Magpie

从三位读者的困惑出发，探讨 AI 逐步取代执行环节后人类还剩下什么，提出判断力、品味与推演三种难以被替代的能力，并反思认知外包对个体主体性的侵蚀。

how did we fix the ai design slop problem in llms - DeepSeek/Kimi/Qwen or Claude/GPT?! i've been thinking about "why do all ai-generated designs look the same?" is it a model problem or a harness problem? context: we're fixing the llm design problem with `/design` for @CommandCodeAI - atm it has …

AI coding works when the stack is right and the developer has the experience to trust the output. Most of the elaborate workflows being shared are people getting in their own way.

Context as infra, taste as config, verification for autonomy, scale via delegation, closing the loop.

how did we make deepseek outperform opus 4.7? i've been thinking about why "open model bad at tool calling" is almost always a harness problem, not a model problem. context: spent the two days looking at billions of tokens in @CommandCodeAI (tb open source ai cli) using deepseek. I ended up writi…

阿里这篇关于 AI Native 时代组织研发的思考非常值得一读。它在思路一个非常重要的底层问题：过去两千年的组织形态，都是围绕人的局限建立起来的。 人会遗忘，会疲惫，会误解，会有情绪。一个人能够稳定协作和管理的人数有限，信息在层级之间传递一定会衰减，所以组织才需要汇报链、部门边界、经理角色、需求评审、流程审批和各种协调机制。很多我们习以为常的管理制度，本质上不是先进设计，而是对人类认知带宽的妥协。 AI Agent 进入组织之后，这个前提开始松动。它不是普通工具。普通工具只是延伸人的手脚，Agent 则开始参与理解、执行、调用系统和交付结果。它没有疲劳，没有情绪，没有传统意义上的沟通损…

Building RAG (Retrieval-Augmented Generation) systems and AI agents in a sandbox is one thing. Shipping them to production and keeping them…

Lessons from my trip to talk to most of the leading AI labs in China.

听《张小珺Jùn｜商业访谈录》上小宇宙。 和你一起，从这里，探索新世界！ 《张小珺商业访谈录》是一档深度访谈节目，以2至7小时的长对话为形式，聚焦真实的人物表达与高密度的前沿信息。由语言即世界工作室出品。 主持人张小珺是财经媒体人，专注中国科技业深度报道，覆盖AI、科技巨头、风险投资与知名人物。曾13次获得国内外新闻奖项，其中2022年至2024年连续三届获得亚洲卓越新闻奖（SOPA）——被称为“亚洲普利策”。她也是语言即世界工作室的创始人。 “如果我的访谈能陪你走一段孤独的未知的路，也许有一天可以离目的地更近一点，我就很温暖。”——张小珺

Corgis, compute, Cold War; Ecclesiastes; ties; Stendhal; humor; Pascal's Wager; deep infrastructure; Germanic obedience; Texas State Fair

For a while now I’ve been hearing from engineers who run multiple coding agents at once—firing up several Claude Code or Codex CLI instances at the same time, sometimes in …

A comprehensive guide to Claude Code from someone who's been using coding AI since 2021. Learn the 5 pillars of agentic coding, custom commands, and advanced workflows.

Three months ago I would have dismissed claims that AI could replace developers. Today, after using Claude Opus 4.5, I believe AI coding agents can absolutely replace developers.

观察到当代码成本趋近于零时，为一次性决策构建专用工具反而是最优策略。提出AI Native的本质是在信息获取成本坍塌后采用全新策略——从直觉驱动转向高分辨率的数据驱动决策。

我学了一辈子的技能，被屠杀了，claude code 已经是远超人类 99% 程序员的编程能力了。Learn the killer techniques of AI-assisted programming.

A practical guide to working with AI coding agents without the hype.

A curated collection of resources, papers, tools, and best practices for Context Engineering in AI agents and Large Language Models (LLMs). - yzfly/awesome-context-engineering

这篇文章分享了Manus通过我们自己的"SGD"所达到的局部最优解。如果你正在构建自己的AI代理，我们希望这些原则能帮助你更快地收敛。

Learn how agents use filesystems for context engineering to store, retrieve, and update context efficiently. Reduce token costs and improve agent performance.

本文将对 Codex 智能体循环进行深度技术剖析，阐述 Codex CLI 如何利用 Responses API 来协调模型、工具、提示，并优化整体性能。

随着模型作为智能体的能力不断提升，我们发现，预先提供更少的细节，反而能让智能体更容易自主提取相关上下文。

Some (well-known) consequences of AI automating work

Some (well-known) consequences of automating work

The nagging feeling that something does not fit

最近关于 AI 辅助编程和 Vibe Coding 的讨论非常激烈，想了想还是写一篇博客来表达我自己的观点。

宝玉的分享，主要在大语言模型、Prompt Engineering、软件工程、工程管理、前端等领域。

Trying to understand the implications of AI for work, education, and life. By Prof. Ethan Mollick. Click to read One Useful Thing, by Ethan Mollick, a Substack publication with hundreds of thousands of subscribers.