bobobo

[Andrew Ng @AndrewYNg]

“Loop engineering” is a hot buzzphrase after mentions of it by Boris Cherny (Claude Code’s creator) and Peter Steinberger (OpenClaw's creator) went viral on social media. Loops are now a key part of how we get AI agents to iterate at length to build software. In this letter, I’d like to share my 3 key loops, shown in the image below, for building 0-to-1 products. These loops guide not just how I build software, but also how I decide what software to build.

Agentic coding loop: Given a product specification and optionally a set of evals (that is, a dataset against which to measure performance), we can have an AI agent write code, test its work, and keep iterating until the code is bug-free and meets its specification. This idea of closing the loop took off around the end of last year, and it has been a game changer in enabling coding agents to work longer productively without human intervention. For example, over the weekend, I was building an app for my daughter to practice typing, and my coding agent could easily work for around an hour, using a web browser to check what it had built multiple times before getting back to me, without needing my intervention.

The engineering loop executes quickly. Every few minutes, the coding agent might build and test a new version of the software. I hear frequently from developers who are finding new ways to engineer more effective engineering loops. This is an active area of invention!

Developer feedback loop: In this loop, a developer examines the current product and steers the coding agent to improve it. Last year, a lot of developers (including me) were acting as the QA (quality assurance) function for our coding agents, manually finding bugs and then asking the agent to fix them. But with coding agents much more able to test their own code, the amount of time we need to spend on this function has decreased significantly. This allows us to make higher-level product decisions, such as what key features to offer, where the UI needs improvement, and so on.

The developer-feedback loop operates over time intervals between tens of minutes and hours — that's how frequently a developer might review a product and give feedback. In the case of the typing app, I changed my mind a few times about the visual design, what cat costumes she can unlock as she learns (she loves cats), and the user flow for a grown-up to log in and steer the child's learning experience.

When a developer has a clear vision for what to build, it is still a lot of work to translate that vision into a specification for a coding agent to implement. Further, after the developer has seen an implementation, they might update (or perhaps clarify) the spec to steer it toward what they want. If you find that the system repeatedly runs into certain problems, building a set of evals for the agent becomes useful.

AI-native teams are increasingly using AI to help shape product direction, for example, automating the gathering and analysis of usage data, summarizing written and verbal customer feedback, or carrying out competitive analysis. However, for pretty much all the products I’m involved in, I see humans as having a significant context advantage over current AI systems — we know a lot more than the AI system about the users and the context the product has to operate in — and thus humans play a critical role. Many people describe this human contribution as “taste,” but I prefer to think of it as humans having a context advantage, since that gives us a clearer path to helping AI systems get better. This also speaks to why this step can’t be automated: So long as the human knows something the AI does not, human-in-the-loop is needed to to inject that knowledge into the system.

External feedback loop: This includes a wide range of tactics like asking a few friends for feedback, launching to alpha testers, or putting the code into production with A/B testing. These tactics are usually slow, rarely taking less than hours and sometimes taking days or even weeks. This data informs the developer vision, which in turn continues to drive the detailed product spec, which in turn drives the coding agent.

With coding agents speeding up software development, more engineers are starting to play a partial product management role. For many engineers who are growing into this role, the hardest part is shaping the product vision and striking a balance between building (bridging the gap between vision and spec) and getting user feedback to evolve the vision. It is important to do both!

I will write more about how to do this in future posts, but for now, I find it encouraging that engineers are playing an expanded role (just as product managers and designers now do more engineering).

