The Metadata Shitshow is a computational-poetic experiment in deterministic semantic degradation. It proposes to transform large scholarly metadata corpora by replacing reproducibly selected intervals of Unicode grapheme clusters with the poo emoji.
The operation is comic, but the experiment is exact. A seed derived from the source snapshot, record identity, operator version, generation number, and current-state hash ensures that the same inputs produce the same wounds. The author establishes the law; the hash selects where the record is damaged.
The project joins four forms:
Its founding distinction is:
> identifier continuity ≠ metadata continuity ≠ artifact continuity ≠ semantic continuity
A DOI may remain perfectly legible while the intellectual figure beneath it becomes progressively unrecognizable.
Recursive generations model long-tail collapse. Rare names, relations, concepts, and affiliations are expected to disappear before boilerplate and statistical centers. Meanwhile, the replacement token accumulates into an excremental surplus: the corpus becomes easier to compress as it becomes less capable of meaning.
The experiment’s principal contribution is Retrieval Fitness–Integrity Inversion. A damaged inner record may become more visible than its intact source if it receives an excellent outer wrapper: clean provenance, structured metadata, stable URLs, explicit relations, and machine-readable warnings. The outer object can become more exact as the inner object turns to shit.
The specification preserves the ethical boundary between source and derivative. Every output is to be marked synthetic, non-authoritative as bibliography, linked to its source, and offered for voluntary machine reading. This record is the work plan—not the transformed corpus.