This is the English version. (Korean original: 내 것들과 요즘의 글쓰기)
It’s been quite an eventful stretch.
There’s a lot I want to say and a lot to write, but for now let me try to get some of it out here, even in a somewhat unorganized form.
The goings-on at the university in China, my research and my writing, AI and tech and development, my new hobbies of photography and making things — and on and on and on…
My things…
Because we’ve entered the age of AI? The phrasing already feels stale.
As I worked my way through AI-powered tools, I found myself drawn more to the personal and the creative — and feeling that they matter more.
Now that humans and machines write in the same code — language — (no matter how much AI detection you run) there is ultimately no way to tell the two apart unless you’re the author of both.
I’ll blabber about that in a separate piece later, but what I want to say is this: out of a greed to write in a way that expresses a more personalized, raw “me,” I got hold of this uncomfortable, awful contraption called a typewriter.
These days, if you ask about my hobbies or interests: “audio,” “writing,” “photography,” “building things through vibe-coding” — well… trivial stuff like that. For all their triviality they cost a lot of money, and they’re the kind of things that mean nothing without the most cutting-edge tech and — beyond analog — the human as the medium… Let me unpack them one at a time later.
I agonized over all sorts of things for this silly little blog. The blog is basically hosted on GitHub, and the problem is the photos and videos. Videos I can handle somehow with YouTube, but photos are a hassle to manage, and hosting good, high-resolution ones is hard. So I uploaded my photos to Flickr and built an app that extracts the URLs to pull them into markdown locally.
The world’s gotten so convenient. Digging through the page source to find the URL for a photo I’d put on Flickr was annoying, so I hacked a few things together and this is what came out — took me about five minutes, maybe…
So, did it make writing fun?
I bought this guy half a year ago.
The one I actually carry around is the Traveler version, but when I leave it at home — or just want to focus on the “writing” itself — ah, it’s a real marvel…
You could call it superficial flair, or just a tech-otaku’s self-indulgence, but doing what you want, the way you want — isn’t that freedom, and shouldn’t I be grateful for it?
Right — so what’s the upshot? Scribbling words out is fun!!!
If you ask whether it’s productive, well… that’s a whole other story…
On writing
People who know me know this: I’m a fundamentally sentimental person, and I write emotionally. My first drafts are always thrashing along on the flow of feeling — that’s the only way the writing comes out at all. Which means I revise it countless times.
After that, it turns into something that’s just about passable.
That a guy like this writes the technical genre called “papers” and makes a living off it — that too is quite the irony…
Sometimes what others call easy is far too hard for me
Late in my PhD and during my postdoc, I got a few commissions to write professional pieces for an HRD magazine.
Ah… damn it… it was brutally hard. If they’d told me to write a research paper on the same topic I’d have knocked it out — but this was a genuine ordeal.
A so-called paper is writing that even a lowly guy like me can produce with good training.
Research itself takes sense and intellectual ability, but I’m talking about the writing style called a “paper.” On top of that, research writing doesn’t need to be very kind (not that it needs none at all).
But professional writing — or writing for a general audience — fundamentally needs a narrative to carry to the reader, and within that narrative you have to dissolve your specialist topic and persuade the person reading it.
And
it has to be kind (in many cases).
So… it’s dreadful…
When you persuade kindly, you slip into this mode of pressing down — “couldn’t you please see it this way?” — and it becomes hard to tell whether I wrote what I wanted to say, or wrote to fit the person.
The two beloved human…s
To borrow the language of the artsy types: in “this scene,” there are two people who heavily shape my writing and whom I always turn to.
Seungsu-hyung and Seongjun-hyung…
I can’t say I learned my writing style from them (I’d feel too bad claiming that), but at least when it comes to the motif and motivation of my writing, those two are always in my head, murmuring like an angel and a devil, saying this and that.
Honestly, Facebook was my emotional trash can in the early days, and only the people who could take it stuck around. But as the crowd grew, I couldn’t keep doing that, and watching the pieces the hyungs wrote, I thought, “ah — this is the way to go.”
