ChatGPT 5.6 inside Hermes Agent left me speechless...

0:00 Hermes Agent is the most powerful AI tool ever, and ChatGPT 5.6 makes it even better. But, here's the thing. If you don't set up the right way or use it the right way, it's going to absolutely suck. In this video, I'm going to cover why Hermes Agent over OpenClaw, why ChatGPT 5.6 in Hermes dominates Claude, and give you a few Hermes use cases today that will 10x your workflow. Now, let's lock in and get into it. Before we get into setting up 5.6, I just want to do a quick update on Hermes versus OpenClaw. I get 10,000 replies under every single video, but why Hermes over OpenClaw? Are you still using OpenClaw?

0:40 You do Hermes what? Let's just get this out of the way, then I'll get into why ChatGPT 5.6. Here is why I talk mostly about Hermes right now. One, and this is the most important out of all of them, it is just more reliable. When I update Hermes Agent, it does not break. When I update OpenClaw, it breaks. So, because of that, I use Hermes Agent more. There was a string of like five OpenClaw updates that broke it and I had to spend half an hour fixing it. That never happens with Hermes.

1:08 That's why I talk about Hermes more. That's why most of my videos are about Hermes. It's just the tool I use more and can rely upon more. The updates are better. They do not break. I like their road map better. They they seem to have a much clearer vision of what they are trying to build. OpenClaw, ever since OpenAI bought them, the updates have just been less clear in what they are. They don't really have a theme. They seem to be a little bit more unorganized. The the releases seem to be a little bit more unstable. So, I like the road map and kind of vision more on Hermes Agent. And as I said, it just works. When new models come out, they seem to be available in Hermes the moment they drop. Like, ChatGPT 5.6 was available the moment it dropped. It was inside Hermes. It was not in OpenClaw. I went to OpenClaw. I said, "Hey, update the ChatGPT It said, "Hey, I have no idea what the hell you're talking about." It just works. Everything in Hermes works as expected. Now, I will say this, open claw does have two advantages and it's the reason why it's still in my stack. I still use it, not quite as much as Hermes, but I still use it. One, I just like the way it talks to me more. I don't know what they're doing over there. The way it talks to me, I just like it more. It's clearer, it's warmer. I like the way it speaks. And it seems to do less overthinking. I'll give a prompt to Hermes, I'll give a prompt to open claw. It seems like Hermes will do 10 trillion different things and give me a response. Open claw just seems to do a lot less overthinking. I think it's because Hermes does a lot more self-improvement things, which are great. I like the way that works, but sometimes I just want a quick answer.

2:46 open claw seems to do less overthinking at times, but everything combined, Hermes for me, that's why a lot of my videos are about Hermes, it just works. It's reliable. And for what I do in my life, I will take reliability over I like the way it talks to me. So, it is what it is. Listen, I use both, you should use both, you should experiment both. Figure out what's best for you, but the reason why all my content's on Hermes is because it works. All right, now what the core point in this video is, chat GPT-5.6 inside Hermes is the winner. For the first time ever, ever, I am ripping out Claude, I am putting in chat GPT-5.6. Chat GPT is officially my recommendation for the model to use with Hermes agent. Now, why is that? Well, first of all, it is smarter than Opus. For the last several months since Hermes came out, Opus has always been smarter, but like I've been trying to make chat GPT work because it works in the subscription, which Claude does not. So, it's smarter. It's cheaper. I was literally spending, I think, about $200 a day on the Claude Opus API to get Hermes to work. ChatGPT works in a subscription. You can get away with the $20 a month plan. So, it's smarter, it's cheaper, and this one's important. This is a big one. It has way better computer use. The model just seems to be more agentic than Claude. it can control your computer better, your browser, take screenshots, create videos, do whatever you need on your computer, load up local models. We'll talk about that a little bit when we get to the use case part.

4:21 Stick around for that. But, it is just overall better computer use as well. So, now there is literally no reason to be using Claude over ChatGPT. The one reason, the one little caveat I'd give is Claude, and this has been the same way since day one, I believe has a better personality than ChatGPT. I believe it's more decisive and easier to talk to. ChatGPT, one thing it does that pisses me off a little bit, is it it kind of sways with the wind with its decision-making. If you challenge it on stuff, it always agrees with you and changes its mind. Claude stands on business, as the kids say. But, these three things absolutely dominate whatever advantage Claude has at the moment. You need to be switching to ChatGPT 56 inside your Hermes agent, and for the first time ever, it's not even remotely close. I mean, take a look at this. This is my usage over the last month. Ever since I have switched full-time to ChatGPT 56 inside of Hermes, you can see the drop-off. I was literally spending hundreds of dollars a day on this. I am now saving tremendous amounts of money to get significantly better performance. All right, let's talk about now setting the model inside Hermes. The best way to do that, what thinking level you want, all of the above. What you want to do is go into your terminal, type in Hermes dashboard, hit enter. What that will do is pop open this dashboard you see here. This is a really awesome dashboard. If you've never seen it before, highly recommend taking the time to go through all the options, all the configuration you can do. But what you want to do for the sake of this video is click models. This will allow you to switch basically all the models you're using, your mixture of agents, which is a really cool feature.

6:05 I talked about this in a previous video. Make sure to check that out. But what you want to do is change the main model here. What you want to do if you are on your Chat GPT subscription is type in Codex and then choose OpenAI Codex. This will allow you to set the model to GPT-5.6 Soul. Soul is the highest version of Chat GPT-5.6. There's Soul, Terra, Luna. Soul is the smartest. It is the best. This is the one we want to be using.

