AI COGNITION
SUBJECT: Knowledge Graphs & TensorNode Architecture
The Tensor Lattice
CubeX³ uses a Knowledge Graph composed of TensorNode objects — one per cell on the
8×8×8 board. Each node holds a 64-float Float32Array representing the full spatial
state of that voxel: occupancy history, threat pressure, material balance, change velocity, and
structural stability. This was the first working prototype of the coordinate-addressable memory that
became the G-ynthetic fractal lattice.
// THE EVALUATOR
The evaluateBoard() function aggregates a board score from three axes:
material value (Risk), tensor synergy scores (Reward), and morale coherence (Relation).
AI personality profiles — Bloodlust, Cunning, Discipline — directly scale these weights
turn by turn, producing tactically distinct opponents.
// SKILL EVOLUTION
Unlockable traits such as Aggressor I and Predictive Change are stored in
SkillData.ts and applied by SkillSystem.ts. Each skill directly
mutates the numeric weights within the Knowledge Graph, making the AI’s strategic
personality a live, evolving data structure.
// ARCHITECTURAL LEGACY
The CubeX³ TensorNode was the first working proof that board state could be represented as a spatially-indexed memory structure rather than a flat list of positions. Its decay functions, occupancy history, and influence scoring directly informed the architecture of the G-ynthetic Memory Cognitive Lattice — which scales the same concept from 512 board cells to unbounded semantic coordinate space.