NEW Agent Operating System is INSANE

Today, I want to be answering some of the latest questions I've had inside the community about agent operating systems, how to build them, how to get the most out of them, the best ways to set them up. Now, this is based on the questions I get inside of the AIPathy community every single day, and each day like this, I just create a video tutorial, answering every single person's questions to help you as much as possible. If you want to ask me questions like this, feel free to post inside the community, link in the description, or go to the AIPathyblog.com. I'm going to get straight into this.

So we've got a great question here from Danish and Danish was asking, you know, what's the best way that you can create like an always on automation agent sequence, right? So exactly could use, could use, could use, for example, you know, Claude or whatever you want. And here's an example. So Danish is asking about the, some folks have agents that work continuously.

Easy, one keeps thinking of ideas, then another one develops it, another tests it and things keep shaping up. And he's tried the goal functionality, but that is the finite activity, which is a very good point. So how do you set this up instead? So the only best way to do this, and this is a great example inside an agent operating system.

So we, for example, have Hermes Oracle and that is running 24 seven. So every 24 hours, it's going to look through the latest headlines, pull that in from Twitter, come up with ideas, and then we can come up with content for our website just by clicking this button, if you draft social media content using this button. And so it's really on a continuous loop, which is really instead of the goal where we view a goal, it doesn't loop, but it loops until the job is done. So how can you set this up?

So one of the best ways you can set this up is by going directly to Hermes or to Claude or Chachipity and asking it to create a scheduled task and what this scheduled task means is that it's running 24 seven. So let's say for example, you want to use an agent that comes up with ideas, develops it, tests it, and keeps shaping it up. So you can ask Hermes agent to actually do that. So you could say, okay, come up with an SEO content idea, then create that idea and test the app and improve it every single day, and that could be locked in as a scheduled task.

So you basically do the same thing inside. Claude has scheduled tasks as well, and it's run on routines. It's called inside Claude. We also have inside Chachipity, something called scheduled as well.

So these are all examples of how you can have this running on a loop, and that's the best way to set this up. Felicia was talking about installing the agent OS. So you may have been in the community for about a week, really happy with their career so far, really enjoyed the vibe so far, so I think it really lets users know that this is people sharing their journey and what they've created. Let me show you another example.

So this is pretty amazing. Someone was asking like how to automate LinkedIn followups the other day. And one of our members, Sheena, who is an absolute rockstar, she created the example for your creative. So she actually built the automation for your creative.

That's probably the best way to describe it to you. So you can automate and build whatever you want. You can see an example right here of the LinkedIn followup agent. And whatever you want to build into your agent operating systems, you can.

All you need to do is just describe it, iterate it, test it, and improve it. And that literally takes way less time than you ever imagined. So setting up this automation, I was dreading setting up the original lead. I thought, hey, it's going to take like an hour or two to sit down and build this and everything else.

And then we did it in Fable 5 and literally Fable 5 built it into the agent OS in like the space of an hour or two, and that was it. I've never really had to touch it since. And it just runs on a loop automatically. So my point here is that if you're thinking about building automation, if you've got something to create, it's going to take way less time than you ever imagined.

Also something that's really cool here, Lee was sharing about how to create your own chief of staff. So someone who starts each day by briefing you, telling you what actually matters, keeps projects from falling through the cracks, you've actually broken down exactly how to do it. So essentially what you're going to do is write your own operating IOU, basically get your DDA to date. Reconnecting the files, the trackers, and everything else.

And then this is the actual prompt that they've set up to create their chief of staff. So you could take, for example, this prompt right here, and you could plug that into Hermes. Just go inside the chat inside Hermes and be like, okay, build this for me. And it can actually do it for you on a schedule, day to day.

You just, you're at that point now where it doesn't even seem to be technical anymore to build these sorts of automations. Like for example, creating an AI chief of staff. You can just go straight into it, take a prompt like this and create it. It's an amazing share though.

Thanks so much for sharing. Another great tip is reducing your Fable 5 token usage. So what you can actually do is use something called the superpower skill, which is available on GitHub. It's a free open source project.

And you can run this prompt. You can see Alan shared the full step-by-step process on how to do it. The other good option is for reducing your token usage. You can use Editing, Kaver, and UnityKaver.

Superpowers as well. You can see that Carl has been using superpowers as well. It's a great way to be more efficient and just reduce the amount of tokens. So if you're hitting token usage limits on chat GPT with GPT 5.6 or Fable 5 or any of these frontier models, this is a really good setup.

So he was asking about building your own customized agent, you know, for whatever you want, for a specific job. Like for example, customer service. So, and this is the approach that we actually take to pretty much everything that's in our business. So for example, here's how we do it.

We have the SEO agent that can do keyword research, then generate the content and actually deploy it across multiple different websites. We did the same with video. So I think this is the way forward, honestly, is like, you look at what you're spending your time on. Then you ask an agent like Cord or Hermes to automate it for you.

And it literally just guides you through the steps so you can build it. That's how we created, for example, Hermes Apollo, which is a voice activated version of Hermes that can literally listen to what I say, build stuff in real time, and even open up tabs or it can open up apps inside my computer. So it's pretty wild what you can do with this stuff. The third question that we got here was from Gabriela and Gabriela was saying, you know, I don't want a super grok subscription, but I do want to use the radar system that we have here for pulling in the latest news.

