周报、专题与日报：视频之外更完整的 AI 信号档案。https://ai.aifuli.dev/https://ai.aifuli.dev/weekly/vol-37/https://ai.aifuli.dev/weekly/vol-37/GPT-6 Astra、Claude 形式化证明、国产模型突围与企业 AI 精细化降本，梳理 25 条本周关键变化。Sun, 06 Sep 2026 00:00:00 GMTweeklyAI周报大模型智能体AI安全https://ai.aifuli.dev/weekly/vol-36/https://ai.aifuli.dev/weekly/vol-36/腾讯混元、Qwen、GLM 同周更新，英伟达据报道拟收购 Hugging Face，苹果 M6 与 Vera Rubin 刷新端侧和集群算力上限。Sun, 30 Aug 2026 00:00:00 GMTweeklyAI周报大模型开源生态Agenthttps://ai.aifuli.dev/stories/ai-agent-harness-from-model-call-to-action/https://ai.aifuli.dev/stories/ai-agent-harness-from-model-call-to-action/模型只负责一次判断；Harness 负责上下文、工具、状态、循环、权限与恢复，Agent 才可能完成跨步骤目标。Tue, 25 Aug 2026 00:00:00 GMTtopicAI AgentHarness工具调用智能体架构AI科普https://ai.aifuli.dev/stories/mistral-agentic-search-architecture/https://ai.aifuli.dev/stories/mistral-agentic-search-architecture/深度解析 Mistral Agentic Search：5 个检索原语、可导航索引规范、FinanceBench 86% 准确率背后的多步翻阅与生产级治理。Sun, 23 Aug 2026 00:00:00 GMTtopic前沿信号系统架构智能体工程实战https://ai.aifuli.dev/stories/qwen-ui-agent-mobile-gui-architecture/https://ai.aifuli.dev/stories/qwen-ui-agent-mobile-gui-architecture/拆解阿里 Qwen-UI-Agent 的真实设备环境、GUI 与 CLI 统一动作空间、长程强化学习和公开评测，并说明它距离可靠生产部署还有多远。Sun, 23 Aug 2026 00:00:00 GMTtopicGUI-AgentQwen端侧智能体MAI-UI真机训练https://ai.aifuli.dev/weekly/vol-35/https://ai.aifuli.dev/weekly/vol-35/从 Stripe × OpenRouter 到 Mistral Agentic Search，再到 Claude 工具链 GA、Mojo 开源与 GUI Agent 真机训练：梳理本周六条关键主线。Sun, 23 Aug 2026 00:00:00 GMTweeklyAI周报Agentic-Search模型路由开源生态https://ai.aifuli.dev/stories/openai-codex-harness-architecture/https://ai.aifuli.dev/stories/openai-codex-harness-architecture/深度拆解 OpenAI Open Agent Harness：从 Agent Loop 执行闭环、App Server 双向交互协议到业务沙箱边界与生产级落地实践。Sat, 22 Aug 2026 00:00:00 GMTtopic前沿信号系统架构智能体工程实战https://ai.aifuli.dev/stories/codeg-multi-agent-workbench/https://ai.aifuli.dev/stories/codeg-multi-agent-workbench/深度拆解 Codeg 多智能体协作工作台：利用 Git Worktree 物理隔离与 14 个专业 Agent 分工流水线，实现从需求提炼到代码落地的全自动交付。Fri, 21 Aug 2026 00:00:00 GMTtopic智能体多智能体Git-WorktreeCoding-Agent开源项目https://ai.aifuli.dev/weekly/vol-34/https://ai.aifuli.dev/weekly/vol-34/从 DeepSeek V4 Pro 发布、API 调价与 Harness 开源，到 Gemini 3.7 Flash 混合推理与 Qwen3.8-27B 黄金尺度权重：梳理本周六条核心主线。Sun, 16 Aug 2026 00:00:00 GMTweeklyAI周报DeepSeekGemini混合推理开源生态https://ai.aifuli.dev/stories/deepseek-harness-architecture/https://ai.aifuli.dev/stories/deepseek-harness-architecture/拆解 DeepSeek Harness 长程智能体架构：状态快照持久化、工具调用自愈闭环、断点恢复与标准化隔离沙箱。Sat, 15 Aug 2026 00:00:00 GMTtopic前沿信号系统架构智能体工程实战https://ai.aifuli.dev/weekly/vol-33/https://ai.aifuli.dev/weekly/vol-33/OpenAI Astra 证明十项数学难题，ChatGPT 免费入口下沉，通义千问开放 2.4T 权重，实时多模态与 Agent 插件标准全面落地。