<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Obelisk Design Bloooooog</title><description>从原理到实践，深入探索技术与设计</description><link>https://obelisk-design.github.io/</link><item><title>AI Agent 共享记忆：从孤岛到协作的架构演进</title><link>https://obelisk-design.github.io/blog/ai-agent-shared-memory/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/ai-agent-shared-memory/</guid><description>深入剖析 AI Agent 记忆系统的设计挑战，从短期记忆困境到共享上下文的工程实践，探索多 Agent 协作的下一代基础设施</description><pubDate>Fri, 29 May 2026 00:00:00 GMT</pubDate></item><item><title>AI Agent Tool Calling：从函数调用到智能编排的深度解析</title><link>https://obelisk-design.github.io/blog/ai-agent-tool-calling/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/ai-agent-tool-calling/</guid><description>深入剖析 AI Agent 工具调用机制的设计原理、实现模式与最佳实践，探索从简单函数调用到复杂工具编排的技术演进</description><pubDate>Fri, 29 May 2026 00:00:00 GMT</pubDate></item><item><title>AI Agent Workflow Orchestration：构建智能工作流编排系统的深度指南</title><link>https://obelisk-design.github.io/blog/ai-agent-workflow-orchestration/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/ai-agent-workflow-orchestration/</guid><description>深入探讨 AI Agent 工作流编排的核心概念、架构模式与实战实践，从单一 Agent 到多 Agent 协作的完整技术路径</description><pubDate>Fri, 29 May 2026 00:00:00 GMT</pubDate></item><item><title>从 Prompt 到 Harness：AI Coding 的工程治理实践</title><link>https://obelisk-design.github.io/blog/ai-coding-harness-practice/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/ai-coding-harness-practice/</guid><description>如何让 AI 从&quot;会写代码&quot;进化到&quot;持续稳定地写出符合项目规范的代码&quot;——Skills、DSL 中间层、三角验证与 Development Harness 的完整实践路径。</description><pubDate>Tue, 20 May 2025 00:00:00 GMT</pubDate></item><item><title>Claude Code 动态工作流：10-100 个并行子 Agent 的工程革命</title><link>https://obelisk-design.github.io/blog/claude-code-dynamic-workflows/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/claude-code-dynamic-workflows/</guid><description>Anthropic 推出 Claude Code 动态工作流，单个会话运行数十到数百个并行子 Agent，自动验证后再交付。Bun 从 Zig 移植到 Rust 仅用 11 天，75 万行代码，99.8% 测试通过率。</description><pubDate>Fri, 29 May 2026 00:00:00 GMT</pubDate></item><item><title>Claude Skills 深度指南：从原理到实战的完整方法论</title><link>https://obelisk-design.github.io/blog/claude-skills-deep-dive/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/claude-skills-deep-dive/</guid><description>解密 Claude Skills 的渐进式披露机制、目录架构设计、frontmatter 优化策略，以及评估驱动的迭代方法论</description><pubDate>Fri, 29 May 2026 00:00:00 GMT</pubDate></item><item><title>用域名邮箱构建隐形护城河：Cloudflare 邮件路由完全指南</title><link>https://obelisk-design.github.io/blog/cloudflare-email-routing-guide/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/cloudflare-email-routing-guide/</guid><description>用一个 Gmail 注册所有账号是单点故障。本文教你用 Cloudflare 免费邮件路由搭建专属域名邮箱，实现账号隔离、无限注册、品牌背书和数据主权。</description><pubDate>Thu, 28 May 2026 00:00:00 GMT</pubDate></item><item><title>离谱的英语学习指南：从苦行僧到热爱驱动的方法论重构</title><link>https://obelisk-design.github.io/blog/english-learning-guide-methodology/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/english-learning-guide-methodology/</guid><description>英语学习不该是痛苦的征程。本文从认知科学、学习金字塔、艾宾浩斯曲线出发，重构一套热爱驱动的英语学习方法论。</description><pubDate>Thu, 28 May 2026 00:00:00 GMT</pubDate></item><item><title>Harness Engineering：从 Prompt 到 Agent 环境的范式转移</title><link>https://obelisk-design.github.io/blog/harness-engineering-paradigm-shift/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/harness-engineering-paradigm-shift/</guid><description>为什么 Agents 难的不是模型而是环境？深入解析三阶段演进与业界实践</description><pubDate>Wed, 20 May 2026 00:00:00 GMT</pubDate></item><item><title>Spec 写好了怎么落地？Harness Engineering 实战</title><link>https://obelisk-design.github.io/blog/harness-engineering-practice/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/harness-engineering-practice/</guid><description>Harness Engineering 是知识沉淀的自动化，让 AI 在正确的约束下执行，人在正确的时机介入。五层架构 + 五个关键时刻 + 24周落地路线。