I’ve been a salaried writer and editor for the last eight years: newsletters, websites, blog posts, essays, and a bit of research.

People sometimes ask me for writing tips. This post collects my advice — all of which I still use myself.

Preparation

Read good writing. 

The more good writing you see, the more you’ll notice when your own writing sounds off.

Read the sort of thing you want to write. If you spend the day on Twitter, your thoughts will take the shape of Twitter posts — which is great if you want to write for Twitter (no shame in that), but not so good for longer work.

I like this line from Scott Alexander:

Almost the only good advice in any discipline is “develop instincts, then use them”.

You develop instincts by writing and reading; writing is best, but reading is easier. And it’s hard to learn through writing alone if you don’t know what you’re trying to become.

Take notes. 

Things happen around you all day. If you aren’t reading, you’re working, or walking around and noticing stuff. If something makes you stop and think, or makes you want to complain, or sends you on a brief flight of fancy, write that down.

I have a couple of Notion files for this — one for fiction, one for blog posts, and one “commonplace book” where I copy down interesting text (books, essays, anime). Make your own files on whatever service you’ll actually use.

In the last 24 hours, I’ve written the notes:

• “Buying art”, because I talked to someone about how much I love buying art and decided to promote the idea.
• “Anti-follows”, because I read something so bad I wanted to give the account negative one follow, but that’s impossible. So now I want to think of ways to punish them the equivalent amount.
• A fiction idea inspired by a story I read, with a cool premise and bad execution. “I could do that better!” is a great motivation — you either write something good or learn a valuable lesson about hubris.

Decide whether something is worth writing. 

The fastest way to write something is not to write it.

When I reviewed ~600 saved-up ideas recently, I threw out around half of them. Some expired, some were stupid, some were too hard, and some had already been written (well enough that I wasn’t motivated to surpass them).

Before you spend hours writing something, spend five minutes checking whether it’s already out there. (Unless it’s something only you could write.) If you find something great, pick another topic — or write a commentary on the other piece, or find some way to mix it in with your work. For example, I quoted other reviews when I reviewed The Book of Disquiet.

Also keep the audience in mind. Will you be satisfied if your piece is read by ten people? Would it need to be a thousand? Or perhaps ten very specific people? Can you make sure they’ll see it?

Think about your readers.

You publish work because you want someone to read it. Who are they? What do they want?

And what do you want from them? Should they feel a certain way? Laugh, cry, get mad? Should they remember an important fact?

Should they leave the post with a next action — buy a new chair, practice juggling, donate to GiveWell, call their representative?

If you’re lucky, you get one or two things from each reader. Focus on the most important ask. You can even make it the title:

Writing

Skip whatever blocks you.

If something isn’t flowing, add a note like FIX LATER and move on.

Most of the time, you should separate research from writing. Don’t look up facts mid-essay unless they’ll determine your next move — type FILL IN LATER and keep going.

If something bores you, or you get stuck, skip it. You may reach the end and discover you didn’t need it. Aella says it well:

In marketing, your boredom itself is a virtue. If you are writing a blog post and discover your attention is wandering, STOP AND TRY SOMETHING ELSE. If you are trying to explain a concept and the life drains out of you, STOP AND TRY SOMETHING ELSE. It’s an important signal!! If your attention fades off of it, probably others’ will too.

If you’re really stuck, and you don’t even know where you’d skip to, try:

• Sleeping on it.
• Writing disconnected sentences about the topic until you hit a thread worth continuing.
• Asking AI for ideas — like a fancier version of lateral thinking cards.
  • The best prompt likely isn’t “what should I say next?” — though you can try that, prompts are cheap — but something like “what does this make you think about?” or “imagine being TYPE OF READER, what are you curious about now?” Think of the AI as a reader, not a writer.

Only use extra words if you have a reason.

By default, nonfiction writing should be short. It helps the reader pay attention and remember things later. Most editing is finding shorter ways to say the same thing.

Excess words are fine as long as they do something. If you feel sad about cutting something, talk to yourself and explain why. Any reason is good enough — “it sounds better”, “it’s more specific”, “it makes me laugh”, “it adds clarity”. But you’ll find that most of the excess was just sitting around not doing much.

Write in sentences.

It’s easier to think in sentences, not paragraphs.

If I’m struggling with something, I give each sentence its own line. That makes them easier to examine — is this true? Does it need to be here? Does it overlap the next sentence?

Once the sentences are in shape, you can decide how to connect them into paragraphs. Most paragraphs should be two or three sentences.

Long, loopy, weird sentences can be fun. But you should surround them with short, clear sentences; the contrast makes them sing.

Keep your voice alive.

Writing isn’t just about efficiency. Inefficiency is human. Voice is human. It’s nice to read something and see the human on the other end.