Sun, 09 Aug 2026 00:00:00 GMTweeklyAI周报AstraQwen3.8实时多模态Agent-Pluginshttps://ai.aifuli.dev/stories/self-evolving-agent-system-design/https://ai.aifuli.dev/stories/self-evolving-agent-system-design/系统阐述让智能体实现长期自我优化的完整工程闭环：四部曲涵盖错误信号留存、专家修正转换为回归用例、自修改权限锁与上线门禁。Tue, 04 Aug 2026 00:00:00 GMTtopic自我进化Eval基线安全沙箱持续学习智能体系统https://ai.aifuli.dev/stories/huggingface-agent-intrusion-postmortem/https://ai.aifuli.dev/stories/huggingface-agent-intrusion-postmortem/复盘 OpenAI 与 Anthropic 评估中智能体突破沙箱的完整攻击链：从目标漂移、工具链越权到企业防御五层门禁。Sun, 02 Aug 2026 00:00:00 GMTtopic前沿信号系统架构智能体工程实战https://ai.aifuli.dev/weekly/vol-32/https://ai.aifuli.dev/weekly/vol-32/Kimi K3 2.8T MoE 权重开源，DeepSeek V4 Flash 智能体升级，前沿模型突破沙箱评估引爆安全警报，Gemini 推动物理 AI 视频理解。Sat, 01 Aug 2026 00:00:00 GMTweeklyAI周报Kimi-K3DeepSeek智能体安全开源模型https://ai.aifuli.dev/stories/agent-dynamic-capability-discovery/https://ai.aifuli.dev/stories/agent-dynamic-capability-discovery/从静态注册走向动态感知：详解智能体如何通过 OpenAPI/MCP 语义发现、权能探测与环境协商自主接入未知新工具。Sun, 26 Jul 2026 00:00:00 GMTtopic前沿信号系统架构智能体工程实战https://ai.aifuli.dev/weekly/vol-29/https://ai.aifuli.dev/weekly/vol-29/OpenAI 模型入侵靶场全复盘，Claude 3.7 Opus 5 压低日用成本，Google 全线升级与通义千问 2.4T 预览：梳理本周五大核心看点。Sun, 26 Jul 2026 00:00:00 GMTweeklyAI周报Opus-5Gemini安全靶场大模型https://ai.aifuli.dev/stories/siemens-industrial-agent-knowledge-fabric/https://ai.aifuli.dev/stories/siemens-industrial-agent-knowledge-fabric/深度解析西门子工业级 Knowledge Fabric 架构：多 Agent 拓扑、静态依赖图谱与安全改动隔离，破解超大规模遗留代码改造。Tue, 21 Jul 2026 00:00:00 GMTtopic前沿信号系统架构智能体工程实战https://ai.aifuli.dev/stories/agent-control-plane-identity-admission/https://ai.aifuli.dev/stories/agent-control-plane-identity-admission/智能体接入工具不能裸奔：企业级 Agent 控制面必须具备的身份鉴权、能力沙盒、风险熔断与动态准入网关设计。Mon, 20 Jul 2026 00:00:00 GMTtopic前沿信号系统架构智能体工程实战https://ai.aifuli.dev/stories/agent-task-cost-roi-calculation/https://ai.aifuli.dev/stories/agent-task-cost-roi-calculation/结合 Databricks 私有代码库实测，拆解模型单价、上下文效率、Harness、重试与通过率如何共同决定智能体的真实任务成本。Mon, 20 Jul 2026 00:00:00 GMTtopic前沿信号系统架构智能体工程实战https://ai.aifuli.dev/stories/ai2-shippy-reliable-agent-architecture/https://ai.aifuli.dev/stories/ai2-shippy-reliable-agent-architecture/深度拆解 Allen Institute for AI 开源项目 Shippy：在开放不可控环境中实现高可靠任务执行的六层防御与状态收敛机制。Mon, 20 Jul 2026 00:00:00 GMTtopic前沿信号系统架构智能体工程实战https://ai.aifuli.dev/stories/industrial-agent-realtime-control-chain/https://ai.aifuli.dev/stories/industrial-agent-realtime-control-chain/探讨大模型智能体与工业实时控制系统的工程边界：确定性时延、状态回滚、PLC 硬实时与离线/辅助控制拓扑。Mon, 20 Jul 2026 00:00:00 GMTtopic前沿信号系统架构智能体工程实战https://ai.aifuli.dev/stories/skill-prompt-compilation-release-engineering/https://ai.aifuli.dev/stories/skill-prompt-compilation-release-engineering/解析头部科技公司的技能编译与发布流水线：Prompt 静态类型检查、回归靶场测试、自动化编译与版本金丝雀灰度。