MEASURED operational plate: formula-tax / I1-I8 / get_kernel (YAML preserved)

MEASURED laptop-Blackwell bench + operational honesty (not LoRA; Λ=Conjecture 1)

MEASURED laptop-Blackwell bench + operational honesty (not LoRA; Λ=Conjecture 1)

stamp GPU cubins UNAVAILABLE (not ROADMAP); CPU import-LIVE stays

stamp import-LIVE CPU get_kernel 7e533ce

honest Kernel Hub README: Python present, not import-LIVE

Add Kernel Hub build/torch-universal + torch-cpu for get_kernel (yarqa-attn, version 1, GitHub 098d19b)

initial commit