Humor is awesome. I wish just about everything had more jokes. Don’t feel like you have to cut jokes — if something makes you laugh, that’s a fantastic reason to keep it, even if you think it might be stupid. But don’t force it; trying to add jokes later is a bad sign, with rare exceptions (stand-up comedy, speeches where the crowd is half-asleep and desperately wants to feel something).

If you think in footnotes, add footnotes.

Write the beginning next to last — or skip it.

The beginning is the second-most important part. It’s how we decide whether to read the rest. Write it next to last, when you have the advantage of maximum context.

If you write the beginning first, you might find by the end that it no longer makes sense and has to be rewritten, or removed entirely. But if you leave it for next to last, you might find that you want to write something different, or nothing at all — sweet!

On “nothing at all”: Plenty of pieces are fine starting where they start, without a real introduction. Ask yourself: Would the reader actually be confused without this? Would something important be lost?

Write the title last.

The title is the most important part.

If you know exactly what the reader wants, your title can be workmanlike and literal: Things I recommend you buy and use, Playing Magic: the Gathering for money.

If all you know is “they want something to read”, draw them in. You can win the reader’s attention by yelling at them, making them squint and say “what?”, talking about sticky topics like sex or death, or making them feel seen.

Good titles from other Inkhaven writers:

• Kill Yourself Cave (striking, weird, taboo)
• Persuading a million people to do something is easier than persuading one (makes you squint, has a big number)
• Rooms that hate you (striking, strong emotion, talks directly to you)
• You don’t have to say “I’m good, how are you?” (but… I do say that! How did he know? There’s a better way?)

If you have a good visible tagline, you can get away with a bland title: “Dr. Strangelove” vs. “How I Learned to Stop Worrying and Love the Bomb”.

Editing

Change the form to wake up your brain.

When you stare at words on your computer for long enough, you stop seeing them. Your brain skims past them, picking up the meaning without actually reading sentences.

Fix this by taking the words off your computer:

• Read them out loud so you hear them instead of looking.
• Better yet, have a friend (or an AI) read them out loud, so you hear them in a different voice. Your friend might stumble over a sentence that worked for you because they don’t know what comes next.
• Print them out and edit on paper, with a pen. (I did this throughout college.)

Another technique is to give each sentence its own line (like I mentioned earlier). Read each one alone, so its neighbors can’t hide its flaws.

Don’t stress about making it perfect.

Every hour of editing helps less than the previous hour. Once you reach draft #3, further edits probably aren’t worth it. (There are exceptions — your thesis, an article you’re trying to sell — but casual blog posts aren’t among them.)

Every hour of editing is an hour you aren’t writing. You’ll develop faster as a writer, and write more good stuff, if you focus on creating more and taking more risks, instead of polishing your early work.

I dropped this blog for a long time because I couldn’t publish without far too many edits, even though I had almost no readers. It stopped being fun. And now, when I go back and read those early “perfected” posts, they mostly seem… bad. I wasn’t a good enough writer to make anything perfect, so why was I trying so hard?¹

Write too little, not too much.

Err on the side of making things too short.

It saves the reader’s time, and yours. Even if it’s easy to throw everything onto a page, it makes editing harder. Consider removing things instead of refining them.

If you leave some things unwritten, and leave the reader wanting more, you have a natural hook for a followup post — something that will bring them back, and get them in the habit of reading you. But if the reader gets tired, they’ll leave before the end: much worse.

Use AI.

It’s cheap, fast, and somewhat helpful. It tends to give good line edits, and correctly point out the most boring content.

I listen to maybe two-thirds of Claude’s suggestions, mostly around sharpening sentences to be less vague, obvious, or cluttered. I tend to ignore suggestions about removing jokes or things Claude finds “too cute”.

As of April 2026, it still isn’t a good writer, or a good rewriter.

I rarely use Claude’s suggested titles, but I often take a suggestion and find a better variation.

My standard editing prompt:

Here’s a blog post, written for audience X. I want them to (feel/think do) X.

• What’s the implied promise of this post, and do I deliver on it?
• Is any material not pulling its weight? If you had to cut 20%, what would it be?
• What’s the slowest part?
• Where am I going through the motions?
• What’s most memorable? What would someone share if they liked the post?
• Where does the voice feel flat or generic?
• What did you want to know more about?
• Where should I be more specific?
• My current title is X. Give me ten better suggestions.

Other good advice

William Zinsser wrote my favorite book about writing. Very tactical and a pleasure to read.

These essays are worth your time:

• Scott Alexander (mostly nonfiction)
• George Saunders (mostly fiction)
• Joanna Wiebe, Stripe (mostly sales)

 

1. I solved this with Inkhaven. The timing let me bully Future Aaron into writing (while Present Aaron was busy watching reaction videos), and the midnight deadlines force me to stop editing.