The stuff I write on Facebook or LinkedIn these days
As I scribbled above…
I’ve fought hard to find my uniqueness at the two extremes.
I’ll get into this in a separate piece later, but the pieces I write on Facebook or LinkedIn go through a few stages.
Basically I don’t have the capacity to write broad, sweeping insight, so I write from the interim conclusions and stray thoughts gathered out of my research (or the research process).
And I write a very boring, stiff report.
Somebody once said… “hey, it’s good, but how’s anyone supposed to get this?” So I gathered up the pieces I’d written and defined a narrative-centered style.
Fortunately, not all the pieces were good — but there weren’t many outright disasters either — so I pulled the strengths from each, defined a style, and documented it.
After writing a piece or two in that style…
And around that time I was fixated on skills and harnessing, so I built a few skill-sets that layer narrative on top of the stiff-report style and applied them in Claude Code. If I had to define it, I’d call it AI-Assisted writing.
(Human writing > AI-Assisted (human-led, AI assisting) > AI-Powered (AI-led, human assisting) > AI-generated writing.)
# "36 Lenses, One Truth" — A Writing Analysis and Practice Guide
**Target text**: final_interim_report_interpretation.md
**Date**: 2026-04-03
---
# Part 1. What makes this piece different
## Genre definition: "Research Narrative"
This piece is neither an academic paper, nor an analysis report, nor an essay. It blends all three.
| Genre | What this piece takes | What this piece drops |
|-------|----------------------|-----------------------|
| Academic paper | Statistics, theory citations, hypothesis testing | Formal structure (IMRD), passive voice, detachment |
| Analysis report | Data-grounded evidence, model comparison | Table dumps, parallel reporting of results |
| Essay | Narrative flow, metaphor, emotional breath | Subjective opinion, unsupported claims |
Result: **a piece that speaks in numbers but delivers as story**. It lets the reader live through "the process of a researcher wrestling with the data."
---
# Part 2. Seven core traits
## Trait 1: Questions build the structure
The spine of the piece is three questions.
1st: "What is happening?" → exploration
2nd: "Why is it happening?" → verification
3rd: "Under what conditions?" → optimization
The 36 analyses are not laid out chronologically. They're rearranged along the axis of "the deepening of the question." The reader experiences the level of the question rising, alongside the researcher.
**Practice point**: Before listing your own results, first write "how many times, and what kind of questions, did I ask of the data?" The shift in the questions becomes the structure of the piece.
---
## Trait 2: It overturns the "obvious conclusion"
The tension of the piece comes from **the betrayal of expectation**.
| Expectation (common sense) | Betrayal (finding) |
|----------------------------|--------------------|
| High DT awareness → high training effect | R²=.005. Six models null. |
| More prepared firms learn better | High-prep Q3_diff = same as low-prep |
| Training has no effect | No — the measurement just can't catch it |
| No linear relationship = no relationship | It revives in the nonlinear |
Each section moves in the order "common sense → data → reversal → reinterpretation." Same structure as a detective novel.
**Practice point**: When writing up a result, first write "what did we expect before seeing this?" Then write "how did the actual result betray that expectation?" The wider the gap between expectation and reality, the more force the writing gains.
## Trait 3: Metaphor turns the abstract into the concrete
> "Like oxygen and hydrogen — neither reacts on its own, but together they become water."
That one sentence makes the statistical concept of "the significance of an interaction term" understandable to anyone.
Metaphors used in the piece:
| Statistical concept | Metaphor |
|---------------------|----------|
| Interaction significant | Oxygen and hydrogen → water |
| Inverted-U curve | Small firm (no capacity) → mid firm (absorbs) → big firm (ceiling) |
| Convergence of 36 models | "The puzzle pieces fall into place" |
| Methodological triangulation | "Lenses" |
| RSM contour | "A prototype of a diagnostic tool" |
The rule for metaphor: a **1:1 correspondence** must hold. Just as "oxygen = DT awareness, hydrogen = smart systems, water = training effect," each element of the metaphor maps precisely onto each element of the original concept. Not a poetic metaphor, but a **structural** one.
