If you were to do angel investing, what types of projects would you look at now? Angel investing is the hardest to do; many funds have a DPI of less than 1, and if you don't invest well, you basically won't be able to raise money. So if it were you, how would you choose? I consulted a friend, and he told a story about a technological revolution cycle. Starting from the last round with Apple, first, researchers had some results, prompting revolutionary progress in underlying technology. Then came the iPhone moment. This was followed by a continuous construction of network infrastructure and the development of some underlying ecosystems, migrating the original web-era scenarios to mobile terminals. As more and more scenarios were migrated, when the momentum slowed down, people began to innovate new scenarios, leading to the emergence of giants like Douyin and Pinduoduo. When the application side couldn't innovate anymore, they started to innovate content, with countless people rushing in to create content. This route can also be referenced in AI. New technological revolutions are like when GPT-3.5 appeared; in the early stages, it was all about competing on models, RAG, infra, so those doing foundational infrastructure will be considered for investment. Because in the long run, the ecosystem needs it. In every track, there will always be 1 or 2 leaders that can continue to develop. However, we have not yet entered the old scenario migration phase; it seems that the current application layer still far from meets the practical requirements, so we won't consider investing for now. It's very likely that once the technological momentum arrives, the migration of old scenarios will happen much faster than now. What entrepreneurs are doing now often becomes technical debt, and it's hard to pivot when the ship is large. So currently, I won't invest in projects in this area. I feel that everyone's thinking is quite similar; my consideration is that AI applications are not mature, so I would choose to directly monetize AI traffic or provide service solutions to B, rather than creating something myself. But actually, AI applications can also be split; there are often underlying application categories. For example, projects that are closely integrated with the infra layer and have strong enough abstraction capabilities to cope with changes are also worth looking at. This thinking also applies to the spatial computing revolution brought by visionOS; it just feels like visionOS has just begun to reach the new technological revolution ---> foundational infrastructure stage.