Concretely, that suggests practices: built-in provenance tracking, explicit uncertainty measures, multiple-option outputs, and human-in-the-loop workflows that make choices reversible and auditable. It suggests cultivating spaces—both physical and virtual—where maintenance and conversation happen together, where music racks sit beside server rows. On a late afternoon in the Unfoxall 54 room, falling light catches dust motes that the program records as incidental telemetry. A human visitor sips tea and scrolls through a reconstruction the system offered: five plausible narratives of a single event, each annotated with likelihood and source fragments. They smile—not because the machine was perfect, but because it trusted them enough to leave the table set for decision.