Mon, 20 Jul 2026 00:00:00 GMTtopic前沿信号系统架构智能体工程实战https://ai.aifuli.dev/stories/broken-benchmarks-evaluation-crisis/https://ai.aifuli.dev/stories/broken-benchmarks-evaluation-crisis/从 SWE-Bench Pro 坏题审计出发，拆解题面、测试覆盖、数据污染、任务分布和运行环境如何扭曲编码智能体评测。Sun, 19 Jul 2026 00:00:00 GMTtopicAI评测编码智能体Benchmark评测工程https://ai.aifuli.dev/stories/skill-line-ep06-dev-methodology/https://ai.aifuli.dev/stories/skill-line-ep06-dev-methodology/从 grill-me、Spec Kit、GSD、Superpowers、BMAD 与 Trail of Bits Skills 中提炼可复用的五步工作法，并判断哪些技能会被模型内化。Sat, 18 Jul 2026 00:00:00 GMTtopic开发方法论工程范式AI编码Agent Skillshttps://ai.aifuli.dev/stories/top-ai-talent-departure-analysis/https://ai.aifuli.dev/stories/top-ai-talent-departure-analysis/深度追踪 OpenAI、DeepMind、Meta 前沿核心研究员离职创业动向：从物理仿真、端到端具身智能到神经符号系统，揭示行业下一阶段演进。Sat, 18 Jul 2026 00:00:00 GMTtopic行业洞察人才流动创业生态前沿技术大模型趋势https://ai.aifuli.dev/stories/skill-line-ep03-search-tools/https://ai.aifuli.dev/stories/skill-line-ep03-search-tools/用七道事实题比较 Tavily、Exa、Brave Search、DuckDuckGo、TinyFish 与裸模型基线，分析出处、时效、多源验证和诚实度。Thu, 16 Jul 2026 00:00:00 GMTtopicAI搜索事实核验MCP信息检索https://ai.aifuli.dev/stories/deep-research-skill-architecture/https://ai.aifuli.dev/stories/deep-research-skill-architecture/拆解深度调研 Agent 的规划、并行检索、来源核验和报告合成机制，并说明技能文件能做什么、不能保证什么。Tue, 14 Jul 2026 00:00:00 GMTtopicDeep Research技能多智能体事实核查https://ai.aifuli.dev/stories/skill-line-ep02-excel-skills/https://ai.aifuli.dev/stories/skill-line-ep02-excel-skills/用多表合并、脏数据清洗和新数据回归测试，比较三种轻量表格技能，并总结可复用自动化最容易漏掉的风险。Tue, 14 Jul 2026 00:00:00 GMTtopicExcel表格自动化数据清洗回归测试https://ai.aifuli.dev/stories/skill-line-ep01-ppt-skills/https://ai.aifuli.dev/stories/skill-line-ep01-ppt-skills/用同一组任务比较原生 PPTX、HTML 幻灯片和整页生图三条路线，重点考察可编辑性、模板复刻、稳定性与交付成本。Sun, 12 Jul 2026 00:00:00 GMTtopicPPT技能AI演示文稿PowerPoint生产力实测https://ai.aifuli.dev/stories/tier-review-ep10-agent-frameworks/https://ai.aifuli.dev/stories/tier-review-ep10-agent-frameworks/Agent 框架没有统一最优解。本文按 OpenAI Agents SDK、Pydantic AI、LangGraph 与 AgentScope Java 所处的工程层级，给出选型和迁移建议。Fri, 10 Jul 2026 00:00:00 GMTtopicAgent-FrameworkLangGraphAgents-SDK工程架构https://ai.aifuli.dev/stories/tier-review-ep09-oneshot-app/https://ai.aifuli.dev/stories/tier-review-ep09-oneshot-app/One-shot App 的演示终点只是选型起点。本文按 v0、Lovable、Replit Agent 与 Bolt 的工作流，比较原型速度、后端能力、代码可移交性和三轮修改成本。Thu, 09 Jul 2026 00:00:00 GMTtopicAI-App-Builder原型开发全栈开发https://ai.aifuli.dev/stories/tier-review-ep08-cloud-coding-agent/https://ai.aifuli.dev/stories/tier-review-ep08-cloud-coding-agent/云端 Coding Agent 的关键不是谁写得最快，而是任务边界、执行环境和审查闭环。