The Roof Rule: Why Some People Are Always Ten Years Ahead
TL;DR: An essay circulating in Chinese commentary this week opens with a claim I can't shake: the difference between people is that some are always ten years early. The mechanism isn't talent or luck. It's timing the overlap between three periods every job, skill, and market passes through — Strangeness, Stability, Decline. The people who transition smoothly start the next thing's strangeness during the current thing's stability. Everyone else tries to fix the roof while it's raining. And in 2026, with AI scheduling the decline of entire skill sets on model-release cadence, most professionals are sitting in a Stability period that feels permanent and isn't.
James here, CEO of Mercury Technology Solutions.
Hong Kong — September 2026
The Essay That's Been Sitting in My Head All Week
It starts with a reader's message. The reader's observation: some things, viewed from a distance, look 命裡注定 — destined. Then the essay walks through the most common financial tragedy of the middle class.
A man reaches middle age. He has capital for the first time — a decade of savings. He has time for the first time — the career is stable, the kids are older. He's a mid-level manager, comfortable. Then the restructure comes. Sidelined. Or laid off.
He cannot start over — taking orders from twenty-five-year-olds at forty-five is a flavor of life most people can't swallow. So he wants one thing: a rapid comeback. Back to roughly where he was, quickly.
可就是這個迅速，要了老命了。
It's the "quickly" that kills him.
Here's the brutal part of the essay. The man has ten-plus years of experience — but every year of it was spent inside a mature machine. He has never once watched a workshop get born. He knows how companies run; he has no idea how they start. So he becomes a boss — even a small boss, in his own industry — and he falls hard, because he's ignorant and in a hurry. Ten years of savings, gone in three. Sometimes worse than gone.
The standard reading of this story is: the layoff ruined him.
Wrong. The layoff was the rain. The ruin was seeded ten sunny years earlier.
The Three Periods
Strip the story down and you get a model worth keeping. Everything you ever do — every job, every skill, every company, every market — passes through three periods:
陌生期 — the Strangeness Period. You know nothing, you're nobody, everything costs more than you expected.
平穩期 — the Stability Period. You're competent, paid, and comfortable. This is where careers live and where they die.
沒落期 — the Decline Period. The market moves, the industry contracts, the model release lands. Whatever made you valuable starts evaporating.
The overlap principle: if you begin the next thing's Strangeness at the *start* of the current thing's Stability, then by the time Decline arrives, the next thing has already stabilized. The transition is 絲滑 — silky. The essay's author: an intern in 2006, investing by 2008, an executive pivot in 2012, an investment career by 2014. His own verdict on the ledger is the sharpest line in the piece: the switching action was meaningless. What mattered was the years of quiet preparation stacked before each switch, inside the previous phase.
Invert the timing and you get the middle-aged catastrophe. Attempt Strangeness during your own Decline and you bring anxiety to a fight that requires patience. The essay's image: you cannot fix the roof while it's raining. The water is coming in everywhere, the losses are visibly compounding, and your mind cannot find calm. You need calm for Strangeness. Rain doesn't sell calm.
The Roof Rule: Strangeness is cheap when it's optional and unaffordable when it's mandatory. Do the next thing's hard part while the current thing still pays.
If you want it as an equation: transition cost is inversely proportional to the stability you have left when you begin. Every quarter you delay the preparation, the same transition gets more expensive — until it costs the savings.
Life is Go, the essay says: 走一步看三步 — play one move, see three ahead. Nobody wins a 19-line board by optimizing the stone in front of them.
Why Nobody Does It
Physics has the answer: water at 99°C is still water. Phase transitions don't announce themselves, and Stability is a phase, not a promise. The Stable period is engineered to feel permanent — salary, title, mortgage, routine. The essay's other killer line describes the trap precisely: some people spend thirty-five to forty discovering they 把一年的工作經驗重複了十遍 — repeated one year of experience ten times. They didn't compound. They looped. The loop only becomes visible when a younger, cheaper version of the loop shows up, or now, when a model does half of it for free.
And it's not just skills. The essay makes the same argument about people: 交情 — the capital of relationships — can only be deposited when you don't need it. Nobody extends a hand to the man who ignored them for a decade and shows up asking. 臨時抱佛腳 — clasping Buddha's feet at the last minute — is not a strategy; it's a prayer with extra steps. Human capital, financial capital, relationship capital: same instrument, same rule. Deposit in sunshine, withdraw in rain.
Yang Wen-li took Iserlohn Fortress — impregnable, fleet-killing, decades of Alliance blood spilled against it — because he'd done the boring preparation nobody valued until it made the impossible look effortless. The fortress fell to a man who had started early. It always does.
Your Decline Is Being Scheduled Right Now
Here's the 2026 translation. The AI transition is not a future event; it's a schedule. Model capabilities compound on release cadence, and each release quietly assigns an expiry date to a class of work — analysis, copy, code, dashboards, workflows. The people treating this as their Stability period — "my job's fine, the tools still need me" — are in a room where the temperature is rising one release at a time.
The audit question is the essay's, updated: are your skills compounding, or is one year of experience being repeated? If your current skill set were posted as a job ad, what fraction of it could this year's models already do? That fraction is your scheduled Decline — and Strangeness in your next domain costs exactly one thing more each quarter you wait.
The Overlap Protocol
Four moves, all of them startable this quarter:
Name your period honestly. Stability or Decline — no third comfortable option. Most people misfile Decline as Stability because the salary hasn't updated yet.
Open a Strangeness account. Ten percent of time, ten percent of capital, into the adjacent domain you'll need in ten years. Small, boring, consistent. The essay's ladder image: be familiar with the third floor, understand the fourth — so that when you enter the building, you start on the fifth.
Deposit 交情 now. Help people you don't need yet. Introduce, review, reply. Relationship capital only counts if it predates the emergency.
Run the loop audit every year. Ten years of compounding or ten loops of one year — the difference is only visible on paper you actually write down.
Stop climbing the ladder you're on. Start building the one you'll need when it ends — because it ends. They all end; the essay's graduation metaphor says it plainly: every stage of life has a parting built in, and your relationship with your current role is no exception.
因果
The essay closes on karma: 因果因果，有些果的確是多年前的因，就已經種下了 — some fruits were seeded by causes planted years ago. The middle-aged collapse was seeded in the comfortable decade. The smooth pivot was seeded in the unpaid evenings. 焉得不敗 — "how could he not fail?" — is the essay's ending. Invert it: prepare in sunshine, and how could you not survive the rain?
It's sunny today. That's not reassurance. That's the deadline.
Mercury Technology Solutions: Accelerate Digitality.