Hacker News Reader

i think anthropic has some internal memorandum at every external facing document/blog MUST anthropomorphise their models

Perhaps they anthropomorphize the models themselves. Is that far-fetched?