技术笔记
从真实问题出发，把设计、实现、评测和复盘写清楚。
这里不追求工具清单，而是记录四类技术如何解决真实问题：Agent 如何行动，机器人如何闭环，论文如何形成证据链，AI 提效如何减少工作中不必要的阻力。
This is not a tool directory. It records how four areas solve real problems: how agents act, how robots close the loop, how papers become evidence chains, and how AI productivity removes avoidable friction.
Agent 系统
Agent Systems
工具调用、规划、记忆、评测与失败恢复，关注从演示走向可交付系统。
Tool use, planning, memory, evaluation and recovery, moving from demos to systems that can ship.
机器人开发
Robotics
感知、控制、软硬件接口、真机调试，以及真实世界里的延迟和异常。
Perception, control, hardware and software interfaces, real-robot debugging, latency and failure handling.
论文阅读
Paper Reading
围绕研究问题、实验证据、失败边界和可迁移机制，建立可以持续比较的论文笔记。
Evidence-led notes that compare research questions, experiments, limitations and transferable mechanisms.
AI 提效
AI Productivity
围绕真实使用场景、实际收益和维护成本，整理可以长期复用的 AI 提效方法。
Practical AI methods evaluated by real use, measurable gains and maintenance cost.