The incident reports, postmortems, and decision memos a technical person writes are all nonfiction at heart: telling others what really happened — accurately, verifiably, credibly. Journalism spent a century forging this craft: most important first, every fact survives a follow-up question, move people with scenes not adjectives, draw conclusions from evidence not stance. This week, four things to learn from reporters.
Principle 01
The Inverted Pyramid: Conclusion in the First Sentence
The Inverted Pyramid — Most Important First
News · Structure
Principle + Master's Words
Put the most important conclusion at the top and order information by decreasing importance — wherever a reader stops, they already have what they most needed.
"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, Ch. 9
Why It Works
The inverted pyramid was born in the newspaper era: editors cut copy from the bottom, so reporters wrote in descending order — chop any line and the core survives. Today the reader's attention is that scissor — in email, Slack, incident reports, most people read two lines and leave. It shares a root with the Pyramid Principle (Day 2), but goes further: not only is the conclusion first, the whole piece is ordered as if the reader might quit at any moment.
Core conclusion · who, what, impact, what you need to do
Key support · data, root cause, scope
Secondary detail · timeline, background
Extras · quotes, appendix
Must readSkippable
The inverted pyramid — cut at any line and the reader still has that layer's full point
Before & After
Last night our monitoring caught some anomalies, the team investigated, and after several hours we confirmed it was a database connection issue; service is now restored.Payments was down 47 min last night, now restored. Root cause: connection-pool exhaustion; scaled and alerts added. No data lost; 3 orders need manual reconciliation (list at end).
After several weeks of discussion across teams, weighing many factors, we have arrived at a decision regarding the framework.We're standardizing on Framework X. Migration starts July 1, done by Q4. Here's why, and what your team needs to do.
When to Use + Common Mistakes
✓ Incident reports, status updates, decision emails, press releases — any time the reader is busy
✗ Not for: suspenseful narrative or stories that need emotional build-up (use the story arc, Day 4)
✗ Mistake: leading with "background" — the reader reaches paragraph three before learning what happened
✗ Mistake: cramming five points into sentence one, so nothing leads; put one heaviest conclusion per sentence
This Week's Exercise + Question
Take a long email or report you recently wrote and flip it: make sentence one "the single thing the reader most needs to know," then order the rest by decreasing importance. Trim until cutting the last paragraph wouldn't hurt understanding.
Question: The inverted pyramid sacrifices suspense and rhythm. When does "lead with the answer" actually backfire?
Principle 02
Verification: Turn "I assume" into "I confirmed"
Verification — The Discipline That Defines Journalism
News · Verify
Principle + Master's Words
What separates journalism from rumor and PR isn't style — it's that every falsifiable claim has been independently verified. Verification is a discipline, not a talent.
"In the end, the discipline of verification is what separates journalism from entertainment, propaganda, fiction, or art."
— Kovach & Rosenstiel, The Elements of Journalism
Why It Works
The core move is separating "I think" from "I confirmed." Three plain rules: named beats anonymous, two independent sources beat one, original documents beat secondhand accounts. For a technical person this is the lifeblood of the postmortem — between "probably caused by" and "logs confirm it was caused by" lies a team's credibility. With every assertion, ask: if someone presses "how do you know?", can I produce evidence? If not, soften the claim or go check.
Before & After
The outage was caused by instability in the upstream service.At 14:02 the upstream API error rate rose from 0.1% to 38% (monitoring link); our retries saturated the connection pool by 14:05 (log link). Upstream was the trigger; un-backed-off retries were the amplifier.
Users were unhappy with the redesign.23 of 40 surveyed users rated the new flow worse (survey link); navigation support tickets rose 3× week-over-week (dashboard).
When to Use + Common Mistakes
✓ Postmortems, root-cause analysis, research reports, external announcements — any conclusion that may be cited
✗ Mistake: hiding behind "reportedly / I heard / everyone feels," writing rumor as fact
✗ Mistake: confusing correlation with causation — two things co-occurring isn't one causing the other
✗ Mistake: concluding from a single source — especially when it conveniently supports what you wanted
This Week's Exercise + Question
Take a report you wrote and tag every assertion by source: ① personally verified ② backed by document/data ③ secondhand ④ pure speculation. For every ③ and ④, either add evidence or rephrase as "to be confirmed."
Question: What source tier is an AI's answer? How do you verify a confident-sounding AI output?
Principle 03
New Journalism: Render with Scenes, Don't Assert with Adjectives
New Journalism — Show, Don't Assert
News · Narrative
Principle + Master's Words
Nonfiction can carry a novel's tension. Tom Wolfe distilled four weapons of the "New Journalism": scene-by-scene construction, full dialogue, point of view, status detail — render instead of label.
"...scene-by-scene construction... the recording of dialogue in full... the third-person point of view... and the recording of everyday gestures, habits, manners — the 'status life' of the people."
— Tom Wolfe, The New Journalism (1973)
Why It Works
An adjective is a judgment the writer makes for the reader; a scene lets the reader reach the judgment themselves — more credible, more memorable. "He was furious" is an assertion; "He set the coffee cup down gently and looked at no one" is a scene — the reader reads the fury. Status details (what someone wears, the words they use, where they sit) quietly reveal identity and standing. This craft is especially useful for the "human" parts of a promo packet, a team story, a postmortem.
Scene by sceneOpen on a concrete moment, not a generalized "usually"
Full dialogueLet people speak; speech beats paraphrase for presence
Point of viewTake the reader inside one person's eyes
Status detailImply identity through objects and gestures, don't state it
Wolfe's four nonfiction devices — make real events read like the scene itself
Before & After
The on-call engineer was under a lot of pressure, very anxious, and the team mood was tense.At 2 a.m. the alert fired for the seventeenth time. She stared at the red line that kept climbing, her hands paused on the keyboard for three seconds, then typed into the channel: "I need someone to look at this with me."