So if you don't want to use the grok to find the latest information, then what you can use is a Google app space API. That's good for the web search. And also you could use a Firecrawl, which is a free API as well. And then you can actually ask the agent that sets up the agent OS for you.

When you install the zip file, you can ask it to change the category of news. So you can use it for any industry. It doesn't just have to be for AI. Another question we got was from Adria and Adria set up the agent OS, but it's looking to use it on loads of different devices, right?

So for example, can I connect it to your phone? So you can use tail scale, another option that I've seen a lot of people use it with, and you can see Adria is actually using it with VPS on tail scale to access the agent OS on their phone. This just means that basically you can use this whole system that you can see right here. But on your phone, instead of on a laptop, so if you use it on a VPS with tail scale, the option you have is you can use it with Cloudflare.

So Cloudflare have an option for this too. But yeah, tail scale and Cloudflare are the two big options here. So you can see, for example, John set up the VPS for the agent OS on mobile. You use it with Cloudflare and that's the other option.

So either of those is pretty decent. The other question we got here from Jason is how to reduce the amount of tokens they use, so they keep getting it with GPT 5.6. So if you're having the same issues, there's a few options for this. We actually have a token minimization playbook below.

So there's a token minimization playbook that you can use. Some of my best tips are testing out quad superpowers. That's pretty good. That's a skill you can use if you use a quad.

But then if you're using GPT 5.6. So then there's a combination of R2K, which is an open source project for reducing the amount of tokens, headroom, caveman, caveman basically means that you reduce the amount of output tokens because it speaks like a caveman and speaks less to you. It's less talkative. You also have options like you can change the effort levels and you could actually use your less powerful nodes for the hard glide work.

So if it's just like a task like create content, you actually don't need a frontier model for that. And you can just use a, an older model like GPT 5.5 and it'll still do a really good job. Additionally, some other stuff that you can do is you delegate, you get tasks to a model you just have a CLI for, so if you're doing something like if you've got a task where you just need to create the website and the backend for it, then you can actually ask GLM 5.2 to do that for you and ask GPT 5.6 to orchestrate it for you. And that way you don't use GPT 5.6 for everything.

It's just like the brain that orchestrates your other agents. Yeah. And token anonymization is quite above. Actually, there's a ton of different options for reducing the amount of tokens you use.

We've got a great question here, which is Hermes or OpenClaw. So Marvel was asking about this. Yeah. What specific tasks?

And you know, if you're doing a big game, which would you go for? So number one, which tasks does Hermes specialize in versus OpenClaw? I would honestly go for Hermes to do everything for you. It's much smoother to use, much easier, much simpler.

It tends to do everything first time round. And I think it's better to just focus on one instead of using OpenClaw and Hermes, because they both do very similar things. Now, if it was a Hermes versus Hermes, I get that question. But if it's just Hermes versus OpenClaw, they're basically the same thing.

It's just that Hermes is a lot smoother and faster to use. So I would just pick Hermes. And then when it comes to connecting Hermes to the engine OS, you can't use your cloud subscription via the CLI, but you can use OpenRooter or you can actually use OSOA, which means that you don't need to pay for API. You can just log in.

As well. That's a good option. Or GLM 5.2. If you want to keep your costs down, GLM 5.2 is pretty cheap on the coding plan.

Then you can just plug that straight into Hermes. There's also some free options. So for example, we actually built out a free AI coder into our system called And that is free API. It has like 90 different providers.

You can just plug that into the system and use it for free to build a code, whatever you want with Hermes too. I use for the GNOME GPT 5.6. So inside Cloud Code, we actually built that out recently. So we've got a system inside the engine OS where you can switch between GPT 5.6 sole or GLM 5.2 inside Cloud Code, which means that you get the best of both worlds because Fable 5 may not be on the subscription forever, but also you run out of tokens pretty quickly on Fable 5.

If caught. So instead, it gives us the most powerful harness, which is gate code. And if they didn't do it, it has GPT 5.6 sole into that. And then if you ever run out of tokens on GPT 5.6 sole, you can switch to GLM 5.2 inside Cloud Code as well.

And you can get that inside the engine OS. And we also have a tutorial on it right here. Let's basically for any questions, if you have ever seen a question, thank the community inside the AI Profit Boarding today. Hopefully that helped you.

Lots of great questions there. And if you want to ask me a question like this and get our agent operating system, you can get that inside the AI Profit Boarding, link in the comments description or go to the AI Profit Boarding dot com. You can also direct message me and, you know, ask me any questions as well. Directly inside the Cloud Suite, you get access to any of our best training.

So if you could hate the GNOME, you don't need to give it to Xperia 8. And to get our agent OS system, you can get it here with a video tutorial and see when it was last updated. You get a full guide for installing it. And then also there's a file to install too.

And you also get access to all of our new best trainings with video tutorials and step-by-step guides, as you can see. Inside the calendar, you can jump all your coach calls, chase, create the other menus, ask questions, inside the app, you can get to any of the AI Profit Boarding systems like the agent OS. We also have over 199 pages of testimonials, wins, people learning, growing, sharing, amazing stuff they're building inside the AI Profit Boarding. So, you know, if they can do it and I can do it, then you can do it too.

You don't need to be technical to use AI. You just need to get started. So OTSI. You can head on over to the sketchy AI Profit Boarding dot com.

Thanks for watching.