There’s a 2,400-year-old test for whether your life is going well, and it goes like this.
Zhuangzi was fishing on the Pu River when two officials showed up to recruit him as prime minister of the state of Chu. He didn’t even turn around. He told them about a sacred turtle the king kept wrapped in cloth and boxed in the ancestral temple. Dead three thousand years. Its bones still honored. Would the turtle, Zhuangzi asked, prefer to be revered as a relic, or alive and dragging its tail through the mud?
The officials said: alive. Of course alive.
Zhuangzi: Then leave. I shall keep dragging my tail through the mud.
A sharp question, and a bit smug. It’s easy to nod along to, and harder to actually pick the mud when someone’s offering you the temple.
I’ve been returning to it because I just listened to a Mandarin podcast called Fire一期一会. This episode’s guest, 大狮姐, is a Tsinghua PhD who went to McKinsey and then walked away from the whole thing to FIRE. Her reframing of the FIRE movement is the cleanest one I’ve heard:
“FIRE is failure studies dressed up as success studies.”
It sounds like a slogan, but it isn’t. There’s an actual academic discipline called Failureology (失败学), systematized around 2000 by Yotaro Hatamura, an engineering professor at the University of Tokyo. He later chaired the Japanese government’s Fukushima nuclear-disaster investigation. His thesis started in engineering but generalizes:
Failures have shapes. They aren’t random. They cluster. The same patterns recur across companies, decades, and domains. Hatamura’s main artifact is a Failure Knowledge Database: anonymized cases that engineers consult before making decisions, treated as just as valuable as a success-case library.
Read FIRE as failure studies, and the math of “save 25× expenses, retire on the 4% rule” is the success-studies packaging. Underneath, what’s usually true is that someone hit a specific shape of failure on a specific kind of success track — McKinsey burnout, post-IPO emptiness, layoff round seven — and reached for FIRE as a dignified exit. The grief is the cause, and the math is its cover story.
The framing also dignifies the grief. Your pain isn’t a verdict on your ability. You collided with a recognizable, classifiable failure mode of the path you were on. Once you can name the shape, you can stop blaming yourself for it.
But the sharpest part of the episode is a distinction 大狮姐 makes almost in passing. I think it’s the most actionable idea in the whole hour.
There are two kinds of FIRE people.
Pain-escape type. Hurt by something concrete in the workplace — meaninglessness, politics, a bad manager, the grind. FIRE is the exit door. They get a 6-to-18-month honeymoon: sleep, exercise, travel, no Slack. Then the pain recedes and a vacuum rises. Without a Second Thing they actually want, they drift into “why did I leave the salary?” regret.
Time-scarcity type. Already had a backlog of things they wanted to do — write, build, parent, study, run a thing — and work was the obstacle. FIRE doesn’t fix anything; it just removes the obstacle. No vacuum, because the to-do list runs three lifetimes long.
Compress that and you get one diagnostic question:
Do you have a backlog, or only a grievance?
Grievances are real and valid. But a grievance is not a destination. If you exit on a grievance with no backlog, FIRE will trade your workplace pain for a more diffuse pain that’s harder to name — the pain of choosing what to do with a Tuesday.
This is useful because most “should I quit” conversations stop at the grievance. They never reach the backlog. People assume a backlog will spontaneously generate once the grievance ends. Sometimes it does. Often, in my limited observation, it doesn’t, and the people who FIRE without one spend the next year writing very thoughtful posts about how they’re “still figuring out what they want.”
In 2026, the backlog question gets harder for a different reason.
大狮姐 names it too. The thing that startled her about the AI era isn’t unemployment. It’s that all of human knowledge has been compressed into a single crystal. Anyone, in five minutes, can summon mid-tier expert competence in any domain via an LLM. Knowledge as a moat is just gone.