The customer was a very demanding and difficult person.He read every line of the contract aloud, twice, and circled the word "best-effort" in red before he would shake hands.
When to Use + Common Mistakes
✓ The human passages of a postmortem, personal-brand essays, team narrative, any place that must move rather than inform
✗ Not for: pure technical specs or conclusion-first emails (use the inverted pyramid, no build-up)
✗ Mistake: inventing or distorting detail for drama — nonfiction's floor is the truth
✗ Mistake: piling on adjectives as "style" ("utterly devastating, incredibly hard") — that's writing nothing
This Week's Exercise + Question
Pick a work moment you lived through (a launch, an argument, a late night). Summarize it in one sentence, then write it as a 150-word scene: only action, dialogue, and visible detail — no evaluative adjectives.
Question: When the real details are mundane, can "showing" fall flat? How do you choose then?
Principle 04
Investigation: Build Accountability on Evidence, Not False Balance
Investigative Reporting — The Best Obtainable Version of the Truth
News · Accountability
Principle + Master's Words
Investigation isn't a longer report — it's closing in on the truth through accumulated evidence. The goal is "the best obtainable version of the truth," not laying both sides side by side and calling it fair.
"Our job is to put together the best obtainable version of the truth."
— Bob Woodward
Why It Works
The key is separating "false balance" from "accountability to evidence." Placing "some say A, others say B" side by side looks neutral but offloads judgment onto the reader. Investigation does the opposite: documents over claims, multiple sources cross-confirming, follow the evidence to wherever it points — even to a conclusion you dislike. Each piece needs a nut graf — one paragraph saying why this matters. For a technical person, root-cause analysis is an internal investigation: keep asking why, chase it down to the commit, the log, the review record; don't stop at "I guess."
Before & After
Some say it was a system design flaw, others say it was an ops mistake; the two sides disagree and it's hard to conclude.Three independent pieces of evidence point to one root cause: a config change shipped without a rollback plan (commit link + missing review record), and monitoring didn't cover that path (alert-config screenshot). It's not about who's wrong — a gate was missing in the process.
There were some concerns about the vendor's reliability.The vendor missed its SLA in 4 of the last 6 months (incident log); two teams independently flagged the same timeout pattern (tickets #2231, #2417). This matters because renewal is due in 30 days.
When to Use + Common Mistakes
✓ Root-cause analysis, due diligence, audits — any report that must name a problem and own it
✗ Mistake: false balance — using "both have a point" to dodge a conclusion you could have verified
✗ Mistake: picking the stance first, then the evidence (confirmation bias), citing only what helps you
✗ Mistake: listing evidence with no nut graf, leaving the reader asking "so what?"
This Week's Exercise + Question
Pick an unresolved dispute on your team. Instead of "hearing all sides," find three independent objective pieces of evidence (data, documents, logs) and see whether they point to the same conclusion. Then write a one-line nut graf: why this is worth solving now.
Question: When the evidence points to a conclusion that offends someone — or overturns your own earlier judgment — how do you write it?
Going Deeper
The inverted pyramid and the story arc (Day 4) are opposites — one drops the conclusion first, one builds layer by layer. Which should a technical person default to?
Default to the inverted pyramid: 99% of daily work writing (emails, reports, status updates) faces a busy reader who may quit any second, and conclusion-first respects their time. The story arc is worth deploying only when you need not to "inform" but to "persuade / mobilize / move," and you're confident they'll read to the end. The cost of misuse is asymmetric — use the arc where the pyramid was needed and the reader leaves during your build-up; the reverse costs only a little drama.
Is "the discipline of verification" easier or harder in the age of AI?
Harder. AI drives the cost of generating "confident-sounding statements" to near zero, while verification cost is unchanged. Rumor once needed people to spread it; now a fluent, self-assured paragraph with fake citations appears instantly. This actually raises the scarce value of verification: people who can tell "reads right" from "survives a follow-up question" become more valuable. Treat AI output as an unnamed secondhand source — useful, but independently verify before publishing.
How does the "show" technique of New Journalism differ between Chinese and English?
The English narrative tradition leans more on precise verbs and concrete nouns (Wolfe, Didion are verb masters); Chinese "showing" draws more on baimiao — the understated, Lu Xun / Wang Zengqi-style arrangement of detail. The shared enemy is adjective-piling, but Chinese carries an extra risk: idioms "package" judgment for you ("焦头烂额," "剑拔弩张") — vivid-seeming, yet like adjectives they conclude on the reader's behalf. Chinese baimiao must dare to use the plainest verbs.
Does investigation's "evidence over claims" clash with psychological safety on a team?
A surface clash, but complementary in fact. Accountability targets "the matter and the process," not "the person" — which is exactly the spirit of the blameless postmortem. Evidence shifts the discussion from "whose fault" to "which gate was missing," lowering interpersonal defensiveness. What truly harms safety is false balance and vague wording: ambiguity makes everyone suspect the finger could be pointing at them. Precise evidence is the greatest kindness to people.
Can these four weapons — inverted pyramid, verification, showing, accountability — coexist in one piece?
They can, and the best nonfiction does exactly this. A great incident postmortem: open with the inverted pyramid (impact + root cause + action), support every assertion in the body with verified evidence, use a scene or two at the key moment to put the reader "there," and aim the whole at the process rather than the person. The four aren't a pick-one — they're four gears: structure via the pyramid, credibility via verification, power via showing, public trust via accountability.