Back to Podcast Digest
AskwhoCasts AI··45m

Half A Month Of Consolation Writing Advice - By Scott Alexander

TL;DR

  • Micro-dishonesty is a writing poison — Scott Alexander says even tiny acts of self-deception, like smuggling your real point into a more “presentable” essay, make prose feel awkward because “the English language hates even the slightest whiff of dishonesty.”

  • Most freshman-writing rules are training wheels, not eternal laws — his recurring joke is to spend “30 years on a distant mountaintop” mastering active voice, no adverbs, no hedging, and the five-paragraph essay, then come back down and break the rules on purpose like Shakespeare or Hemingway.

  • Explainers fail when the writer has no real question or thesis — his example is the generic AI explainer that drifts into prompt engineering, data centers, or transformer layers because the true motive is just “I feel like I should write an AI explainer,” not answering a specific constrained question.

  • The antidote to stale blog discourse is first-hand contact with reality — instead of recycling Silicon Valley memes about Seeing Like a State, he argues good writing comes from moving closer to the world: direct experience, primary sources, or at least actually reading the book rather than a tweet about a review of it.

  • You usually don’t need a genius-level idea; you need to do the obvious work nobody did — on topics from immigrant crime to COVID lockdowns to racial bias in the justice system, he claims even in 2026 the low-hanging fruit is still there because few bloggers seriously gather details or read the old scientific papers.

  • In an AI age, non-fiction has to earn its keep by saying what readers can’t trivially regenerate from a prompt — his closing test is brutal: if your essay just lists obvious pros and cons that anyone could produce in 30 seconds, why should anyone read you?

The Breakdown

The opening case against “micro dishonesty”

Scott starts with the sharpest claim in the talk: writer’s block and ugly prose often come from lying to yourself. He tells stories of mentees who wrote around the thing they actually wanted to say — including one autobiographical essay whose “recovery” section rang false because the recovery hadn’t really happened yet. His line is memorable and brutal: English punishes even tiny dishonesty by making your writing worse.

Cliches, passive voice, and the 30-year mountaintop

From there he moves into classic writing rules and reframes them as disciplines rather than commandments. “Avoid cliches like the plague” becomes less about deleting stock phrases and more about noticing missed opportunities for something vivid, like his Jim Cramer stock-pick joke. The passive voice rant gets the same treatment: freshman English is simplistic, but beginners still benefit from strict practice until they develop an ear for the deeper “indescribable bad thing.”

Untangling sentences instead of worshipping rules

His examples here are very Scott: playful, obsessive, and concrete. He walks through five versions of “Bob hit the ball,” including a deliberately absurd blindfolded-baseball anecdote to show that even a tortured sentence can be right in the right context. Then he revises clunky phrases from his own drafts and past Inkhaven posts — like changing “John was the person who we were looking for” to “We were looking for John” — to make the real problem obvious: tangled attention.

Why generic explainers feel hollow

One of the most relevant sections for AI readers is his attack on explainers that exist only because the writer feels they ought to exist. His AI example is perfect: if you don’t know whether your piece is about prompt engineering, data centers, or transformer layers, you probably don’t have a real reason for writing it. His fix is simple but demanding — turn the piece into an honest answer to an actual question a curious person might ask.

The five-paragraph essay and the blogosphere trap

Scott unexpectedly defends the much-hated five-paragraph essay as a scaffolding tool: thesis, topic sentences, evidence, conclusion. It’s cliched and soulless, he admits, but he keeps returning to it because so many weak essays still fail at those basic structural promises. Then he zooms out to complain about the “bloggerosphere,” where bloggers read blogs, get their opinions from blogs, and serve warmed-over versions back to readers who already saw them on Substack last week.

Competence versus brilliance, and how to find ideas

He clarifies a remark that he finds only about one new blogger per year he truly loves: that’s the bar for brilliance, not for usefulness. Most writers don’t need to be Matt Levine, whose finance newsletter he reads despite having no desire to trade credit default swaps; they just need to write competently about something people genuinely care about. And for anyone asking where ideas come from, he basically says: everywhere — the world is still scandalously under-written if you’re willing to do the work.

Controversy, poetry, and startup “runway”

The human centerpiece is his grudging admiration for Mikhail Sam, who arrived at Inkhaven and immediately published an attack on Light Cone while living on its campus and eating its food. Scott thinks it was wrong and awkward and still can’t help respecting the “true blogger spirit” of kicking the hornet’s nest for real rather than faking controversy. He follows that with a lovely section on poetry as a machine for escaping cliche — culminating in his delighted line “it rises up like auction bids” — and a practical startup metaphor: readers give you only a little runway, so don’t waste it on blue-curtain coffee-shop preambles.

Conflict, curiosity, and the final AI-era standard

In the last stretch, he argues that even non-fiction needs narrative propulsion, usually conflict or curiosity. His favorite example is an Inkhaven post on Hammurabi that hooks the reader by asking why scholars are uncertain about exact dates that are somehow still weirdly specific. He closes with the most modern test in the piece: if your argument can be trivially regenerated from a prompt — obvious pros and cons anyone could list in 30 seconds — then the burden is on you to offer something earned, specific, and actually yours.