I built a tool that lets me ramble at an AI system until it hands me back something readable. So before I make my case, let’s get the obvious disclaimer out of the way: yes, I built the thing, yes I think it’s good, and no, that doesn’t automatically make me full of shit.
Here’s the argument. For most of human history, there’s been a toll booth standing between a good idea and the people who’d actually benefit from hearing it, and the toll was eloquence. If you couldn’t say the thing well, it didn’t matter how right you were — the room moved on without you. I think that toll booth just got torched, and I think almost everybody arguing about AI and writing is arguing about the wrong fire.
The Eloquence Tax
Somewhere along the way we decided that how well someone talks is a reasonable stand-in for how well someone thinks. Sit an intelligent person in a room full of other intelligent people, and if they haven’t also put in the reps on sounding intelligent, that room will quietly write them off. That’s just how it works, and to be fair, it’s a genuinely useful skill face to face — I’m not saying vocabulary and sentence structure don’t matter.
What I am saying is that if fluency is the only toll we accept, we’re taxing a whole category of people who never signed up to pay it. Some people are gifted with words. Some are gifted with numbers, or materials, or color, or systems — and are genuinely mediocre at narrating any of it out loud. Intelligence isn’t one axis, it’s a whole messy cluster of them, and we’ve spent decades pretending one narrow lane — can you sound smart on command — is the whole highway.
Here’s the version of that argument that actually sticks with me: there are smart people who sound stupid, there are smart people who say genuinely stupid shit anyway, and then there are stupid people who sound smart. That last group splits in two — stupid people who fool other stupid people, and stupid people polished enough to fool smart ones too. Passing the eloquence test and passing the rationality test are not the same test. We’ve been grading everyone on the first one and calling it the second, and collecting the tax the whole time.
What “AI Slop” Actually Is
Every time a new model ships, somebody posts the same funeral notice: writing is dead, why bother with an English degree, the machines do it for you now. I think that panic is aimed at the wrong target, and the “AI slop” conversation is proof.
Large language models don’t invent anything. They’re trained on other people’s writing, so functionally they act less like an author and more like a filter — a genuinely democratizing one, since it hands people across every skill level the ability to write at least competently. That part isn’t controversial. Where people fuck up is assuming everything a model produces is slop by default.
It isn’t. Slop is uninspired output. Full stop. Doesn’t matter if a human typed it or a model spat it out — it’s the same failure wearing a different tool, whether it’s AI video slop, AI image slop, or the next wave you should probably brace for: AI app slop. The question was never “did a machine touch this.” It’s whether inspiration showed up anywhere in the process. A human cranking out a templated LinkedIn post with zero thought behind it is doing the exact same thing as a lazy one-line prompt into a chatbot. Same crime, different weapon.
The LinkedIn Bros Did This to Themselves
For the better part of a decade now, we’ve watched LinkedIn “thought leaders” post two-sentence paragraphs in ALL CAPS, with a canyon of white space between each one, writing that reads like it was composed by a caffeinated fourth grader — and rack up hundreds of thousands of likes doing it.
I’m not saying those guys are wrong. I’m not saying their information is bad. What I’m saying is: what exactly was so precious about the form of what they were producing, if the bar was already sitting on the floor to begin with? Form was never carrying the weight everyone assumes it was. What’s being said matters more than how it’s said — especially in thought leadership and opinion writing specifically. Caveat, because I’m not trying to get dragged into a fight I didn’t start: I’m not talking about fiction here, where craft and voice are the product. I’m talking about the stuff people get paid actual money to have opinions about.
And if that’s true, then laboring over literary style from scratch, on every single piece, forever, is a loss-leading habit for most people. Not because style doesn’t matter — it does. But once a style actually exists, executing it should be easy. It shouldn’t require re-litigating your voice, your structure, your opener, your closer, every single fucking time you have something worth saying.
Vibe Scribing
Which is the whole reason I’ve started calling this thing “vibe scribing” — a deliberate rip on vibe coding. Vibe coding turns prompts into technical output. Vibe scribing turns spoken ideas into written ones.
The model I keep coming back to is the medieval scribe. A lord talked, a scribe wrote it down — that was the entire job description. We don’t have scribes anymore, but we’ve built something that can functionally be one: a system that takes your best ideas as you say them and distills them into lean, working thought, or into finished content, whichever shape you actually need.
It scopes the same way vibe coding does, too. You need a style. You need context. You need a defined output — unless you’re just collecting raw material, in which case none of that matters yet, you’re just capturing. Get those three in place and there’s no real reason the system can’t do the writing for you the same way it does the coding for you.
Where this actually goes, eventually, is bigger than single pieces — a growing pile of captured thoughts that starts finding its own connections, surfacing content angles, business ideas, without me asking for a damn thing. Apply the same mechanism to fiction and you get world-building: characters, settings, history, developed the same conversational way, feeding stories generated off a world I actually built instead of one prompted out of thin air. That part’s deliberately further down the road — I want the foundation solid before I build the parts that go hunting for connections on their own.
For now, this is where it starts: talk into the void, let the system distill it, see what it turns into. This piece is what that process produced.