longform thoughts
Why does almost no one publish their AI research reports? Four overdetermined reasons, and a much older tradition that suggests the form already has somewhere to live.
Expert wine judges can’t reliably score the same wine twice. Readers can’t reliably distinguish AI text from human text. The structural parallel reveals something about how labels shape experience itself.
A mode of authorship where human direction and model elaboration produce work neither achieves alone, and why it deserves its own name.
--- title: "proximasan" subtitle: "longform thoughts" listing: contents: posts sort: "date desc" type: default categories: true feed: true fields: [date, title, description, categories] page-layout: full ---