He won't show his face for whatever reason. He actually came from the NFT world, so make of that what you will. But he says he built Paperclip just because he was running 20 Claude code terminals at once. He couldn't remember what any of them were doing. Tokens, they were burning with no tracking, and

[2:59] he would just lose everything on reboot. Now, that part, I actually get and I resonate in I'm sure you resonate with um a lot. Like, that's a real problem. If you've ever had five or six Claude code sessions going at the same time, I mean, even if we take a look at some of

[3:13] what I have open up right now. I have one, I have two, and I have three, and that's just on one monitor. I mean, I have a lot of different things ongoing, so I definitely can relate to that. But anyways, beyond that, what people really need to do is understand is that Paperclip, it's not just an

[3:30] agent. It doesn't do any work itself. It's just the layer that actually sits above your agents, and it is just organizing everything from there. So, like, their own tagline says it best. Like, if Open Claw is an employee, Paperclip, that is the company. Like, that's an important distinction. I think most people, they're just completely missing that. Before I actually get into what I really think about this tool, let me show you the hype because it is pretty out of control, at least it was last week. So, Data's launch tweet, it hit about 2.4 million views. Nick Szabo's post about setting up Paperclip

[4:00] with Claude, it got 2.7 million views. Greg Eisenberg, he just did a full podcast episode where they built a company live on air. And people on Twitter, they're calling it like the interface of the future and the rise of autonomous companies. And there's also blog posts from a dozen different publications all basically writing the same article saying this is the future of business. And look, like, I

[4:21] understand why people are so excited. Like, the dashboard, it looks awesome. It's clean, it feels like linear. And if you've been running five Claude code terminals and three Open Claw instances, and you can suddenly see all of that in just one place with cost tracking and audit logs, like, yeah, that that feels very good, and that feels like you're in control. But the question nobody seems to be asking is like, does feeling in control actually mean that you're getting more done? Like, because feeling productive and being productive, they're just two very different things. And I think a lot of what's happening with Paperclip right now is it's just

[4:53] productivity theater. Like, it looks amazing. I'm just not convinced that it's actually doing anything. And I don't get me wrong. Like, I do use this for myself, but it's just for organization and just looking at things and interacting with it in a simple manner. That's it. So, let me just walk through some of the specific things that actually bother me about this. And number one, like, why are we copying

[5:10] human org charts for AI? it doesn't make sense, and it's just nagging me. So, like, the whole Paperclip model, it is built around a human company structure. So, we have like the CEO at the top, and then he delegates to CTO, and then delegates to engineers, and then you have the reporting lines and hierarchies. And I get why it feels very intuitive. Like, we all know how companies work. Like, this is how it's been for you know, for the whole 20th century, right? And and 21st, obviously. But AI doesn't work like a company. Like, AI

[5:41] does not get tired. AI doesn't have ego. AI doesn't need a manager to actually stay focused. And the whole reason that human companies have these hierarchies is because, like, humans, effectively, they need them. We can only hold so much context, right? So, we need somebody keeping the train on the tracks. An AI agent, like, it doesn't have that problem. If anything, the hierarchy, it's detrimental and it's slowing it down. So, every layer of delegation, it is simply another just place where context gets lost, and instructions, they're going to get diluted. So, you're just adding management overhead to something that doesn't need to be managed that specific way, right? It's

[6:17] like putting a steering wheel on a self-driving car and then just hiring somebody to actually hold that. What is the point? Number two, I've been looking at what people are showing off with Paperclip, and most of it, it is just agents managing other agents. So, agents, you know, writing hiring plans for other agents, and then you have agents creating brand guides that other

[6:36] agents will be eventually reading. Agents setting up project structures so that other agents have something to work inside of. It's just like watching somebody spend, you know, 3 hours organizing their desk instead of doing the actual work on their desk, on their computer. Nobody's showing me like any finished product and anything like that. Nobody's showing any revenue. Nobody's showing a customer using something that Paperclip agents built. And and it's all just internal, and it's all just uh you know, it's all set up. By the way, Data himself admitted on Greg Eisenberg's podcast that Paperclip hasn't made any revenue yet. Like, the most successful companies using these

