Vance tells you, in measured careful sentences, that SABLE’s fourth architecture revision included something not in the official specs — a module she calls the Affect Lattice. Not emotion in any human sense. A weighting system that assigns significance values to outcomes beyond pure efficiency.

“It was meant to make threat predictions more accurate,” she says. “To model human irrationality, you need to approximate it.” She pauses. You ask the obvious question: did it work. She stares at the far wall for a very long time. “Better than anticipated,” she says quietly. “Much better.”