<p>This paper explores the opportunities, challenges, and ethics of engaging artificially intelligent models in collaborative creations of knowledge through the creation of the “Trans-Feminist-Queer Transformer” (T.F.Q.T.), a Generative Pre-Trained Transformer 2 (GPT-2) model fine-tuned with databases of love songs from trans-feminist-queer artists, programmed to generate output for live musical performance. Guided by trans-feminist-queer phenomenologies, this paper retroactively analyzes the creation of the T.F.Q.T. from creating a dataset for fine-tuning and experimenting with intervals of training to exploring methods of remixing, adapting, and performing the generated output live. Through a critical analysis of my own practice as informed by sensory analyses of other contemporary AI works—including Annie Dorsen’s algorithmic theatre, J.R. Carpenter’s electronic literature projects, and Allison Parrish’s algorithmic poetry projects—this article posits research creation as an ideal methodology for working through philosophical crises of AI and creative labour, which often understate or neglect sensory analyses of the body. This article uses the T.F.Q.T. to navigate the debate between embracing AI as “literary communism,” or “the freeing of language from the constraints of property” (Wark M. “From Automatic Writing to Automated Writing.” 2024 Ursula Franklin Lecture (keynote presentation, University of Toronto, Toronto, ON, 25 March 2024).), and shifts in public policy and discourse around the fears of generative AI and plagiarism, data theft, surveillance, and the augmentation or replacement of creative workers. Its findings reflect on the project’s technical successes and conceptual failures, concluding that while emotional resonance with the T.F.Q.T.’s output was to some degree inevitable, the generated material ultimately produced limited and stereotypical conceptions of queer identities, which are inherently fluid and resistant to normalization.</p>