US semiconductor sanctions did not stop Chinese AI. Rather
US semiconductor sanctions did not stop Chinese AI. Rather, it produced a "monster" of bottom layer optimization, such as training DeepSeek V3 with 2,048 pieces of H800 with reduced performance. Behind Cursor and Harvey, there is also a Chinese open source model at work. Chips may be decoupled, but wisdom continues to flow. Official website…
Redesigned the pieces with LLM help.
Then wired a watcher to the GUI such that now I can play a game with an LLM. Here I'm playing a game with deepseek v4.1 flash.
I saw someone using DeepSeek's new Japanese voice package to test speaking Japanese. Don't tell me, this female voice looks like a bean bag.
I can only imagine how cute the beanbag would be when he speaks Japanese (*/∇＼*) Tested speaking Japanese with GPT before I don’t think it’s as attractive as DeepSeek😍
Anthropic is already running ~30,000 AI agents at once inside its r&d stack.
And the biggest story is NOT the number. It is the architecture keeping that fleet from collapsing. Anthropic just published one of the clearest looks yet at how a frontier AI lab is actually using
You have reached the end of the archive
All of deepseek