I’m starting to realize the biggest risk to US AI dominance is indeed China.
Anthropic and OpenAI doesn’t have to worry about Google/Meta/xAI. They haven’t been competitive
I haven’t seen US startups or mid size companies able to create a competitive models from scratch.
The most competitive model: Composer 2.5 from cursor is train on Kimi 2.5.
But China is releasing so much capable models (many open) last 2 months: Qwen, GLM, MiMo, Minimax, Kimi.