「他是个全栈工程师，前后端都能写，也懂点产品和设计，什么都能聊。」听的人礼貌点头，然后什么都不会发生。因为没有人心里的问题长成「我要找个什么都能聊的人」，他们心里的是「我这个下单流程太难用了」。
「他专门给咖啡店改点单小程序，做过十来家，最典型的是把下单从七步压到三步。」——能力没变，可提取性变了。谁正被点单流程折磨，这句就是钥匙；没这个问题的人本来也不会成为你的客户。
「我帮 [窄到能想象出一张脸的谁]，解决 [他自己会用原话说出来的那个问题]，做法是 [一个别人复制起来有成本的方式]。」
自检：念给一个圈外朋友听，他能原样复述就过关；他要问「所以你具体做什么」，就还没窄够。
反面清单：赋能、连接、一站式、多年经验、专业团队——共同点是，删掉它们句子的信息量不变。
EN Nobody searches for you. What actually happens is that a problem shows up and someone asks themselves, "who do I know for this?" A personal brand is nothing more than the probability that your name surfaces in that three-second window. That reframes the work entirely: the goal is not reach, it's retrievability. Memory is cue-dependent — what gets recalled is whatever was bound tightly to the cue that's now present — and activation spreading from a concept is divided among everything attached to it, so the more things your name is linked to, the weaker every single link becomes. This is why "I do a bit of everything" is not versatility in the market's eyes; it's invisibility. The move is to weld yourself to one concrete problem, stated in the words the other person would actually use, and to accept that you're deliberately becoming unfindable for everything else. Narrowness feels like giving up opportunity. In practice it's the only thing that makes any opportunity reach you at all — because the alternative isn't six chances, it's six links too thin to carry anything.
机制是记忆的取出方式，不是别人的注意力多少。人脑不按名字检索，是按线索检索：当下的线索（一个具体问题）和当初存进去时的线索重合越多，越容易被捞出来。而一个概念挂的东西越多，激活就在这些连接之间摊得越散——你多标注一个身份，其余每个身份被提取的强度就掉一点。所以「窄」不是谦虚，是把有限的激活集中到一条线上。
边界：窄定位是还没被想起的人的杠杆；一旦你在某个位置已被牢牢想起，窄就从杠杆变成天花板，那时该往相邻一格扩，而不是继续钉死。另外，窄的是对外的标签，不是能力边界——私下什么都做没关系，但对外每多讲一件，钩子就钝一分。
"I'm the person you call when your checkout flow is a mess." —— 用「什么时候找我」定义自己。
"That's adjacent to what I do, but it's not what I'd want to be hired for." —— 礼貌地把自己排除在别的格子外。
"If you remember one thing about me, make it this." —— 指定别人该记住哪一条。
「我的设备更好，经验也更多，出片质量不一样。」这句默认接受了「我们在同一列里比」这个前提。而「更好」是买方下单前无法验证的属性——他能验证的只有价格，于是你输在价格上。
「5,000 那种是通用商拍，拍得漂亮。我做的是另一件事：只拍详情页第一张图，因为那一张决定点进来的人走不走。我有三个同类目的前后数据，最近那家的首图点击率从 2.1% 到 3.4%。所以你要判断的不是 12,000 贵不贵，是多出来的那点点击值不值 7,000——这个我们可以一起算。」
EN Claiming you're "better" quietly accepts the buyer's frame: a list of five names to be ranked. The trouble is that the only column an outsider can actually compare before buying is price, so every "better" argument eventually loses to the cheapest row. Claiming you're the only person who does one specific thing removes the list. Now the question isn't who's cheapest, it's whether the thing is worth doing — and that's a question you're qualified to answer with them. The mechanism is evaluability: attributes people can't judge in isolation only start driving decisions when there's something to compare against, which is why the same person is valued completely differently in joint versus separate evaluation. Two guardrails. Differentiation isn't inventing a category nobody wants — if nobody is already spending money on the problem you named, you haven't positioned yourself, you've hidden. And the category has to be one they can hold in their head: recognisable enough to be filed somewhere, narrow enough that you're the obvious occupant. The test isn't whether the label sounds clever. It's whether someone can repeat it and someone else immediately knows what it's for.
