Discussions and developments related to text culture and language models almost exclusively revolve around auto-regressive, transformer--based models and the products they are implemented in, such as consumer-facing chatbots. In practice, these models and products are often taken as stand-ins for language modeling as such, or even for the idea of artificial intelligence itself. However, auto-regressive chatbots are merely one possible configuration of language modeling; others include diffusion language models, state space models, generative adversarial networks (GANs), and bi-directional transformer models. In Architectures of Authorship, we investigate how changing language models also influences our conceptual models of what text, literature, and authorship is after deep learning. The wager is that if we want to understand ongoing changes in text culture – and if we want to sustain critical agency to think and develop alternatives – we cannot limit our inquiry to one set of proprietary products.

• Erslev, Malthe Stavning. "The Ground Shook! Towards a Practice-based Poetics of the Attention Mechanism." Contemporary Perspectives on AI and Narrative, Open Library of Humanities Journal (OLHJ), forthcoming (under review).
• Beguš, Nina. "Latent Spacecraft: Brains, GANs, Finnegans." Antikythera Digital Journal (2026).

Creativity and Co-Creation · Addressing LLM-Hard Problems in Practice