gaoyaze
AI-Era Reading Method
Is a book worth reading for you? Not based on others' ratings, but on the sparks it strikes with your existing knowledge. AI scans the entire book first, highlighting which sections deserve deep reading and which can be skipped. After you finish, it quantifies exactly how many new concepts you learned and which old beliefs changed. It reads your Board memory and gets more accurate with use. About "getting more accurate with use": This skill reads your Board memory (MEMORY.md) as a matching baseline. New users don't need to prepare in advance—on first use, the skill makes a rough estimate based on your stated areas of interest in the current conversation, then suggests at the end of the reading map that you write key information to memory. As you continue using it, your areas of interest, known concepts, and tracked topics naturally accumulate in MEMORY.md, and the accuracy of cross-matching and three-pronged prediction gradually improves. The simplest way to build memory: after each book, tell your AI assistant about the new companies, people, technologies, and core insights you encountered, and have it append them to MEMORY.md. This habit itself is what Adler called "syntopical reading"—knowledge forms a network through accumulation, not scattered fragments.
📱 只读版 · 💻 电脑编辑
从千形万象的流变中抽出三个不变原型，再用可组合的修饰覆盖整个连续谱
领域：气象学
地域：英国伦敦 · 1803
抽象模式：原型分解——为流变之物建立结构语法
具象的原点
十九世纪之前，天空是一片无法言说的混沌。云转瞬即逝，每一秒都在变形、消散、重生，没人能为它们命名，更没人能记录“今天天上究竟发生了什么”。气象日志里只剩下“多云”“阴沉”“晴朗”这类模糊的形容词，画家则凭直觉把云当作情绪的布景，画下的是各自的想象而非同一个对象。问题不在于观察不够细致，而在于缺少一套共同的语言——没有名字，就无法记录；没有记录，就无法比较；无法比较，气象学就永远停留在“今日天气”的日记层面。
这不是没有人尝试过。1802年，法国博物学家拉马克（Jean-Baptiste Lamarck）几乎同时提出了一套云的描述术语，列出了五个法语词。但他犯了三个致命错误：用法语而非通用学术语言，没有配图，发表在一本冷门的学术期刊上。结果他的分类法在法国之外无人知晓，拿破仑一纸命令让他停止气象研究，整个方案彻底湮没。云的命名权，悬而未决。
抽象的洞察
路克·霍华德（Luke Howard, 1772–1864）是伦敦的制造业化学家，贵格会教徒，业余气象学家。他从十岁起就坚持记天气日记，三十年如一日地仰望天空，记录每一片云的形状、高度、走向和变化。正是这种近乎偏执的长期观察让他看到了别人看不见的东西：云虽然千形万象、瞬息万变，但所有的变化都围绕少数几个“原型”展开，其余形态不过是原型之间的过渡与组合。
1802年12月，他在阿斯克西安学会（Askesian Society）宣读了一篇论文，次年以《论云的形态变化》（Essay on the Modification of Clouds）为题正式发表。他借鉴了林奈的生物分类思想，但做了关键改造：他定义三个不可再分的基本原型——cirrus（卷云，源自拉丁语“一缕头发”）、cumulus（积云，“堆叠”）、stratus（层云，“水平铺展”），再加一个功能性状态 nimbus（雨云）。然后他用四个复合修饰词——cirrocumulus、cirrostratus、cumulostratus、cumulonimbus——描述原型之间的过渡。关键的一步是，他坚持用拉丁语命名。拉丁语是全欧洲学者的通用语，这意味着一个伦敦人看到的“cirrus”和一个维也纳人看到的“cirrus”指向同一片天空。霍华德没有消灭云的变化，而是为变化本身建立了坐标。
封装的成果
霍华德封装出的不是一张静态的分类表，而是一套生成语法：三个不变原型作为词根，可组合的修饰作为前缀或后缀，任何一片云都能被分解为原型的组合，任何一种过渡都能用词缀关系表达。这是一种形式化的结构语言——“原型 × 修饰”覆盖了整个连续谱。
$$\text{云} \in {\text{cirrus},\ \text{cumulus},\ \text{stratus}}^{\otimes} \cup {\text{nimbus}}$$
其中 $\otimes$ 表示原型的可组合叠加，复合形态由原型的邻接关系生成。他亲自绘制水彩素描为每种云配图，让命名有了视觉锚点。至此，“无法定义的东西”获得了可命名、可记录、可比较的结构。
涌现的可能
一旦云有了名字，整片天空就从不可言说变成了可读取的数据流。霍华德的命名法在几十年间被欧洲各国气象站采纳，1887年英国气象学家阿伯克龙比和瑞典的希尔德布兰德森以它为基础制定了国际通用的十种云形分类，1929年国际气象委员会正式采纳，1951年移交给世界气象组织（WMO）沿用至今——你现在手机天气 App 里看到的“积雨云”“卷层云”，词根全部来自1803年那个伦敦化学家的手稿。这套语法还具备扩展能力：2017年 WMO 在国际云图典中新增了 volutus（滚云）、asperitas（波涛云）等新云种，只需追加新词缀，整个体系无需重构。更深远的是，它启发了整整一代风景画家——霍华德的云素描让画家第一次有“标准云形”可对照，歌德甚至为他写下长诗，赞叹“他用理性的手抓住了任何手都抓不住的东西”。而“为流变之物建立结构语法”这一范式本身，也成为应对一切连续变化现象的方法论原型。
Daily Abstract Thinking
What does this Skill do? Every day, it introduces a real-world case: something that was once explored through experience and hands-on intuition, but was later distilled into a general method or standard (such as Braille or paper size standards). This process of turning specific experience into a general method is called abstraction. How can it help you? It builds one key ability: turning experience into methods. When you encounter a challenge, you can extract reusable patterns instead of starting from scratch every time. How do you use it? After installing it, just say "Give me one". Each day, it gives you a new case, along with an image and a reflection question that you can apply to your own work. Cases don’t repeat, and they come from a range of industries. Who is it for? Anyone who wants to turn experience into lasting knowledge instead of letting it become increasingly fragmented—useful for product work, investing, writing, management, and more.
AI Delivery Check
Check whether AI-generated deliverables (web pages, apps, documents, etc.) fully satisfy all user requirements from the conversation. Review each item across functionality, visuals, and content. If any fall short, trigger AI corrections to ensure the final output aligns with user expectations.
Info Trust Check v7
Check the trustworthiness of AI responses. Supports @-referencing documents or directly verifying current conversation content. Results are shown in the chat by default; only prompts for generating a work when the content is lengthy or important. v7 adds: flexible input sources and smart output strategies. v7 is the result of multiple rounds of refinement, overcoming many edge cases, making it ideal for verifying critical information. Multi-factor verification: source independence, quantity, and authority; detection of logical inconsistencies in outputs; domain relevance checks; cross-reference of Chinese and English sources.