<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>AI Agent on Peng Tan's AI Blog</title><link>https://77f3dc70.hobbytp-github-io.pages.dev/zh/tags/ai-agent/</link><description>一个关注 AI 各领域的专题博客</description><atom:link href="https://77f3dc70.hobbytp-github-io.pages.dev/zh/tags/ai-agent/index.xml" rel="self" type="application/rss+xml"/><item><title>NVIDIA AVO 满分之后：AI 的基本单位正在从模型变成“持续学习系统”</title><link>https://77f3dc70.hobbytp-github-io.pages.dev/zh/my_insights/nvidia_avo_arc_agi_3/</link><pubDate>Sun, 23 Aug 2026 00:00:00 +0800</pubDate><guid>https://77f3dc70.hobbytp-github-io.pages.dev/zh/my_insights/nvidia_avo_arc_agi_3/</guid><description>NVIDIA AVO 在 ARC-AGI-3 公开演示集取得 100 分，真正值得关注的并不是又一个满分，而是 Harness 如何把短时推理转化为跨上下文持续积累的系统能力。本文拆解 AVO、RHAE 的计量盲区，以及模型 Scaling、测试时计算与系统时扩展之间的新边界。</description></item><item><title>DeepSeek Harness 真正冲击的不是 Coding Agent，而是“模型能力”这笔账</title><link>https://77f3dc70.hobbytp-github-io.pages.dev/zh/my_insights/deepseek_harness/</link><pubDate>Thu, 20 Aug 2026 10:00:00 +0800</pubDate><guid>https://77f3dc70.hobbytp-github-io.pages.dev/zh/my_insights/deepseek_harness/</guid><description>DeepSeek Harness 的真正价值，不是再造一个 Coding Agent，而是把模型、工具、上下文、沙箱、Agent Loop 与运行轨迹变成可声明、可比较的工程变量，迫使行业重新思考 Agent 能力究竟应该如何归因。</description></item><item><title>Agent OS 的诞生：从 MCP、A2UI 到 A2A，开放智能体栈的协议拼图渐露端倪</title><link>https://77f3dc70.hobbytp-github-io.pages.dev/zh/my_insights/agent_os_protocols/</link><pubDate>Wed, 07 Jan 2026 23:00:00 +0800</pubDate><guid>https://77f3dc70.hobbytp-github-io.pages.dev/zh/my_insights/agent_os_protocols/</guid><description>深度解析2026年Agent协议生态。我们正从单体智能走向群体智能，但谁将定义未来的标准？本文将从计算机解剖学视角，深度拆解 MCP、A2UI、A2A 以及记忆与优化机制如何构成开放智能体栈（The Open Agent Stack），并探讨其与封闭生态的终局博弈。</description></item><item><title>G2 2025年AI代理洞察报告：和MIT报告相反的关于AI代理的5个惊人发现</title><link>https://77f3dc70.hobbytp-github-io.pages.dev/zh/reports/g2insightreport_aiagent_2025/</link><pubDate>Thu, 09 Oct 2025 23:03:00 +0800</pubDate><guid>https://77f3dc70.hobbytp-github-io.pages.dev/zh/reports/g2insightreport_aiagent_2025/</guid><description>G2 Insight Report: AI Agents in 2025</description></item><item><title>Nanobot 深度剖析：如何用 4000 行代码构建超轻量级个人 AI 助理</title><link>https://77f3dc70.hobbytp-github-io.pages.dev/zh/projects/auto/nanobot/</link><pubDate>Fri, 15 Mar 2024 00:00:00 +0000</pubDate><guid>https://77f3dc70.hobbytp-github-io.pages.dev/zh/projects/auto/nanobot/</guid><description>Nanobot 是一个超轻量级的个人 AI 助理框架，仅用 4000 行代码实现了 ReAct 循环、多渠道消息总线、工具/技能系统、长短期记忆、子智能体协同等功能。可以看成是 OpenClaw 的极简版，是一个非常好的学习项目。</description></item></channel></rss>