This is wild. How is it able to do it in real time when flagship LLMs take several seconds to generate text? I assume not LLMs (world generation isn’t really language), but then why are LLMs so (relatively) slow?
add a skeleton here at some point
4 months ago