Mixfont's Ghost Font encodes messages in motion, noise, and decoys so people can read them while leading AI models often latch onto the decoy or fail on a still frame. It updates the old OCR-proof ZXX idea for the multimodal age, with ripples for CAPTCHAs and AI perception benchmarks.