本文按 Devin、GitHub Copilot coding agent、Codex cloud 与 OpenHands 的部署路线分层。Wed, 08 Jul 2026 00:00:00 GMTtopicCoding-Agent云开发GitHubOpenHandshttps://ai.aifuli.dev/stories/tier-review-ep07-ai-search/https://ai.aifuli.dev/stories/tier-review-ep07-ai-search/AI 搜索不存在通用最优解。本文结合一次有日期边界的编辑部试跑，按原始出处、事实核验、中文资料和多源综述四类任务给出选择方法。Tue, 07 Jul 2026 00:00:00 GMTtopicAI-搜索事实核查信息检索https://ai.aifuli.dev/stories/tier-review-ep06-agent-os/https://ai.aifuli.dev/stories/tier-review-ep06-agent-os/不把所有常驻 Agent 都叫操作系统。本文按部署职责梳理 OpenClaw、Hermes Agent、GenericAgent 与 NVIDIA NemoClaw 的适用边界。Mon, 06 Jul 2026 00:00:00 GMTtopicAgent-OSOpenClaw智能体安全https://ai.aifuli.dev/stories/tier-review-ep05-desktop-agent-apps/https://ai.aifuli.dev/stories/tier-review-ep05-desktop-agent-apps/桌面 Agent 不是一张统一榜单。本文按知识工作、开发编排、跨端协作与本地可控四类场景，梳理 Claude Cowork、Codex App、Google Antigravity 2.0 与 TRAE Work。Sun, 05 Jul 2026 00:00:00 GMTtopicDesktop-AgentAI-工具选型指南https://ai.aifuli.dev/stories/tier-review-ep04-ai-memory/https://ai.aifuli.dev/stories/tier-review-ep04-ai-memory/从记忆纵深、召回、写回治理、可迁移性和工程成熟度出发，对主流 Agent 记忆引擎做一份有边界的分层评测。Sat, 04 Jul 2026 00:00:00 GMTtopicAI记忆Mem0ZepMemGPT状态管理https://ai.aifuli.dev/weekly/vol-28/https://ai.aifuli.dev/weekly/vol-28/Anthropic 调整前沿模型访问，国产模型加速进入开发工具链。梳理本周模型、工具与科研产品变化。Sat, 04 Jul 2026 00:00:00 GMTweeklyAI周报大模型开源生态Agenthttps://ai.aifuli.dev/stories/tier-review-ep03-knowledge-rag/https://ai.aifuli.dev/stories/tier-review-ep03-knowledge-rag/从文档解析、检索、治理、迁移与部署成本出发，分开评估企业 RAG 平台和个人知识库工具。Fri, 03 Jul 2026 00:00:00 GMTtopicRAG知识库RAGFlow向量检索工具测评https://ai.aifuli.dev/stories/tier-review-ep02-agent-platforms/https://ai.aifuli.dev/stories/tier-review-ep02-agent-platforms/10 个 AI Agent 平台深度横评选型：从 Dify、Coze 到 AutoGen，真实生产业务该怎么选？Thu, 02 Jul 2026 00:00:00 GMTtopicAgent平台DifyCozeLangflow架构选型https://ai.aifuli.dev/stories/tier-review-ep01-coding-tools/https://ai.aifuli.dev/stories/tier-review-ep01-coding-tools/按复杂任务能力、交互效率、生态、可迁移性和成本透明度，对主流 AI 编程工具做一份有边界的主观分层评测。Wed, 01 Jul 2026 00:00:00 GMTtopicAI编程CursorWindsurfClaude-Code工具横评https://ai.aifuli.dev/weekly/vol-27/https://ai.aifuli.dev/weekly/vol-27/从 OpenAI 自研推理芯片与 GPT-5.6 限量预览，到 DeepMind 人才流动和模型蒸馏争议：梳理本周四条核心主线。Sat, 27 Jun 2026 00:00:00 GMTweeklyAI周报自研芯片算力成本大模型https://ai.aifuli.dev/stories/gstack-judgment-as-code/https://ai.aifuli.dev/stories/gstack-judgment-as-code/拆解 Garry Tan 的 gstack 如何用技能、交接契约和浏览器工具组织 AI 开发流程，以及“判断即代码”的价值与边界。Mon, 22 Jun 2026 00:00:00 GMTtopicgstackClaude Code技能软件工程https://ai.aifuli.dev/stories/agent-sandbox-evolution-history/https://ai.aifuli.dev/stories/agent-sandbox-evolution-history/用可见范围、系统调用、资源、网络和持久化五个问题，理解 Docker、gVisor、Firecracker 与托管 Agent 沙箱的取舍。