Agent Engineer
Building autonomous agents that code, reason, and operate computers.
Coding Agent · GUI Agent · Computer Use · Human–AI Collaboration.
Projects
What I'm Building
Autonomous agents, developer tools, and experiments — each one a step toward machines that can truly reason, code, and act.
Thinking
Explorations & Ideas
Notes on coding agents, GUI agents, computer use, and the path toward AGI.
Computer Use Agent 技术分析报告
从 API 到 GUI 的范式跃迁：深入分析 Computer Use Agent 的核心技术架构、OpenAI Codex 实现、能力边界与 AI IDE 集成路径。
Computer Use Agent：当 AI 学会操作你的电脑
从 cua-driver 出发，解析 Computer Use Agent 的核心机制：AI 如何感知桌面 UI、执行操作，以及这对 AI 编程工具意味着什么。
Google I/O 2026 特别速报与深度整理
基于 2026 年 5 月 19 日 Google I/O 开幕 Keynote，深度整理 Gemini 3.5 系列大模型、Antigravity 工具链拆分、会员体系调整及应用功能演进。
认识 Composer 2.5：Cursor 最新的 Agent 编程模型
介绍 Cursor 自研 Agent 编程模型 Composer 2.5：定位、核心能力、训练技术画像、计费与系列位置。
Journey
Evolution
From frontend engineering to AI IDE — a path driven by curiosity and shipping.
Agent Engineer @ Ant Group
Building CodeFuse's autonomous coding agents — agentic modes (Plan / Agent / Spec), harness engineering for code-gen, and GUI Agent / Computer Use exploration.
Frontend Architect & AI Engineer @ INTSIG
Designed multi-stage AI Agent workflows with Dify + RAG, built streaming AI interaction SDK, and led the international 3D brand website rebuild with Nuxt3 + Three.js.
Frontend Engineer @ SI-TECH
Led frontend development for China Unicom's enterprise platform. Hybrid architecture, performance optimization, and mobile-first responsive design.