</description><pubDate>Thu, 21 May 2026 00:00:00 GMT</pubDate></item><item><title>Hermes + Polymarket：搭建自学习 BTC 交易代理的完整指南</title><link>https://obelisk-design.github.io/blog/hermes-polymarket-self-learning-trading-agent/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/hermes-polymarket-self-learning-trading-agent/</guid><description>从 100 美元起步，用 Hermes + Atomic + Claude Opus 搭建能持续复盘、持续调参、具备自学习能力的 BTC 交易代理。</description><pubDate>Thu, 28 May 2026 00:00:00 GMT</pubDate></item><item><title>Intel Mac 抹掉启动磁盘完全指南：从原理到实践的终极解析</title><link>https://obelisk-design.github.io/blog/intel-mac-erase-guide/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/intel-mac-erase-guide/</guid><description>不只是点几下鼠标——深入理解 Mac 磁盘抹掉的底层机制、APFS 文件系统、恢复模式架构，以及 Intel 与 Apple Silicon 的本质区别</description><pubDate>Fri, 29 May 2026 00:00:00 GMT</pubDate></item><item><title>从 LLM 到 Agent：理解 AI 智能体的底层逻辑</title><link>https://obelisk-design.github.io/blog/llm-to-agent-fundamentals/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/llm-to-agent-fundamentals/</guid><description>深入剖析 LLM 的本质机制，从 Token、Prompt、Context 到 Scaling Law 与涌现，理解为什么&quot;预测下一个词&quot;的模型能推理、写代码、做 Agent</description><pubDate>Fri, 22 May 2026 00:00:00 GMT</pubDate></item><item><title>Matt Pocock 的 AI Agent Skills：工程师的实战武器库</title><link>https://obelisk-design.github.io/blog/matt-pocock-ai-agent-skills/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/matt-pocock-ai-agent-skills/</guid><description>深入解析 10 万 Star 的 AI 编程技能库，从入门到精通</description><pubDate>Wed, 20 May 2026 00:00:00 GMT</pubDate></item><item><title>Matt Pocock 的 Agent Skills 系统：用工程纪律驯服 AI 编程助手</title><link>https://obelisk-design.github.io/blog/matt-pocock-ai-agent-skills-deep-dive/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/matt-pocock-ai-agent-skills-deep-dive/</guid><description>从沟通对齐到测试驱动，一套可组合的 AI Agent 技能系统如何解决 AI 编程的真实失败模式。</description><pubDate>Tue, 02 Jun 2026 00:00:00 GMT</pubDate></item><item><title>MCP 协议深度解析：AI 时代的 USB-C，如何重塑模型与世界的连接方式</title><link>https://obelisk-design.github.io/blog/mcp-protocol-deep-dive/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/mcp-protocol-deep-dive/</guid><description>从 M×N 集成困局到标准化协议，深入剖析 Model Context Protocol 的设计哲学、核心规范、工程实践与未来演进</description><pubDate>Thu, 28 May 2026 00:00:00 GMT</pubDate></item><item><title>中台蜕变：从代码供应商到 AI 积木供应商</title><link>https://obelisk-design.github.io/blog/middle-platform-ai-building-blocks/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/middle-platform-ai-building-blocks/</guid><description>把中台改造成 AI Agent 的积木库，做三件事：积木说明书（Tool Description）+ 拼装规则（约束声明化）+ 参考图纸（样板间），让搭积木这个十年愿景第一次真正成立</description><pubDate>Wed, 20 May 2026 00:00:00 GMT</pubDate></item><item><title>OpenCLI：把任意网站变成 CLI，让 AI Agent 操作你的浏览器</title><link>https://obelisk-design.github.io/blog/opencli-browser-to-cli/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/opencli-browser-to-cli/</guid><description>OpenCLI 是一个强大的工具，可以将任何网站转换为 CLI 命令，并让 AI Agent 通过你已登录的浏览器进行自动化操作。支持 100+ 网站适配器。</description><pubDate>Tue, 19 May 2026 00:00:00 GMT</pubDate></item><item><title>OpenCLI：把任意网站变成 CLI —— 让 AI Agent 拥有「命令行级」的互联网操作能力</title><link>https://obelisk-design.github.io/blog/opencli-website-to-cli-tool-deep-dive/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/opencli-website-to-cli-tool-deep-dive/</guid><description>从 Chrome 扩展到站点适配器自修复，OpenCLI 如何用工程化手段解决 AI 操控网页的不确定性问题。</description><pubDate>Tue, 02 Jun 2026 00:00:00 GMT</pubDate></item><item><title>OpenSpec 入门：AI 时代的规范驱动开发</title><link>https://obelisk-design.github.io/blog/openspec-spec-driven-development/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/openspec-spec-driven-development/</guid><description>AI 编码助手指令模糊时会靠猜，OpenSpec 通过规范驱动开发（SDD）让 AI 先搞清楚要做什么，再动手写代码。</description><pubDate>Thu, 21 May 2026 00:00:00 GMT</pubDate></item><item><title>TikTok 不露脸内容工厂：从 0 到月入 $3000 的 AI 自动化流水线全解析</title><link>https://obelisk-design.github.io/blog/tiktok-faceless-pipeline/</link><guid isPermaLink="true">https://obelisk-design.github.io/blog/tiktok-faceless-pipeline/</guid><description>深度拆解一套 TikTok Faceless 自动化内容生产流水线——从赛道选择、AI 工具链编排、数据驱动调优到变现策略，89 天跑通的实战 SOP 与技术架构。</description><pubDate>Tue, 02 Jun 2026 00:00:00 GMT</pubDate></item></channel></rss>