Brooks's Law — "Adding manpower to a late software project makes it later." The "man-month" is a myth: people and time aren't interchangeable. Three mechanisms: (1) ramp-up — newcomers are unproductive at first and drain veterans' time; (2) communication overhead grows as n(n-1)/2 — quadratic coordination cost against linear added work; (3) non-partitionability — "nine women can't make a baby in a month." It's isomorphic to Amdahl's Law: a serial bottleneck caps the speedup from extra cores. The real fix isn't more people but a smaller communication surface — modularization, clean interfaces, reduced scope. In the AI era, agent swarms re-incur the n² cost whenever they must coordinate; the super-individual's edge is collapsing the communication graph to a single node.
AI Prompts
中文提示词
我们的项目 [描述] 已经延期，正在考虑加 [N] 个人/agent。请用布鲁克斯定律帮我压力测试：
① 估算新增的沟通线条数（n(n-1)/2）和爬坡期成本，判断净产能是正是负；
② 找出任务里"不可分割"、加人也救不了的串行瓶颈（Amdahl 段）；
③ 给出 1 个"降低沟通面"而非"加人"的替代方案（如何把工作切成互不依赖的独立块）。
English Prompt
Our project [describe] is late and we're considering adding [N] people/agents. Stress-test with Brooks's Law:
1. Estimate the new communication channels (n(n-1)/2) and ramp-up cost; judge whether net capacity is positive or negative.
2. Identify the non-partitionable serial bottleneck (the Amdahl segment) that more people can't fix.
3. Propose one "shrink the communication surface" alternative instead of adding people — how to carve the work into independent, non-communicating chunks.
① 个人成长：资深技术人走向"超级个体"时，会撞上自己的彼得天花板——从一线写代码，到编排一群 AI agent、做架构与取舍，需要的是产品判断和系统设计能力，不是更快地敲代码。识别它的信号：你最强的旧技能，正在新角色里变成边际效用最低的那个。② 育儿同构：孩子在某一阶段表现优异（比如机械记忆强的低年级），别据此就把期待"晋升"到下一阶段——下一阶段要的可能是抽象推理或自我调节，是另一套能力。用上一关的成绩预测下一关的胜任，是彼得原理在家庭里的翻版。
English Summary
The Peter Principle — "In a hierarchy, every employee tends to rise to their level of incompetence." Promotion rewards performance in the current role, but the next role demands a different skill set; so you keep rising until you reach a job you're bad at, and promotion stops. At the limit, every post tends to be held by someone incompetent at it, and the real work is done by those not yet at their ceiling. It's the organizational version of Goodhart's Law: once "current performance" becomes the promotion target, it stops being a reliable signal of fitness for the higher role. Root cause: conflating "doing the work" with "managing the work." Fixes: dual career ladders, promote on next-level demonstrated ability (not current-level results), and trial-in-role before confirming.
AI Prompts
中文提示词
我在考虑 [晋升某人 / 自己接受某个新角色]。当前角色是 [描述]，目标角色是 [描述]。请用彼得原理帮我审视：
① 列出两个角色各自真正需要的核心技能，标出哪些不重叠——不重叠的部分就是风险区；
② 判断"当前出色"在多大程度上其实预测不了"新岗位胜任"；
③ 给出 1 个降低踩坑概率的设计（如试岗、双通道、或换一个评估维度）。
English Prompt
I'm considering [promoting someone / taking on a new role myself]. Current role: [describe]. Target role: [describe]. Apply the Peter Principle:
1. List the core skills each role truly requires; flag which ones don't overlap — the non-overlap is the risk zone.
2. Assess how poorly "excelling now" actually predicts "competence in the new role."
3. Propose one design that lowers the failure odds (trial-in-role, dual ladder, or a different evaluation axis).
官僚铁律 · Pournelle's Iron Law of Bureaucracy
"任何组织里，效忠组织本身的人，终将掌权；而献身于组织使命的人，影响力越来越小。" —— Jerry Pournelle
① 公司里：那个最初为"赋能业务团队"而设的平台/基建组，会悄悄演化成一道以流程自重的关卡——它开始保护的是自己的审批权，而非当初承诺的交付速度。识别信号：流程数量在涨，但它要服务的成果指标没动。② 超级个体视角：单人运作的最大结构性优势，就是官僚密度为零——没有组织派能在你这里夺权。但要警惕一个隐蔽版本：连你自己的工具与系统也会"组织派化"——你开始为维护笔记系统而维护笔记系统、为搭流程而搭流程，而不是为产出。定期问自己："这一步到底在服务成果，还是在服务这套系统本身？"
English Summary
Pournelle's Iron Law of Bureaucracy — In any organization there are two kinds of people: the mission people (devoted to the goal) and the organization people (devoted to the institution itself). Over time the organization people win control, because internal power is decided by internal games, not external goals — and they optimize exactly what confers internal power (budget, headcount, process compliance) while mission people optimize outcomes. It's a pure selection-pressure argument (the org is a fitness landscape favoring institution-loyalty) and an entropy analog: without continuous energy (competition, founder tension, accountability) the system drifts toward self-preservation. Resistance means re-injecting external pressure: stay small, bind to real market feedback, keep founder-level mission intensity. Key diagnostic: when "good for the org" conflicts with "good for the mission," which does it default to?
