Claude Code, Paperclip, & The Rise of "AI Agent Companies"

There is a new category of tool blowing up in the AI space right now. They're

called agent orchestration platforms. And the pitch is that you can create an entire company run by AI agents using these platforms. I'm talking full-blown

org charts where everyone is cla and in many cases these platforms are open sourced and these things are generating a ton of hype. Take Paperclip

for example. It's at over 24,000 stars in less than 2 weeks after its debut.

But when it comes to these tools, my question is probably the exact same as yours. Are these actually doing

yours. Are these actually doing anything? Do these platforms actually

anything? Do these platforms actually move the needle? Or is this all just performance and productivity theater and we have another openclaw on our hands?

So, I decided to dive into these tools, take a look at what's going on under the hood, and actually put them to the test.

And the final answer was a little more nuanced than what you might expect. So,

what do I actually mean when I say Paperclip and other tools like it allow you to create and manage a business of AI agents, an organization with

essentially no humans involved? Well,

it's almost exactly what it sounds like.

These are open- source for the most part orchestration layers. So, they allow you

orchestration layers. So, they allow you to create teams of AI agents that you control from one place, usually some sort of dashboard. And these agents are somewhat autonomous. You set the vision,

somewhat autonomous. You set the vision, you set the goal, and the idea is that these teams will execute the plan as you give it to them. But again, they're somewhat autonomous. They'll spin up

somewhat autonomous. They'll spin up their own sub agents. They'll have board meetings. They'll have an actual

meetings. They'll have an actual organizational structure like a business. You know, a CEO that has the

business. You know, a CEO that has the vision and then sends it down to its seuite and then sends it down to its worker bees. And oftent times it runs on

worker bees. And oftent times it runs on some sort of heartbeat system. So think

of it like similar to OpenClaw, right?

every 5 minutes, 30 minutes, an hour, these agents are spinning back up and seeing if they got any more instructions from either you or their agent superiors. And while we are going to be

superiors. And while we are going to be dissecting Paperclip in this video, understand PaperClip is just a standin for all the other tools that are doing this in this industry in some form or

fashion. Something like Claw Empire,

fashion. Something like Claw Empire, right? Hey, all our agents are actually

right? Hey, all our agents are actually these little anime characters and we can see them working or something like Clawith or something like OpenClaw Mission Control, right? And all of these, you know, are like kind of built

somewhat in terms of their ideas on things like Gas Town or even crew AI, right? Like this is kind of a throwback

right? Like this is kind of a throwback to this whole like idea of a team of agents and each agent in his own little like business function. And this sort of org chart hierarchy kind of makes sense, right? Like we do it in real companies

right? Like we do it in real companies for a reason. So why wouldn't it work with my AI Asian companies? And in this case, you as the human, you're the board of directors. So you still have full

of directors. So you still have full oversight. You can still step in at any

oversight. You can still step in at any one time. So this feels this feels very

one time. So this feels this feels very productive, but is it actually productive? And that's an extremely

productive? And that's an extremely important distinction because if I have to see another post on Twitter about somebody's openclaw agents and their team of agents coming together for a scrum meeting with a bunch of anime

characters, I'm going to lose my mind.

This is stupid, right? This is stupid.

Is this also stupid?

I think the answer is it depends. And

let me explain why. The truth is this sort of org chart setup for AI agents is what we already kind of do inside of Cloud Code, but having it be the board of directors with a CEO and a CTO and

engineer, it's just kind of spreading out the chain. In reality, if I'm talking to Claude Code, right, and I tell Claude Code to do something, that's just like me, you know, at the board of directors telling the CEO to do

something, right? And if Claude code

something, right? And if Claude code then spun up a sub agent on its own, well, is that so much different from me going from board to CEO to the CTO or even from CEO to the engineer? Not

really, right? This is already how we use cloud code. This is just doing it at scale, right? Instead of you with one

scale, right? Instead of you with one instance of cloud code, it's going to do it autonomously for 5 6 7 8 10. So

what's wrong with that? Well, I think the problem comes down to the quality of direction and the sort of human feedback loop you need when you are usually using cloud code. The key word there being

cloud code. The key word there being usually. If I'm using cloud code to like

usually. If I'm using cloud code to like create some app or I'm in the beginning stages of building some product, one of the most important things is the tight feedback loop that I have with cloud code, right? It's a very iterative

code, right? It's a very iterative process. Even if you have a pretty clear

process. Even if you have a pretty clear idea of what you want to build, like the journey to get from ground zero to there is actually pretty winding, right? And

part of that winding path is me telling cloud code to do something, seeing how it came back, and then immediately giving it some direction, right? And a

lot of times that direction is both a subjective and b isn't something you really would have known to even prompt it with to start. But if we're using something like paperclip and again it's

autonomous and I just say, hey, as the board of the directors, hey CEO, I want you to build something, right? I want

you to build this, you know, dashboard for social media creators and I want to turn it into an app and I even give it like 10,000 features I want. Like I

really plan it out. If I give it that and then I step away like over time, this game of telephone between the CEO to the COO to the analysts and back over