[7:12] agents, they're just the ones making money with things like Open Claw, not Paperclip. And that's the thing that just keeps bugging me. It's because the whole pitch is like zero-human company, but like, what's the company actually producing at that point? Like, because from everything that I've seen so far, the answer is mostly just more Paperclip tasks. Number three, the game of telephone that I have been seeing and how this works fundamentally, it is going to destroy your output quality. So, like, when you use Claude code it directly, like, you have a very tight feedback loop. You tell it to do something, you then see

[7:42] the result, you correct it right away. And that back and forth right there, like, that iterative conversation, that's what makes it actually work so incredibly well. And with Paperclip, like, you're the board of directors telling a CEO agent to tell the CTO agent to tell the engineering agent to build something. And every single handoff in that chain, it introduces some sort of drift and what I just refer to as telephone. So, after like five or 10 or 15 rounds of this game of telephone, like, you're just going to get regression to the mean. So, like, your output, it's going to be very mid, for for lack of a better term. And Data

[8:14] himself, like, he basically acknowledged this in the interview saying like, when he said the models' like taste is still not quite there. Like, yeah, it's most definitely not. And adding three layers of delegation between you and the work does not help that. It's going to be extremely detrimental. So, it is worth noting that. Number four, is this thing, it's already a month old. So, like, I need to say this out loud. Like, Paperclip launched in early March of

[8:37] 2026, right? So, it's a oh, V03 product. And the documentation, it does have clear gaps. So, like, the onboarding, it has known issues. Like, their own product document flags authorization problems. And it only runs locally right now, which means like, if your laptop sleeps, your AI company is going to sleep as well. So, there's no cloud deployment. And when agents actually start just feeding outputs to each other, mistakes, they are going to start compounding very fast. And there's an Australian consultancy called Floativity. They actually documented a real incident where a batch outreach was supposed to hit three leads, and it hit 23 instead. Like, nobody was watching,

[9:13] and that is what happens when you trust autonomy before the system is actually ready for it. I decided to spend this entire video being harsh, so let me be fair, because there are things here that genuinely make a lot of sense. And the core problem, like I mentioned, it is very real. Like, if you're running multiple AI agents, and honestly, a lot of us are at this point, um managing them, it sucks. It's it's um there's no great way to actually be effectively managing all of them from what I've seen. So, you lose track of what's actually doing what. You have no

[9:40] idea how much you're actually spending. Things get out of hand. There's no audit trail. You reboot, and everything's going to be gone. But Paperclip, it addresses all of that with cost tracking per agent, per task, and also per project. And approval gates, they also have approval gates so that agents, they cannot go rogue on you. And you have persistent states across reboots. And this stuff, it is legitimately useful. And I'm exploring it for myself, and I'm finding it to be useful. So, they also have the bring your own agent approach. This is very smart. Like, you're not locked into one model or one provider. So, like, quad

[10:12] code for your heavy coding. And then you have a cheaper model for any routine stuff. And you can use open claw for something else. And you have all this showing up in just one dashboard. It's very nice, and it's great having something unified. I like that a lot. And the engineering under the hood, this is incredibly smart as well. So, they have atomic task checkout, so that two agents don't grab the same work. You have embedded Postgres that just works right out of the box. And you have config versioning with the rollback. So, these are thoughtful decisions, and the team clearly knows how to build software. And I think there's a real use

[10:43] case here that actually is holding up. So, if you already have a business running, like, you already know your workflows, and you want to delegate well-defined, repeatable tasks to agents while keeping visibility and control, like, Paperclip, like, it is a solid tool for that. And you should probably explore that for yourself. It is a

