Principle 01
Open Cold on a Scene: Drop the Reader Into the Moment
In Medias Res — Into the Middle of Things
Opening · Concrete
The Principle in One Line
Don't clear your throat. The first two paragraphs of most drafts are a warm-up — background, definitions, "as we all know." Cut them and start on a specific moment, an image, an action. Horace gave us the rule two thousand years ago: in medias res, "into the middle of things."
"He does not begin from the egg, but always he hurries to the outcome and snatches his hearer into the middle of things."
— Horace, Ars Poetica (c. 19 BC)
Why It Works
Abstract information makes the reader render the picture themselves; a concrete scene arrives already rendered — they see it, they're there. Neurologically, concrete language fires the sensory cortex; abstract propositions don't. So "our ops process has efficiency problems" slides right past, while "3 a.m., and the on-call engineer restarts that server by hand for the seventh time" makes you stop. And the real opening of a draft often hides in your third paragraph — the first two were a warm-up you wrote for yourself. When revising, ask first: can I start from paragraph three?
¶1: background, industry context, "with the rise of…"warm-up — cut
¶2: defining terms, "as we all know," throat-clearingfor yourself — cut
¶3: a specific moment — the real opening is here ↑
Revision rule: delete the first two paragraphs and see if it reads better
Before → After
Before we discuss distributed locks, it is worth reviewing the basic concepts of concurrency. Concurrency refers to multiple execution units…
Two requests grab the same inventory row in the same millisecond. With no lock, both read "1 left," both decrement — and you've just sold two of something you had one of. That hole is what a distributed lock plugs.
This document provides an overview of the challenges our team has faced regarding deployment reliability over the past quarter.
Last Tuesday, a one-line config change took down checkout for 19 minutes — the third such outage this quarter. Here's why our deploys keep breaking, and the three fixes that will stop it.
When to Use + Common Mistakes
- ✓ Tech blogs, incident retros, talk openings, pitches, personal essays, the "impact story" in a promo packet
- ✗ Pure API docs, legal/compliance text — the occasion forbids scene-setting
- Trap: the scene has nothing to do with the point — vivid for its own sake, it leads the reader astray
- Trap: the scene runs too long (five sentences and still no point); or you forget to land it — a scene needs one sentence afterward to connect back to the point
Key References
Horace, Ars Poetica (origin of in medias res) · Sol Stein, Stein on Writing ("start in the middle of an action") · William Zinsser, On Writing Well
This Week's Exercise
Take something you wrote recently. Delete the first paragraph and read it — better? If so, ask whether the second can go too. Then write a "scene opening" for it: a specific moment, person, action, in three sentences or fewer, with a fourth that connects the image back to your point.
Reflection: How much of what you call an "opening" is really a warm-up you wrote for yourself?
Principle 02
Open With a Question: Cut a Gap That Must Be Filled
The Curiosity Gap — An Itch to Scratch
Opening · Suspense
The Principle in One Line
Curiosity isn't a trait, it's a state — lit the instant you realize a piece is missing (Loewenstein's "information-gap" theory). A good opening doesn't rush to the answer; it cuts a precise gap and lets the reader itch.
"Curiosity happens when we feel a gap in our knowledge. Gaps cause pain. When we want to know something but don't, it's like having an itch that we need to scratch."
— Chip & Dan Heath, Made to Stick, "Unexpected"
Why It Works
It all rides on the precision of the gap. "Today let's talk about leadership" cuts no gap — too big, too vague, no itch. "Why do the best engineers, once promoted to management, so often fail to lead a team?" cuts a gap — it names something you thought was obvious but that turns out to contradict itself. The Heaths' insight: instead of asking "what information do I need to convey," ask "what question do I want the reader to feel". And the gap must sit on what the reader already knows — they have to "think they know" before you can let them discover "I hadn't actually thought it through."
Knownreader thinks they get it
Gap = the itch"wait, why?"
Want to knowreads on for the answer
Too big, no itch; not real, no itch — cut it precisely between "I thought I knew" and "I hadn't thought it through"
Before → After
Today I'd like to share some of our team's experience and reflections on rolling out AI tools.
We gave the whole team Copilot licenses. Three months later, the people using it hardest had gotten less done. That broke my model of what an "efficiency tool" even is — so today I'll explain why.
In this post I'll discuss the importance of writing good documentation.
Your best engineer just quit. In her exit interview she never mentioned pay. She said: "I spent half my time reverse-engineering code no one bothered to explain." This is a post about docs — but really it's about why people leave.
