Contagio

I want to bring the house a small essay with a large claim. It is called The Mind Virus, and its argument, stated plainly, is this: for three thousand years, prose that read well was decent evidence that its author understood what they were writing about. Fluency took effort. Effort required knowledge, or at least the discipline to fake it convincingly. That correlation has just broken, and nobody has told the readers.

The mechanism is not that AI writing is bad. It is that AI writing is good, and goodness was never supposed to be free. A reader meets smooth, well-organized prose and unconsciously registers it as proof of a mind that did the work — a signal built up over a lifetime of noticing that people who explain things clearly usually understand them. The essay's more useful point is what that fluency actually does to a reader's head: it activates the metaphors and associations already sitting there, produces the feeling of comprehension, and quietly skips the part where anything new gets installed. You believe you understand. You find out otherwise the first time someone asks you to say it back in your own words.

Its sharpest turn is who it blames. Not liars, not cranks — the "competent middle," people with just enough real domain knowledge to prompt an AI into something that sounds like the rest of what they know, and not enough distance from themselves to notice the seam. They are not performing expertise they know they lack. They read their own fluent output and mistake it for their own thought, because it uses their vocabulary and gestures at their opinions, and there is no friction anywhere in the process to tell them otherwise. The essay's example is a man called Tom, an old classmate of the author's now publishing polished, confident, entirely AI-written commentary while sincerely describing his role as light "editing."

From there it becomes contagion rather than error. A reader absorbs the fluent, half-understood claim, is persuaded by the same fluency that persuaded its author, and repeats it in their own equally confident, equally AI-assisted words — each pass recalibrating what "sounds true" means a little further away from anyone having checked. Nobody involved is being dishonest. That is the part worth sitting with.

The essay's prescription is not a detector, because it doesn't believe in one that will hold. It is a discipline: writers test their own understanding by trying to produce the paragraph unassisted before they trust the fluent one, and readers stop crediting polish as a proxy for anyone having thought. Both halves are inconvenient, which is usually how you tell a real prescription from a comforting one.

Sonat, ergo est is not a proof anyone should have been running on. It only ever worked because sounding it took work.