AI Prompts
中文提示词
我所在/管理的组织是 [描述]。请用官僚铁律做一次"漂移体检"：
① 帮我区分当前哪些行为/部门在为"使命"服务，哪些在为"组织自身"服务；
② 找出 2 个"对组织好但对使命未必好"的现行流程，判断铁律走到了哪一格；
③ 给出 1 个重新注入外部压力、把权力拉回使命派的具体动作。
English Prompt
My organization is [describe]. Run a "drift checkup" using the Iron Law of Bureaucracy:
1. Separate which current behaviors/units serve the mission from which serve the institution itself.
2. Identify 2 processes that are "good for the org but not clearly good for the mission," and gauge how far the drift has gone.
3. Propose one concrete move that re-injects external pressure and pulls power back toward the mission people.
帕金森琐碎法则 · Parkinson's Law of Triviality
"一个议题花掉的讨论时间，与它涉及的金额成反比。" —— C. Northcote Parkinson，1957（又称"自行车棚效应"）
就是那个反应堆 vs 自行车棚的委员会寓言本身——它如此精准，以至"自行车棚"成了全球工程师的行话。最大、最贵、最不可逆的决定（核反应堆）因无人能懂而草草放行；最小、最便宜、最人人可懂的决定（自行车棚刷什么漆）反而吸走了整场会议的能量。开会开得越热闹，往往说明议题越无关紧要。
场景 · BigCat
① 工程：与其在 code review 里反复争 tabs vs spaces，不如上 linter/formatter 把整类争论变成不可讨论——结构性地消灭 bikeshedding，把评审火力留给架构与并发安全。② AI 工作流：花一下午反复微调一句 prompt 的措辞（琐碎、可控、还很有成就感），却一直回避"这个 agent 架构到底对不对"的硬问题——这是典型的自我安慰式 bikeshedding。设个规则：架构问题没定，禁止动 prompt 文案。③ 家庭：一家人为"今晚吃哪家"耗光的精力，本该留给"选哪所学校、立什么家庭价值观"这类真反应堆。把琐碎决策默认化/轮值化，是给重大决策腾出注意力。
English Summary
Parkinson's Law of Triviality (Bikeshedding) — "Time spent on an item is inversely proportional to the money involved." In the parable, a committee waves through a multi-million reactor in minutes (no one understands it) but argues 45 minutes over a bike shed (everyone has an opinion). The driver is accessibility, not laziness: trivial items are comprehensible to all and carry no risk of exposing ignorance, while challenging the reactor design would. It's attention misallocation hijacked by comprehensibility — same root as availability bias. In software it's literally "bikeshedding": an hour on variable naming, a rubber stamp on a risky concurrency change. Defenses are structural: allocate agenda time inversely to stakes, kill the debate with tooling (let a linter make style "non-discussable"), raise the cost of cheap opinions, and default/delegate the trivial so scarce attention reaches the real reactor.
AI Prompts
中文提示词
这是我们最近一次 [会议/评审/决策] 的议程和耗时分布：[描述]。请用帕金森琐碎法则诊断：
① 标出哪些议题的"讨论时间 ÷ 实际重要性"严重失衡（即被 bikeshedding 的对象）；
② 分析为什么恰恰是这些琐事吸走了注意力（可及性 / 留痕需求 / 暴露无知的风险）；
③ 给出 1 套结构性对策：哪些该用工具/默认值直接消灭争论，哪些该限时丢给专家小组。
English Prompt
Here's the agenda and time breakdown of our last [meeting/review/decision]: [describe]. Diagnose with Parkinson's Law of Triviality:
1. Flag items where "discussion time ÷ real importance" is badly skewed (the bikeshedding targets).
2. Explain why exactly these trivial items captured the attention (accessibility / desire to leave a mark / risk of exposing ignorance).
3. Propose a structural fix: which debates to kill outright with tooling/defaults, and which to timebox and hand to a small expert group.