If your backlog used to be “learn X” — write a book, switch into ML, finally master Chinese literature, get good at financial modeling — the payoff structure has changed. Knowing X used to be load-bearing. Now it’s a commodity. The new moat is having lived through something with a particular taste and a particular emotional history. Experience, judgment, taste, the felt sense of having actually done the thing. AI cannot mass-produce that yet.
This slips a heavier definition into FIRE. It moves the goal from “buy your way out of bad work” to buy time to compound experience. Which is a much more demanding job description. Compounding experience needs intentional friction — real relationships, projects with real consequences, communities, parenting, art that risks being bad. It isn’t the same thing as having freedom.
The other metaphor in the episode that’s been rattling around my head is the bamboo raft.
Picture human society as the Titanic, steering toward the AI iceberg. Three responses are available.
One: sit in the cabin and be anxious.
Two: climb from third class to first class — work harder, dodge the next layoff round, hope the iceberg hits someone else’s deck first.
Three: step off the rails and build a small bamboo raft alongside the ship.
The raft is not a lifeboat. It can’t save everyone. It may not save you. What it does is change your relationship with the ship. You no longer depend on it, but you can still trade with it.
That last clause is the part I want to underline. People talk about FIRE as a binary — working or not working. The raft framing turns it into a question of relationship instead. “Do I show up to the office today?” becomes am I a passenger on this thing, or a neighbor of it?
A passenger has no leverage. A neighbor has every option a passenger has, plus the option to walk away. That extra option, even unused, changes how you negotiate.
The turtle, I think, is the same point in older words.
The temple sacrifice has standing: it is honored, photographed, written about. It is also dead. The mud-tail turtle has nothing — no title, no shrine, no ceremonial cloth. It only has the one thing the relic doesn’t: it gets to keep being a turtle.
The trap is that we measure ourselves on the temple’s KPIs. Salary band, title, prestige of the company logo, the LinkedIn headline that doesn’t need explaining. These are all metrics about you, not of you. A great temple metric can sit on a corpse.
So here’s the test I want to actually hold myself to, year over year:
Am I more turtle this year than last year, or more relic?
More turtle: I made things. Took some risks that cost me. Had relationships that demanded something of me. Did work I would still do if no one was keeping score.
More relic: my titles got fancier. My photo went on a few more lists. My calendar filled up with things I’d be embarrassed to describe to a stranger from outside my industry.
It’s possible — common, even — to be progressing in relic-units while regressing in turtle-units. You can be on the success template and accumulating sacrifice cloth at the same time. McKinsey + first class + Monday morning crying is exactly that scene: high relic-velocity, near-zero turtle-velocity.
I’m not FIREd. I’m not planning to FIRE. I have an actual job that I do most days and largely enjoy. But I’ve found these two questions to be the most useful self-checks I have:
Turtle or relic — annually. Backlog or grievance — weekly.
They’re cutting at the same thing. Alive is not the same as honored, and the difference compounds. A small grievance compounds into a big one. A small backlog compounds into a life. The people I respect most are quietly running a backlog far in excess of anything their job requires. The people I quietly worry about are maintaining a beautifully decorated relic.
Twenty-four centuries on, Zhuangzi still cuts. The temple is welcoming. The cloth is flattering. The mud is honest.
Drag your tail in the mud.
