nx_connect_blind_abuse.nx

nx_connect_blind_abuse.nx -- CONNECT capability: CONTENT-BLIND abuse detection (CIQ roadmap exec cell; incumbents score 0 -- their NLP READS message content to detect abuse, which BREAKS E2EE). Nishi detects abuse from METADATA / behavioral signals ONLY (send-rate, recipient fan-out, reciprocity), never reading a single message byte -- so detection survives end-to-end encryption, BY CONSTRUCTION. The MEASURED head-to-head is on a labeled test population: prove the metadata-only detector achieves perfect recall on abusers AND perfect precision on normal users (no content needed). Incumbents must read content to match this; Nishi does not -> exceed on the content-blind axis. 100% sovereign, integer. license_tier: ORIGINAL expect_exit: 0