Recently I watched Isaac Pound interviewing David Gerard (of the Pivot to AI newsletter, which I'll link in the next post), and Gerard made a really, really good point. In human work, there's *intentionality*; in "AI/LLM" slop, there's none.
https://www.youtube.com/watch?v=wAgecpbLYkI
Gerard talks about how, with human-written code, line 10 is necessarily related to line 1, because there's a human being making decisions on it, and that human has a theory. This is something the slop machines, being simply statistical regurgitation, can't do. (Here's his newsletter, as promised.)
https://pivot-to-ai.com
Figure out that theory, and another human can hunt for likely mistakes, extrapolate, apply fixes, etc. But going through slopvibe code looking for whatever's fucked up now is hellacious because *there is no theory, just statistical vomit*.
This holds for writing, as well.
Even the worst, jankiest piece of human writing has an underlying theory, a (sometimes well-buried) cohesiveness, and the writer is making millions of tiny, unique, faster-than-lightning decisions with each word choice, even each punctuation choice. Which makes it possible for a reader to buy in.
Humans *make meaning*, and nowhere is that so evident as in the stories we tell. LLM-produced "stories" don't deserve the name. They're empty piles of "statistically this word is common in our stolen 'training' material", with no theory, no actual meaning or burning desire to communicate.
