Starting a business in China, the most important thing is still 2G. In Shanghai, there is a GPU company whose founder's background and product quality far exceed that of Moore Threads and Muxi, but because the founder is not familiar with 2G dealings, they have never been able to ring the bell. Pure science students find it hard to attract investors. If a founder focuses on how amazing their technology is rather than their business model and ecosystem layout, it is highly likely that their startup will fail. Pause for a second and think about why your technology needs these accelerations. Is the acceleration meaningful? This question is very similar to the many L2s in blockchain; stop and ask, do we really need these TPS and programming in different languages? Speaking of AI hardware, Looki is particularly popular in the U.S., and there are many ads on Xiaohongshu in China, but the focus of AI hardware is still on concepts first. Why do we need a separate, explicit little box to record and create smart minutes? Once OpenAI's AI PIN is launched and popularized, it will usher in an iPhone moment. In the future, Apple Intelligence will be the natural AI hardware, and this type of segmentation will eventually be replaced by systematic ecological upgrades.