Transparent language
AI outputs use careful phrasing like “might relate to” and “seems connected to” so users always remain the final interpreter.
No invented memory
The system never fabricates events, emotions, or relationships that are not grounded in the saved fragment set.
Gentle resurfacing
Resurfacing is framed as an invitation, not a demand. The product avoids aggressive urgency and manipulative prompts.
User choice stays central
Every resurfaced fragment supports multiple paths: act, reconnect, archive, or let it go.
Prototype settings mock
These switches are rendered as read-only prototype preferences so judges can inspect the trust model without guessing whether a control is interactive.
Read-only prototype preference
Read-only prototype preference
Read-only prototype preference
Read-only prototype preference