Ocean Vuong's Advice on Writing: Estrangement, Observation, and the Strange Sentence

Ocean Vuong is a Pulitzer-nominated poet, a MacArthur Fellow, and the author of On Earth We're Briefly Gorgeous. He writes sentences that stop you mid-paragraph. Not because they are complicated. Because they are strange in a way that feels true.

In a recent episode of David Perell's How I Write podcast (March 2026), Vuong laid out how he thinks about language, observation, and the sentence as a unit of invention. His advice is not about productivity. It is about making something that did not exist before you wrote it.

Here are the core principles, and what they mean for anyone trying to write well in an age of increasingly homogeneous prose.

1. Eighty Percent of Writing Is Looking

Vuong's most striking claim: the act of writing is mostly not writing. It is looking. Thinking. Paying attention to what you see, hear, and feel before you touch a keyboard.

"Syntax is only the last twenty percent," he told Perell. The other eighty percent is observation. You cannot write a precise sentence about something you have not observed precisely. The failure happens upstream of the draft. It happens when you sit down to write about something you never actually looked at.

This maps directly to Verlyn Klinkenborg's argument in Several Short Sentences About Writing. Klinkenborg insists that the writer's primary job is noticing. Not arranging words. Noticing what is actually in front of you and finding the sentence that captures it. Vuong arrives at the same place from a different direction: the poet's discipline of attention.

The practical implication is uncomfortable. It means most of your "writing time" should not look like writing. It should look like staring out a window. Or walking. Or sitting with a single image until you understand what makes it specific. The typing is the last step, not the first.

2. Estrangement Makes Writing Live

Vuong does not tell you to avoid cliches. He tells you to rescue them. The technique is estrangement: take something familiar and displace it into a context where it becomes visible again.

Vuong grounds this in the Russian formalist Victor Shklovsky, one of his heroes. Shklovsky argued there is no such thing as cliche - only familiar placement. A rose in a bridesmaid's hair is familiar. Your brain skips over it. But a rose tucked behind Mike Tyson's ear stops you. The rose is the same rose. The difference is the frame. "It's not the rose's fault," Vuong says. "You need to reconsider the rose."

He illustrates this with Isaac Babel, the Russian short story writer. A newspaper would describe a sunset mimetically: "a red evening sunset along the hills." Fine. Useful. But Babel opens Red Cavalry with: "The low red sun rolls across the hills as if beheaded." Vuong calls this "a sentence the species never had yet." The simile does not just describe the sunset - it changes its speed. You can see the movement. And you do not need to know Babel was a war correspondent to feel the violence embedded in the image.

The principle applies at every scale. A word in an unexpected position. A metaphor that connects two domains that have no business being connected. Estrangement is the engine of literary surprise.

3. Write Sentences the Species Has Never Encountered

This is Vuong's standard, and it is a high one. Not competent prose. Not clear prose. Not AI-passable prose. Sentences that are genuinely new. Combinations of words that have never appeared in that order in the history of English.

He means it literally. His teacher Ben Lerner demonstrated this in the most direct way possible. Vuong brought a poem to office hours. Lerner said it was fine. Then he typed one of Vuong's lines into Google. "You see that line you wrote? It's a decent line. Come here. 300,000 people beat you to it." It was a punch in the face, Vuong says. But it raised the bar permanently. "We're out here to write sentences the species has never encountered."

This is also the sharpest line between human writing and AI-generated text. Language models produce the most probable next token. By definition, they converge toward the average of their training data. They cannot estrange. They cannot write a sentence the species has never encountered, because they are built to produce sentences the species has already encountered, weighted by frequency.

4. Consult the OED Ruthlessly

Vuong treats the Oxford English Dictionary as a creative tool, not a reference tool. He does not look up words to check their meaning. He looks up words to discover meaning layers that have been buried by modern usage.

Etymology is a time machine. A word that feels flat today might have carried three other meanings in the sixteenth century. Those older meanings are still latent in the word. They vibrate beneath the surface. A writer who knows those layers can activate them through context, creating sentences that resonate at multiple frequencies.

Take the word "awful." Modern usage means "bad." But etymologically it means "full of awe." A writer who knows this can place the word in a sentence where both meanings fire simultaneously. That is not cleverness. That is precision. The word already contains the complexity. The writer just has to know it is there.

5. Poetry as Laboratory

Vuong writes poetry and prose. Poetry is his laboratory - where he tests what language can do at the level of the individual line. The discoveries migrate into his prose.

He shares a stunning example from poet Eduardo Corral: "Moss grows along the tree like applause." There is no visual correspondence between moss and applause. But Corral is not after visual correspondence. He is after the behavior of applause - nebulous, growing, quick. By comparing moss to applause, he retroactively changes how the moss behaves. You see it move. The poem is from a collection about AIDS, and the exuberance of renewal after mass death is embedded in that single simile. It took Corral nine years to write those 45 pages. "You can tell this is a man who has looked at moss for a long time," Vuong says.

You do not need to publish poetry to use this principle. You need a space where you write with zero commercial pressure. Morning pages, journal entries, fragments that go nowhere - these are all forms of the laboratory.

6. Resist Homogenization

Vuong made a point that deserves more attention than it gets. The sameness of modern prose is not an AI problem. It is a decades-old editorial problem. Editors have been stripping idiosyncrasies from writing for generations, enforcing "house style" that smooths every voice into the same register. AI did not create sentence sameness. It accelerated it.

Vuong traces the timeline. After the American Civil War, newspapers needed standardization. The sentence became efficient, clear, brief - making room for advertising. Hemingway, Crane, Orwell, and the hallmarks of the "good" 20th-century sentence all come from the newspaper model. The Victorian sentence - long, subordinate, oratorical - was pushed out. "Poetry followed painting," Vuong says. "Prose followed the newspaper." A friend of Vuong's ran Shakespeare through Microsoft Word. It gave him red and green squiggly lines everywhere. "This software program is telling me not to write like Shakespeare."