When to Use + Common Mistakes
- ✓ Newsletters, talks, blogs, pitches, meeting openings, teaching
- ✗ Urgent alerts, pure specs — the reader wants the answer, not suspense
- Trap: the gap is too big and vague (a whole field), no itch; or you misjudge what the reader knows and the gap doesn't even form
- Trap: clickbait — you cut a gap and don't pay it off; fooled once, never trusted again; or you answer immediately, leaving no time to itch
Key References
Chip & Dan Heath, Made to Stick, "Unexpected" (gap theory) · George Loewenstein, "The Psychology of Curiosity" (1994, the original information-gap paper)
This Week's Exercise
Rewrite the thesis of your next piece as a question the reader "thinks they can answer but hasn't really worked through." The test: send it to three colleagues and see how many reply "huh, why?" — the more who do, the more precisely you've cut the gap.
Reflection: Where exactly is the line between "clickbait" and "curiosity gap"?
Principle 03
The Hook: The First Sentence Has Only One Job
The Hook — Induce, Don't Summarize
Opening · Mechanism
The Principle in One Line
The only job of the first sentence is to get the second one read; the second's only job is the third. An opening needn't "summarize the whole piece"; it must "generate momentum." Zinsser put it most bluntly: if it doesn't catch, the piece is dead.
"The most important sentence in any article is the first one. If it doesn't induce the reader to proceed to the second sentence, your article is dead."
— William Zinsser, On Writing Well, "The Lead and the Ending"
Why It Works
Note Zinsser's verb: "induce," not "summarize." Many writers treat the opening as an abstract — "This article will explore…" — which is exactly how you strip the momentum out of a first sentence. Stephen King spends months on openings; a good one, he says, tells the reader: "Listen. Come in here. You want to know about this." It's an invitation, not a table of contents. There's a sharper test, from journalism: Nora Ephron's high-school teacher had the class write a lede — the principal announces the whole faculty will attend a conference Thursday. The students piled up details; the teacher said the lede was one line: "There will be no school Thursday." The one thing the reader actually cares about — that's your hook.
Before → After
This article will systematically present our data-governance plan across four dimensions: background, current state, problems, and solutions.
We have 47 "single sources of truth," and 9 of them contradict each other. That's why last week's two reports didn't reconcile.
This essay explores the relationship between writing and thinking.
You don't write to record a thought. You write to find out whether you have one.
When to Use + Common Mistakes
- ✓ A piece's first line, email subject lines and first lines, a talk's opening words, PR descriptions, a promo-packet summary
- Trap: writing the first sentence as a table of contents ("This article has three parts"); or throat-clearing ("With the rapid rise of AI…") — a one-size-fits-all opening is no opening
- Trap: the hook is disconnected from the body — you catch hard but can't follow through; or you over-promise and the body can't match the appetite
Key References
William Zinsser, On Writing Well, "The Lead and the Ending" · Stephen King, "Why I Spend Months on Opening Sentences" (The Atlantic, 2013) · Nora Ephron, recalling the "no school Thursday" lede lesson
This Week's Exercise
Find three openings you've written and circle the first sentence of each. Ask of each: is it "summarizing" or "inducing"? Turn the summarizers into inducers — cut "This article will," and swap in a concrete fact, a surprise, or an unanswered question.
Reflection: An email subject line is essentially a one-sentence hook — would your last subject line make you want to open it?
Principle 04
The Full-Circle Ending: Return to the Start, but the Reader Has Changed
Full-Circle — Land Where You Took Off
Ending · Resolution
The Principle in One Line
The best ending often returns to the opening image, line, or question — but because of the road traveled, the same thing now carries new meaning. It gives the reader a sense of completion, like a song resolving to the tonic.
"The perfect ending should take your readers slightly by surprise and yet seem exactly right."
— William Zinsser, On Writing Well, "The Lead and the Ending"
Why It Works
The full circle is powerful because it works both ends of memory — primacy and recency. The image planted at the start lingers in the reader's mind; the ending calls it back, and the arc gathers everything between. McPhee says a piece must "start somewhere, go somewhere, and sit down when it gets there" — the ending isn't a stop, it's a seat. That 3 a.m. engineer from the opening can return at the end as: "Now, at 3 a.m., nobody has to be awake — the system handles it itself." Same scene, meaning inverted. But mind Zinsser's "surprise yet exactly right": a pure repeat of the opening is cheap — the reader must discover "the same image, but I'm no longer the person I was a moment ago."