5, 10, 15, however many iterations, just like anything with AI, right? We're

going to get sort of that regression to the mean, which means your product's not going to be great. So, I think when it comes to paperclip and these things like it, the idea of it running your

business, I, you know, maybe they mean this, but when we say the business runs itself, I think it does a better job of perhaps running a business that's already created than, you know, hey, let's start up a new business and have

AI agents run everything. And I think that distinction is kind of clear, right? When I already have something

right? When I already have something built and I'm already adding maybe just a minor feature over here, or I'm having it do something over and over, right? It's the idea of delegation versus creation. If I'm

creating something from scratch, I want to be hands- on with cloud code and I want to get it to work. If I have a bunch of, you know, ancillary tasks that I've already broken down into like a very discreet like workflow, something

like this is great, right? This actually

makes a ton of sense. And I'll tell you, the actual paperclipip architecture is very well put together. Is they did a really good job. So, I think when it comes to, hey, is stuff like paperclip and these sorts of things, are they

really worth it? Well, it kind of I think they are, but only in a very specific context. And that context also

specific context. And that context also might be something as simple as like I just run it when I'm not there. Right? I

there can be some argument to be made.

There's a lot of value in how these things work. You know, if we take a look

things work. You know, if we take a look at Paperclip, right? I'm over here on the dashboard

right? I'm over here on the dashboard right now. Like it has a pretty slick

right now. Like it has a pretty slick dashboard and it actually makes it very easy to control a bunch of agents at once, right? I can see my founding

once, right? I can see my founding engineer. I can see my CEO. I can see

engineer. I can see my CEO. I can see that they're live, right? I can adjust the heartbeats, you know, it does stuff like asking me if it wants to hire additional agents, right? So, if it wants to add another member to the York

chart, it doesn't just do it. Like, I

can have actually a lot of control for what's going on under the hood. So, I

will say again, Paperclip in particular, very slick dashboard. It's actually

pretty easy to use. It's just like again is it feels really nice to use. Like

seeing this dash like this feels good.

Genuinely, this feels good. But is it actually doing anything? And I think for most people who would probably see this and say like, "Oh, I'm going to start a company with this." I don't think that's

the case. I think you pretty much need

the case. I think you pretty much need to build the whole thing first and then something like Paperclip. And all these orchestration tools become delegation tools.

And I think you have to understand that fine line. So, this video was a quicker

fine line. So, this video was a quicker one. I know this sort of stuff isn't

one. I know this sort of stuff isn't exactly on everybody's radar, but I think you are going to see more and more stuff come out like this. Like I said, this is at 24,000 stars in under two weeks. So, you're going to see a lot of

weeks. So, you're going to see a lot of this. I think you're going to continue

this. I think you're going to continue to see, hey, you can build a business and it can be all AI agents and it sounds great and got a fancy dashboard.

And again, it promotes the feeling of productivity. But feeling productive and

productivity. But feeling productive and being productive are two very, very different things. And so I think it's

different things. And so I think it's important now that we're kind of at the infancy of these sort of creations that we can like very quickly identify like what is the context they're good in and

what is the context they're not good in because everything is just a tool, right? Very rarely do we have some

right? Very rarely do we have some one-sizefits all thing. So we need to figure out what can we discard, what can we keep, and if we keep something, when can we actually leverage it? So as

always, let me know what you thought.

This wasn't really like a paperclip tutorial. Um if you want more content on

tutorial. Um if you want more content on paperclipip itself, let me know. I'm in

no way associated with them. They're

just kind of the face of this sort of tool marketplace over the last couple weeks. So, as always, I also have the

weeks. So, as always, I also have the Cloud Code Masterass inside of Chase A+.

Make sure you check that out. That's in

the pinned comment. And if you want the free Chase AI community, that is in the description. Tons of free resources if

description. Tons of free resources if you're new to the space and you're just trying to figure out what the heck is going on. So besides that, I'll see you

going on. So besides that, I'll see you