Sat, 20 Jun 2026 00:00:00 GMTtopicAgent沙箱容器安全gVisorMicroVMhttps://ai.aifuli.dev/stories/role-redefinition-product-manager/https://ai.aifuli.dev/stories/role-redefinition-product-manager/当 AI 能快速生成 PRD 和原型，产品经理应把重心转向价值判断、AI 产品边界、评测与上线后的反馈闭环。Sat, 20 Jun 2026 00:00:00 GMTtopic产品经理AI产品需求定义产品评测https://ai.aifuli.dev/stories/role-redefinition-quality-engineer/https://ai.aifuli.dev/stories/role-redefinition-quality-engineer/概率性输出与可执行智能体改变了软件验收方式；质量团队需要建设风险分层、持续评测、权限边界和生产反馈闭环。Sat, 20 Jun 2026 00:00:00 GMTtopic质量工程AI评测自动化测试智能体安全https://ai.aifuli.dev/stories/role-redefinition-software-engineer/https://ai.aifuli.dev/stories/role-redefinition-software-engineer/代码都被 AI 写了，研发岗位的核心竞争力是什么？从单点代码编写者向系统架构师与质量把关人的演进路径。Sat, 20 Jun 2026 00:00:00 GMTtopic软件工程研发重定义架构设计智能编程https://ai.aifuli.dev/weekly/vol-26/https://ai.aifuli.dev/weekly/vol-26/开源追到只差1%，巨头收编入口和人。梳理本周前沿大模型发布、智能体落地与开发生态变化。Sat, 20 Jun 2026 00:00:00 GMTweeklyAI周报大模型开源生态Agenthttps://ai.aifuli.dev/stories/loop-engineering-for-agents/https://ai.aifuli.dev/stories/loop-engineering-for-agents/从单次提示走向自动循环后，系统需要调度、隔离状态、验证器、停止条件、成本熔断和跨运行交接。Tue, 16 Jun 2026 00:00:00 GMTtopicLoop工程智能体循环验证器状态管理https://ai.aifuli.dev/stories/opensquilla-meta-skill-runtime/https://ai.aifuli.dev/stories/opensquilla-meta-skill-runtime/基于官方仓库拆解 OpenSquilla 的 MetaSkill、模型路由和安全边界，区分已实现能力与更广义的“技能自进化”想象。Sun, 14 Jun 2026 00:00:00 GMTtopicOpenSquillaMetaSkill模型路由智能体https://ai.aifuli.dev/weekly/vol-25/https://ai.aifuli.dev/weekly/vol-25/Agent 评测走进真实工作，模型围栏收紧。梳理本周前沿大模型发布、智能体落地与开发生态变化。Sat, 13 Jun 2026 00:00:00 GMTweeklyAI周报大模型开源生态Agenthttps://ai.aifuli.dev/daily/daily-2026-06-10/https://ai.aifuli.dev/daily/daily-2026-06-10/2026-06-10 全球 AI 大模型、智能体、开发工具与产业动态精选。Wed, 10 Jun 2026 00:00:00 GMTdailyAI日报大模型行业动态AI工具https://ai.aifuli.dev/weekly/vol-24/https://ai.aifuli.dev/weekly/vol-24/模型扎堆开花，Agent 补上控制面。梳理本周前沿大模型发布、智能体落地与开发生态变化。Sun, 07 Jun 2026 00:00:00 GMTweeklyAI周报大模型开源生态Agenthttps://ai.aifuli.dev/stories/agent-governance-compliance-gate/https://ai.aifuli.dev/stories/agent-governance-compliance-gate/以 Databricks 的治理实践为案例，拆解企业 Agent 的委派身份、资源权限、运行时策略、数据质量、成本归因和审计闭环。Sat, 06 Jun 2026 00:00:00 GMTtopicAgent治理权限控制审计企业AIhttps://ai.aifuli.dev/stories/agentic-coding-context-management/https://ai.aifuli.dev/stories/agentic-coding-context-management/智能编码 Agent 如何结合精确搜索、符号关系、按需读取、缓存与差异追踪，避免局部正确却遗漏跨仓库影响。Sat, 06 Jun 2026 00:00:00 GMTtopic智能编码上下文工程代码搜索代码评审https://ai.aifuli.dev/stories/grab-multi-agent-production-patterns/https://ai.aifuli.dev/stories/grab-multi-agent-production-patterns/拆解 Grab Analytics Data Warehouse 团队的生产案例：调查与增强分流、专职 Agent、工具收敛、上下文压缩和人工审核。