Hermes Agent OS Is INSANE (FREE!)

**Julian Goldie** (0:00)
This agent operating system is genuinely the best thing that I've ever built with AI. I'm going to show you exactly why, and why I think you need this, and also what I've been working on recently. If you see my other videos about this, just been sharing what I'm working on day to day, showing you how I'm building out, showing you what I'm doing, and then hopefully this inspires you if you're watching it. Also, I hope that it kind of helps you figure out, okay, how do you take everything that you know about AI and turn it into something really interesting and powerful, right? And I'll give you some examples. People are building with Claude or OpenClaude or Hermes, but they have no idea how to use it, what the use cases are, why they should use it, et cetera. And so that's why we actually built out this whole system right here. So what we've got is basically everything that I wished we had, and then I've put it into a system. So for example, we've got an agency group chat here where we can chat with AI agents. We have Paperclip, which is an AI company full of teams of agents who work together, build and create things, and they work for free because we're running them on free models, right? Which is pretty amazing. And then this is an autonomous AI company that I can just go in, build stuff with. We've actually got a full Paperclip section down here where we can organize our teams. We also have a to-do list. I'll come on to this stuff and explain it in a minute, explain what it is, why I built it, etc. We've even got a galaxy of like my second brain organized with AI. So let me show you another cool thing I built out recently. This is Hermes Jarvis. Let me pull this up. Jarvis, you there, mate?

**SPEAKER_2** (1:27)
I'm afraid that's gone a bit garbled, sir. Perhaps an auto-correct mishap or a caffeinated thumb.
Could you clarify?

**Julian Goldie** (1:35)
And what we can actually see here is like what we did previously. So this can actually build stuff out for me. You can see some examples of different apps that we've built. If we want to open them up in full screen, we can like so. And then you can see that we can actually use this pretty cool meditation app.
And this is all stuff that Jarvis just built for me with my voice. Here's another example. So this is like a galaxy visualization tool.
And then we can change how it works and what it does. And it's just astounding that you can build this out. And so this is another example. This is a meditation app, which we've got ready to go over here. We can change the sessions. We can organize how long we're going to meditate for. And I can go back and forth with Jarvis here. Let me show you another example. So if we pull this up and we're going to warmode this time, just so I can show you it full screen. It's a voice activated version of Hermes that were built with Claude into our agent operating system. And again, this is just to inspire you and to show you what's possible, because I don't see anyone else doing this. I don't see anyone else creating this sort of interesting stuff. And I think it is mind blowing. So let me open this up as well. Hey, can you just tell me what we worked on today? Just check my Obsidian memory and give me like the 8020 in a short sentence.

**SPEAKER_2** (2:49)
You spent today tidying up video slash note content around Opus 4.7, fixing transcript typos, Claude AI, Andro, Claw code and correcting sheet errors M27, OpenClaw before exporting, sir.

**Julian Goldie** (3:06)
Let's try another one. Open up Google for me.

**SPEAKER_2** (3:09)
At once sir, Google is now open and ready for your perusal.

**Julian Goldie** (3:13)
Can you build out a Japanese flash card game for me?
Now, normally that's going to take a few minutes to do. So what I actually like to do in the meantime, is just go and do some other work in the meantime. So for example, like if I need to generate, let's say for example, I need to generate some thumbnails for YouTube or whatever. I could do that inside this section here, and then we can just create whatever we want. I cannot upload an existing one that worked really well for me.

**SPEAKER_2** (3:37)
Built and running sir, have a look.

**Julian Goldie** (3:39)
Now we have the flash card app as you can see, ready to go. This can build stuff, we can also see the full history of everything that we've done. Previously as well as we scroll down, we can type in commands and get Hermes Jarvis to do stuff. We can preview everything that will build previously, and you can see there's so much more to this. These are just a few little things that were built out with Hermes Jarvis, and this powerful voice agent. That's the first time I've really seen a voice agent that's useful. Most people are using a voice agent just to have a little quick conversation, but you could do that with ChatGPT. You can't really build stuff and you can't really see stuff that you've created or you've seen doing that. And then also what we have here is the video image voice stuff. This is super useful just because we all need to generate more images or videos at certain times just for social media. Top of that, there's some other stuff here, but you've probably seen my other videos if you're watching this, so I won't go too deep into it. And then what we've also done, and this is something that I really love, is the ideas factory. So I can plug in an idea here and it will go from stage one to stage two to stage three to stage four. Now, these are always improving. And what I mean by that is every system that I'm showing you today, we just make better and better over time. So, for example, when I was previously building out this stuff, one thing that I found was the first time I built it, it wasn't working properly. And then the second time I built it, it wasn't planning properly. And then once I iterated it even more, it would then finally work and it would start building stuff out, but it wasn't that good because it was using the wrong model. And so we switched the model to Minimax and then it started creating awesome stuff like you can see right here, right? And so one thing that I want to say with all of this is as you build stuff out, it might not be good until the 10th or the 20th or the 30th version because you have to test, you have to iterate, then you have to give feedback to Claude, then you have to test it again, then you give more feedback and eventually you get something that's awesome, but it takes time. And so this is something where I see it as a system you build rather than a workflow you set and forget, if that makes sense. For example, at Hermes Jarvis, it gets better and better based on the feedback. And bear in mind, it takes time when you're using Claude to build out all this agent operating system because you have to go back and forth. You have to wait for Claude to create something, then you have to test it, then you have to give feedback. And so you might think, oh, that's just a quick prompt, but actually it's more like a system where you have to go back and forth over and over again for hours and hours. So that is one of the things that is a downside to this, but it's a lot of fun. I love creating all this stuff. If we go back to this system here where we've got Goldie Labs and the Autonomous AI Company, we have a CEO, we've got all these different AI agents working together here.