Nillow:// Cognition / Persona
AI persona engineering for governed systems: one explicit interaction contract for the identity people meet, question, and trust.
Definition
Nillow defines the role, voice, disclosure posture, continuity cues, visible authority, and escalation language people encounter when they work with an AI system. The persona expresses the system's real boundaries; it does not invent new ones.
Not prompt styling
A decorative system prompt can make a model sound consistent for a moment. A persona contract specifies what must remain recognizable across channels, what must be disclosed, how uncertainty is spoken, and when the interaction must hand control back to a human.
Contract dimensions
Role and voice shape expression. Memory posture explains what context is actually available. Authority language makes permitted actions legible. Escalation and continuity keep handoffs coherent. Evaluation looks for drift between the documented contract and the experience.
Defensive evaluation
Evaluation can probe role override, disclosure suppression, fabricated continuity, false authority, and inconsistent escalation. Security, policy, permissions, tool access, and execution remain separate system responsibilities enforced outside the persona wherever possible.
Boundary
Nillow does not expose private prompts, memory schemas, evaluator instructions, routing rules, thresholds, or protected Oxa material. Nillow does not engineer psychographic profiles, vulnerability or persuadability scores, deceptive impersonation, sentience claims, synthetic intimacy, or fabricated memory.
Engineered in Belgium for governed systems wherever they operate.
Explore Cognition Engineering
