find-your-margin by cdeistopened/content-os
npx skills add https://github.com/cdeistopened/content-os --skill find-your-margin通过评估知识工作者的技能栈与人工智能能力,找出那些人类判断力与人工智能吞吐量相结合、能创造出他人无法触及的盈余的狭窄领域。基于边际主义观点:智力如同资本,会流向回报最高的领域——以及一个对立观点:个人制胜之道不在于分配效率的竞争,而在于找到符合人类尺度的丰厚边际。
边际智力陷阱: 在竞争性的分配经济中,盈余趋于零。每个人都使用相同的模型、相同的提示、相同的工作流程。再多分配一小时给人工智能的边际回报趋近于 0 美元。
逃脱之道: 你的边际优势存在于这样一个地方:你特定的知识、品味和背景使得人工智能对你而言的价值,远大于任何其他使用相同模型的人。Naval 说:"成为你所做领域的世界最佳。不断重新定义你所做的事情,直到这成为现实。" Adams 说:"平庸技能的组合可以让你变得出奇地有价值。"
数学表达: 丰厚边际 =(你的人工智能增强输出的价值)减去(普通操作员使用相同工具能产出的价值)。如果这个差距很小,你就陷入了商品化陷阱。如果差距很大,你就找到了你的边际优势。
五个步骤。每一步都产生一个具体的成果。以互动方式运行——展示步骤,提出问题,等待答案,综合后再继续。
列出用户所知道、所做或经历过的一切可能相关的事情。范围要广。
逐一提出以下问题,等待完整回答:
输出: 一份包含 8-15 个知识领域/技能/经历的原始列表。先不要筛选。
将步骤 1 中的每项技能归入以下三个类别之一:
| 类别 |
|---|
广告位招租
在这里展示您的产品或服务
触达数万 AI 开发者,精准高效
| 定义 |
|---|
| 示例 |
|---|
| 商品化 | 人工智能在这方面做得和你一样好甚至更好。任何拥有提示词的人都能匹配你的输出。 | 总结文章,基础文案写作,数据格式化,代码样板 |
| 增强 | 人工智能让你快 5-10 倍/好 5-10 倍,但你的输入才是使输出有价值的关键。人工智能需要你的判断力、品味或背景才能产出好东西。 | 用你的声音进行编辑,针对你特定领域的战略分析,创意指导 |
| 未触及 | 人工智能完全无法做到这一点。它需要物理存在、人际关系、具身经验或无法通过文本传递的隐性知识。 | 客户关系,物理工艺,本地/社区知识,察言观色,数十年培养的品味 |
对于每项技能,问:"如果我给一个没有领域经验的聪明人同样的 AI 工具,他们能在一周内在这个领域匹配你的输出吗?" 如果能 → 商品化。如果需要几个月 → 增强。如果需要几年或永远不能 → 未触及。
输出: 所有技能的三列地图。
丰厚边际存在于增强技能和未触及技能的交集处。具体来说:
浏览用户的"增强"列,针对每一项提问:
寻找以下组合:
输出: 2-4 个"边际优势赛道"——未触及输入 + 增强输出的具体组合,用户在其中拥有可防御的、高盈余的位置。
现在设计用户应该如何实际分配他们的时间。三个区域:
区域 1 — 完全委派(低注意力,高自主性) 来自"商品化"列的任务。建立系统、模板或工作流程,让人工智能在最低限度的监督下运行。目标:在此处花费接近零的注意力。释放时间。
提问:"你还在手动做哪些人工智能已经做得足够好的事情?你会把哪些工作委派给一个称职的实习生并且从不检查?"
区域 2 — 轻度验证(中等注意力,中等自主性) 来自"增强"列但与你的核心边际优势赛道不相交的任务。人工智能做繁重的工作;你进行质量检查。足够好就是好。
提问:"你在哪些地方过度打磨已经达到 80% 的 AI 输出?你的完美主义在哪些地方耗费了你的时间却没有为你赢得边际优势?"
