Tedium: archival exhibition
at “Hype! Studies” Barcelona

Is there anything inherently valuable in using large language models (LLMs)?

Do workers engage with these systems differently than managerial narratives of optimisation and cost reduction suggest? And how do they navigate the capacities of these tools, making sense of their results?
To answer these questions, we shift our attention away from the perspectives of developers and toward the lived expertise, values, and tactics of workers. We attempted to establish the conditions under which the use of LLMs can be described within the specific professional norms and material conditions—what we refer to as their ecologies. We explore where the need to prompt arises, how outputs are handled, and what it means to engage with LLMs not as tools, but as technologies-in-use. We designed a participatory, experimental research protocol that is open to reproduction and adaptation through collaborations among sociologists, STS scholars, and design researchers. Over six- to eight-month cohorts, we worked with 40 professionals as co-inquirers. Through regular meetings and collaborative exercises, we introduced intentional pauses into everyday routines, creating space for reflection on the role and impact of LLMs. These encounters generated a multimodal archive—comprising audio, images, drawings, and chat logs—that captures the textures of LLMs-in-use.
Tedium is a curated exploration of this archive. Composed of five scenes, it follows the arc of our inquiry from the first meeting to the closing conversation. Each scene has a dual structure: a description, featured in the video installation, evokes the rhythms of our weekly sessions; and a commentary, presented in the following pages, offers interpretations, conceptual links, and analytical reflections that extend beyond participants' formulations.
What emerges departs sharply from narratives of automation or seamless delegation. LLMs do not simply accommodate users: they require accommodation. They fragment workflows, produce verbose outputs, and introduce new layers of interpretive and evaluative labour. Rather than streamlining work, they often multiply it, flattening nuance and amplifying tedium. While many co-inquirers found LLMs helpful, the promised gains in productivity proved elusive: more work, not necessarily better work. Yet amid these tensions, a quieter form of value appeared. The LLM, alien yet patient, provided a space for “silly work,” for emotional scaffolding, for low-stakes experimentation. Its judgment-free presence created rare permission to hesitate, to play, to think aloud. Far from the hyperbolic claims of disruption and efficiency, this value was grounded in practice, not promise. Anything but hype!