① 设计 AI 系统：你亲手画的那套 agent 架构，会被你系统性高估——因为它是"我的"。对治不是更努力地客观，而是请一个不在乎你面子的人做红队（red team），或让另一个 LLM 专门扮演"拆台者"列出这套设计会怎么崩。② 育儿：父母对"我最懂我的孩子"几乎全员过度自信——而这恰是反馈最慢的领域。把对孩子的判断当成带概率的假设，留出"我可能读错了"的空间，比笃定更接近真相。越笃定的领域，越要怀疑那只是没有反馈而已。
English Summary
Excessive Self-Regard Tendency — humans systematically overrate themselves and anything tagged "mine": my judgment, my plan, my child (the Lake Wobegon effect: ~90% rate themselves above-average drivers). The mechanism isn't vanity but endowment — once an idea becomes "mine," the brain over-weights it. Overconfidence isn't shameful; it's the fuel for action — the problem is always calibration, not elimination. It's deadliest where feedback is slow and noisy (investing, strategy, parenting); fast-feedback domains (chess, poker) self-correct. Munger's antidote: "I'm not entitled to an opinion until I can state the other side's case better than they can." Separate confidence-in-a-decision from confidence-in-being-right, attach probabilities, and keep a calibration ledger.
AI Prompts
中文提示词
我对 [某个判断 / 方案 / 计划] 很有信心，把握大约 [X成]。请用过度自信模型给我做校准：
① 我这份信心里，有多少来自证据，有多少来自"这是我的"（禀赋效应）？
② 这个领域的反馈快还是慢？我上一次被现实纠正是什么时候？
③ 请扮演拆台者，把"我错了"这一面的论证讲到比我自己还强，并指出我最可能高估的那个环节。
English Prompt
I'm confident about [a judgment / plan], roughly [X]% sure. Calibrate me using the Excessive Self-Regard model:
1. How much of this confidence comes from evidence vs. the fact that it's "mine" (endowment effect)?
2. Is feedback in this domain fast or slow? When was I last corrected by reality?
3. Play devil's advocate: argue the "I'm wrong" side better than I could, and name the single step I'm most likely overrating.
① 作为追求"AI 超级个体"的技术人：很容易陷入和同行比 GitHub star、比评测分、比融资额的内耗——而这些参照系多半是算法喂的。换成"去年的自己"，嫉妒立刻失去燃料。② 育儿里最隐蔽的陷阱：很多"鸡娃"的真正驱动，是父母对其他父母的嫉妒，借孩子成绩来比拼——孩子成了大人嫉妒的代偿工具。先察觉这层，才能把"别人家孩子"请出教育现场。嫉妒治不好，但参照系可以换。
English Summary
Envy / Jealousy Tendency — Munger insists it's envy, not greed, that drives most human irrationality. Envy is the one deadly sin with zero upside: pure suffering, no pleasure. It's relative and local — we don't envy billionaires, we envy the peer one rung up (Russell: "Beggars don't envy millionaires; they envy other beggars who do better"). Your pain is set not by your absolute condition but by your chosen reference group — and social media industrializes envy by making everyone's highlight reel your reference group. Envy is invisible to the envier, disguising itself as "unfairness." Its direct antidote is the Buddhist muditā (sympathetic joy), a trainable capacity. Practice: treat the sting at others' good news as an envy alarm, then shrink/swap your reference group and convert "why them?" into "what did they get right that I can learn?"
AI Prompts
中文提示词
我最近因为 [某个人 / 某件事] 持续感到不舒服或不甘心。请用嫉妒驱动模型帮我拆解：
① 我现在用的参照系是谁？这个参照系是我主动选的，还是被环境/算法喂给我的？
② 剥掉"这不公平"的外衣，底下有多少其实是嫉妒？
③ 给我一个"随喜 + 学习"的改写：从对方身上我具体能学什么，以及如何把参照系换成"过去的我"。
English Prompt
I keep feeling uneasy or resentful about [a person / event]. Use the Envy Tendency model to dissect it:
1. Who is my current reference group? Did I choose it, or was it fed to me by my environment/algorithms?
2. Strip away the "this is unfair" framing — how much underneath is actually envy?
3. Give me a "muditā + learning" reframe: what specifically can I learn from them, and how do I reset my reference group to "my past self"?
