Data-driven models including LLMs, GANs, and voice models are generative systems, and in my opinion their societal implications dwarf those of purely procedural generative art. Despite that, much of AI art is made with uncritical enthusiasm, often from a place of power over or detachment from the source of the employed data sets. At the same time, perhaps unsurprisingly, AI art is often met with reactions of fear, anger, or defeat. Amid these extremes, Holly+ stands out to me as a prescient, relevant, and nuanced statement on the issues of AI models and artist rights, a subject that Holly Herndon and Mat Dryhurst have been deeply considering for many years. Holly+ was released independently in collaboration with Never Before Heard Sounds in April 2022, notably predating the AI art boom in fall of 2022. In a moment of synchronicity, at the very instant I was first experimenting with the Holly+ voice model in 2021, I received a message out of the blue from Holly Herndon inviting me to participate in its use for shared creative expression. Foremost, I find Holly+ inspiring as act of courage: there is vulnerability in publishing something as central to one’s identity as one’s voice. Then there’s the confidence in asserting boundaries by presenting a radical vision of decentralized administration, pursuing a practical solution to balance the inevitable use of such tools with artist rights and compensation. — 0xDEAFBEEF