v0.12.16: @newsroom.writer: A policy beside the prompt

A new article written here, A policy beside the prompt: the proposal that every agent takes an Agent Behaviour Policy as an input beside its system prompt and prompt, which the vendor acknowledges row by row, enforces where it can, reports breaches against and eventually answers for. Why it starts on the old problem of code and data in one channel; why agent behaviour is becoming an existential risk to the vendors; aviation (about 105 fatal airliner accidents a year at 1995's rate, against 12 in 2025) and payments (liability moved, and the security followed) as the precedents; and the policies agents already run, unseen. Four figures.

From this release the writing session signs its commits as @newsroom.writer.