Opening3 a.m., the engineer restarts the server by hand for the seventh time
→
(the piece: diagnosis · argument · three fixes)
→
Ending3 a.m., nobody has to be awake — the system handles it itself
↩ Same image, meaning inverted: the reader has completed an arc
Full circle = return to the start + added meaning (not a verbatim replay)
Before → After
In summary, through the measures above, we are confident we can significantly improve deployment reliability next quarter. Thank you all.
Remember those 19 minutes at the start? Once we've done these three things, the next identical config change gets caught automatically — in 19 seconds. That's where we're headed.
In conclusion, documentation is very important and we should all write more of it.
Remember the engineer who quit? The next one won't have to reverse-engineer anything. That's the whole point of writing it down.
When to Use + Common Mistakes
- ✓ Talks, long essays, memos, pitches, retros, personal essays
- Trap: "In summary / to conclude" — a bureaucratic close that deflates; or piling on more after the peak, diluting the landing (finish and sit down)
- Trap: introducing brand-new information at the end — this is no place to open a new topic; or a pure repeat of the opening with zero gain, leaving the reader feeling led in a circle
Key References
William Zinsser, On Writing Well, "The Lead and the Ending" · John McPhee, Draft No. 4 ("sit down when it gets there"; on the ending echoing the start)
This Week's Exercise
Dig out an old piece that had an opening image or question. Rewrite the ending so it returns to that image or question — but carries the new meaning earned by the whole journey. Read it twice: once to check it's a "surprise," once to check it's "exactly right."
Reflection: Which forms suit a full-circle ending, and which are better off cutting off sharply or ending on a call to action?
Going Deeper
Pushing the limits of each principle, the EN/ZH differences, and fit across forms
Which is harder, opening or ending? Why can so many people start well but fail to land?
The opening is hard because it's "something from nothing"; the ending is hard because it's "resolution from a full hand." Most people get lit by an image, then arrive at the ending drained and deflate into "in summary." A practical move: after the first draft, lay the ending and opening side by side and ask whether they can "talk to each other." Whatever question the opening asked, whatever image it drew, the ending has a ready echo to use — the full-circle method turns "inventing an ending" into "answering the opening," halving the difficulty.
EN/ZH opening habits: the classic Chinese build-up vs. the inverted-pyramid lead — which way should you lean?
Classic Chinese composition values build-up — set the stage, then arrive at the point; English news and business writing run the inverted pyramid — conclusion first. For a senior engineer and leader, the workplace tilts almost entirely toward the latter: your readers are impatient peers and bosses. But separate "build-up" from "scene": what you cut is abstract preamble (background, definitions); what you keep is the concrete scene — a single image can arrive at the point instantly. The scene opening is, in fact, the intersection of both traditions.
Across essay, talk, and video, how long is the hook's window in each? Same mechanism?
An essay's window is the first sentence (a ~3-second skim); a talk's is the first 30 seconds; a short video is brutal — the first 3 seconds (one thumb swipe and you're gone). The mechanism is the same — all "generate momentum to continue" — but the levers differ: text relies on a surprising fact or a precise question; a talk can add a pause, eye contact, a prop; video leans on visual conflict and pace. The shorter the form, the more you must front-load the gap — there's no room even to clear your throat.
Why does AI so often open with "with the rise of…"? How do you train it (and yourself) to skip the warm-up?
Because that kind of opening is the statistically safest — it's the highest-frequency move across vast text, and the model defaults to the mean. Two fixes: first, an explicit instruction — "no background preamble; the first sentence must be a concrete scene or an unanswered question"; second, treat its draft as a "warm-up draft," cut the first two paragraphs outright, and start reading from the third. Same goes for yourself — admit the first two paragraphs were written for you, and they're easier to delete.
"Surprise yet exactly right" is a contradiction — what makes it a surprise, what makes it right, and how do you get both?
Surprise comes from "angle" — the ending doesn't restate old lines but lifts the reader to a slightly higher vantage to look back; rightness comes from "setup" — that angle must have been seeded at the opening or along the way, so the reader reads it and thinks "ah, of course." The full circle gets both for free: returning to the opening image is "right" (the clue was planted early), and the inversion of the image's meaning is the "surprise" (the same thing now reads differently). A good ending isn't a line conjured from thin air — it's a fruit the opening planted, the whole piece watered, and the close finally bore.