PewdiePie's Odysseus VS Hermes: Who wins?

**Julian Goldie** (0:00)
Odysseus AI Agent versus Hermes Agent. Who wins? So today, we're gonna be testing them out side by side. If you weren't aware already, Odysseus is the new AI agent that's just dropped from PewDiePie this week, and it is like an agentic operating system, as you can see right here. So you've got an AI agent, but it's built into a full UI like, and today we're gonna be testing them side by side, to see who outperforms, whether it's Hermes or whether it's Odysseus, which one is the best. Now, if you've never used this before, you can get access to Odysseus, it's free and open source, and you can just follow the setup instructions here. If you've not got the technical expertise to set up yourself, no problem at all, you can actually just get Claude Code to set it up for you, just give it the GitHub instructions. And these are two different open source AI agents. Now we ran both, testing them out side by side. And here's what I would say, is that Odysseus is a polished single app, you run and manage yourself. Our Hermes is a stand alone AI agent. You can run it inside a desktop app, like you can see right here, or we've actually plugged it into our agent operated system. That's one of the first things that I would say is, if you look at the way Odysseus is set up, really it's designed to just be used alone, right? This is not designed to be set up and used inside a system where all your AI agents orchestrate together. It feels like you've got an AI model underneath, and this is more like a workspace where you can use an AI agent, which is Odysseus. The other thing that I would say here is, when you're using it, one thing that I found is it's a little bit messy, for example, when you're using the UI. So for example, if you're using Hermes Agent, it's going to be nice, it's going to be clean, whether you're using that in the terminal or inside the Agent OS, you can keep everything organized. Whereas for example, if you click on the email section here, you can see that actually comes out in a pop-up. The problem with that is if you're switching between all these different tools, you can quickly have three or four different pop-ups layered on top of each other, which is not that useful. Additionally, the other thing that I've seen between both of these is that if you actually look at the brain, the brain of Odysseus is built into the app. What I mean by that, for example, is if you look at Hermes Agent, they've designed it so you can plug any memory system that you want into it. Whereas if you look at the memory section and the skills inside Odysseus, you can add memories manually, but you wouldn't link this directly to Obsidian. You can import stuff, you can import your Obsidian, but it's not going to be linked to live. So for example, if you look at Hermes Agent, they've basically focused on having whatever memory you want to plug in there, it's optional and it's configurable. So you can basically plug in different memory. They have a full setup for the memory status here. Whereas if you look, for example, at Odysseus, it's more like something you type in, and then you manually import stuff. So it's not like a living memory that you share across all your AI agents. And that's the other thing that I've noticed as well. For example, if you look at Odysseus, it's really designed for people who just want one agent, one place, everything stuck in. If you look at Hermes Agent, it's basically something that can be orchestrated across multiple different AI agents. It's designed to work together with different ones, right? So for example, if you're using an agent operating system or you're using, for example, paperclip, Hermes Agent released a add-on for paperclip, so that you can integrate it with other AI agents. If you look, for example, Odysseus is more designed to just be used in isolation. Because for example, the brain that you plug in, it's not shared across multiple agents, it's just really plugged into Odysseus directly. It's one standalone app, it's not designed to be configured with everything else. It just feels like something that's totally isolated from everything else. Now, the other thing that I've seen here is, it seems to be designed for using local models inside Odysseus. Whereas Hermes Agent, obviously they have News Portal, you can use free APIs, they just released another one today, which is News Portal with NemoTron 3, so they've partnered with them. And so the actual memory system that you plug into Hermes is designed to be usable for anyone who's running local or cloud-based models.