And so we called it, we started with Managed Inference and we're actually calling it Managed AI because it's not just inference. So we started by offering what we call tailored deployments. There were these deployments that were suited for large customers that had very bespoke performance needs. And then we extended that. We now have a pretty rich product offering. We have a model marketplace backed by our serverless inference service where you can come in, you can experiment with new models. We have self-serve deployments where you're able to pick a shape, a throughput or a latency optimized shape and test different models. Typically in the user journey, we see people start with serverless, then they want to move to self-serve because they wanted some kind of SLA.

Yeah, it's sort of interesting from a startup standpoint. Like you mentioned sort of GPT-3 or like early ChatGPT, those closed source solutions were starting to become critical to like a small group of people around that time. And open source existed, but it was a little bit of a curiosity or sort of an enthusiast thing. And as the frontier has expanded, particularly with closed source models, more and more open source has been dragged in as kind of like critical behind it, like if that makes sense. Like at any given point in time, including now, I think models from OpenAI and Anthropic are kind of more widely used and more critical kind of in general than open source models.