The Grift Economy Ate the Internet — and LLMs Will Make It Worse

Stephen Diehl argues that online predation has shifted from a fringe abuse of the network to its core operating logic. Where a scammer once had to hunt for a victim, the platform now profiles the weakness, tunes the pitch, takes the payment, and queues up the next mark. Recommendation engines are framed as vast reinforcement-learning loops that continuously experiment on attention, retention, and conversion — a system in which calm accuracy loses and outrage, threat, and impossible promises win. The social harm never shows up in the objective function; it lands as an externality nobody optimizes against.

The more original claim is that the grift economy is not just large but participatory. Ordinary users are conscripted into the downline: aspiring influencers, affiliate marketers, dropshippers, and life coaches become unpaid distributors for whoever sits above them, selling worldviews and courses they often believe themselves because belief makes the selling bearable. Diehl notes this collapses the clean line between predator and prey — many grifters are also marks, too invested to admit the product is worthless. He points to the structural reason legitimate business and grift diverge: a real trade ends when a need is met, while a grift must preserve the need, so wellness gurus can’t let you feel well and trading gurus can’t let you get secure. Satisfaction is churn; misery is recurring revenue.

He casts crypto as the purest expression of the pattern — a recursive machine where promotion drives price, price is sold as proof of adoption, and every holder becomes a volunteer publicist for their own position. The rest of the internet, he says, has learned the same lessons: empty promises are the cheapest product, addicts are the most scalable labor force, and marketing works best when hidden inside identity so criticism feels like a personal attack. The piece closes on a warning that language models will make all of this cheaper and worse — mass-producing the persuasion, personas, and content the extraction machine runs on.