The Grammar of Arrival
The Grammar of Arrival grew out of a research question from my thesis The Loss of a Body: what happens when situated writing—shaped by a body, a history, a habit of speech—meets a system that has already decided which sentences are ordinary? A language model doesn't read meaning; it assigns probability.
Visitors write one sentence, starting only from the word I, on a local model kept offline. It scores every word by how surprising it was and keeps only what it found most improbable—the rest, including the sentence itself, is deleted on the spot. What remains joins a shared, asymmetrical map projected across the gallery wall, alongside a printed receipt of the visitor's own resistance to what the model expected of them.
Developed from the artistic research thesis The Loss of a Body: The Measurable Improbability of the Situated Sentence (rMA Art and Performance Studies, University of Amsterdam, 2026), which asks what a language model's loss metric can and cannot register about the body that wrote the sentence it is scoring.