**Practice point**: Write your core finding in a single sentence. Then make three metaphors that begin "this is like ___." Pick the one whose 1:1 correspondence is most exact. If the metaphor isn't precise, it's better not to use it.
## Trait 4: Numbers, selectively, and in context
This piece has a lot of numbers. But **not every** number is in it. Rather than the full results table of all 36 models, only "the numbers needed to tell this story" were chosen.
The pattern in which numbers appear:
[claim] → [numerical evidence] → [interpretation]
"DT awareness is useless" → R²=.005, B=-0.017, p=.879 → "consistently null across six models"
What it never does:
- Paste a whole table wholesale
- List numbers and move on without interpretation
- Report only as "significant / not significant"
**Practice point**: Pick just three numbers from an analysis table and write one paragraph. Rule: (1) one sentence of claim, (2) two or three numbers, (3) one sentence of "what this means." Drill this three-part structure.
---
## Trait 5: The narrative of convergence
The most powerful device in this piece is showing that **"different methodologies arrive at the same conclusion."**
The structure of Section 3:
1st (T4): repeated participation → DT awareness d=.55 (sig.), Q3_diff n.s.
2nd (T10): DID → training B=.636 (p<.001), Q3_diff n.s.
3rd (RSM): smart systems R²=.517, Q3_diff R²=.136.
→ "Five methodologies converge on the same conclusion."
A single methodology's result can be doubted. When three point the same way, persuasion emerges. When five converge, it becomes hard to refute. By repeatedly showing "convergence," the piece narratively proves the robustness of its conclusion.
**Practice point**: Pick one of your core findings. Gather three or more different analysis results that support it. Arrange them not chronologically but in the order "in methodology A → in methodology B → in methodology C → convergence."
---
## Trait 6: Theory is written as "connection," not "explanation"
The way Chapter 3 handles theory:
[theory name (author, year)] → [original meaning] → [reinterpretation in this data] → [supporting analysis number]
Example:
> Edwards & Cable's (2009) P-E Fit is originally a theory at the **person-organization** level. This study extends it to the **firm-infrastructure** level. Person = DT awareness, Environment = smart systems, Fit = the interaction term.
It doesn't "explain" theory in a textbook way. It **connects** the data finding to the theory. Not "this theory says so, and our data shows exactly that," but a reinterpretation: "**twist the theory this way** and our data is explained."
**Practice point**: Pick one theory and one analysis result, and write these three sentences. (1) "Theory X originally explains ___." (2) "In this data, that appears as ___." (3) "This reinterprets/extends theory X from a ___ perspective."
---
## Trait 7: The end is not an "answer" but the "next question"
> "This is not an end, but the start of the next question: 'How do we create that condition?'"
The final sentence is a question mark, not a period. The conclusion doesn't tidy everything up. It honestly reveals the remaining question (Chapter 4) and lets that question become "the motivation for the next study."
This is **strategy, not humility**. Write "we figured everything out" and the reader shrugs, "oh yeah?" and closes it. Write "we got this far, but that over there is still open" and the reader follows: "what's over there?"
**Practice point**: After writing the conclusion, delete the last paragraph. Instead, write three "questions this analysis could not answer." Bring the most provocative one up as the final sentence.
---
# Part 3. Sentence-level techniques
## Technique 1: Short sentence → long sentence → short sentence
> "The answer was baffling. Even high DT awareness doesn't mean high training effect. More prepared firms don't learn better. There are many roads to success."
All four sentences are short. But right after:
> "What looked like 'no relationship' in the first round was refined to 'none linearly' in the second, and finally resolved to 'yes, nonlinearly (synergy)' in the third."
One long sentence compresses three stages. There's **rhythm**. Short-short-short-long. That cadence won't let the reader go.
**Practice**: Find a stretch in your writing where sentence lengths are all similar. Cut three consecutive sentences short, then fuse the next one long.
