MindForm — Cognitive Patterns: A Stateless Interpretive Layer for Experience-Driven Personality in LLM Agents

Abstract

The two preceding papers in this series document personality as state: a temperament and trait vector a character is born with, and the values, moral outlook, beliefs and habits they earn by living. Both described a feedback loop in which that state colors how future experience is read, and both deferred the mechanism to this paper. This paper documents the Cognitive Patterns node, which implements it — and which, uniquely in the architecture, holds no state of its own. It owns no vector, persists nothing between turns, and appears in no slot of the personality structure. What it does is read every other node's state together and decide what an incoming experience means before any node is permitted to form from it. Five bounded tilts are applied in fixed order: a behavior gate that governs how much of an experience gets in at all by scaling intensity alone, followed by trait, memory, drive and self tilts that color what it means. Each tilt exists in two encodings — vector arithmetic on the deterministic path, and a short natural-language brief appended to the prompt on the model-primary path. We explain why concentrating these tilts in a stateless node is preferable to distributing them across the nodes that supply their inputs, and why the resulting perception-formation loop compounds without diverging. We close with the node's principal limitations: every gain is hand-authored, the two encodings are never checked against each other, most of the Character substrates — beliefs above all — do not participate in the lens, and the node has never been evaluated in isolation.