How I Write

an FAQ in dialogue form

Q: Sarah, how do you write so well?

A: I don’t write all that well, actually.

Look at a New Yorker article sometime — just pay attention to the writing mechanics, the sentence structure and narrative flow. It leaves me in the dust. I don’t even really edit my blog posts!

Not to mention the fact that I don’t even try to do the hard kinds of writing, like fiction or poetry or humor.

I just blog and post stuff I think. There’s not a lot of writerly art to it.

Q: Quit being modest and give me a real answer. Whatever you may say about it, I like your writing. I wish I could write like you, but my own blog is languishing. What’s your secret?

A: Well, let me put this in perspective a bit.

I have spent nearly every free waking moment either reading or writing since I was three.

I write at work (marketing materials, research, emails, etc), I blog and do other writing projects on the side, and then for fun and relaxation I write some more (social media, instant-messaging, my paper journal and my public and private Roams, my perfume blog.) Apart from cooking and exercise, I don’t really have non-written hobbies.¹

I have probably written more words, lifetime, than I have spoken aloud.

I am a basically textual creature, is what I am saying.

I also had some opportunities growing up that not everybody gets. My mom was in the publishing industry², and she inked up my school assignments when I was a kid with brutal, spot-on edits. And there were always plenty of books at home; my childhood apartment had a “TV room” that was a bit of a misnomer given that every wall had built-in bookshelves with my parents’ books, clear to the ceiling.

Sometimes I’ll see somebody talk about the challenges of writing, like “How do you avoid getting distracted from writing a blog post to play a game of chess?”

I cannot relate to this. I’d be more likely to get distracted from playing chess and write a blog post.

Or somebody will talk about how she manages the guilt of feeling like she ought to publish more posts.

Again, that’s counterintuitive for me. I’ve spent far more time feeling guilty for not being able to shut my yap and stop posting so much.

I adopted a weekly blog post schedule this past fall, but that’s far less “discipline myself to write more” and more “treat myself to weekly dedicated writing time so I don’t go insane.”

I cannot stress enough that this is not an impressive feat of self-mastery, this is how I am wired. I am a loudmouth. (Or “loud-finger”, I suppose?) I have a lot to say, and it itches to get out.

I have been blogging since college, and have several times tried to quit because I thought it was “unprofessional” or “unserious”, but I always found myself going back to the old addiction, until I met enough legit professional connections through my writing that I finally admitted “ok, the Universe clearly rewards me for doing this and I should feel free to run with it.”

I’m probably the last person on earth who can teach you how to get motivated to write more.

Q: Ok, but how do you write? What actually is your process for writing a blog post?

I do exactly what your English teacher told you not to do.

I don’t really write multiple discrete drafts.

I do a bunch of in-line micro-edits, rephrases, etc, and I sometimes move paragraphs and sections around partway through. This is supposedly Bad Practice but it’s what feels natural and I haven’t tried to force myself to write drafts since I was a teenager.

I also am “rephrasing” things in my head all the time; on my morning walk today I was running through most of this post, and this naturally involves a fair amount of reworking individual sentences and phrases.

Apparently not everybody mentally subvocalizes when they read or write; I do. As I’m writing this line, my “mind’s ear” is simulating the sound of my own voice. So I can kind of go sentence by sentence and think about how it would sound spoken out loud, and whether the flow sounds right, whether it’s grammatically correct, etc. My “writing voice” is very much like my conversational speaking voice, except of course that I can go back and fix all the little problems as I notice them.

When it comes to thinking about what to write, I always have a bunch of “to-write” ideas floating around, some written down, some just in my head.

Mostly my “to-write” backlog is full of research questions, not conclusions. Stuff like:

• are those follistatin gene therapies for real?

• what about those people who say they have machine learning models to read thoughts off an EEG?

• can I collate all the case studies of striking personality changes that occurred as side effects of deep brain stimulation?

Usually these are natural follow-up questions I become curious about in the process of learning about something else; sometimes somebody asks me a question.

