I mean, you've lived through both the Linux and Kubernetes transitions. Looking at AI today, where do you see the biggest risk of enterprises recreating the same kind of proprietary silos that we got rid of with open source and that we were eliminating over the last two decades? 

Yeah, well, certainly there's a lot of conversations around just the models themselves, right? Like right now, particularly in the US, the most capable models are, the proprietary API based models that sit behind these, you know, these sort of walls, right. 

Is that you can't deploy yourself. You can't see what, what, you know, what's going on there and so forth.