Sure. Good to talk to you. It's a great question. And so bear with me as I answer it because I have a lot of thoughts on this. So when we think about both the user perspective in terms of what we're doing for the user and also how we run the company and how we actually build product that is related to that as well, right? Because the changes that are happening things generally in how code is written, for example, are incredible. I had a demo for example, of a product this week, in which NAI synthetic computer is basically operating at the same level and you would expect a junior or senior year intern, software engineer to operate. And that's pretty incredible, right, in terms of how much effectiveness it's going to add to an engineer [ activity ] now. We're not fully there yet, but we had [ a election ]. So the way we tend to think of it is in the following way. Right now, we believe software in general, software development and products themselves are in the same stage kind of in early days of cell phones, we were in cell phones versus landline. And you could go after and kind of do a lot more investment into mainline, right, because that's what was present then as the main way of communication. You could say, "Hey, no, actually, I'm going to go invest in software, in cell phones and build a cell phone driven environment". First, for example, because it might not be a good connectivity on online life. And so that's the kind of decision we need to make as well as we think about building products. So when we think about dating, for example, what we -- the way we are putting to say, what is the AI first base dating product like? Rather than what is popular today in dating. And do we try to replicate and mimic in our application because, again, we don't have dating features. I think it forms a much better approach because that's what everybody else is going to go. And so since we have an opportunity to build from scratch, whether those for that future rather than what might be working today, and that's how we tend to think about it. So where I think applications for AI are significant in our products, it's mainly threefold. One has to do with communications, chatting, companionship, or support for the user, whether it's coaching or mental health in either direction, that kind of whole bucket of communications, general component, I think, is pretty significant, and then there's a lot we can do there, and we are working on that. The second one is around matching and creating better connections between users, right? We have an incredible amount of information about our users. As you probably know, our users generated $121 billion chat messages last year. That's an incredible amount of information about what they like, what they don't like, what they're into, and using all that as a way to understand the use of better and potentially create good matches between users, we think it's a huge opportunity. So that's the second kind of component. And then thirdly is application health or quality to trust and safety is another big component. And there, we're already actually doing a bunch of things using AI, and we think we can improve as an application like us a social media app or dating app, there is an idea that is not allowed, right? It's against a [indiscernible] guidelines or it is illegal. The more we can do to that type of behavior before it even happens or before anyone reports it and get it when we move from the product and from the community, the better we are, right? Because that means we're creating a better experience for users. So actually using quite a bit of AI already. We have done, I think, a really good job in proactively finding that behavior and getting it removed to the point where our user reports on illegal activity in the app or activity against the community has dropped 50% over the last 3 years. And that's because we're actually able to get ahead of it and get it off the half before it becomes a pump for anybody. And so that's the third bucket of where we think opportunity is significant with.