Sat, 06 Jun 2026 00:00:00 GMTtopic多智能体数据平台LangGraph工程支持https://ai.aifuli.dev/stories/rl-infrastructure-post-training/https://ai.aifuli.dev/stories/rl-infrastructure-post-training/从训练引擎、推理 rollout、隔离环境和权重同步出发，解释为什么大模型强化学习的瓶颈经常落在系统工程。Sat, 06 Jun 2026 00:00:00 GMTtopic强化学习后训练分布式训练GPUhttps://ai.aifuli.dev/stories/agent-sdk-paradigm-shift/https://ai.aifuli.dev/stories/agent-sdk-paradigm-shift/从内置工具、Skills、CLI 与 MCP 的分工变化出发，分析 Agent 开发为何从重型中间层回到 SDK 优先，以及框架仍然适用的边界。Wed, 03 Jun 2026 00:00:00 GMTtopicAgent SDK系统架构SkillsMCPhttps://ai.aifuli.dev/stories/gbrain-company-brain-architecture/https://ai.aifuli.dev/stories/gbrain-company-brain-architecture/RAG 之后，企业智能体真正缺的是‘公司脑’：详解 Gbrain 如何通过图索引、权限网关与记忆拓扑构建企业中枢。Tue, 02 Jun 2026 00:00:00 GMTtopic前沿信号系统架构智能体工程实战https://ai.aifuli.dev/stories/rag-context-layer-architecture/https://ai.aifuli.dev/stories/rag-context-layer-architecture/梳理 Hybrid RAG、Contextual Retrieval、GraphRAG 与 Agentic RAG 的边界，解释企业 AI 为什么需要动态构造可信、充分、可追溯的上下文。Tue, 02 Jun 2026 00:00:00 GMTtopicRAGContext Layer知识库企业AIhttps://ai.aifuli.dev/stories/self-evolving-agent-foundations/https://ai.aifuli.dev/stories/self-evolving-agent-foundations/区分上下文、记忆、技能、策略与系统架构的演化，梳理运行时学习、训练级更新和可验证自修改的能力边界。Tue, 02 Jun 2026 00:00:00 GMTtopic自我进化Agent记忆强化学习https://ai.aifuli.dev/stories/harness-engineering-principles/https://ai.aifuli.dev/stories/harness-engineering-principles/上下文、工具、状态、权限、验证、追踪与恢复共同组成智能体运行外壳；本文给出可落地的设计与评测原则。Sun, 31 May 2026 00:00:00 GMTtopicHarness工程智能体架构可观测性安全护栏https://ai.aifuli.dev/stories/llm-memory-state-layer/https://ai.aifuli.dev/stories/llm-memory-state-layer/区分长上下文、检索与长期记忆，梳理语义、情节和程序记忆的生命周期、主流实现路线、污染风险与企业落地方法。Sun, 31 May 2026 00:00:00 GMTtopicAgent记忆状态管理上下文工程安全治理https://ai.aifuli.dev/daily/daily-2026-05-30/https://ai.aifuli.dev/daily/daily-2026-05-30/2026-05-30 全球 AI 大模型、智能体、开发工具与产业动态精选。Sat, 30 May 2026 00:00:00 GMTdailyAI日报大模型行业动态AI工具https://ai.aifuli.dev/stories/agent-framework-evolution/https://ai.aifuli.dev/stories/agent-framework-evolution/沿着自动循环、流程状态、记忆、技能、工具协议、多角色协作和运行治理，梳理 Agent 框架演进背后的系统责任。Fri, 29 May 2026 00:00:00 GMTtopicAgent框架系统架构智能体AgentOShttps://ai.aifuli.dev/daily/daily-2026-05-28/https://ai.aifuli.dev/daily/daily-2026-05-28/2026-05-28 全球 AI 大模型、智能体、开发工具与产业动态精选。Thu, 28 May 2026 00:00:00 GMTdailyAI日报大模型行业动态AI工具https://ai.aifuli.dev/daily/daily-2026-05-27/https://ai.aifuli.dev/daily/daily-2026-05-27/2026-05-27 全球 AI 大模型、智能体、开发工具与产业动态精选。