Non-research posts, like this one, are different; these are “soapbox rants” that I’ve given verbally or on social media several times, or essays that have been brewing in me for ages and need to get out. They take less time to write when I’m pretty satisfied with my message (which is why I’m doing one today, since my Friday is busier than usual).

But my non-research posts are more likely to be duds that linger as unpublished drafts, since they stray more into the realm of opinion and I wind up feeling disgusted with my not-quite-right lens, especially on more general-interest, non-STEM, emotionally fraught topics, where I worry that I’m not saying objectively true things so much as airing my personal biases.

When I do a research post, I usually have Google Scholar and Sci-Hub open in other tabs, maybe a spreadsheet if I’m trying to be systematic, and I’ll do keyword searches and summarize & footnote what I find in each source as I go.³

I’ll typically also have ChatGPT open, mainly for asking questions. Recent examples:

• “what does “mass properties” refer to in this engineer’s bio?”⁴

• “how much does OPM-MEG cost?”

• "does the attached document refer to any piece of hardware that could be expected to contain optoelectronic components?”

• “what side effects occur from excessive deuterium consumption?”

I have kind of a sense of what I expect ChatGPT can and can’t do in these contexts. I expect, for instance, that the list of side effects of drinking too much deuterium won’t be complete or fully correct, but that it’ll mostly overlap with what I’d find from a Google or Google Scholar search, and I’ll get it faster, with less cruft to dig through.

There’s not really that much to it; I come up with questions, and then I search for answers.

I probably do more searching than most people; I occasionally get Google Scholar accusing me of being a bot for going through too many pages, which I’ve heard is not common.

I also have some heuristics like “try to find a literature that’s reasonably small and then go through all of it.” It is satisfying, for instance, that I know of literally every experiment where they’ve done something to extend lifespan in a mammal.

In a world where contradictory and low-quality studies are common, and where I’m a rando layman who doesn’t have access to the private data or sidechannel academic gossip that would reveal non-obvious info about which studies are fatally flawed, it’s not obvious that I can learn a real thing at all from sitting on my butt in front of a computer screen. Completionism combined with baseline skepticism is one of my few available tactics, and I use it as much as I can.

Q: You always have such a great way of putting things. What are your principles of writing style?

I am very unoriginal on this point.

Just do what it says in Strunk & White. Or Orwell. Or Ogilvy. Or Safire.

If you had a good English teacher, or a newspaper editor, do what they told you to do.

1. Never use a metaphor, simile, or other figure of speech which you are used to seeing in print.

2. Never use a long word where a short one will do.

3. If it is possible to cut a word out, always cut it out.

4. Never use the passive where you can use the active.

5. Never use a foreign phrase, a scientific word, or a jargon word if you can think of an everyday English equivalent.

6. Break any of these rules sooner than say anything outright barbarous.⁵

This is the near-universal consensus on “good” modern English prose style.

To write good English, since the early 20th century, has meant to make your point using strong, short, plain words.

This is what it means to write good journalism, good advertising copy, good explanations of technical topics, good arguments, good speeches — anything where the goal is to communicate important truths to your readers without confusing or boring them.

Writing this way is craft, not art. You don’t need genius to do it. It’s very normal.

Some writers get bored of “good plain English” and do more exotic things — and sometimes they succeed in making something beautiful. But I’m not even trying for that. I mostly write about science and tech for a lay audience, and I have to compete with culture-war ragebait for their attention. I need every ounce of clarity and vividness I can get.

Most of the reworking I do, at the clause/sentence/paragraph level, is for concision and density. I want the greatest amount of vivid and important stuff, and the least amount of boring stuff, that I can fit in each space without the overall flow sounding unnatural.

I could probably get my “ratio” even higher if I spent more time editing, but I also want to have weekly posts that say nontrivial things, so I accept the tradeoff.

