The slop before AI
Call it medieval slop, if you will.
Francesco Petrarca, born in 1304, certainly saw something rotten in the intellectual output of his own century, even if he lacked the vocabulary to name it that way.
He hunted down manuscripts of Cicero and Livy while watching university faculties churn out volumes of something entirely different, and the difference disgusted him. The term medieval was born out of that contempt.
His target were the dialecticians, the logic specialists dominating fourteenth-century faculties in Paris and Bologna - the ‘LLMs’ of the day.
In one famous letter, he described his contemporaries as a “black army of ants” swarming from a rotten oak, devastating the fields of sound learning. That’s not gentle criticism. That’s a man genuinely furious about what passed for scholarship around him. And it’s so mid-2020s.
Petrarca’s specific complaints, scattered across letters and treatises, targeted scholastic methodology, the obscurity of technical language, and an excessive confidence in pure logical rationality. These men, he said, “condemn Plato and Aristotle, and laugh at Socrates and Pythagoras,” while lacking the knowledge or virtue to understand any of them. They scorned Cicero, whom Petrarca called the bright sun of eloquence, and dismissed Seneca, Livy, and Sallust as crude and unfinished.
What galled Petrarca wasn’t the quantity of commentary exactly, which was also part of the larger humanist complain. It was misplaced confidence, technical competence mistaken for wisdom.
Scholastic logicians spent enormous energy refining syllogism after syllogism while treating classical eloquence as beneath serious attention.
Petrarca thought this exactly backward. Style reflected moral seriousness for him, and language stripped of eloquence meant thought stripped of virtue.
Another leading Italian humanist, Leonardo Bruni, extended this critique in the early fifteenth century, attacking medieval translations of Aristotle’s Ethics for mangling meaning through clumsy technical vocabulary. The resulting debate turned on a genuinely hard question, one we still wrestle with: does precision matter more than accessibility, does faithfulness to a text outweigh whether anyone can actually read it well.
The parallel to today writes itself, almost too well.
AI-generated text, slop in the eyes of post-Petrarcan critics, now floods social platforms and search results with prose that’s grammatically flawless, structurally correct, and utterly forgettable, full of the same predictable phrases and hollow structures repeated endlessly. Not the average of human output, but the mediocre, medieval even, point. Mediocritas yes, just not aurea.
Researchers have even given this a technical name, model collapse, describing what happens when AI systems train increasingly on their own previous outputs and gradually degrade, producing content that grows less accurate and less diverse with each cycle.
One estimate puts the current share of AI-generated content online at 57 percent, with projections suggesting it could reach 90 percent within the next few years.
Perhaps another humanist revolution is in order.