New Hermes Agent Update is INSANE!

There's a brand new update from Hermes today, and it is v0.17. So I'm going to talk you through exactly what it means for this AI super agent. Bear in mind like Hermes agent is now the most popular AI agent that's open source in the world. It's free to download, free to set up.

And you can see that v0.17 just dropped a few hours ago. This is called the Reach Release. And I'm going to explain exactly what it means, how to use it, give you some actual use cases so you're not just looking for a changelog and actually make this very, very actionable for you so you can get the most out of this. Now, you can see there's a bunch of stuff.

If you use iMessage, then you would start plugging this in. Basically, you can use Hermes on iMessage now. Most people I personally don't know don't even use, sorry, most people I know are not using iMessage. But if you are, you can now plug in Hermes into it.

Let's talk about the stuff that's actually useful right here. So the first thing that I can see on the changelog that's really, really useful is sub-agents. So you can now kick off sub-agents and have a team of AI agents working together asynchronously. What does this actually mean?

So basically, you can fan out and delegate a task to multiple sub-agents and Hermes will do it automatically with its delegate tool. This is probably the most important part of the release. So what that means essentially is, for example, if I'm like, OK, I need to do some SEO. And this is just an example.

You can do whatever you want with this. So we could say, OK, here's my keyword. Create five articles at once and we'll have five sub-agents, which you actually use right here to create five different pages at once. Using this process.

And so you can have five sub-agents working in parallel. So basically how this works, you go into Hermes, you like, right. I need keywords for this. It deploys five sub-agents and then they go live to your website, as you can see right here.

So you don't need to control or manage the sub-agents. Your main Hermes profile will manage it and then it will get published. And the great thing about that is you can have like five workers that finish the work at different times. They don't just finish together.

And then each result comes back inside the chat once it's done. Pretty powerful stuff. There's also an editing images thing here. So basically you can edit images now with Hermes 2, which is pretty cool.

So that means image to image editing, as you can see. So you can do text to image, you can do image to image and you can basically edit anything inside one place using your Hermes agents as long as you have the image model plugged in. So, for example, if you've got Minimax or Grok or ChatGPT plugged into Hermes, they can now edit images for you. If you want to see an example of that, if you look at this YouTube thumbnails tool that we've created inside our agent operating system inside the thumbnail studio, we can plug in an image, we can tell it edits and then it will make like three or four different versions of that image at the same time, which is pretty amazing.

So you can see all these different variations of the same image so that we get multiple options to choose from. And they're all generated automatically with AI. We've even set up a timer inside here so we can see how long it took for each image generation just to make sure it's not taking too long. Now, another big thing is the automation blueprint.

So basically what this means is you can get templates that you can plug into Hermes to automate stuff. Let me show you an example of that. So this is something I call the Hermes automation engine. And essentially, as you can see inside this video, you can take a template from Hermes News Research and have loads of templates on the website.

And then you can take those templates and plug them into your Hermes. So here's an example of a scheduled task that we could create. And you can see the details here. And so basically what this means is these are like pre-coded templates for automating quite a lot of different workflows.

And then you can just copy and paste that plug into Hermes and boom. You've got the automation set up. It might ask for a couple of APIs, but basically the 8020 is fully automated. So it could be like competitive research, scouts.

It could be, for example, researching the latest AI news digest. You can just copy them, paste them. Now, why would you use this? Because it's been engineered by the founders and creators of Hermes agent.

So, you know, the automation blueprint is going to be really powerful and it saves you time. Now, also, what's pretty cool here is you can use Curse's new model through your Twitter or Grok subscription. So what this essentially means is you can use something called Grok Composer 2.5 Fast. And if you already subscribe to Twitter or if you already have a Grok account, well, you don't need to pay extra and you don't need to pay for APIs.

You can just log in with OAuth. Here's an example. So if you go to Hermes here and then we go to manage, we can go to our models. Then we can change the model.

We can type in Grok and then you'll see Grok Composer 2.5 Fast. Now, why is that? It's because Grok have just acquired Cursor. And so you can use their fast model inside here, which is pretty awesome.

And then you just click on that, switch model, boom, shakalaka. We can reload the model. And that is now set to Grok with Composer Fast. Perfect.

And we don't need to pay for that model. So now if we say, OK, what API are you using inside the chat? You see, it says this CLI session is running on Grok OAuth with Grok Composer 2.5 Fast. Now, there's also something called the full profile builder.

What this means essentially is you can have multiple different agent profiles. So if you look over here at the top, inside our agent operating system, we have multiple different agent profiles depending on the API and depending on the skill. So we've got an SEO lead, we've got a game developer. We have GLM 5.2 as a separate profile.

A local agent called North Mini Code plugged in. And these are all different agents. So we can call upon a chat in parallel whilst we are using Hermes. And so it's really easy to create multiple different profiles.

Now, how do you do that? You just go back to your dashboard and then you would go from here to profiles. You can create a new profile and you just add in the name. Where are you going to get the config from?

So you can actually clone a previous profile, add in the description, add in the model that you want to use. And that's basically it. Then you hit create and you can manage all your different profiles inside this section right here, which is really useful because then you can have different profiles for APIs. You can have different profiles with different skills.

You can have different sole MD files for each agent. And you can have them all working together on a Kanban board. Let me show you an example of that. So if we go over to the Kanban section and this like a dashboard that we've created for Hermes and all our other agents, it just makes it way easier to organize everything.