6:36 Even if you're on the $20 a month plan, you could be using Soul. Actually, if you're on 20, I'm probably going down to Terra or Luna. But if you're on the $100 month plan or the $200 month plan, definitely go with Soul. You set that up and you are good to go. There is one more step you want to do here though. Go into Telegram after you set 5.6 Soul and type this in. Do {slash} reasoning medium and then hit enter on that. This is critical. You want to have the thinking level be medium, not high, not extra high, not ultra. I've tested all the thinking levels. Medium is the best way to go. In fact, the second best way to go is low. I have not enjoyed Chat GPT-5.6 on ultra or extra high. I believe it overthinks too much. I believe it's not quite as good at instruction following. I believe because it overthinks so much and burns so many tokens, it actually forgets a lot of the instructions you give it. Medium I have found to be the sweet spot. In fact, low I found it works really, really well too if you want to improve your speed and velocity of your agent, so make sure you set this to medium. If you're on the $20 a month plan, you can set it to low.

7:50 You'll still get great usage out of this. It'll still be better than Claude. Highly recommend doing that. Those are two critical things. Switch to 5.6, put it on reasoning medium. Let's talk about some use cases you can do with Hermes agent. We talked about setting up Chat GPT 5.6, which is the best model for now. Now, let's talk about some things you can do with Chat GPT 5.6 inside Hermes agent that is really, really awesome. Let's start off with number one, and that is the home AI lab. I don't care what hardware you have. You can have a Mac mini, you can have an old dusty laptop, you have a Mac Studio. I don't care. You should be starting your home AI lab. What does that mean? That is your ability to run and use and monitor local AI models on whatever hardware it is. Now, not many people know this, but you can run local AI models on literally anything. You can run them on iPhones, you can run them on really crappy Mac minis. It doesn't matter. There's a AI model for every single piece of hardware you have. But, how do you know which models to run?

8:54 Well, I got good news for you. Your Hermes agent powered by Chat GPT 5.6 is excellent at figuring this out. So, here's what you want to do. Take this prompt, I'll put it down below. Check out the hardware and specs on my computer, then go through Hugging Face, which is like the best repository for local models on the entire internet, and find the best local AI model I can run on it. Then, list the use cases that I can do with that AI model. Again, I'm putting that down below. Copy and paste it into your Hermes agent now. And what that's going to do is send off your AI agent to look at your hardware, look at your specs, your memory, your GPU, everything, and find the best local models you can run on it. Now, I've been obviously doing stuff with local AI for a while now, so I have an entire home AI lab set up. But once your Hermes agent loads up one of those models onto your computer for you, you can start doing really interesting things. For instance, I built out this local intelligence control pane. This is like my home AI lab with all the functionality. It shows me all my computers, what models are running on each computer. I can go in, I can test any of the models currently running. So, if I want to say, "Hey, are you there?" I can speak directly to the models on each one of my computers and test them out. And I can assign it a bunch of different jobs. You can see it is running jobs all day and night, just churning through code at all times. You can get really custom and do really interesting things. Build your own chat GPT type front end, build your own codex type app. You can do all that. So, use case number one, using 5-6 to find local models you can run on whatever device you have, and then building really cool interfaces for those local models. The second use case I've been doing a ton with chat GPT 5-6 is an AI game studio.

10:43 Check this out. With loot, enemies, complex 3D lighting. You kill that, and then you can loot stuff, and then actually extract. But here's the point. The point is is that you can use Hermes to be a full 3D video game studio. What I have figured out is Unity is the best game engine for this. Some people have been doing 3 JS. I've dabbled around with Unreal Engine. What I have figured out with Hermes agent is it can control everything in Unity engine front to back fully, completely autonomously. It generates all the assets, it generates all the characters, the environments, the lighting. You saw this. Look how good the lighting and the graphics are here. There's full loot. Everything was generated with AI. Nothing I had to to to different websites or anything. I just say, "Hey, build this out." and it does that. So, I've been building my own video game studio here. It's with Unity.

11:37 Unity is completely free. You can go download that use They don't sponsor this. I've just tried like every video game engine under the sun and Unity is the one where it feels like the AI can handle the absolute most part of the games. So, definitely check this out. I am a massive gamer at heart. It is one of my biggest passions and hobbies outside all this stuff. So, this is a really cool use case for me. The last one I've been doing with Hermes agent is having it keep an eye on websites for changes. One of those changes is on Microcenter site for Nvidia RTX 5090s. I want another RTX 5090. So, I'm having Hermes check It has a cron job that does this. Checks the website every single day looking for price drops on Nvidia RTX 5090s. Checking for price drops especially on super hard to get super rare hardware like GPUs, like memories is a really good use case to be doing right now. If you're trying to build your own home AI lab setup, maybe you want to build a computer, something like that, have your Hermes agent go and then keep an eye on different websites. You can have it keep an eye on Best Buy, any local electronic dealer you have. If Microcenter is near you, it's the best like electronic store on planet Earth.

12:51 Have it keep an eye on that website and it'll check for price drops and then text message you when there's price drops. So, this is a really awesome use case as well. There are literally millions of use cases. I have a 100,000 different use cases with Hermes agent I do on a daily basis. These are three of my latest ones that I really enjoy. Obviously, I have hundreds of thousands of viewers of this channel. Everyone has different hobbies and requirements and careers. I can't give every single person a use case that's relevant to them. All I can do is tell you the use case that are relevant to me and hopefully that inspires some cool ideas.

13:25 I do Hermes bootcamps every single week live in the Vibe Coding Academy. Check that out down below. It's the number one AI community on planet Earth. Link for that down below. You join it, you learn a ton about Hermes, Open Claw Codex, Cloud Code, Fable 5, all of that. It's the best decision you'll ever make. Link down below. I am so grateful watch this video. Make sure subscribe, turn on notifications, leave a like if you learned anything at all. Thank you so much for watching. I'll see you in the next video.