Thanks, Felix, this is Alex. Let me take this one on. I will actually take a step back a little bit. We've been talking about the AI, and I just want to go back to something that I've discussed maybe a year or even two ago. I think it really has multi-dimensional impact on us, and look, we really welcome every single advance of this in AI capabilities from the foundational models. I think every single advance is welcoming use to us and let me explain why. So, first of all, we think AI has a huge, huge impact on educational content creation. The entire creation process is benefiting tremendously from AI. I think number one, it has tremendous implication in terms of efficiency gain, in terms of speed gain, and we're witnessing that every single day in the last two years. I think that -- and it also by the way has made things that previously seemed impossible to be possible. Leveraging this capability, you can develop multi-language, multi-curriculum, adapted learning content material. So, I think that's just something that we didn't think was quite possible. I think we are getting very close to be able to use AI to generate interactive, high-quality content in real time. Again, that wasn't even a cost issue. I think that was just basically not possible before. So, these, I think, offer really raising up the top the ceiling of what's possible while doing it, it was a tremendously different cost structure than before, right? So, I think that's just the first thing I want to really lock with everybody on the call. I think look, secondly, with every single advance of AI, we just see more possibilities of leveraging on service, it goes from instruction, to customer service to after sales service and across the board. I think in these areas, again, we have -- it's not even the possibility. I think we're pushing those every day in terms of both raising the level of customer service, while gaining more efficiency. So, I think that's also been tremendously helpful. And again, I think we're just expecting more, as the capability of AI, it gets even more obvious. I think number three in terms of research and development, we're really seeing so much more code generated by AI, that is advancing speed, gaining efficiency, we are a large research and development and education technology, and I think AI is really giving us a very different way of developing this. It's faster. It's nimble. It's you get very fast prototyping, and you get more efficiency not only from using AI to code by itself, but that nimble and fast prototyping is actually removing a lot of friction in the R&D process versus before. I just want to log those three. Those are things that we probably didn't discuss in a whole other detail in the last few calls, but I think those are actually very tangible, meaningful, and we expect more benefit, and more impact from those. So, if going back to the integration in learning scenarios, look, I think we're increasingly seeing AI becoming, a learning companion. I think the shelf to chat assistant, that really embodies that fact that it is there. More and more students who are native in this AI age are getting comfortable to get on the spot help and learning support from AI as they learn. So, currently, DeepSeek V3 service is one of the foundational models for our Mass GPT or in Chinese. Beyond its general intelligence capabilities, we've also finding that on learning content, enhancing its ability to support really subject specific tasks. And also, let me add this. I think we need to fine-tune it and to support those pedagogical scenarios. For our model to be able to produce the answer is one thing. For it to be able to actually make the solution more accessible, easier to understand, liked by students, these are actually, we believe are fundamentally vertical capabilities to an education player and that's how we look at it, that we're going to continue to work and invest on those, really lowering the barrier to accessing these AI capabilities, improving the user interface, making it easier for user to provide their input to the large language model, but also working really hard to make the output from AI, not the raw straight output from these large language models, but make the output a learning purpose and serve it well. And lastly, I would just say, look, we'll also remain committed to contributing to the broader intelligent learning ecosystem, advancing new educational paradigms, supporting future schools and third-party institutions, as we've done in the past and we really look forward to reimagining the future of learning, and shape a more dynamic for thinking education ecosystem as a whole. Felix, I hope that answered your question.