机制两层。一、可评价性：有些属性人单独看根本判断不了好坏，只有摆在一起才有意义——所以一旦进入被并排比较的状态，那些难评价的优点（品味、判断力、经验）自动失效，只剩最好比的那一列在起作用。二、类别归属：人按「最像这一类的典型」归类和记忆，你要争的不是旧类别里名次靠前，是成为一个新的窄类别里唯一的典型。两层合起来，解释了同一个人为什么报价能差两倍。
边界：差异化不等于生造一个没人要的类别。检验只有一条——你划出的类别里，现在有没有人正在为它花钱？没有就不是定位，是隐身。另一种常见失败：窄到对方脑子里没格子可放，他不知道把你归到哪儿，于是既不记住也不转述——要落在一个他已有的概念旁边。报价怎么被接住，「太贵了怎么办」那期的拆分与对比直接可用。
"That's a different service, not a cheaper version of mine." —— 拒绝被并排。
"The question isn't whether I'm expensive — it's whether this is worth doing at all." —— 换掉他要回答的问题。
"I'd rather be the obvious choice for a few people than an option for everyone." —— 一句能说出口的定位宣言。
EN Trust built one meeting at a time is linear: each new person costs you the whole process again. Content is the only mechanism that makes it compound — one investment, delivered indefinitely. But the compounding doesn't come from posting often; it comes from hitting the same narrow spot repeatedly. Three mechanisms do the work. Repeated exposure raises liking on its own, even when people don't consciously register the exposure. Repetition also makes things easier to process, and the brain misreads that ease as familiarity and truth — which is exactly why this is a tool with an ethical edge, not a neutral one. And neurally, repeated stimuli produce weaker, cheaper cortical responses over time; the subjective feel of that efficiency is "I know this person". Two honest caveats. Repetition does not make a wrong claim right, and the same effect works on you: a positioning line you've repeated for three years starts to feel true to you too, long after it's expired. Second, the effect isn't unbounded — past a point, more exposure produces tedium rather than warmth. Say the same thing from new angles, not the same sentence again.
机制三层叠加。一、单纯曝光：反复见到同一个东西，好感会自己上升，不需要理由。二、加工流畅性：见得多了处理更省力，大脑会把这份「省力」误读成「熟悉」和「这话大概是真的」。三、重复抑制：同一刺激反复出现时，皮层的神经反应减弱、处理成本真的下降——这是前两层的物理底座。结果是：你反复出现在同一个位置，别人不必读完你的全部作品，也会先觉得「这人应该是懂的」。
伦理边界必须写清楚：重复只能让一句话更容易被接受，不能让错的变成对的。用在你没验证过的主张上，就是拿别人的认知捷径换曝光——这正是说服与操纵的分界，「影响力六原理」那期讲的是同一条界。真相错觉对你自己一样起作用：重复了三年的定位你自己也会越来越信，哪怕它早已过期，所以每年要做一次外部检验，问真实客户而不是问自己的读者。失效场景：复利的是「同一个位置」被重复，重复同一句话只会腻。
"I've been writing about this one thing for two years." —— 用时长本身当资历。
"Same idea, different angle — here's what changed my mind since." —— 重复位置而非重复句子。
"Everything I know about this is public; hiring me is just faster." —— 把公开内容变成信任凭据。
他记得的是「我认识一个挺厉害的工程师」。这句话在饭桌上不会被说出口——说了会被追问「厉害在哪」，他答不上来，显得随便。为了不尴尬，他选择不说。你损失了一个机会，而且永远不会知道。
他记得的是「我认识一个专门给咖啡店改点单流程的，说是能把下单从七步压到三步」。这句话他敢说：具体、有画面、被追问也答得上（「他做过十来家」），说出来还显得他消息灵通。让转述者不丢脸，是你能给他的最大帮助。
见完一个人，临走补一句：「以后碰到有人 [具体问题]，你就说 [一句话定位 + 一个数字]，记这一句就够。」直接把稿子给他，别指望他自己总结。
三条缺一不可：不超过二十个字、含一个能追问的细节、说出来对他有面子。
EN Opportunity rarely arrives through the handful of people you actively maintain — they know roughly what you know. It arrives through people who only occasionally think of you, in a conversation you'll never hear about, and all they can carry is one sentence. That makes the sentence, not the portfolio, the thing worth polishing. The bottleneck is social, not artistic: a vague compliment ("she's really good") won't get said out loud, because saying it invites a follow-up question the speaker can't answer, and looking careless in front of others is a cost they'd rather avoid. A concrete line with a number in it is safe to repeat — it survives the follow-up and makes the speaker look well-connected. So hand people the script instead of hoping they'll compose one: keep it short enough to carry intact, specific enough that it doesn't drift, and flattering enough to the person repeating it that they want to. And when someone does pass your name on, close the loop and tell them what happened. That's not politeness; it's the only signal that tells them the referral was worth making.
机制两条。一、信息的新鲜度来自距离：密友和你活动在同一个圈子，知道的机会和你高度重合；真正的新信息来自那些不常联系、因而处在别的圈子里的人。二、转述有社交成本：替你说话的人在意的不只是帮不帮得上你，还有说这句话会不会让他自己显得没谱。模糊的夸奖会招来追问，答不上就丢脸，所以他宁可不说。你把话写具体，本质上是在替他消除风险——这才是他愿意开口的原因。
边界：弱连接的价值来自「关系还在」，不是「关系很浅」——五年没说过话、也从没帮过对方的人不会替你转述，前置是长期的、不带即时目的的给予（「networking 不油腻」那期专门讲）。也别把递转述句做成群发模板。最后一条常被忽略：有人替你转述之后，一定要回头告诉他结果——闭环一次，他下次才会再开口。
"If it ever comes up, here's the one line to use." —— 主动递稿。
"That intro you made turned into a project — thank you, here's how it went." —— 闭环回报。
"I'd rather be easy to describe than impressive to describe." —— 可转述优先于漂亮。
九十分钟，四件事。做完你手上会有一句能被别人原样搬运的话。