Volume 20 of the Studies in World-Formation series examines the relationship between artificial intelligence, World-Binding, and epistemic form. Its point of departure is the thesis that technical capability is not identical with World-Binding. The three studies brought together in this volume investigate this difference on three interconnected levels: the scalability of intelligence, the reach of World Models, and the question of authorship under conditions of generative AI.
The first part develops, through the concept of the Sweet Spot of Intelligence, the thesis that intelligence cannot be understood as an indefinitely scalable property. It appears instead as a relational capacity to integrate difference into a coherent and actionable world. Too little difference does not generate sufficient world formation; too much difference can undermine a system’s capacity for integration. Intelligence therefore does not lie at the maximum of complexity, but within a limited range of maximal World-Binding Capacity.
The second part applies this perspective to World Models. It distinguishes between world modeling and World-Binding and argues that a model can describe relations, recognize patterns, and simulate consequences without generating the irreversible embeddedness through which relevance, Eigenzeit, and responsibility emerge. The description of a binding is not identical with the binding itself; the simulation of a consequence is not identical with the consequence. Responsibility therefore does not necessarily arise from maximal predictive capability, but from dealing with non-closure, uncertainty, and real consequences.
The third part shifts the same question into the domain of authorship and generative AI. It distinguishes text genesis from work genesis and understands authorship as an Eigenzeit relation: as an irreversible connection between historically situated form formation, a setzung, and the future possibilities of the instance that carries this setzung. A machine can generate, transform, or reorganize language without thereby occupying the same authorial position as a human being whose work emerges from an ongoing Eigenzeit.
Neurodivergent authorship acquires particular significance in this context. Generative AI can function as an externalized foreign Raster or as an epistemic probe through which an already existing form is transformed, misunderstood, tested, and returned. Under certain conditions, this transformation may not weaken authorship but protect it, insofar as it increases communicative accessibility without eliminating the difference that made mediation necessary in the first place. At the same time, smoothing, normalization, and premature closure create new risks for the integrity of epistemic form.
Volume 20 connects these three investigations through a common question: When does technical capability become a genuine relation to the world? The answer is similar across all three fields. More complexity is not automatically more intelligence. A more comprehensive model is not automatically more world. More generated language is not automatically more authorship. What is missing in each case is irreversible relational embeddedness, within which differences acquire meaning, consequences generate history, and a system is itself changed through its setzungen.
The volume therefore does not develop a general theory against artificial intelligence. Rather, it investigates its ontological and epistemic conditions. Artificial systems can model, transform, simulate, and irritate epistemic forms. Their limits emerge where these operations are confused with World-Binding itself. The central concern of the volume is therefore not how powerful artificial intelligence can become, but under what conditions capability passes into world, relevance, responsibility, and authorship.
Dowload: https://zenodo.org/records/22812186/files/Speed_2026_World-Binding-Vol_20_en.pdf?download=1