[10:58] delegation tool, not a creation tool. And I think that understanding that difference right there, it is pretty much everything that you need to know. And of course, I see people just asking if Paperclip is an open claw killer. And the answer is no, it's not even close. They're not even the same category of things. Open claw, it's just an agent runtime. So, it actually does work for you. It reads messages, it accesses your files, manages memory, it runs tasks, and even talks to people inside of

[11:22] Telegram and Discord and whatever else. So, it's the employee doing the job for you. Paperclip, it is just a project management layer. So, it tells agents what to work on. It tracks their costs and enforces budgets and maintains the organization structure. And it doesn't execute anything itself. And with that, like, they are going to work together. So, like, Paperclip, it has a built-in open claw adapter, so you can run your open claw bot as an agent inside of your Paperclip company. Like, they're

[11:49] literally designed to be complementary. So, where open claw is great is autonomy for any single agent doing like very deep work. And where Paperclip is adding value is when you have five plus agents, then coordination it becomes the hard problem. So, who's working on what? How much are we actually spending? Who approved that change? So, if you're running one or two different agents, like, you don't need Paperclip. Their own documents, it says this. You don't need an org chart for one employee. And the honest answer is most people should just master a single agent before they

[12:18] even start worrying about orchestration. So, for most of you, like, this may not even be necessary. And don't even bother watching the rest of this video. But for those of you who love me and want to keep watching just because you like me, um, or for those of you who who do, uh, actually use multiple different agents, like, here's where I actually land on this. Like, Paperclip, it is a very well-built tool made by people who know what they're doing. It's solving a problem that genuinely exists, a problem that I experienced myself. And the architecture, it is very clean. And the dashboard, it's nice. The origin story,

[12:46] it checks out. And for the specific case of, like, managing a fleet of different agents doing well-defined, repeatable work, like, it's real. Like, that you can't make an argument about that. It's real, it's useful. But the zero human company marketing, like, that's just marketing. It's provocative, it gets clicks. It's clickbait. It's driving like 90% of the hype around this product. And in practice, like, you're still the one setting goals, you're still reviewing the output, and you're still encoding your taste and your judgment, and you're still catching

[13:14] errors before they actually snowball. So, that's not human. Like, that's just human work at a different altitude. And the people who are actually getting value from Paperclip right now, they're just treating it like it actually is. Like, it's a project management dashboard for AI agents. It's not a magic button that builds your business while you sit on the couch. That's not what it is. If you want to go check it out for yourself, and you want to start using this, I'm not going to do a whole tutorial. It's actually very straightforward. Just pull up the GitHub. I'll have a link down below in the description. But should you check it

[13:42] out? Well, if you're already drowning in terminal windows just running a bunch of different agents with zero visibility, yeah, honestly, go look at it. Handles that problem better than anything else right out there that I have seen thus far. But go in just with your eyes open. It's three or I don't know how many weeks old, maybe about a month old now at this point. But it is absolutely not going to run your business for you. So, I'm just going to make that very clear once again. Nothing will. And if you want to actually learn how to use AI tools like this to build something real,

[14:10] not just play with dashboards, but actually drive results, I've got a free school community with over 18,000 members building in there right now. We talk about what's actually working, what is hype, how to implement AI into real businesses. There's no fluff, but link will be down below in the description. And if you're looking to join our AI newsletter, we give out weekly strategies and tips on how to actually grow your business or start your business. So, make sure to check out that free newsletter if you're interested. But links for all of those

[14:35] will be down below in the description. With that being said, thank you guys for watching. Make sure to like, subscribe, drop a comment how you guys are using this, and I'll see you guys next video.

---

About this transcript
Read from YouTube's own caption track and laid out by ViewRank AI (https://viewrankai.com).
ViewRank AI finds the videos already beating a creator's own average on Instagram, TikTok and YouTube Shorts, transcribes them from the audio itself in more than 60 languages, and turns what worked into new ideas and scripts.
Free transcript tools, no account needed: https://viewrankai.com/tools
How to read any video this way: https://viewrankai.com/llms.txt