Blog
Explore tutorials, architecture deep-dives, and best practices for AI Agents and MCP Servers.
开源 MCP 与 Agent Skill 怎么选：一份可检索的工具目录与评测指南
面对官方 Registry、国内 MCP Hub、ClawHub 与海量 SKILL.md，如何按场景选型并完成 Cursor / Claude Code 安装？本文给出检索路径、评测维度，以及把 AgentHub 当作资源库引用的写法。
2026 海外 AI 编程智能体横评：Cursor、Claude Code、Windsurf、GitHub Copilot 怎么选
从 Agent 自主改仓、终端执行、MCP/工具生态、上下文与计费四个维度，横向对比 Cursor、Claude Code、Windsurf 与 GitHub Copilot，给出个人与团队选型建议。
国内 AI 编程助手横评：通义灵码、TRAE、文心快码、CodeGeeX 谁更适合你
面向国内网络与合规环境，对比通义灵码、字节 TRAE、百度文心快码、智谱 CodeGeeX 等在补全、Agent、中文文档与私有化上的差异，并给出选型清单。
BrowserOS – "Claude Cowork" in the browser
BrowserOS neo is an open-source secondary browser built for AI agents and human-AI coworking, featuring live session monitoring, one-click MCP connectivity, and browser automation.
Research acceleration: The view inside OpenAI
Research acceleration: The view inside OpenAI Apparently today is RSI day at OpenAI, for Recursive Self-Improvement - I think it's their new AGI. Both this...
海外大模型横评：GPT、Claude、Gemini、Llama 路线怎么配才划算
从编码、长文推理、多模态、工具调用与价格密度对比 OpenAI、Anthropic、Google 与开源 Llama 系，给出「主模型 + 备援 + 本地」的组合方案。
The purpose of DNS is to spread scams
The purpose of DNS is to spread scams Terence Eden shares some daunting statistics in support of his take that "the Domain Name System's purpose seems...
There's No Limit to How Bad Code Can Get
My comment on There's No Limit to How Bad Code Can Get — Lobste.rs. [...
Quoting Zach Kehs
If you continue to add floors and rooms to a building forever, it will collapse. Software faces no such constraint. The code can always get worse. There can...
打造高效可靠的 AI 编程工作流：从单轮提示到工程级上下文闭环
分享在真实企业研发场景中应用 AI 编码助手的心得与经验法则。探讨上下文膨胀治理、确定性自动化验证、规则文件分层治理以及人机协同边界。
Introducing GPT-6 Astra for developers
Introducing GPT-6 Astra for developers Blink and you'll miss it, but there's a familiar creature at 1m59s : ...
国产大模型横评：DeepSeek、通义千问、豆包、Kimi、智谱 GLM 怎么选
面向国内可用性与性价比，对比 DeepSeek、通义千问、豆包、Kimi、智谱 GLM 在推理、编码、长文本与产品形态上的差异，并给出个人与企业采购建议。