I once had a conversation with Professor Yao Dong, who put forward a very interesting viewpoint—open source is essentially another form of talent monopoly. He gave the example that the existing general compression algorithms are all open source, while the researchers capable of developing compression algorithms are concentrated in just a few countries, which in turn leads to other countries lacking the ability to continue research and innovation in compression algorithms. Currently, it seems that researchers in large model studies are also mainly concentrated in China and the United States, with no other country having such a large group of researchers. (Mistral is a rare case, with 65% of its capital still being French capital.) Moreover, the leading open-weight models have all been domestic models in the top 5 since last month. The only new largest model is openai-oss-120B. From this perspective, the future realization of AGI will almost certainly only emerge from China and the United States.