Chang Che

I get alot of requests from Chinese tech companies to profile them "like Colossus." It's b/c their coverage is very similar to the way the censored Chinese media ecosystem works.

Moonshot K3.1 release has the strongest commitments to AI safety I've seen in the Chinese labs—a sign of some self-regulation and a sign that Chinese labs are also responding to increasing AI capabilities.

They removed training data that had strengthened K3's cyber capabilities. They've committed to not open-source frontier cyber capabilities, but said they will continue to improve it. Nothing on bio risks, and general misuse a la OpenAI-Hugging Face.

Kimi K3.1 just came out... kimi.ai/blog/kimi-k3-1…

Very cool, wish you all the best!

Just out: my interview with Yinfi, a former tech worker in Shanghai turned YouTuber, on the inner workings of Chinese tech companies.

The rise of China debunked a myth in the West that freedom of speech was critical to innovation—or so we thought. What if China’s tech rise resulted not in spite of freedom of speech, but from its selective accommodation?

Yinfi tells us about the hidden costs (and hidden liberties) of China's tech rise as well as the story of how the tech juggernaut he worked for, miHoYo, abandoned projects involving him as soon as he began raising uncomfortable questions about the political system.