Writing with AI

Frank Chimero asks the right question: Where do I stand in relation to the machine — above it, beside it, under it?

Each position carries a different power dynamic. To be above is to steer, beside is to collaborate, below is to serve.

I stand beside it. I think with AI.

Not everyone does. Oliver Burkeman, in an interview for sublime, shared how he sees AI in his writing process. Spoiler: he doesn’t use it. I respect that. His voice is singular, and perhaps AI would mess with the magic.

David Perell sees it differently. On Feb 24, 2025, he wrote this on X:

“If you have an idiosyncratic way of looking at the world, you don’t have much to worry about.”

“When it comes to discovering ideas, I’ve also found that jamming with an LLM is more productive than doing it with most people I know (save for a few genius-level conversationalists). And I’m not the only one. Microsoft CEO Satya Nadella says: ‘The new workflow for me is I think with AI and work with my colleagues.'”

If you want to go deeper, Perell posted The Ultimate Guide to Writing with AI a month later (Mar 26, 2025).

Both perspectives matter — Burkeman’s skepticism and Perell’s embrace. I land somewhere between them, though closer to Perell.

A week doesn’t seem to go by where I don’t get asked if I use AI for writing, and, if so, how I use it, and what models I prefer. This is my answer.

⦿

Chimero’s framing, from his recent talk Beyond the Machine, offers useful language for anyone feeling uncertain about AI:

“I find the instrument framing more appealing as a person who has spent decades honing a set of skills.”

“Thinking of AI as an instrument recenters the focus on practice. Instruments require a performance that relies on technique — the horn makes the sound, but how and what you blow into it matters; the drum machine keeps time and plays the samples, but what you sample and how you swing on top of it becomes your signature.”

“In other words, instruments can surprise you with what they offer, but they are not automatic. In the end, they require a touch. You use a tool, but you play an instrument. It’s a more expansive way of doing, and the doing of it all is important, because that’s where you develop the instincts for excellence.”

AI as instrument, not tool. That distinction matters.

So, how do I use AI?

Like Perell and Nadella, I think with AI. I operate “beside it” using Chimero’s framing.

This is how I started using AI in 2022. It’s not hyperbolic to say that, without AI as my collaborative sparring partner, TDW wouldn’t have the depth and nuance it does.

This, for me, is where I derive the most value from AI — not the writing or editing help, but the thinking.

If you want an excellent, actionable perspective on using AI to be a better thinker, read “Think First, AI Second” by Ines Lee on Every (Dec 2, 2025). Lee’s essay closely mirrors how I “think with AI.”

So, thinking aside, yes, I use AI to help with writing.

Being dyslexic means I don’t get spelling “for free.” Words don’t lock themselves into letters — I have to wrestle them into place.

I still spell “geography” via an absurd mnemonic — Gregory eats old grey rats and paints houses yellow — burned into my brain since I was ten.

I used Grammarly for years, but it fought my voice, constantly trying to formalize and flatten my writing. AI is better — it recognizes my fingerprint and mostly keeps it intact.

Writing, of course, is a way of organizing our own thinking.

I write all Draft Ø’s (the vomit on the page) and 1’s (less vomity) myself. It’s important for me to extract my messy thoughts void of any outside influence or interference.

Only at this point do I expose AI to my words and thinking.

I start by using AI the way a book publisher would use a developmental editor — someone who looks at the big picture: structure, argument, coherence.

Writing has a hierarchy of meaning: word → phrase → sentence → paragraph → essay.

As a discovery writer, I write drafts forward, discovering the territory just ahead as each sentence reveals new thoughts. I never plan a piece structurally; rather, the entire thing emerges on the page, sentence by sentence.

“Writing is a navigation tool in some sense. You think about where you’re going so you can test out the root before you implement it. That’s all associated with thought, and the deepest form of thought is writing.”

— Dr. Jordan B. Peterson (source)

Discovery writing feels like moving forward through fog (generative). Developmental editing begins at the destination and redraws the road (architectural).

When sparring with AI on the structural flow of a piece, I’m always asking the AI “why” it suggests what it does.

I learn by understanding where I fell short, and why — my North Star is building understanding when collaborating with the model.

I push back at times. It does, too. We “fight” our corners until there is a consensus where, in a way, we’re both happy. We’ve reached a shared understanding.

Then I go back and make changes in Ulysses.

I return for another round of toing and froing.

Once the “developmental” component is done, I instruct the AI to offer line edits where necessary. We start at the top of the piece and work our way through, sentence by sentence.

This is almost always a process of simplification.

I sometimes (often!) pack too much into a thought — too many arguments or messy thinking in a single sentence or paragraph. AI is excellent at simplification.

My early drafts are raw, spikey, not yet smoothed out. AI helps with that smoothing, normally by removing noise to reveal clarity.

I sometimes choose a word or phrase AI suggests, even when I know it’s not a word I would have picked myself. But when I look at the piece as a whole, and I feel the suggestion is an overall improvement (there is a “taste” I can feel across a piece, and that’s where I’m aiming), I’ll keep it.

More often than not, I don’t use its suggestions verbatim, but rather as examples from which I further shape myself.

Between what I wrote as a first raw expression and the smoother version we landed on is a discussion I always find fascinating and meaningful. Here’s an excerpt from our discussion for the two examples above:

Lovely instinct. The “geography” story is more than ornament — it reveals the mechanism. What you want is to fold it in without bloating the line or turning it into an anecdote dump. The trick is to hint at the method rather than recount the joke in full (leave the full mnemonic as a later Easter egg if you like).

Here are several versions at different “altitudes,” from haiku-tight to slightly more expansive. Pick based on how lyrical vs. explanatory you want the moment to be.

If this appears in an essay or manifesto context (which it smells like), I’d go with #3:

“Being dyslexic means I don’t get spelling “for free.” Words don’t lock themselves into letters — I have to wrestle them into place, smuggling them in disguised as stories.”

Then later — after you’ve earned it — reveal the geography mnemonic as a lived example.

Structure:

Principle → Mechanism → Reveal → Delight

Exactly how good nonfiction works.

In the end, the expression I landed on was the simpler version you see in this piece.

I never prompt AI to generate writing from scratch. That would miss the point entirely — my leverage as a writer is my ideas, ideas that emerge from an idiosyncratic way of looking at the world. AI can’t produce those from a prompt.

What models do I use?

I use ChatGPT Plus (5.x) as my default sparring partner and writing editor. But I also use Claude Pro — mostly Sonnet 4.5 — and often find myself flip-flopping between the two as I work through edits of a piece.

I don’t use Gemini for writing help. Never have.

I’m grateful I get to use AI to help with my writing — not in coming up with ideas, but in smoothing the expressions of the ideas I’m shaping for you.

I don’t think the write-with-AI process is that different from working with a New Yorker editor — the back and forth, the developmental suggestions, the fight for consensus. The editor happens to be silicon instead of carbon, but the relationship is the same: beside, not below.

And that’s where I intend to stay.

~ André