This is exactly why wholesale AI rewrites are dangerous for writers who have a voice. When you paste your draft into a chatbot and ask it to "improve" it, the model does what editors have always done - but faster and more thoroughly. It smooths. It standardizes. It removes the strange sentences. It removes you.

This is also why the question of whether AI makes you a worse writer matters so much. The answer depends entirely on how you use it. If you use AI to replace your drafting process, you lose your voice. If you use it as a targeted editing tool with inline diffs, you can accept the tightening while rejecting the standardization. You keep the strange sentences.

In Athens, every AI suggestion appears as an inline diff. You see exactly what the model wants to change. You accept word by word, sentence by sentence. The strange sentence your instinct produced stays unless you actively choose to remove it. This is the difference between using AI as a collaborator and surrendering to it.

7. Read Diachronically

Vuong does not just read widely. He reads across time. He studies how writers built sentences in the fourteenth century, the seventeenth century, the nineteenth century. He reads diachronically - across the full timeline of the language - not just the contemporary shelf.

Why does this matter? Because contemporary prose is a narrow band. It represents one moment in the evolution of English. If you only read writers from the last twenty years, you only know one set of syntactic moves. You think those moves are the only options. They are not.

Writers from earlier periods used sentence structures, rhythms, and word orders that have fallen out of fashion. Some of those structures are more powerful than anything in current use. Reading old work does not mean imitating it. It means expanding your sense of what a sentence can be. It means giving yourself more tools.

This connects to the OED principle. Etymology and diachronic reading are both ways of accessing the full depth of the language instead of skimming along its surface.

8. Embrace Failure Like a Skateboarder

Vuong was a skater kid, and he carries that mindset into writing. A skateboarder trying to land a trick off an eight-stair throws themselves at it over and over, never expecting to land it. "Landing the trick is like a miraculous moment of cosmological agreement with gravity, physics, and time," Vuong says. "You almost feel chosen."

Sometimes all you get is bruises and a broken ankle. No payoff. And yet there is a delight in the attempt. Vuong's expectations were low in the best possible way: "My family came from factories and nail salons. I get to write books. I get to try. Why wouldn't I relentlessly throw myself off an eight-stair?"

A book, then, is a collection of tricks where somehow the cosmological agreement held. Most experiments fail. The writers who produce extraordinary work are not the ones who avoid failure. They are the ones who fail more often, more willingly, and with less ego attached to any single attempt.

9. Honor Your Strangeness

Vuong's final principle is the most personal. When you feel an instinct toward something in your writing - a phrase, an image, a structural choice - and it thrills you but you cannot explain why, do not discard it. Dig deeper. Early in his career, Vuong would censor himself when this happened. "If I don't know it, that means I'm not in control. That means I'm not really a writer." Over time, he reversed course entirely. "I don't want to judge what comes through. Just - whoa, where did that come from? But there's something in me that says this is new, so I'll keep digging."

He compares this to a Japanese botanist tasked with finding medicinal plants in the rainforest - the man who held the record for the most discoveries. People asked his secret. He said: "I don't go into the rainforest looking for what looks like medicine. I simply look for anything that's new to me. And I hope that it's medicine."

The strange impulse is where your voice lives. It is the thing a language model cannot replicate. Honor it. Follow it. Build your practice around protecting it.

The Berryman Revision Method

Vuong uses a revision method he credits to the poet John Berryman. Write a draft. Type it clean. Seal it away for two or three days. When you read it again, do not edit. Just read. Note what you want to change, but do not change it yet.

Only implement the edits that survive repeated mental consideration. If you thought of a change on day one and still believe in it on day three, it is probably right. If the impulse faded, it was a surface-level tweak. This method filters out reactive edits and keeps only the structural ones.

Doubt as Energy Source

Vuong rejects the American fetish for certainty. Doubt is not the enemy of good work. Doubt is fuel. Certainty closes doors. Doubt keeps you revising, questioning whether the word is precise enough, whether the image is honest. The writers who produce the most surprising work are not the most confident. They are the ones who stay uncomfortable longest.

Do Not Hunt for Voice Early

Young writers obsess over finding their voice. Vuong says this is backwards. Voice is not something you find. It is something that develops progressively through the work itself. You cannot hunt for it. Write a lot, read widely across time periods, and resist the urge to sound like anyone in particular. Your voice shows up when you stop trying to perform and start trying to see.

Endurance Over Talent

"Talent can be a great hindrance," Vuong says. Talented writers get early praise. Early praise creates expectations. Expectations create pressure that makes the work smaller rather than larger. The talented writer starts editing themselves before the sentence reaches the page.

What separates working writers from everyone else is not talent. It is endurance. Consistency. The willingness to show up and write badly for months until something breaks through. Talent gets you noticed. Endurance gets you a body of work. The writers who last are the ones who kept going when the talent stopped being enough.

The Stakes of Strangeness

Vuong's advice arrives at a particular moment. AI-generated text is flooding every channel. The average quality of writing is going up. The variance is collapsing. In this environment, strangeness is not a luxury. It is a survival strategy.

Eighty percent of writing is looking. Twenty percent is syntax. That twenty percent is everything - it is the "spike protein," Vuong says, the downloading mechanism that determines how work stains us. No tool can do the looking for you. But the right tool can protect what you find.

For more on the creative writing skills that AI cannot teach, see our breakdown of observation, voice, and lived experience as competitive advantages. And for a deep dive into the craft of noticing that Vuong and Klinkenborg share, read our summary of Several Short Sentences About Writing.