区域 3 — 全心投入(高注意力,低自主性) 来自步骤 3 的你的边际优势赛道。这是你的注意力能获得最多回报的地方。人工智能处理吞吐量;你提供判断力、品味、背景和创造力,使输出不可替代。
提问:"如果你每天只能在人工智能增强的工作上花费 2 小时,哪些任务能为你带来最多的回报——金钱、声誉还是满足感?"
输出: 一个三区域组合,包含每个区域的具体任务和预估的时间分配。
将所有内容综合成一页文档:
寻找你的边际优势 — 结果
===========================
姓名: [用户]
日期: [今天]
你的技能栈
核心技能: [来自未触及列的 2-3 项]
AI 增强技能: [来自增强列的 2-3 项]
商品化技能(委派这些): [列表]
你的边际优势赛道
赛道 1: [未触及输入] + [增强输出] = [创造的具体价值]
赛道 2: [未触及输入] + [增强输出] = [创造的具体价值]
你的注意力组合
完全委派(约 X 小时/周): [任务]
轻度验证(约 X 小时/周): [任务]
全心投入(约 X 小时/周): [任务 — 你的边际优势赛道]
丰厚边际测试
"如果一个拥有相同 AI 工具的聪明通才试图在 [赛道 1/2] 与我竞争,
他们会失败,因为:_______________"
NAVAL 检查
"我正在成为 [根据你的边际优势赛道,而非你的职位头衔,重新定义你所做的事情] 的世界最佳。"
第一步行动
本周,将 [具体任务] 从区域 3 转移到区域 1 或 2,
并将节省的时间重新投入到 [具体的边际优势赛道活动] 中。
输出: 完成的边际优势地图。
productize-yourself — Naval 关于构建特定知识 + 杠杆的完整框架。当用户从零开始并需要找到他们的利基市场时使用。find-your-margin 假设他们已经拥有技能,并想知道如何利用人工智能部署它们。content-product-fit — 分析内容是否自然地导向产品。在找到边际优势赛道后使用,以验证商业模式。每周安装数
1
代码仓库
GitHub 星标数
2
首次出现
今天
安全审计
安装于
zencoder1
amp1
cline1
openclaw1
opencode1
cursor1
Audit a knowledge worker's skill stack against AI capabilities to find the narrow lanes where their human judgment + AI throughput creates surplus nobody else can touch. Based on the marginalist insight that intelligence, like capital, flows to the area of highest return — and the counterpoint that individuals win not by competing on allocation efficiency, but by finding fat margins at human scale.
The Marginal Intelligence Trap: In a competitive allocation economy, surplus converges toward zero. Everyone uses the same models, the same prompts, the same workflows. The marginal return on one more hour of AI allocation approaches $0.
The Escape: Your margin lives where YOUR specific knowledge, taste, and context make AI dramatically more valuable than it would be for anyone else running the same model. Naval: "Become the best in the world at what you do. Keep redefining what you do until this is true." Adams: "A combination of mediocre skills can make you surprisingly valuable."
The Math: Fat margin = (value of your AI-augmented output) minus (what a generic operator could produce with the same tools). If that gap is small, you're in a commodity trap. If it's large, you've found your margin.
Five steps. Each produces a concrete artifact. Run interactively — present the step, ask questions, wait for answers, synthesize before moving on.
Surface everything the user knows, does, or has experienced that might be relevant. Cast wide.
Ask these questions one at a time, wait for full answers:
Output: A raw list of 8-15 knowledge areas / skills / experiences. Don't filter yet.
