I’ve been down a rabbit hole for the past few months, obsessed with a single question: “What makes a writing human?”
It started, ironically, because I was trying to automate my reading list.
I’m a tinkerer. So naturally, I wrote a bunch of Python scripts to feed podcast transcripts and long-form essays into LLMs. The goal was simple: extraction. I wanted to bypass the fluff, get the “key takeaways,” and inject the knowledge directly into my brain. Efficiency.
Technically, it worked. The scripts spat out perfect, bulleted lists. They captured every logical argument Seth Godin made. They summarized Elizabeth Gilbert’s advice on creativity with 100% accuracy.
But reading them felt… dead.
It was like eating nutrient paste instead of a meal. The calories were there, but the flavor—the thing that actually makes you want to eat—was gone. The summaries were “dry.” Sterile. Smooth.
That’s when I realized that for great writers, the “fluff” isn’t fluff. The inefficiency is the point.
The magic wasn’t in the logical conclusion of the argument; it was in the meandering path they took to get there. It was in the friction. It was in the specific, weird metaphors (John Mayer talking about a “feedback loop” inside his head). It was in the rhetorical devices—like antithesis (“Hunger is cheap; the palate is expensive”) or merism (“We searched high and low”)—that acted like velcro for my brain.
LLMs are fantastic at saber (knowing facts/data). They can summarize the “what” all day long. But they fail hard at conocer (knowing a person/place/feeling). They lack the “tacit knowledge” that comes from having a body that has actually felt cold rain or awkward silences.
When I asked the AI to summarize these human moments, it stripped away the texture. It ironed out the wrinkles to make the text “efficient.” It removed the moistness of the human experience and replaced it with the dryness of information.
And that scared me. Because I realized I often try to write like that—optimizing for clarity, stripping away the “unnecessary,” trying to sound smart and polished.
Dry writing, as opposed to…. moist writing. XD
So I’m shifting my goal posts.
I just tweeted this, and I think it’s going to be my new north star:
“My writing goal now is to convey ideas through words that will lose their power when transformed by LLMs.
If my writing can be summarized without losing its essence, I am merely conveying information, not an idea.”
Perhaps call it the Reverse Turing Test.
If I write something and an AI can summarize it perfectly without losing the vibe, then I haven’t really written anything. I’ve just dispensed data.
I want to be a texture-maker, not an information-dispenser.
I want to use analogies that only make sense if you’ve lived my specific life. I want to use caveats—”It depends”—because I know reality is messy. I want to speak to you (I-Thou), not at a demographic (I-It).
I’m done trying to be smooth. Smooth is slippery. Smooth is forgettable.
I want to be sticky. I want to write things that break the summarizer. I want to write in a way that makes the machine hallucinate because it just can’t compute the weight of the words.
If the summary feels like a betrayal? Then I know I’m on the right track.
