Esmée Colbourne - Holding Sloppy Orbs

From CTPwiki

In 2025, as a Data Rescue volunteer, I became a data holder of US federal HIFLD datasets. Data loss, such as what Data Rescue was trying to prevent, is the result of failures in infrastructural politics and the inherent fragility of digital saving. Rescue moved data through sign-in portals, spreadsheets and repositories, thinning and re-articulating the context of the original data points (Rao, 2019). Post rescue and upload into DataLumos the copies of data stayed with me, on my desktop. Struggling with the ethics of holding US data, I decided to visualize and play with the data as a way to keep it in motion, and theoretically keep it public through use (Gray, 2025).

After much thought I turned the rescued data into digital maps, tiles and 3D-printed orbs. HIFLD data arrived as points: coordinates for paths, pipelines and boundaries, stored as the potential for vectors in GeoJSON format. Data points here are precise and machine-readable, and can be redrawn at different scales. The data points ended and sloppy and abstracted shapes took their place. No longer machine readable their meaning transitioned to something more symbolic.

The transformation process to create orbs made the data over processed, becoming slop. I needed to put 2D source positions and spikes around an orb. I struggled with this for months, trying to use multiple solutions and code on my own. I was encouraged by makerspace volunteers to troubleshoot using AI and so leaned heavily on it eventually to make viable code and printable shapes. AI assistance supported the abstraction of material, the orb's shape, spikes and indents, iteratively simplifying the object until it was printable, and addressing an opened mesh and closed the seam where the last longitude meets the first. The tile misprints showed too much detail for the printer, and the orb misprints that data wasn't right for a three-dimensional spherical mesh, yet I pushed forward. The result was three levels of abstraction, all in the process of breaking: the map being the most legible, the tile being readable but decontextualized, and the orb being the least connected visually to an expected map. The object flowing through map, tile and orb reduced the power of its original cartography (Harley, 2011).

Materially these objects are layered. 3D printing builds objects extruded layer by layer, and the orbs carry these layers on their surface. Grooves and supporting scaffolding corrupt the quick prints. The layers have become delicate and unstable. Travel and handling have already snapped off spikes. The larger prints proved the most fragile, and chunks of their spikes have already fallen away. This cycle of breakdown and repair have been increased through handling. Like born digital content, storage and memory holding materials are not fixed in their states, always becoming something different, in part through degrading and in part through context (Ingold, 2007).

The orb currently give no counts, provenance or scale. My orbs have no legend, but do have paper labels underneath to indicate what they are. It is less a map, more a souvenir, authenticating the experience of rescue by rematerializing and making digital data more sensory in output (Stewart, 1993). The souvenir does not feel like a hallucination, just a worldly object, its meanings and presence emerging from interaction or from a glance to where it is placed on a shelf. Holding the spikes of data, I feel less alienated from data I could only see on a screen (Lupton, 2017). I can feel the sharp corners on the palm of my hand, the indents being made deeper if I squeeze it. These objects have been rolled around, and thrown, and dropped.

Here I gained insight into ownership, and the ethics of holding. Passing around the objects made people ask more questions about their sources and their textures, and gave them increased interest in the Data Rescue Project as a whole. Each conversation the object creates keeps the memory of data in circulation. Without memory, storing data does not save it; it pushes loss into the future (Chun, 2008).

Slop is usually discussed as a degenerative flood that degrades the internet (Kendall & Markelj, 2026) and now sits inside all dying platforms. Generative culture, like quick forms of making can look like a gimmick but I think in this case, using data to make slop repoliticises it by creating conversations about how data survives and what it means for all of cultural production (Galip & Lupinacci, 2026). Holding the data raises a question of ethics in relation to slop: is moving public data from a repository into AI a legitimate use? Gray (2025) argues for public data cultures that go beyond releasing data, towards using it in inventive and purposeful ways. Making the orbs is a form of use, and I worked on the assumption that HIFLD data had already been scraped into training sets. Yet I struggle because the ethics change once feeding a model more training data. Offering data to AI changes data flow and its politics. Data becomes raw material, valued for what it can train rather than what it records. It takes in public data without the care, context and collective labour of the rescue.

The orbs cannot do much, and I still feel ambivalent about them. They may be slop. Yet as slop they made other things tangible, by slowing the fading of memory of rescue. The sloppy orbs extend this process of transformation, introducing another layer of interpretation and potential loss. They do not restore the data to an original state, nor do they offer a solution to the ongoing problems of digital preservation. Instead, they make visible one aspect of the fragile stack of memory.

I think this work was valuable to me because of how it highlighted the incompleteness of AI assisted transformation. The orbs cannot communicate everything the datasets once contained, but they can invite people to engage with theories around saving, the crisis of loss and the responsibility for keeping it. Their fragility makes holding a delicate practice. To hold data gently and ethically requires attention, constant repair and continued political engagement. I am still unsure whether turning public data into slop is an ethical act, particularly when AI is involved. The orbs specifically are not the data, and they are not an archive. They are just a sloppy souvenir.

References

Chun, W. H. K. (2008). The enduring ephemeral, or the future is a memory. Critical Inquiry, 35(1), 148–171.

Galip, I., & Lupinacci, L. (2026). Degenerative AI: AI slop, brainrot, and the gimmick of generative culture. European Journal of Cultural Studies. https://doi.org/10.1177/13675494261468638

Gray, J. W. Y. (2025). Public data cultures. Polity.

Harley, J. B. (2011). Deconstructing the map. In M. Dodge (Ed.), Classics in cartography (pp. 273–294). Wiley-Blackwell.

Ingold, T. (2007). Materials against materiality. Archaeological Dialogues, 14(1), 1–16.

Kendall, T., & Markelj, J. (2026). Brainrot, AI slop, and the enshittification of the internet. European Journal of Cultural Studies. https://doi.org/10.1177/13675494261468648

Lupton, D. (2017). Feeling your data: Touch and making sense of personal digital data. New Media & Society, 19(10), 1599–1614.

Rao, V. (2019). The extended internet universe. In Y. Strickler (Ed.), The dark forest anthology of the internet (2024). The Dark Forest Collective.

Stewart, S. (1993). On longing: Narratives of the miniature, the gigantic, the souvenir, the collection. Duke University Press.