[Original text: The Batch]

[老鬼 @laogui]

React Doctor 这个代码检测工具和 Codex 的 Goal 简直是绝配！

今天用一条命令跑了 2 小时，直接干掉了 300 个代码质量与性能隐患，最后拿到了 100 分满分的健康分数，成就感满满。

顺便安利一下天才少年 Aiden Bai 和他打造的三个 React 质量与性能神器。他 16 岁独立开发 Million.js；18 岁带队入选 Y Combinator，为公司融资 1410 万美元。如今他做的这三款工具，刚好串联起了 AI 时代人机协同的完美闭环：

1. React Scan：浏览器里直接跑，哪个组件在重复渲染就用彩色边框高亮+闪烁，一眼就能看到性能浪费在哪。。

2. React Grab：网页上按 Cmd/Ctrl+C 点元素，直接复制出对应代码的文件路径、行号和组件栈，扔给 AI 让它精准修改。

3. React Doctor：专门扫描 AI 写出来的 React 代码问题（state、effect、性能、架构等），然后给项目打一个 0-100 的健康分数。能直接集成到 Codex、Cursor、Claude Code 里，或接入 CI 里自动运行。

现在最好的用法是直接在 Codex 里输入这条 Goal 指令：

/goal run "npx react-doctor@latest" and fix issues until you get a score of 100. do it properly without taking any shortcuts.

让它自己跑、自己修、自己验证，直到 100 分。

[Stanford NLP Group @stanfordnlp]

The “problem” with CS336 is not the ~22 hours of videos but the larger number of hours it takes to do the assignments.

But that is where most of the real learning occurs.

We’re reminded of @karpathy’s seminal tweet:

https://t.co/fvSeE2bDkE

2026 site: https://t.co/E1pzUSC6Tr

[Movez @0xMovez]

Ex-Google engineer explained AI agent loops, harness, evals in 20 minutes - better than 500$ courses.

trace every run → judge it with an LLM → diagnose → fix → ship.

That loop is how agents self-improve over time.

Agent loops + memory + harness + evals - thats the stack.

Watch it, then save the framework below.

[bobobo @hibobo233]

@0xJoveXu 网页有 pro 模型用

[梭哈.AI @SUOHA_AI]

如果你想加入世界顶级的AI公司，这篇笔记能帮你少走很多弯路

Alisa Liu 是华盛顿大学（UW）的 NLP PhD 学生，她最近拿到了 OpenAI Research Scientist 的 offer，并分享了一篇非常实用的求职笔记

她是怎么备战的？

先建立广度：把 Stanford CS336《Language Modeling from Scratch》全部 lectures 看完，这门课帮她把散落的知识点串成一个清晰的整体框架

再深度突破：一个概念一个概念深挖 —— 读 blog + paper + 大量和 ChatGPT/Claude 对话 + 从零实现代码

最关键的是：Transformer 的实现与调试要练到 muscle memory，并且完全关闭 AI 辅助练习 coding（因为真实面试时你必须自己写）

持续做结构化笔记（她有公开的 LLM Notes 可参考）

每个面试前做针对性突击复习，面试当天必须睡够觉（她第一次技术面试只睡 2 小时，结果发挥失常）

学习路径总结：广度先行（CS336）→ 逐个概念深度 + 动手实现 → 针对性面试前 cramming

OpenAI面试主要考什么？

ML Coding（出现频率最高）：用 PyTorch 实现架构、decoding 策略、Transformer 等

General Coding：LeetCode 风格题目

Technical Discussion：实验设计讨论 + 快速概念问答（positional encoding 的不同方式、parallelism、PPO vs GRPO 等）

