Comment by rfoo

Let's face it. Chinese labs made some of the biggest advancements. AND training on Claude (or GPT) output IS unreasonably effective. The two sentences are true at the same time.