Anti-slop

It started as hygiene. As AI writing flooded every feed, the tells hardened into a public checklist: the em dash, "delve," "it's not X, it's Y," the suspiciously tidy closer. Reading became forensics, and "rewrite this so it doesn't sound like AI" became one of the most common instructions ever given to an AI.

Then it became an identity, and the identity became a flex. The anti-slop practitioner does not hide the AI use; they showcase the countermeasures. Custom anti-slop skills and system prompts get posted and quote-shared like gym routines: four hundred lines teaching the model to vary its sentences, hedge less, be slightly wrong sometimes. The confession is the brag: everything I ship is generated, and you cannot tell, because I engineered it that way. Being anti-slop does not mean abstaining; it means generating with a superiority complex.

The war's front line stays the em dash, so notorious a tell that innocent human em-dash users became collateral damage and the fix went to the top: Sam Altman announced, as a "small-but-happy win," that ChatGPT would finally obey the instruction to stop using them. A punctuation mark got a patch note. The whole posture is downstream of taste is the moat: if anyone can generate, taste means generating undetectably. This compendium runs the same racket, and admits it: the house style bans the em dash outright, enforced by an automated test, and yes, that is bragging. See clanker for what you call the ghostwriter, and grindslop for slop's other franchise.