Kevin Xu @ kevinsxu's article is very interesting: Why do some Chinese AI laboratories need distillation? -The Anthropic report named three Chinese AI laboratories, DeepSeek, Moonshot, and MiniMax, as conducting distillation, but Kevin believes that the main reason behind this is the lack of high-quality data from these independent laboratories, which is a desperate move. In contrast, the internal laboratories of large Chinese technology companies, such as Alibaba Qwen and ByteDance, can obtain real-world data through internal business units and have a lower dependence on distillation (so currently Anthropic has not caught their hands on them on the spot); -Kevin Xu proposed that China does not have a mature data industry to support AI laboratories - this is very counterintuitive, as the common view is that China has a data advantage, after all, China has a large population, generates a large amount of data, and lacks data privacy. Kevin's view is that for some vertical fields, such as manufacturing and supply chain, this view is correct, because China has a large number of manufacturing and supply chain, but for the training of large-scale cutting-edge models useful general knowledge or information, most of which come from the Internet or structured knowledge work, the Chinese team has no advantage. (Many Chinese teams have given him feedback that using some domestic data suppliers has poor quality and is a waste of time) -He believes that the anti distillation policy has the greatest impact on independent laboratories in China, but it will not fundamentally change the competitive landscape, as smart students will eventually surpass teachers, regardless of whether distillation is allowed to exist.