Structural Entanglement in the Informative Subspace: Eight Experiments on Why Every Direction Carries Every Concept

Eight experiments across four transformer architectures (GPT-2 124M to Qwen-7B) establish structural entanglement as a geometric phenomenon. Random Gaussian projections to d>=448 reproduce learned entanglement intensity (EI=1.50), while projections to the 7-dimensional informative rank show baseline EI (0.18). PCA reverses entanglement. Superlinear amplification: triple EI exceeds mean pairwise by 2x. Bootstrap 95% confidence intervals (2,000 stratified iterations) confirm significance across all four models. Entanglement is geometric, not learned.