AI 教程与实战

Anthropic 的 AI 滥用报告称，2 个与伊朗安全组织关联的单位使用 Claude 构建服务同一个监控系统的工具。其中一个发布了伪装成祷告时间工具的 Firefox 扩展来收集社交媒体身份；中央系统存储国民 ID、信仰、犯罪记录和社交账号。报告图片还提到另一调查中伊朗关联威胁行为者用 Claude 收集分析公开数据，针对美国海军编制目标建议手册，并被 Anthropic 封禁账号。

OpenAI 的 Colin Jarvis 接受采访，讲述如何组建 FDE（前沿部署工程师）团队、日常工作内容与团队价值观：先让东西存在并跑通，再考虑规模化。团队如今更重视领域专家，成功标准是在每个客户问题中最终取代自己。推文附有采访链接和 FDE 招聘页面。

Picture a solid mid-level engineer joining your team. Nobody gives them access to the issue tracker....

The MCP instructions field reaches the model before any tool call, 66% of live servers send it, and a shared cache can serve it to a caller who...

Recently, I saw a LinkedIn post discussing "MCP vs API" and felt like we needed to choose one. But I...

One Claude Code session driving a browser through chrome-devtools-mcp works fine. Open a second...

Why you don't need a separate $400/mo vector database for RAG. How to use Postgres with pgvector and HNSW indexes for sub-10ms similarity search.

Earendil 工程师发文讨论如何量化 LLM 生成代码的粗糙程度，指出 AI as a judge 很难有效评估代码质量，而 LOC 变化、verbosity 和 erosion 是更可行的指标。

A backend engineer boarded a late train with a half-finished agent run still sitting in a terminal....

作者转发并提炼了一篇写作 Skill 搭建教程，核心观点是不要在生成阶段纠正 AI 文风，而要在构思阶段通过 Brief 确认闸门约束模型。三步为把 3-5 篇代表作锁进只读 raw/ 文件夹让模型反向提炼风格规则、用 【F】 事实 【W】 观点 【I】 推论 【？】 存疑四标签对齐素材、模型先交出含核心冲突和叙事骨架的 Brief 经人类确认后才写正文。

TL;DR I run up to four Claude Code agents at the same time on a single repository by...

Hace unos días, tuve que hacer algo que hacía mucho tiempo no consideraba necesario en mitad de una...

Read on: the A/B test this corrects · 繁體中文版 Two weeks ago I published an A/B test: I cut 41 AI...

开发者 luispoveda93 对 MiniCPM5-1B 在西班牙语、加泰罗尼亚语、巴斯克语和加利西亚语上做了零样本评测，覆盖推理、NLI、复述识别、QA 和翻译任务。该模型还被转换为约 1.1GB 的 INT8 LiteRT-LM 模型，可在 Android、iOS、桌面和 IoT 设备上离线推理。面壁智能表示 MiniCPM5-2B 已发布，希望更多开发者探索其多语言与端侧应用。

Fourteen parts ago, this series made a promise: build a real app with AI and hide nothing — every...

You've probably watched an AI think through a problem step by step, nod along with the logic, and...

When an agent needs help from another service, there is an architectural question to answer first:...

With MCP tool search on, which is the default, the first request of a fresh Claude Code session on...

One is a component you build. The other is how you wire many together. Confusing them costs months. Here's the architecture distinction.

作者提出"Waymo效应"，类比 LLM 正以无摩擦体验取代科研协作，使去协作化成为激励体系下的理性选择。文章结合团队科学研究与生成式 AI 早期证据指出个体生产力上升但想法多样性收窄，主张资助者和机构将协作视为基础设施，资助工作坊、访问与共址等"摩擦"，并援引 Dashun Wang 的"指挥科学"与 Bainbridge 自动化悖论强调人类须保留主导权。

Yesterday afternoon a runtime I wrote created 70 paragraphs on the same draft article. It was...

ExecCritic found that weak generated tests reduced repair success. A runnable Python example shows how to catch a test that approves the wrong fix.

Every founder knows the feeling: you share a new startup idea with friends, and they all smile and...

Here is a sentence that will break most computers. “I went to the bank.” Did you go to a...

Is that webshop legit? Most of us have pasted a URL into a reputation checker at least once. The...

Text-to-SQL looks great in a demo. Then someone ships a conversion rate that JOINs without ON, a...

OpenAI 发文（系列上篇）讲述其在线存储平台 Habitat 的演进：现每秒处理超 7000 万请求、服务每周超 10 亿用户、管理超 500PB 数据，覆盖近 40 个地区。

WatchMe is a small macOS menu-bar app I built: it lists your running processes, you tick the ones you...

通义千问发布《中小学老师千问使用指南》，收录一线教师的 14 个高频教学场景，拆解为 45 轮连续实操和 193 张操作画面。指南覆盖课前预习、课堂互动与课后组卷，例如用"千问小讲堂"讲解《观潮》、生成可交互 HTML 电路实验，以及按课题智能组卷并标注考点、答案与解析。

AI coding tools have made it easy to go from an idea to working code. You describe what you want,...

In discussions around AI-assisted programming with LLMs, I often see two opposing camps. And since...

Believe it or not, it's harder to be a software engineer in 2026 than it was five or ten years ago....

当你安装了 dr eggbot 的视角

I deleted a working function today. It ran. It was clean. It was better written than the version in...

I ship client sites fast on a simple free-tier stack: Next.js for the app, Supabase for...

A production bug I shipped, why the obvious fix was incomplete, and what actually closes it. ...

Many All of us are coding with AI these days. There is a lot of momentum; code is being produced at...

A WordPress function documented in five words quietly deleted the non-English part of every URL my crawler analytics recorded — turning hundreds of distinct pages into a single confident-looking row.

A small TypeScript experiment comparing AgentInspect and Promptfoo for local agent trajectory checks.

Intro AI Model Upgrades Aren't Patches. They're Silent Breaking Changes Wearing a Version...