庄子在濮水边钓鱼，楚王派了两位大夫来，请他出山管国家的事。
庄子拿着鱼竿，头也没回，反问他们：听说楚国有只神龟，已经死了三千年。楚王把它的骨头用布包好，装进盒子，供在庙里。你们说，它是愿意死了以后被人供着，还是愿意活着，在泥里拖着尾巴爬？
两位大夫说：当然是活着，在泥里爬。
庄子说：那你们回去吧，我还是在泥里拖着尾巴爬。
（《庄子·秋水》，白话转述）
《庄子·秋水》原文（简体）
庄子钓于濮水。楚王使大夫二人往先焉，曰：“愿以境内累矣！”
庄子持竿不顾，曰：“吾闻楚有神龟，死已三千岁矣，王巾笥而藏之庙堂之上。此龟者，宁其死为留骨而贵乎？宁其生而曳尾于涂中乎？”
二大夫曰：“宁生而曳尾涂中。”庄子曰：“往矣！吾将曳尾于涂中。”
挺锋利的一个问题。但也说得有点轻巧。嘴上点头容易，真有人把“庙堂”递到你面前的时候，能不能选泥潭，是另外一回事。
这就好比金庸笔下的扫地僧。少林寺藏经阁里那位老人，一待四十多年，玄字辈的高僧没一个认得他，扫地、烧火、干粗活。可真到萧远山、慕容博内伤发作的时候，整个少林七十二绝技加在一起也救不了——只有他能。他不在庙堂上，所以他还是他。
换个角度：少林寺方丈玄慈，香火不断，地位拉满。然后呢？然后当年犯的戒被翻出来，当众挨了二百棍，自绝经脉。庙堂的位置，从来不保人活着，它只保人被供。
我最近又想起这道题，是因为听了一期叫《Fire一期一会》的中文播客。这一期请的嘉宾是大狮姐——清华博士，去了麦肯锡，干到能主动 FIRE 退休了。她对 FIRE 的重新定义，是我听过最干脆的一个：
“FIRE 是包装成成功学的失败学。”
听着像句口号，其实不是。“失败学”（しっぱいがく）真的是一门正经学科。2000 年前后，东京大学的机械工程教授畑村洋太郎把它系统化了。他后来当过日本福岛核事故调查委员会主席，把这门学问从工程领域推向了制度层面。
他的核心命题是：失败是有“形状”的。 失败并不随机，能分类，也能预判。同样的失败模式，会跨公司、跨年代、跨行业反复出现。所以畑村做了一个失败知识数据库，把案例匿名整理好入库，工程师做决策前先翻一遍。在他看来，这个库和成功案例库一样重要。
把这个框架套到 FIRE 上，那句话就立住了。“存够 25 倍年支出、用 4% 法则退休”——这一层成功学的语法是壳。 壳底下，大多数 FIRE 的人是先有了“我跑不动了 / 我不想跑了”的失败感，再去找 FIRE 这套数学，给自己一个体面的退场理由。
这个框架还让“失败”这件事变得体面了。它说：你痛，不是因为你能力不行。你撞上的是一类成功路径的系统性失败模式。它有名字，有形状，别人也踩过。一旦你能把它叫出来，就可以不再为它内疚。
但真正让我觉得有用的，是大狮姐顺嘴提的一个分类。她说，FIRE 的人其实分两种。
第一种，逃避痛苦型。 被职场里某个具体的东西伤到了——无意义、政治、PUA、内卷、一个烂老板。FIRE 是逃生票。退休后大概率有 6 到 18 个月的蜜月期：睡觉、运动、旅行、没有 Slack。然后痛感消退，真空升上来。如果手里没有一件特别想做的事，他们会进入一种很难命名的反悔螺旋：我为什么要放弃那份高薪？
第二种，时间不够用型。 本来就有一堆想做的事（写作、研究、创业、陪娃、读书），工作只是占着时间。FIRE 不解决任何问题，它只是把障碍挪开。没有真空期，因为想做的事排到了三辈子以后。
把这个分类抽象成一个问题，就是：
你有 backlog，还是只有怨气？
怨气是真实的，也是合理的。但怨气不是目的地。如果只带着怨气走，没有 backlog，FIRE 会把你的“职场之痛”换成更弥散、更难命名的另一种痛——周二上午要决定干什么的痛。