You could just create this yourself. But if you want to get minus inside the AI Profit Boarding, link in the comments description or go to the AI Profit Boarding dot com. And if we go to the content section here, we have, for example, a content judge, a content editor, video builder. These are all separate profiles that we have to complete tasks, which is going to lead me to my next point, which and probably my favorite part, which is agent loops.

So most people are using AI by prompting it. But the problem with that is that if you are prompting everything yourself, well, then you have to quality control everything yourself. You have to actually think of the prompt. You have to manage the models.

You have to check everything and then give feedback. That costs a lot of time. The way that we use it instead is that you set the goal. The builder improves.

Then you have a judge that grades it and that goes round in a loop until it's finally done. Now, there's two ways that we've actually done this and two systems I've built around this whole idea that's just dropped from Hermes. So method number one is that we can use Hermes like so. We can give it a goal.

We use a free API for the builder. We can set how many rounds it gets a chance to complete the work. And then after each round, the judge will score out of 100. And if it doesn't hit 90 or more based on the goal and the target.

And this is, by the way, a separate impartial adversarial judge. We can change the model here and they can judge the work, which means that the loop goes round and round until finally the quality standard is hit. And that is a self-improving loop. So you can see how that works.

You got the builder and you have the judge. And for example, if I was doing this for SEO, we would go round and round and round until the content passes the quality score and then it would get published. But you see how I don't have to get involved. I don't have to manage the agents.

I don't have to give feedback. Don't have to quality control everything because the judge is handling all that. Basically, this profile right here, which is a separate agent profile, manages the whole process that you would normally step in for. And so you don't need to prompt, no feedback.

And if it fails, it fixes based on the notes the judge gives the builder and it loops round, round and round, which is really, really powerful. So we actually ran a test with this. And you can see that round one scored 62, then 71, then 84, then 93. If you want to see an example of a blog post we actually create with this, you can check it out right here.

So this is the blog post that we fully published on autopilot with our SEO agents. And again, we don't have to quality control this because we've got our agents actually checking if it's good or not. And you can see that the quality of work is pretty nice. This is way nicer than a human could write it.

It looks good. It's formatted beautifully. It's got a nice video embedded and it's just created exactly how I want it. That's what we need.

Now, another option for doing this, and this is method number two, is that you can have a Kanban board. So you can run a whole team of Hermes profiles with a judge. And this is really a way to scale the whole process for loop engineering, but with a team of agent profiles because of that new agent profile update that came from Hermes this week. And so the Kanban board runs many Hermes profiles at once.

And that could be a researcher, a writer, a video producer, a judge profile, the grades work before it gets done. And so that loops round and round and you drop a goal into triage inside your Kanban board. So you can just drop a task here. It'll get triaged, sent to to-do lists.

And then the work from the researcher, the writer, the video producer goes to the judge. If it's good enough, great. It gets shipped. If it's not good enough, it loops round.

Now, if you're wondering, okay, what's the quality of that? Well, you've already seen the blog post and that looks pretty good. But also you can see that we've generated a video too. And that video looks pretty nice too.

So this is completely AI generated video that looks super nice, nicely formatted, et cetera, nicely edited. And that was completely done using the same sort of system. So loop engineer is probably one of the biggest announcements this week in terms of what you can do with Hermes and how powerful it is. And there's two separate ways to do it, which is the Kanban board system and the loop engineering system.

We've got both inside the agent operating system and that's basically it. Now there's a bunch of other stuff, as you can see right here, you know, there's providers, there's new model support, but I would say that is the 80 20 of what you need to know. There's some security updates as well, which is good, but that's pretty much it when it comes to using Hermes agent and everything else. We've also set up some new stuff.

Like for example, you can actually run Hermes with local models, which we tested out earlier this week. And we've also got a voice activated version of Hermes is called open Jarvis. And that means you can just voice activate your Hermes agent. It can do daily briefs for you, pull stuff up on the screen, do stuff, remember stuff, build stuff, and actually talk to you.

So as an example of that, if we go over to the Hermes section here, go to Hermes Jarvis, we can speak to this. You can see all my previous transcripts over here. You can see everything that we've built in preview over here. And then we can also put this in war mode and then we can have a wake word.

And if we have the wake word on and it's just running in the background, we've got Hermes ready to go whenever we need it. And we just have to say, Hey, Hermes go off and do this, whatever. So really powerful stuff. Hermes agent is getting more powerful all the time.

Self improving loops are basically, it feels almost like we're at AGI because it's a self improving autonomous agent. Really fun to play with as well. If you haven't already tested it. And if you want to get all of this stuff from me with the full setup that we've created, which is custom made for Hermes, you know updates like this land every few weeks.

The agent operating system inside the Air Proffit Boarding turns Hermes, Claude and OpenCourt into one system with shared memory. So every new feature like background sub-agents or iMessage makes a whole thing more powerful automatically. You get the full agent operating system zip file, the setup walkthrough, four weekly coaching calls, daily tutorials, a 30 day roadmap, 3,600 members, a member map and a 24 seven community. And loads of people are building agent operating systems based on the zip file that we give away.

So you can see all this cool stuff here in terms of what people are built, how powerful it is, how it works, et cetera. And you can get that all inside the AI Proffit Boarding link in the comments, description, or go to the AIProffitBoarding.com. And you can connect with me personally. You can ask questions inside the community.

I cover them in daily video tutorials. You can get the agent operating system that we're showing you today inside this section here with a daily update with new updates and a zip file so that you can download and install this yourself. Plus we have new tutorials based on all the cool stuff that just drops. So feel free to get that link in the comments, description or go to the AIProffitBoarding.com.

Cheers for watching.