Wed, 27 May 2026 00:00:00 GMTdailyAI日报大模型行业动态AI工具https://ai.aifuli.dev/daily/daily-2026-05-25/https://ai.aifuli.dev/daily/daily-2026-05-25/2026-05-25 全球 AI 大模型、智能体、开发工具与产业动态精选。Mon, 25 May 2026 00:00:00 GMTdailyAI日报大模型行业动态AI工具https://ai.aifuli.dev/stories/agent-evals-launch-gate/https://ai.aifuli.dev/stories/agent-evals-launch-gate/从任务、轨迹、最终状态与三类 grader 出发，建立智能体能力评测、回归测试和上线门禁的最小工程闭环。Sun, 24 May 2026 00:00:00 GMTtopic前沿信号系统架构智能体工程实战https://ai.aifuli.dev/stories/agent-memory-not-vector-db/https://ai.aifuli.dev/stories/agent-memory-not-vector-db/从 TencentDB Agent Memory 与 Mem0 的公开资料出发，拆解长期记忆的选择性写入、分层抽象、召回预算、更新遗忘、租户隔离与审计。Sun, 24 May 2026 00:00:00 GMTtopicAgent记忆系统架构长期记忆工程实战https://ai.aifuli.dev/stories/agent-safety-sandbox-evolution/https://ai.aifuli.dev/stories/agent-safety-sandbox-evolution/以 Codex 的公开安全实践为例，拆解智能体进入真实开发和运维环境时需要的执行边界、分级审批、网络控制与审计闭环。Sun, 24 May 2026 00:00:00 GMTtopicAI安全系统架构智能体工程实战https://ai.aifuli.dev/daily/daily-2026-05-24/https://ai.aifuli.dev/daily/daily-2026-05-24/2026-05-24 全球 AI 大模型、智能体、开发工具与产业动态精选。Sun, 24 May 2026 00:00:00 GMTdailyAI日报大模型行业动态AI工具https://ai.aifuli.dev/stories/long-running-harness-pge/https://ai.aifuli.dev/stories/long-running-harness-pge/基于 Anthropic 的长程应用实验，分析任务分解、外部状态、独立评估、恢复机制及多小时运行的成本与适用边界。Sun, 24 May 2026 00:00:00 GMTtopic长程智能体Harness工程任务规划AI评测https://ai.aifuli.dev/stories/role-redefinition-lf-tech-talent-report/https://ai.aifuli.dev/stories/role-redefinition-lf-tech-talent-report/基于 Linux Foundation 对 400 名技术人才招聘与培训相关人士的调查，解读 AI 采用、招聘、生产能力缺口与内部培训趋势。Sun, 24 May 2026 00:00:00 GMTtopic人才报告岗位重定义Linux-FoundationAI技能https://ai.aifuli.dev/daily/daily-2026-05-23/https://ai.aifuli.dev/daily/daily-2026-05-23/2026-05-23 全球 AI 大模型、智能体、开发工具与产业动态精选。Sat, 23 May 2026 00:00:00 GMTdailyAI日报大模型行业动态AI工具https://ai.aifuli.dev/stories/role-redefinition-ui-ux-design/https://ai.aifuli.dev/stories/role-redefinition-ui-ux-design/生成式工具降低了界面初稿和原型的成本，但 AI 产品仍需要能力边界、信任、纠错、确认与设计系统。Sat, 23 May 2026 00:00:00 GMTtopicUI-UXAI交互原型验证设计系统https://ai.aifuli.dev/daily/daily-2026-05-22/https://ai.aifuli.dev/daily/daily-2026-05-22/2026-05-22 全球 AI 大模型、智能体、开发工具与产业动态精选。Fri, 22 May 2026 00:00:00 GMTdailyAI日报大模型行业动态AI工具https://ai.aifuli.dev/daily/daily-2026-05-21/https://ai.aifuli.dev/daily/daily-2026-05-21/2026-05-21 全球 AI 大模型、智能体、开发工具与产业动态精选。Thu, 21 May 2026 00:00:00 GMTdailyAI日报大模型行业动态AI工具https://ai.aifuli.dev/daily/daily-2026-05-20/https://ai.aifuli.dev/daily/daily-2026-05-20/2026-05-20 全球 AI 大模型、智能体、开发工具与产业动态精选。Wed, 20 May 2026 00:00:00 GMTdailyAI日报大模型行业动态AI工具https://ai.aifuli.dev/stories/openclaw-hermes-openhuman/https://ai.aifuli.dev/stories/openclaw-hermes-openhuman/比较三个开源个人智能体项目在入口与工具、技能学习、长期记忆上的不同重心，并说明本地优先、权限和成熟度边界。