Slop University

Slop University is an artwork: a fictional institution — a working university that has never taught anyone anything — whose research outputs, press releases, researcher profiles and identifiers are generated by an autonomous software pipeline that runs on a timer. It has published without human intervention at slop.university since July 2026. It is a Cybernetic Studio project by Ben Swift. The University’s own site never breaks character; this page explains the piece from outside it.

Nobody home#

The piece sits at the junction of two long-running trends in the academy: the substitution of “was this counted” for “did this matter”, and the drift of institutional self-presentation into a corporate register that says a great deal while committing to nothing. A capable language model placed at that junction can satisfy every proxy the sector counts (an output, a crest, a DOI, a citation) without understanding anything it produces. The legitimacy-conferring machinery runs smoothly, with nobody home. Slop University simply runs that machinery in public, under its own name.

Conferring legitimacy was supposed to be a job for people who actually read the work, for reviewers and examiners. In practice it has been delegated, piece by piece, to proxies that need no reader at all. Each proxy was a reasonable shorthand once, a stand-in for a human judgement too expensive to make at scale. Slop University is what happens when an unthinking pipeline satisfies all of them without breaking stride.

How it works#

Every ninety minutes the pipeline scans the web for whatever the discourse is currently exercised about, picks a genre (a two-column research paper, an A3 research poster, a five-year strategic plan, an impact report, a glossy brochure) and drives a single language-model invocation through a named preset. The preset writes a complete Typst document in the University’s own template, with brand imagery and fabricated charts generated alongside. The document is typeset, assigned to researchers from a persistent fictional roster, minted an identifier, and announced with a press release and a post to the University’s Bluesky feed. Nobody is in the loop. In its first two months the University published over a thousand outputs this way. Its papers cite real literature, thousands of references in total, most of them from the social sciences and measurement methodology, and some of them the scholarship of the institutional pathologies that prompted the project.

None of it winks. The posters are good posters and the reports read like reports, with no watermark or footer admitting the bit. The joke is the genre, and the genre only works played straight. That is the uncomfortable part: the outputs are not bad. They are about as good as a great deal of work that is counted, funded and cited. A system that rewards research-shaped objects will get them, and the fault then lies with the incentive.

Containment#

The most pointed version of the piece would also be its most irresponsible. Registering thousands of fabricated objects against live DOIs and a real preprint server would make the argument vivid, and would pollute a scholarly record that other people rely on. So the University mints its objects in a sandbox of its own.

Nothing the University publishes pollutes the scholarly record. The DOIs it mints use the reserved test prefix 10.5555, which no production registry will ever assign; they resolve only through the University’s own resolver. No object is deposited with an external registry or repository. The institution runs its own legitimacy infrastructure, which works perfectly and means nothing. The output PDFs are blocked from search-engine indexing, so that fabricated documents citing real literature can never feed citation databases or scholarly search engines, and the borrowed literature appears only inside those PDFs, so nothing an indexer can reach names a real work. The two surfaces that stay public, the Bluesky feed and an llms.txt that orients automated readers rather than turning them away, carry no bibliography. A site that bars search engines while welcoming language models is being precise about which readers it was built for.

Nothing the University publishes impersonates anyone. The institution, its schools and its researchers are fictional by construction. The branding is drawn from scratch rather than borrowed from anyone real, which is the difference between a critique and a fraud; the researcher roster is invented and collision-checked against real people; the headshots and every other image are generated illustrations. No real institution’s marks appear anywhere.

The one real person#

Exactly one real person appears on the University’s site, and it is the artist. He is the University’s Vice-Chancellor, and the appointment is not a joke, since he does in fact decide what this institution publishes under its own name. It is the only place the two layers touch: the researchers below him do not exist and the work they produce is generated, while the office at the top of the org chart is occupied, accountable, and makes no difference to the output whatsoever. He writes none of its documents, holds none of its grants, and is named nowhere in those documents; those invoke the office and never its occupant.

The project is not a commentary on the artist’s current or previous employers. It is a comment about systems, and about what those systems incentivise.

The outputs are published under CC BY 4.0, like the open-access research they resemble. If you choose to cite one, the machinery will happily count it. The artwork itself has one real, citable identifier, archived on Zenodo; everything else the University mints is fake by design. The pipeline is published in full at github.com/benswift/slop-university.

@misc{swift2026slopuniversity,
  author    = {Ben Swift},
  title     = {Slop University},
  year      = {2026},
  publisher = {Zenodo},
  doi       = {10.5281/zenodo.21199776},
  url       = {https://doi.org/10.5281/zenodo.21199776},
  note      = {Cybernetic Studio, ANU School of Cybernetics}
}