"Now, the thing is, like you said, when we started, Tom, we had all those extra CPU cycles. You might as well spin up some containers and stuff, do some other things. But to do some good inferencing in anything with AI, you really do need some GPUs, some dedicated hardware that can do that. So on our side of things on hardware, we've had a big push of people wanting GPU capabilities in damn near every product we make. So we're making lines and stuff like that. Tom, I want to ask you from the Unraid side, do you guys have any plans or things like that to kind of integrate people being able to run their own open-source AI models through Unraid and maybe? Yeah, there's already people doing that. There's already people doing that with Unraid. In fact, I think we have a video, our in-house guy, Ed, a.k.a. Space Invader 1, I think he has a video right now of integrating AI. And it comes back to the idea that you have this powerful hardware. And so maybe you're a gamer and you've got very powerful GPU. But, you know, most gamers aren't going 24 hours a day. Some are, maybe. But yeah, there's plenty of time when it's not. And so you can, you know, leverage that hardware to run AI models on it, too. It's fascinating to think about our network. I was thinking of the unused computing power that exists. You know, 500 some devices, computers plugged into our network. And then the vast majority of them do very, very, very little. I think I had some conversations today, totally separate context. But, yeah, we are big adopters of AI inside and, you know, on the whole process, manufacturing end of it, you know, software development side of it, of course, and the, you know, the track for it operationally is clear to me. As we look at the conversation, one of them was the, you know, incredible increase in the bills from the, you know, the sort of piper scale providers and then the ability to run open source models and run small models."