这个诊断之所以有用，是因为身边大多数“我要不要辞职 / 我要不要 FIRE”的对话，全停在怨气这一层，从来没走到 backlog。大家默认怨气消失之后，backlog 会自动长出来。有时候确实会。但更多时候，没有 backlog 就 FIRE 的人，下一年都在很认真地写“我还在寻找自己想做什么”。
到 2026 年，backlog 这件事变得更难了。原因是 AI。
大狮姐这段也讲得很狠。AI 时代最让她震撼的不是失业，是人类过去所有的知识被压成了一颗“晶体”。任何人，五分钟，通过一个 LLM 就能调出任意领域中等专家的水平。“知识获取”这条护城河，整体失效了。
如果你的 backlog 原来是“学 X”（为了写一本书、转行做 ML、读完中国古典文学、把财务模型搭得漂亮），那这个 backlog 的回报结构已经变了。“懂 X”以前是承重的，现在是大路货。新的护城河，是带着自己的品味和情感，亲身经历过 X。是体验、判断、品味、亲手做过的那种身体记忆。这些 AI 暂时还量产不了。
这等于悄悄给 FIRE 换了个定义：目标从“花钱买掉烂工作”，变成了花时间复利体验。这份工作可难多了。复利体验需要主动的摩擦：真实的关系、有后果的项目、社区、养育、可能失败的创作。它和“获得自由”不是同一件事。
这一期里另一个一直在我脑子里转的比喻，是小竹筏。
画面是这样的：人类社会这艘“泰坦尼克”，正撞向 AI 这块冰山。我们有三种姿态可选：
一：坐在船舱里焦虑。
二：从三等舱往头等舱爬——加班、晋升，希望冰山先撞别人那层甲板。
三：跳出轨道，在泰坦尼克边上，做一个小竹筏。
竹筏不是逃生艇。它救不了所有人，未必能救你自己。它的价值是改变你和那艘船的关系：你不再依赖它，但还能和它共存。
这后半句是我想划重点的地方。大家讲 FIRE 时习惯把它讲成二元——上班 / 不上班。竹筏这个比喻把它变成了一个关系问题。该问的，也就从“我今天要不要去办公室”，换成了“我是这艘船的乘客，还是它的邻居”。
乘客没有筹码。邻居有乘客的所有选项，加上一个“随时走开”的选项。这个额外的选项就算从不行使，也已经改变了谈判结构。
绕回到乌龟。
庙堂上的乌龟是有“地位”的：被尊敬、被记录、被供奉。它也是死的。泥潭里的乌龟什么都没有——没有 Title、没有神龛、没有盖身的布巾。它只有一样东西：它还在做一只乌龟。
陷阱在于：我们用庙堂的 KPI 衡量自己。薪资档位、Title、公司名声、不需要解释的 LinkedIn headline——这些都是关于你的指标，不是你的指标。一个非常漂亮的庙堂指标，可以挂在一具尸体上。
所以我想给自己留一道年度自检题：
这一年，我过得更自在了，还是只是履历更好看了？
过得更自在：做了点自己想做的东西，承担了一些有代价的风险，有几段需要付出的关系，做了即使没人计分也会做的事。
履历更好看：Title 更花了，照片多上了几张榜单，日历里塞满了我没法对一个圈外人解释清楚的事。
这两件事未必同步。职位升了，履历更漂亮了，却可能越来越没空做自己想做的事。麦肯锡、头等舱、周一早上崩溃大哭，就是这种落差：别人觉得你过得很好，自己却越来越难受。
我没 FIRE，也没打算 FIRE。我有一份正经工作，大部分时候还挺喜欢的。但这两道题在我手里是真好用：
过得更自在了，还是只是履历更好看了——每年问一次。 Backlog 还是怨气——周度自检。
两道题切的是同一块东西：活着不等于被尊敬，二者的差距会复利。 一个小怨气会复利成一个大怨气。一个小 backlog 会复利成一种生活。我最佩服的那些人，几乎都在静悄悄地经营一个远超工作要求的 backlog；我隐隐担心的那些人，都在精心维持一个体面的样子。
两千四百年过去了，庄子还是那么锋利。庙堂在招手。布巾很体面。泥巴很诚实。
就这样吧。去泥里拖尾巴吧。
Comments留言