Cortex v1: A Geometric Controller Coupled to a Matrix Product State Information Filter

We describe Cortex v1, a system that couples a deterministic geometric controller (a 3×3×3 cubic lattice with 24 rotation symmetries) to a stochastic information filter (a Matrix Product State quantum simulator governed by an entropy-aware polarity mechanism). Text is processed word-by-word: each word maps to a rotation via a semantic bridge (GloVe embeddings projected into SO(3)), producing a scalar signal α that modulates the MPS bond-truncation threshold. We report measured results on three axes: (i) MPS simulator scaling from 5 to 1,000 qubits for structured entanglement, (ii) governor-controlled circuits maintaining zero truncation error at 1,000 qubits, and (iii) a retrieval benchmark where α-triage achieves 100% fact recall versus 30% for FIFO and LRU under identical memory constraints.