① 技术决策：你在全公司会上力推过某套架构/某个技术选型，半年后数据说它不行——一致性偏见会让你越辩护越用力，因为推翻它等于推翻"当初拍板的我"。解法：把它从"我的方案"重命名为"v1 假设"，公开宣布"我们当时基于 X 选了它，现在 X 变了"，给自己一个体面的下台阶。② 育儿：脱口而出定了一条规矩，事后发现不合理，却因为"不能在孩子面前显得善变/没权威"而硬撑——这是用孩子的成长给你的一致性买单。坦白"妈妈/爸爸想得不周到，我们改一下"，反而示范了最珍贵的能力：基于证据更新自己。
English Summary
Commitment & Consistency Tendency — once we take a position, especially publicly or effortfully, the brain locks it in to avoid the pain of appearing inconsistent, distorting later evidence to defend it. Munger: "The human mind is a lot like the human egg — once one sperm gets in, it shuts down to others." The more public and costly the commitment, the tighter the lock — which is exactly how cults and pyramid schemes work (front-load a costly, public act). Beliefs fused with identity make changing your mind feel like a small ego-death. Defenses: delay public commitment; hold beliefs as hypotheses with pre-written "kill criteria"; and reward yourself for changing your mind — people who are right a lot tend to change their minds a lot.
AI Prompts
中文提示词
我在 [某个观点 / 决策 / 立场] 上越来越难松口，但隐约觉得证据在变。请用一致性偏见模型帮我解锁：
① 我当初是在多公开、多费力的情况下做这个承诺的？锁得有多死？
② 我现在的辩护，有多少是为了"事情本身对"，多少是为了"不否定当初的我"？
③ 给我写 2-3 条具体的"杀死条件"，并设计一个体面的、不丢面子的转身说法。
English Prompt
I find it harder and harder to back down on [a view / decision], yet the evidence seems to be shifting. Use the Commitment & Consistency model to unlock me:
1. How public and effortful was my original commitment? How tight is the lock?
2. How much of my defense is "the thing is actually right" vs. "not negating my past self"?
3. Write me 2–3 concrete "kill criteria," and craft a face-saving way to reverse course.
① AI 赛道的某些融资/产品狂热，正是教科书级的 Lollapalooza：社会认同（人人都在做 agent）+ 稀缺/FOMO（错过这波就没了）+ 权威背书（大佬都投了）+ 一致性（已公开下注就得追加）同向叠加，把整个行业的估值推到相变点。识别它，能让你在狂热里保持一份冷静的仓位。② 反向善用：想让一个好习惯在自己或团队里"自然发生"，就主动把多个正向倾向叠到一起——公开承诺 + 同伴社群 + 即时反馈 + 身份认同，四力同向，行为改变会从"靠意志硬撑"跳变为"顺水推舟"。Lollapalooza 是中性引擎：看你把哪些力对齐到哪个方向。
English Summary
The Lollapalooza Effect — Munger's capstone: when several psychological tendencies act in the same direction at once, the result isn't additive but multiplicative — they resonate and tip past a critical threshold into extreme outcomes (manias, panics, bubbles, cults, buying frenzies). Most big real-world events are multi-bias confluences, not single-bias errors — which is why single-factor analysis fails: each bias checked alone looks tolerable, but you miss their product. Structurally identical to complexity science's critical points / phase transitions / positive feedback. A Lollapalooza can be accidental (market bubbles) or deliberately engineered (casinos, auctions, sales scripts). Key defense: you can't catch a confluence by introspecting one bias at a time — you need a whole-checklist scan (Munger's latticework). When something feels overwhelmingly compelling, suspect a Lollapalooza and list which 3–4 biases are stacking.
AI Prompts
中文提示词
我正面对 [某个让我感觉"压倒性地诱人/正确"的决策、机会或潮流]。请用 Lollapalooza 效应给我做合流扫描：
① 此刻有哪 3-5 个心理倾向正在同向叠加（如社会认同、稀缺、权威、承诺一致、过度自信、嫉妒）？
② 它们是自然合流，还是有人刻意工程化把这些钩子挂在一起？
③ 假如逐一拆掉每个钩子，这个机会还剩多少真实价值？给我一个能在狂热中保持冷静的行动建议。
English Prompt
I'm facing [a decision / opportunity / trend that feels overwhelmingly compelling]. Run a Lollapalooza confluence scan:
1. Which 3–5 psychological tendencies are stacking in the same direction right now (social proof, scarcity, authority, commitment/consistency, overconfidence, envy)?
2. Is this a natural confluence, or has someone deliberately engineered the hooks together?
3. If I removed each hook one by one, how much real value remains? Give me an action that keeps me calm amid the frenzy.