## Technique 2: Bold only at "the reversal point"
Look at where bold is used across the piece:
- useless **on its own**
- when high **together**
- **fits exactly**
- **the limit of measurement**
All the bold falls on "the word that overturns the reader's expectation." Emphasis is used not on "what's important" but on "what's the reversal."
**Practice**: Remove all bold from your writing. Reread, and reapply it only where "the reader should be surprised here." Cap it at five.
## Technique 3: Quotation marks for "what the data is saying"
> We **nearly** concluded that "DT awareness is a useless variable."
The quotation marks aren't a real quote; they mark **"what we thought at the time."** A device to show the gap between the past interpretation and the present one.
**Practice**: Find a moment in the analysis where "at first I thought X, but later it changed to Y." Put the first thought in quotation marks, then write the changed interpretation after it.
---
# Part 4. Five practice tasks
## Task 1: One-sentence summary
Write your research in this form:
> The [phenomenon] of [subject] is not "[common-sense expectation]," but "[actual finding]."
Original example:
> The effect of SME digital-transformation training is not "there or not there," but appears "when awareness and infrastructure meet."
## Task 2: Three puzzle pieces
Pick one core finding. Gather three pieces of evidence that support it from different analyses/methodologies. Write in this form:
[methodology A (analysis no.)]: [result, one line].
[methodology B (analysis no.)]: [result, one line].
[methodology C (analysis no.)]: [result, one line].
N methodologies converge on the same conclusion: [conclusion].
## Task 3: Build a metaphor
Pick your hardest statistical result. Make three metaphors. For each, draw an element-by-element 1:1 correspondence table. Pick the most exact one.
## Task 4: Three sentences of theory connection
(1) [Theory name] originally explains [original meaning] in [original context].
(2) In this data, that appears as the [pattern] of [variable A] × [variable B].
(3) This reinterprets [theory name] from a [new perspective].
## Task 5: End on a question
Write your conclusion. Then write three "questions this analysis could not answer." End the piece with the most provocative one. The last sentence must be a question mark, or a call to the next action.
---
# Part 5. Why this way of writing works
Academic writing usually demands **detachment**. Passive voice, third person, no emotion.
This piece goes the opposite way. **"We asked. The answer was baffling. It was obvious."** The researcher's emotions and thought process are laid bare. And yet the data and figures are accurate to the level of an academic paper.
Why the combination works: **credibility is secured by the numbers, and the pleasure of reading is secured by the narrative.** Narrative without numbers is fiction; numbers without narrative is a report. Fuse the two and you get "research you want to read."
This also touches the "research storytelling" current drawing attention in academia lately. Nature and Science's News & Views, and Harvard Business Review's research summaries, are all this format. They deliver, at once, the accuracy of data to the expert and the pull of story to the non-expert.
---
# Summary: checklist
After writing, check the following:
- [ ] Is the spine of the piece built on "the deepening of the question"?
- [ ] Does each section have the reversal of "expectation → betrayal → reinterpretation"?
- [ ] Is there a 1:1 metaphor for the core concept?
- [ ] Are numbers inside the three-part "claim–evidence–interpretation" structure?
- [ ] Does it show the convergence of different methodologies?
- [ ] Is theory written as "connection/reinterpretation," not "explanation"?
- [ ] Is the last sentence a "next question," not an "answer"?
- [ ] Is there variation in sentence rhythm? (short-short-short-long)
- [ ] Is bold used only at "reversal points"?
- [ ] When read, does it feel like "living through" the research process together?
It was good… and it was sad…
That my own thing — learning from my own writing to reflect it into another of my things — creates a cognitive dissonance, as if what’s mine is no longer mine… and yet the fact that this is fairly effective comes down, right now, to two reasons:
First, I can pull the good points out of my own writing, formalize them, and I can learn from them. Through that I can establish a style.
Second, in many cases I can easily produce writing in these styles using stiff pieces as the source (I’m an otaku with my own philosophy, so I’m not satisfied unless I make the original source myself — and seeing something an AI just churned out end to end gives me a slight twinge of disgust…).
Since this isn’t research, I didn’t open it up for ethics’ sake, but I hope it might, just maybe, cut down on misunderstandings down the road.