> Every few weeks they leapfrog each other. There is variation within those capabilities, it’s possible to drop off the curve (Meta, for now) or fail to get onto it (Apple, Amazon, Microsoft, for now), or remain six months behind the frontier (China), or rely heavily on other people’s work (China, again)

I really dislike this narrative where it's always China = bad, and US companies = good.

These labs all copy from each other. OpenAI and Anthropic have "distilled" each other models too and routinely poach key researchers from competitors. Not only that, there's evidence Sonnet 4.6 has heavily distilled Deepseek R1 too, in fact, if you ask Sonnet 4.6 in Chinese who it is, it will tell you it's a Deepseek model.

Chinese are the only ones publishing papers on their models non stop.

The whole AI race is entirely based on blatant copyright infringements and copying each other.