Generative AI tools have altered the digital landscape through which design practice, pedagogy, and research now move. Like the ocean, this new environment is vast, powerful, and in constant motion—characterised by disruptive swells and hidden currents. Patterns exist—waves of innovation, tides of expectation—but they are often difficult to anticipate. Amidst this volatile sea, educators, researchers, and learners attempt to orient themselves. Tools surge and recede in popularity, policy frameworks shift with headlines, and institutional narratives rise and fall. The water holds patterns—waves of hype, undercurrents of resistance, moments of stillness—but it rarely settles into certainty.

Anchored in this oceanic metaphor, this piece traces a practice-led journey through shifting technological currents—where generative AI mediates knowledge production, shapes cognitive labour, and contributes to both insight and technostress. This autoethnographic inquiry explores what it means to think, create, and reflect in conversation with generative AI—where cognition, self-awareness, and interaction emerge as relational, situated, and negotiated practices.

Rather than offering a polished framework or fixed conclusion, this narrative moves through disorientation, cognitive overload, and relational recalibration. Grounded in the principles of Inquiry-Based Learning, it treats uncertainty not as a flaw, but as a generative condition of learning. Through a series of tidal movements—advances, retreats, recalibrations—it traces how authorship, attention, and agency are shaped in relation to the interface.

This work invites a situated, relational approach to learning with AI—one that values resonance and rhythm over resolution, and embraces the unfolding nature of inquiry as practice. The task is not to master the tools, but to practise within the unpredictable swells—with care, attunement, and a willingness to work with AI as a flawed yet valuable collaborator.

The AI in the Library project aims to develop innovative AI literacy programmes for Australian librarians and library clients to understand the benefits and risks of using popular text generators (LLMs) such as ChatGPT. The first output of the project was The Making of Misbehaving Machines (2024) exhibit.

This paper outlines the design-led process for the co-design of this exhibit with library staff. This involved a series of workshops with library and academic stakeholders, giving hands-on experience with the creation of a novel dataset, on which the project was further developed. This dataset was used to manipulate a large language model via a custom coded interface, taking advantage of system prompts – which are made apparent in the final design prototype. This design prototype formed the basis of the final, in-library exhibit and we will discuss how the various elements of this outcome were designed around sustainable practices and principles of slow computing.

By merging collaborative dataset creation, interface design and public exhibition outcome into one workflow, the project demonstrates how DH methodologies—particularly critical making, slow computing and infrastructural inversion, along-side design-led practices—can critically address AI in library contexts. At the same time, critical co-design facilitated not only the design of the misbehaving machine with librarians, it also created a space for collaborative investigation of the materialities and contingencies of machine learning practice, discussion and engagements with ethical questions about generative AI, as well as the limits and possibilities of LLMs in information search.

The growing availability of artificial intelligence tools for humanities research continues to raise important, typically practical questions, about appropriate use, the ethics surrounding model training data used, issues around authorship, et cetera. However, notions of artificial intelligences participating in humanistic inquiry into art raise further existential questions about the phenomenologies of critique and aesthetic judgement. This presentation considers some metaphysical implications of AI-enhanced research methods upon aesthetic criticism, particularly by developing a phenomenology of computer-assisted vision deployed toward methods of visual analysis fundamental to fields like art history and visual culture. Specifically, I examine Lauren Tilton and Taylor Arnold’s Distant Viewing Toolkit, comprising not only open-source tools for large-scale computational analyses of massive image datasets, but also a ready-made research epistemology predicated on the incorporation of machine learning into the human critic’s “trained judgement.” The particular philosophical interest of the distant viewing framework lies in its augmentation of existing data analysis pipelines with a recursively adaptive annotation process. This proposes, in turn, human-computer interoperation rather than mere tool use in AI-augmented critique. What, then, does it mean to practice aesthetic evaluation with an other, artificial “intelligence”? Especially against the historical backdrop of scepticism toward tech-enhanced humanities scholarship, I question whether AI concepts represent only negative implications for knowledge work in manifestly subjective, intuitive, and corporeal topics like aesthetics. This presentation takes seriously, and in good faith, ideas of working with AI as potential critical collaborators.