The handful of people and pages actually worth your time — sorted easy to hard, each explained like you’re five.真正值得你花时间的那几个人、几个页面——从易到难排好，每一个都用讲给五岁小孩的方式解释。
jrforeal — the Concepts deckjrforeal — 概念篇
Before anyone lets you drive a car, someone points at the pedals and says what each one does. This is that — the plain-words tour of what AI even is, before a single button gets pressed.在让你开车之前，总得有人先指着踏板告诉你每个是干嘛的。这个就是这样——用大白话带你看 AI 到底是什么，一个按钮都还没按之前先搞懂。
Your own starting point. No code, no jargon — just the mental model everything else on this page assumes you have.你自己的起点。没有代码、没有术语——只是这页其它所有东西都默认你已经有的那套基本认知。
platform.claude.com/cookbookplatform.claude.com/cookbook
A recipe book, except every recipe is one thing you can get an AI to cook for you — read a document, answer from your own files, check its own homework — with the exact steps laid out.一本食谱，只不过每道菜都是你能让 AI 帮你做的一件事——读一份文件、从你自己的资料里找答案、检查它自己的作业——而且每一步都写得清清楚楚。
~80 official recipes. The fastest jump from “what is this even for?” to “oh — I can just copy that.”约 80 个官方示例。从「这到底能干嘛」跳到「哦，照抄就行」最快的一条路。
Personal-AI-assistant setups个人 AI 助手搭建
A guy on YouTube who teaches you to get a robot helper to do your chores. You say what you want; it works out how. He teaches you to be the boss giving orders, not the worker doing the job.YouTube 上的一个人，教你怎么让一个机器人助手帮你干活。你说你要什么，它自己想办法怎么做。他教你当那个下命令的老板，而不是亲自干活的员工。
A smart model bossing cheaper ones around, asking the AI what to do next, skills that patch themselves. Practical operator stuff — no maths.用一个聪明的模型去指挥便宜的模型、反过来问 AI 下一步该干嘛、会自我修补的技能。全是能上手的操作——没有数学。
Chamath · Sacks · Calacanis · FriedbergChamath · Sacks · Calacanis · Friedberg
Four rich friends sit around a poker table every week and argue about what’s really going on with money and technology. You listen in, and slowly you start to think the way the grown-ups think.四个有钱的朋友每周围着一张扑克桌，吵关于钱和科技到底在发生什么。你在旁边听着，慢慢地，就开始像那些大人一样思考了。
A weekly reality check on where AI and money are actually heading — who wins, what it costs, where the capital flows. Orientation, not a tutorial.每周一次的现实校准：AI 和钱到底往哪走——谁赢、成本多少、资本流向哪。是帮你找方向的，不是教程。
Andrej Karpathy · Sequoia AI AscentAndrej Karpathy · 红杉 AI Ascent
One of the people who actually built AI explains the difference between letting the robot just wing it, and engineering it properly. It’s the grown-up way to build real things with AI.真正参与造出 AI 的人之一，来解释「让机器人随便瞎搞」和「把它认真工程化」之间的区别。这是用 AI 认真做出真东西的那种成熟做法。
The contract-first build loop this whole workspace runs on: decide what “done” means before you build, verify against evidence, never declare done on faith. Highest-value idea here — once you’re already building.这整个工作方式所依据的「先立契约」循环：动手前先定义什么叫「做完」、拿证据去验证、绝不凭感觉说做完了。这页里价值最高的一条——前提是你已经在动手做了。