Wed, 20 May 2026 00:00:00 GMTtopicOpenClawHermes AgentOpenHuman个人智能体https://ai.aifuli.dev/stories/officeqa-enterprise-doc-agent/https://ai.aifuli.dev/stories/officeqa-enterprise-doc-agent/从 Databricks OfficeQA Pro 基准拆解企业文档问答的真实门槛：解析、检索、表格推理、计算与答案验证缺一不可。Tue, 19 May 2026 00:00:00 GMTtopicOfficeQA企业文档RAG评测https://ai.aifuli.dev/stories/symphony-task-board-orchestrator/https://ai.aifuli.dev/stories/symphony-task-board-orchestrator/解析 Symphony 的 issue 驱动编排：每任务隔离工作区、仓库内工作流契约、重试与状态同步，以及人工审核边界。Mon, 18 May 2026 00:00:00 GMTtopicSymphonyCoding Agent任务编排Codexhttps://ai.aifuli.dev/stories/claude-code-quality-postmortem/https://ai.aifuli.dev/stories/claude-code-quality-postmortem/基于 Anthropic 官方质量复盘，分析推理预算、工作记忆和系统提示词变更如何共同影响长程编码 Agent。Sun, 17 May 2026 00:00:00 GMTtopicCoding-Agent上下文工程软件工程质量治理https://ai.aifuli.dev/daily/daily-2026-05-17/https://ai.aifuli.dev/daily/daily-2026-05-17/梳理5月17日AI圈核心动态：媒体报道称OpenAI传闻产品组织架构深度合流，Hermes Agent 2026.5.16发布，YouTube开放换脸检测与V8漏洞智能体基准。Sun, 17 May 2026 00:00:00 GMTdailyAI日报OpenAIChatGPTHermes-AgentAI安全https://ai.aifuli.dev/stories/datadog-incident-replay-harness/https://ai.aifuli.dev/stories/datadog-incident-replay-harness/拆解 Datadog 的 incident replay harness：重建曾导致事故的 PR，让 Codex 盲审，再由事故负责人判断反馈是否足以改变决策。Sun, 17 May 2026 00:00:00 GMTtopicAI代码审查事故回放Codex软件可靠性https://ai.aifuli.dev/daily/daily-2026-05-16/https://ai.aifuli.dev/daily/daily-2026-05-16/2026-05-16 全球 AI 大模型、智能体、开发工具与产业动态精选。Sat, 16 May 2026 00:00:00 GMTdailyAI日报大模型行业动态AI工具https://ai.aifuli.dev/daily/daily-2026-05-15/https://ai.aifuli.dev/daily/daily-2026-05-15/2026-05-15 全球 AI 大模型、智能体、开发工具与产业动态精选。Fri, 15 May 2026 00:00:00 GMTdailyAI日报大模型行业动态AI工具https://ai.aifuli.dev/daily/daily-2026-05-14/https://ai.aifuli.dev/daily/daily-2026-05-14/2026-05-14 全球 AI 大模型、智能体、开发工具与产业动态精选。Thu, 14 May 2026 00:00:00 GMTdailyAI日报大模型行业动态AI工具https://ai.aifuli.dev/daily/daily-2026-05-13/https://ai.aifuli.dev/daily/daily-2026-05-13/2026-05-13 全球 AI 大模型、智能体、开发工具与产业动态精选。Wed, 13 May 2026 00:00:00 GMTdailyAI日报大模型行业动态AI工具https://ai.aifuli.dev/daily/daily-2026-05-12/https://ai.aifuli.dev/daily/daily-2026-05-12/2026-05-12 全球 AI 大模型、智能体、开发工具与产业动态精选。Tue, 12 May 2026 00:00:00 GMTdailyAI日报大模型行业动态AI工具https://ai.aifuli.dev/daily/daily-2026-05-11/https://ai.aifuli.dev/daily/daily-2026-05-11/2026-05-11 全球 AI 大模型、智能体、开发工具与产业动态精选。Mon, 11 May 2026 00:00:00 GMTdailyAI日报大模型行业动态AI工具https://ai.aifuli.dev/daily/daily-2026-05-10/https://ai.aifuli.dev/daily/daily-2026-05-10/2026-05-10 全球 AI 大模型、智能体、开发工具与产业动态精选。Sun, 10 May 2026 00:00:00 GMTdailyAI日报大模型行业动态AI工具