Take each skill from Step 1 and sort it into one of three buckets:
| Bucket | Definition | Example |
|---|---|---|
| Commoditized | AI does this as well as you or better. Anyone with a prompt can match your output. | Summarizing articles, basic copywriting, data formatting, code boilerplate |
| Amplified | AI makes you 5-10x faster/better, but YOUR input is what makes the output valuable. The AI needs your judgment, taste, or context to produce something good. | Editing with voice, strategic analysis of your specific domain, creative direction |
| Untouched | AI can't do this at all. It requires physical presence, relationships, embodied experience, or tacit knowledge that can't be transmitted through text. | Client relationships, physical craft, local/community knowledge, reading a room, taste developed over decades |
For each skill, ask: "If I gave a smart person with no domain experience the same AI tools, could they match your output in this area within a week?" If yes → Commoditized. If they'd need months → Amplified. If they'd need years or could never → Untouched.
Output: A three-column map of all skills.
The fat margin lives at the INTERSECTION of Amplified and Untouched skills. Specifically:
Walk through the user's Amplified column and ask for each one:
Look for combinations where:
Output: 2-4 "margin lanes" — specific combinations of Untouched input + Amplified output where the user has a defensible, high-surplus position.
Now design how the user should actually spend their time. Three zones:
Zone 1 — Delegate Fully (Low Attention, High Autonomy) Tasks from the Commoditized column. Set up systems, templates, or workflows where AI runs with minimal supervision. Goal: spend near-zero attention here. Free up hours.
Ask: "What are you still doing manually that AI already does well enough? What would you delegate to a competent intern and never check?"
Zone 2 — Verify Lightly (Medium Attention, Medium Autonomy) Tasks from the Amplified column that DON'T intersect with your core margin lanes. AI does the heavy lifting; you do a quality pass. Good enough is good enough.
Ask: "Where are you over-polishing AI output that's already 80% there? Where is your perfectionism costing you time without earning you margin?"
Zone 3 — Full Presence (High Attention, Low Autonomy) Your margin lanes from Step 3. This is where your attention earns the most. AI handles throughput; you provide the judgment, taste, context, and creativity that makes the output irreplaceable.
Ask: "If you could only spend 2 hours a day on AI-augmented work, which tasks would earn you the most — in money, reputation, or satisfaction?"
Output: A three-zone portfolio with specific tasks in each zone and estimated time allocation.
Synthesize everything into a one-page document:
FIND YOUR MARGIN — RESULTS
===========================
Name: [User]
Date: [Today]
YOUR SKILL STACK
Core skills: [2-3 from Untouched column]
AI-amplified skills: [2-3 from Amplified column]
Commoditized (delegate these): [list]
YOUR MARGIN LANES
Lane 1: [Untouched input] + [Amplified output] = [specific value created]
Lane 2: [Untouched input] + [Amplified output] = [specific value created]
YOUR ATTENTION PORTFOLIO
Delegate Fully (~X hrs/wk): [tasks]
Verify Lightly (~X hrs/wk): [tasks]
Full Presence (~X hrs/wk): [tasks — your margin lanes]
THE FAT MARGIN TEST
"If a smart generalist with the same AI tools tried to compete
with me in [Lane 1/2], they would fail because: _______________"
NAVAL CHECK
"I am becoming the best in the world at [redefine what you do
based on your margin lanes, not your job title]."
FIRST MOVE
This week, shift [specific task] from Zone 3 to Zone 1 or 2,
and reinvest that time into [specific margin lane activity].
Output: The completed Margin Map.
productize-yourself — Naval's full framework for building specific knowledge + leverage. Use when the user is starting from scratch and needs to find their niche. find-your-margin assumes they already have skills and wants to know how to deploy them with AI.content-product-fit — Analyze whether content leads naturally to a product. Use after finding margin lanes to validate the business model.Weekly Installs
1
Repository
GitHub Stars
2
First Seen
Today
Security Audits
Gen Agent Trust HubPassSocketFailSnykPass
Installed on
zencoder1
amp1
cline1
openclaw1
opencode1
cursor1
头脑风暴技能:AI协作设计流程,将创意转化为完整规范与实施计划
83,800 周安装