Research Discussion：讲自己的项目、insight 和未来方向

Behavioral：提前把 PhD 经历整理成故事（她第一场 behavioral 直接翻车，因为没准备）

Math + Job Talk（聚焦自己最核心的方向）

如果你想准备 OpenAI / 类似 lab 的面试，必须精通这些资源

以下是她实际使用的学习资源：

1. 斯坦福大学的“从零开始的语言建模”课
https://t.co/IQrm8EuuoS

2. The Illustrated GPT-2（Jalammar） https://t.co/pZq2BCEhta
用可视化方式快速理解 GPT-2 的内部机制，适合建立直觉

3. Self-Attention & Transformers（CS224n PDF） https://t.co/s7lgCF6j4p
深入理解自注意力机制的核心原理

4. Backpropagation（CS231n） https://t.co/jmEcCRSgtf
手写 backward pass 的基础

5. Introduction to Policy Gradient for LMs https://t.co/hgAIO1cXeC 理解语言模型的策略梯度方法

6. Lightweight Guide to understanding GRPO and RL principles
https://t.co/OGDv7tZBsB
快速掌握 GRPO（近期 RLHF 相关的重要概念）

7. How to Scale Your Model（JAX scaling book） https://t.co/nm1ExirsrF
理解模型 scaling 的工程与理论要点

额外高频练习：
LeetCode（常规 + ML 相关题）

反复从零实现 Transformer（无 AI 辅助）

她的 LLM Notes（学习方法参考）：https://t.co/hsLkZv3NtF

[Line @0xLinehigher]

非常建议每一个选择计算机系的大学生，在大学时期将cs336啃完，不开中文字幕，只开英文字幕。啃完之后，你对LLM的理解和英语能力至少在国内前百分之1%。

这门课超过国内任何一所大学里面计算机的课程。

《Stanford CS336: Language Modeling from Scratch》 是一门斯坦福大学计算机科学系的课程，这门课每一年春季都会有，每一年都会在油管免费开放。

里面的内容非常前沿而且实用，目前油管上的版本是25年的版本，第一节课提到的模型已经是GPT-4。而与此同时，国内的计算机课程还在用着至少5年前的PPT。

依稀记得，我的第一节C语言课程，PPT的里面的演示系统是XP...

啃完CS336，你会收获LLM的全栈技术栈，从数据收集，清洗，模型训练，优化，评估到部署，Transforemer架构，注意力机制，各种目前最新的机器学习训练方法

这门课的目标就是从头让你手搓一个LLM，亲手实现完整的pipeline。

两位主要讲师：
1、Percy Liang
MIT 本科 + MEng，UC Berkeley 博士（2011）。曾在 Google 做 post-doc，后加入 Stanford 任教。

2、Tatsu
Harvard 本科（统计与数学），MIT 博士（co-advised by Tommi Jaakkola 和 David Gifford），之后在 Stanford 做 Percy Liang 和 John Duchi 的 post-doc

耐心啃完这个，所有计算机课程几乎都不需要上了...因为在过程中必然会遇到很多不懂的点，边学边查，然后顺带就把学校关于机器学习和深度学习的专业课给学完了。

如果刚上大一，可以先看cs50，然后再到cs336，先稍微打点计算机的基础。

与此同时，一定要把线性代数学好，将高数丢了都要把线性代数学好，线性代数是机器学习的底层DNA。

如果线性代数没有学好，会比较难理解神经网络，相当于没学好加减乘除就想去求导。

[Jesse @jesse_vermeulen]

just published a writeup on how we achieved this

https://t.co/ylUbgOBbdS

[mousepotato @iluciddreaming]

Google 又干掉了一个创业公司……

Google AI Edge Eloquent 现已支持 Mac，完全本地的 Wispr Flow 替代品。

基于最新 Gemma 模型，支持实时语音转录 + 语音命令编辑文本。

免费、无订阅、无需联网，隐私全本地。

[Julian Garnier @JulianGarnier]

Anime.js 4.5 is out and it's a fun one:
Introducing the @threejs adapter 🎉

- Up to 50% less code for 3D animations
- CSS transform-like API for 3D objects (rotate, skew…)
- Simpler material color animations
- Easy instanced mesh animations
- Stagger 3D

And so much more! ⬇︎

[陈大黄 @realchendahuang]

分享给大家我一年多的血泪独立开发踩坑经验。

2026 年，独立开发者想开发 APP，最佳实践是：

MVP 阶段先做 Web➕PWA。开发快，调试快，迭代快。

Web 技术栈：

React➕TanStack Start➕Vite➕TypeScript➕shadcn/ui➕Base UI➕Tailwind CSS➕better-auth➕Vercel AI SDK➕oxfmt / oxlint

