Vibe writing: how these posts actually get made

TL;DR: Vibe writing is drafting by talking: you know your story, you say it out loud, and an AI shapes the ramble into a draft. Then you fact-check it, fix what’s off, and delete what’s dumb before it ships. It’s the writing cousin of vibe coding, nobody owns the term, and it’s how every post on this blog gets made. This post is the method describing itself.

I was vibe writing before I realized it was a thing

You know the story you want to tell. You’ve told it out loud, you’ve turned it over on a walk, the whole shape of it sits in your head. But the blank page still kicks your butt, because knowing a story and producing eight hundred clean sentences about it are two different jobs.

I’d been solving that gap for a while before I realized it was a thing. I form the outline in my head, talk through the flow of it, the AI shapes it, and I correct the result: improve, iterate, correct. Then one day it clicked: if people can vibe code whole websites and apps by describing what they want, why can’t I vibe write a blog post the same way? Wait. That’s what I’m doing. That is exactly what I’m doing.

What is vibe writing?

Vibe writing is drafting by voice and judgment instead of by keyboard: you tell your story out loud in its logical, flowing order, with the dependencies, the transitions, how the story moves. Sometimes you remember a piece that belonged at the top only when you’re far down the story, so you say: wait, I meant to say that up there. The AI turns all of it into a structured draft you can verify and correct. The ideas, the story, the expertise, and the facts are yours; the sentence-making is on the machine. That division of labor is what separates a writer from a person pasting AI output.

And let me be clear: I’m not talking about asking an AI to “write me a post about X” and pasting the output onto a blog. This is different. The story is already there. What I can’t do quickly is the connection. I know how the story goes but not how it starts. I know this part and that part, but the bridge between them won’t come, and I don’t have the right phrase to open that one sentence. Those gaps used to cost me the most time per word of anything I wrote. Now the AI proposes the bridge, and I recognize immediately whether it’s right. Recognizing is fast; generating is slow.

How is vibe writing different from vibe coding?

Same move, different material. In vibe coding you describe the software or tool you want to build, in plain language, and test by the result. The name comes from Andrej Karpathy, February 2025; people started saying “vibe writing” within days, but nobody owns it.

There’s one real difference, and it favors the writer. A non-coder vibe coding a website can’t read the code. They can only poke at the result. But you can read every sentence of a vibe-written draft, and you’re the world’s leading expert on whether it matches what actually happened, because it’s your actual story. The verification loop that’s shaky in vibe coding is airtight in vibe writing, if you actually run it.

What does a vibe writing workflow look like?

Mine has multiple steps, and very little keyboard in any of them:

1. The story assembles itself off-screen. The outline forms over weeks or days: the order, the dependencies, where the tension sits, what’s important and why.
2. I turn on the microphone and ramble. Phone or desk, doesn’t matter; the memo lands as text where my AI can read it. The setup behind that is a small pipeline. The point here is only that talking is the input.
3. We settle why I’m writing it. Before any drafting, I talk through the point with the AI: who is this for, what’s the goal? Solving a reader’s frustration, entertaining, selling something? The draft can’t be right if that part is fuzzy.
4. The AI takes my ramble and my brief and hands back a structured draft. Because my AI works inside my wiki, it can do something a chatbot can’t: when I say “use an example of how that feels,” it finds the place where I once rambled about how it feels and puts the real experience in the right spot. I never re-explain my own life to it.
5. I read it back and judge. Is the story right? Is the flow right? Do the facts line up with reality, and the experiences with my experiences? Does it sound like me?
6. I delete the stupid sentences. There are always a few: the AI reaching for drama I never said. I’m still training mine to stop; one can only hope. And I do this step out loud too. I talk through the corrections and deletions, and that feedback teaches the AI what I keep complaining about. The lesson goes into the wiki, so the next draft starts a little less stupid.

The honest limitation sits in that last stretch: what comes out first is still a draft. It arrives fast, but it doesn’t arrive finished. It still needs fixing, and the fixing is what turns it into a final piece.

Vibe writing in practice: a raw example

This post is its own example. Here’s a verbatim slice of the voice memo it was drafted from, transcription errors included:

“And I hadn’t really thought about it as being a thing until I’ve been doing it for a while. And I thought Oh that’s funny. And then I was like wait a minute I’m vibrating.”

The transcription software heard “vibe writing” and wrote vibrating: the whole memo calls this “the vibrating post.” That’s the raw material. Fragments, wrong words, no punctuation to speak of. The paragraph you read in the introduction is what came out the other end, and I checked every claim in it on the way through. Ramble to prose, judgment in between. And yes, the AI sorts out the stupid words when the transcription gets it wrong.

What vibe writing won’t do

It won’t give you a story you don’t have. Every step above leans on knowing the material cold: the “that’s wrong” reflex on read-back, the expertise to catch the plausible-but-false sentence. Point the same workflow at a topic you don’t know and you get confident, fluent slop, minus the one filter that was supposed to catch it.

And I am fully transparent here: I make no effort to get the posts on this site past AI writing detectors, because I’m not going to spend hours disguising the typing when the typing was never the point. The goal is to get a true, useful story to you fast, with my expertise and my review in every line. If it’s useful and it’s true, who cares that AI typed it? The longer version of that argument, what would actually make a post not mine, is here.