karpathy

LLM Wiki

A pattern for building personal knowledge bases using LLMs.

This is an idea file, it is designed to be copy pasted to your own LLM Agent (e.g. OpenAI Codex, Claude Code, OpenCode / Pi, or etc.). Its goal is to communicate the high level idea, but your agent will build out the specifics in collaboration with you.

The core idea

Most people's experience with LLMs and documents looks like RAG: you upload a collection of files, the LLM retrieves relevant chunks at query time, and generates an answer. This works, but the LLM is rediscovering knowledge from scratch on every question. There's no accumulation. Ask a subtle question that requires synthesizing five documents, and the LLM has to find and piece together the relevant fragments every time. Nothing is built up. NotebookLM, ChatGPT file uploads, and most RAG systems work this way.

dianoetic

Send Webmentions with GitHub Actions

Action that uses curl to send your webmentions. It gets the most recent addition to a JSON feed (ideally, your latest post), looks for the value of a key called uri (ideally, your post's permalink) and sends webmentions for it.

The services currently in the file are webmention.app and brid.gy, but any service that can be triggered with a POST request will work fine with this.

This workflow file is run once another GitHub Action named "Build" completes successfully.

Usage

rvanzon

// just an example. A cleaner way is to wrap the showLine-stuff in a dedicated component
<template>
  <div>
    <my-line v-if="showLine" :data="lineData" :options="options">
  </div>
</template>

<script>
export default {
  data () {

chantastic

Hi Nicholas,

I saw you tweet about JSX yesterday. It seemed like the discussion devolved pretty quickly but I wanted to share our experience over the last year. I understand your concerns. I've made similar remarks about JSX. When we started using it Planning Center, I led the charge to write React without it. I don't imagine I'd have much to say that you haven't considered but, if it's helpful, here's a pattern that changed my opinion:

The idea that "React is the V in MVC" is disingenuous. It's a good pitch but, for many of us, it feels like in invitation to repeat our history of coupled views. In practice, React is the V and the C. Dan Abramov describes the division as Smart and Dumb Components. At our office, we call them stateless and container components (view-controllers if we're Flux). The idea is pretty simple: components can't

GiaoGiaoCat

// 检测浏览器的 User Agent 应该是非常简单的事情
// 微信在 Android 下的 User Agent
mozilla/5.0 (linux; u; android 4.1.2; zh-cn; mi-one plus build/jzo54k) applewebkit/534.30 (khtml, like gecko) version/4.0 mobile safari/534.30 micromessenger/5.0.1.352

// 微信在 iPhone 下的 User Agent
mozilla/5.0 (iphone; cpu iphone os 5_1_1 like mac os x) applewebkit/534.46 (khtml, like gecko) mobile/9b206 micromessenger/5.0

// 通过javascript判断
// 很容易看出来，微信的 User Agent 都有‘micromessenger’字符串标示，我们判断是否含有这些字符串就OK了
function isWeixinBrowser(){

koistya

Files

The basic structure of a React+Flux application (see other examples)

 - /src/actions/AppActions.js     - Action creators (Flux)
 - /src/components/Application.js - The top-level React component
 - /src/constants/ActionTypes.js  - Action types (Flux)
 - /src/core/Dispatcher.js        - Dispatcher (Flux)
 - /src/stores/AppStore.js        - The main store (Flux)

kejun

mathclub是最近做的一个个人项目，帮助考SAT的同学通过在线做题、回顾、问答提高成绩。用户功能有：计次/计时做题、成绩单、错题分布、错题回顾、提问、汇总以及注册登录。管理后台主要是题库管理、学员管理、成绩单管理、问题回复。怎么看都像学校里的课设，的确项目本身并不出奇，开发上选用的一些方案或许更有意思。

整个项目一个人从产品需求、原型设计、前后端开发到部署历时2周左右。可以从截图上感受一下：

技术选型上服务端是Node.js，应用框架选了老牌的Express（4.x变化挺大不少中间件都废了），数据服务用的是MongoLab（MongoDB的云服务平台），图片上传用的是又拍云，程序部署在Nodejitsu上。模板引擎没选主流的Jade或ejs，而是用Express React Views它实现了在服务端渲染React组件。前端框架是用React，这次有意想追求前后端全部组件化的组织。之前是用Webpack实现CommonJS模块打包，这次用Browserify配置更简单，它有丰富的transform很赞，其中的reactify转换React的JSX很完美。CSS用Sass＋autoprefixer让人省心。将这一切串起来的自动构建工具是Gulp。我其实崇尚用最精简的工具组合开发，上述组合在我看来比较精简了。（帖纸留念）

![image](http://satexam.b0.upaiyu

anhulife

document.addEventListener('WeixinJSBridgeReady', function(){
	//如果执行到这块的代码,就说明是在微信内部浏览器内打开的.
	alert('当前页面在微信内嵌浏览器中打开!');
});

yefuchs

#The Great Firewall (GFW) Contributors List

注：数据来源为 dblp 和 cndblp, 下面括号中的数字表示 dblp 中显示的跟方滨兴合作论文的数量

###Binxing Fang (方滨兴)

中国工程院院士,北京邮电大学教授,中国科学院计算技术研究所网络方向首席科学家
http://en.wikipedia.org/wiki/Fang_Binxing

murtaugh

<figure class="quote">
	<blockquote>It is the unofficial force—the Baker Street irregulars.</blockquote>
</figure>