后端统一：

Hono➕Cloudflare Workers

Web 调试：

Chrome DevTools MCP➕Codex Chrome 插件

等产品真的跑起来，再做移动端。

移动端技术栈：

React Native➕Expo➕Expo Router➕NativeWind➕React Native Reusables➕Reanimated➕Gesture Handler➕EAS Build / Submit / Update

核心思路很简单：

先用 Web➕PWA 验证需求。

真的有用户，再上 React Native➕Expo 做 APP。

[小盖 @xiaogaifun]

https://t.co/sshHthPn3R

[Paidax @xin_pai88825]

设计师必备 Design Skill 推荐，整理了我日常高频在用的 5 款实用 Skill，亲测实用性拉满

1、Text to Lottie Skill
上传 SVG + 填写动画描述，快速生成可编辑 Lottie 动画文件。

2、GSAP Skill
一键生成专业网页交互动效，自动输出高性能代码，支持 SVG 动画、鼠标跟随、滚动视差文字等效果。

3、three-scope-map-skill
输入地区即可生成数据大屏 3D 交互地图，多款配色主题，支持缩放、平移、点位、飞线等大屏常用特效，开箱即用。

4、Web to Design md
输入网址一键扒取网页配色、字体、CSS、动效等全套设计规范，输出可被 AI 使用的 design.md 文档。

5、Shadcn/ui Skill
适配前端项目，自动对齐项目配置、统一组件规范、支持多主题，一键生成规整后台页面，省去大量调试工作。

[Jesse @jesse_vermeulen]

finally cracked the code for liquid glass on the web

[Mengke Wang @MengkePM]

https://t.co/ghUgtm1M1w

[Matt Pocock @mattpocockuk]

Announcing mattpocock/skills v1

- Achieved a 63% reduction in token cost for skill descriptions
- Split skills into model-invocable and user-invocable skills, adding /codebase-design, /domain-modeling, and /grilling
- (UPDATED) /writing-great-skills - rewritten from the ground up, encoding my skill-writing best practices
- (UPDATED) /diagnose -> /diagnosing-bugs - now model-invocable, awesome for fixing hard bugs
- (NEW) /ask-matt: a router skill that teaches you how all the engineering skills work together

[Khairallah AL-Awady @eng_khairallah1]

https://t.co/4boc8dTGGh

[Tw93 @HiTw93]

假如你的朋友最近需要更新简历，一定要把 Kami 推荐给他，我单独细致优化了一个版本，单独让 Kami 写简历变得非常好用好看清晰，让他把他的原生素材 md 准备好，然后对着 AI 说 /kami 帮我产出一个简历，然后调1-2下差不多就好了。
https://t.co/XB5mfwU9rX

[爱丽丝呀！ @BTCqzy1]

AI 生成动画的质感短板，终于被补齐了！

以前 AI 吐的动画又僵又平、毫无灵魂……

分享一个轻量级开源神器 anime.js ，直接提升你的动画质感～

一行代码搞定丝滑复杂动画：CSS 属性、SVG 变形、Timeline 时间轴、Stagger 交错、Scroll 触发、Draggable 拖拽、Spring 物理…全都能轻松驾驭。

为什么 anime.js 是动画库天花板？
超轻量：仅几 KB 模块化，不臃肿
API 极简：直观易读，零学习曲线——特别适合 AI 生成代码
功能全面：SVG Morph + Motion Path + Scroll Observer + Spring 物理 + Timeline…复杂场景也不怂

完美适配 Cursor / Claude 等agent智能体，感兴趣的可以看看～

[Daniel @DanielWhit21874]

You know what is so satisfying to use ?

Slot-text

https://t.co/NLjNwFO8I3