Tools like Glaze and Nightshade popularized the idea of poisoning images before they hit machine learning systems, embedding noise that causes models to misread what they see. Signal Jammer runs that same logic live. Adversarial noise gets applied to a live camera feed in real time, before the signal reaches the generative model, so what the model sees is not you and what it produces is not you. A ghostly white edge detection pass from the original feed gets layered back over the output, just enough of the human figure left to register. Built in TouchDesigner the piece is about staying hidden in the age of AI surveillance.