What drives me crazy is seeing walls of redundant, vague text in other people’s writing, as though people aren’t even trying to communicate. Maybe they’re trying to sound “respectable” or “professional”, hiding behind a mountain of generic extruded word-product that says nothing. But nobody will read that shit.

As David Ogilvy says,

American businessmen are not taught that it is a sin to bore your fellow creatures.

Don’t Be American Businessmen.

The other thing I try to do, which you won’t find in the style manuals, is to write for both logic and vibe.

Logic: is what I’m trying to say coherent and true? Are my arguments valid? Have I made my case?

Vibe: are the associations and emotions likely to be evoked by my words the ones I intend?

A common failure mode among logical thinkers is to write stuff that works on a logical level and not on a vibe level.

For instance, you never want to describe something as “not sugary” or “not fraud.”

In general, you want to avoid saying “not X” where X is the most vivid, high-emotional-intensity word in the sentence. People’s attention will be drawn to the vivid word first, and much less to the neutral word “not” denying it. They’ll come away with an implicit feeling that maybe the thing you’re talking about is sugary or is fraud. You’d want to rephrase any such sentence completely — maybe you’d use an antonym.

There’s a level on which you read a passage as a sequence of words structured into clauses, and clauses structured into sentences, and claims structured into arguments, connected by logical, meaning-bearing relationships. If you fail on that logical level, you get word salad.

And there’s also a level that’s more like a bag of words — the order of the words doesn’t matter, relationship words like “not” and “but” don’t matter, some part of the mind is just scanning for vivid words and phrases that evoke associations and memories and feelings.

And your writing will be easier to understand and more emotionally resonant and satisfying to read if the vibes evoked from a “bag-of-words” skim match, or at least don’t contradict, the overall point you’re trying to make.

If you fail on the vibe level, your writing comes out dull, confusing, or unsettling. Readers don’t get it and don’t like it.

Robin Hanson’s recent Overcoming Bias post this week is an example of a post that fails on the vibe level, I think.

His main point is that cryptocurrency itself is less valuable to the world than the kinds of people who get into crypto and the beneficial things they do with their newfound wealth.

But the post does not feel like praise of crypto folks, even though in the second-to-last paragraph he speaks of their “great potential to improve the world”.

Most of the rest of the post describes newly rich people, including the crypto rich, as “sensitive to criticism” (sounds bad), says they “resent” skepticism (sounds bad), and that they prefer “flattering narratives” (sounds bad), and so on.

His argument gives a reason to be pro-crypto; his tone is not consistently tuned to be pro-crypto.

A lot of smart people write like this. Sometimes they do it on purpose.

I vaguely remember, years ago, a friend sending me a piece by an academic historian about the Enlightenment origins of science, that described science as “Luciferian.” I was rather put out and asked my friend why he thought I would like a piece condemning all of science as devilish! It turned out that I had misunderstood, and “Luciferian” was meant to have neutral or good connotations, reflecting transgression of the established order in the pursuit of (human) power.

I assume this confusion wouldn’t have come up with her intended audience of historians, but I don’t think this kind of “sophistication” is a good idea when writing for ordinary lay readers.

No matter how clear you think you’re being, somebody is liable to miss your point. So don’t try to play clever tricks.

If you think something is good, you should make it sound good — and not compare it to Lucifer.

1. Would I like to play piano or knit more? Yes. Does my toddler tolerate it? No.

2. she worked at the University of Chicago Press.

3. Once upon a time, I wondered if I should be more “professional” and use PubMed searches, the way the “systematic reviews” in journal articles do. This was a terrible idea. PubMed does not find nearly all the papers associated with your search topic, and includes tons of extraneous, utterly unrelated ones. I don’t know why PubMed is the standard for academics’ lit reviews, but not being an academic I am free to use a search engine that works. God help us all if the team maintaining Google Scholar ever quits.

4. a mass properties engineer manages how mass and weight are distributed across an aircraft, spacecraft, or satellite.

5. https://www.orwell.ru/library/essays/politics/english/e_polit