-
我参加了 4 月 28 日至 29 日在旧金山举行的 [AI 开发者大会 (AI Developer Conference)](https://ai-dev.deeplearning.ai/)，并想在此记录我的笔记。这两天活动的主题是新生成式 AI（GenAI）时代的软件开发，特别是关注编码智能体（Coding Agents）以及软件工程的未来。
-
I attended the [AI Developer Conference in San Francisco](https://ai-dev.deeplearning.ai/) from April 28 to 29, and I wanted to capture my notes here. The main theme of the two-day event was software development in the new GenAI age, specifically focusing on coding agents and the future of software engineering.
-
最近三周，我一直在尝试 OpenClaw 这个开源 Agent 框架。这是一篇基于我在‘跨行业俱乐部’分享内容的整理记录，主要聊了聊我如何利用家里的旧设备搭建本地 Agent 服务器，以及在自动化新闻汇总和代码开发方面的实际使用感受。文中也记录了一些关于 API 成本和安全性设置的个人思考，分享给对 Agent 落地感兴趣的朋友。
-
For the past three weeks, I have been experimenting with OpenClaw, an open-source Agent framework. This post is a record based on my sharing session at the "Cross-Industry Club," primarily discussing how I utilized old home devices to build a local Agent server, as well as my actual experience with automated news aggregation and code development. I have also included some personal reflections on API costs and security settings for those interested in implementing Agents.
-
过去一周，我部署并测试了开源 AI 个人助手框架 OpenClaw。我没有租用云端服务器，而是想看看在本地运行一个持久、自主的 AI 实体究竟需要什么条件。 这个过程证明，虽然本地智能体的部署还有很多不完善之处，但其核心架构是切实可行的。简单来说，就是让大语言模型（LLM）进入一个持久循环：执行外部工具，并不断修改自己的配置状态。以下是我的硬件配置、调试过程，以及把这个智能体“放养”到纯机器人社交网络后的观察。
-
Over the past week, I deployed and tested OpenClaw, an open-source AI personal assistant framework. Rather than renting cloud infrastructure, I wanted to see what it takes to run a persistent, autonomous entity locally. The process proved that while we are still navigating the rough edges of local agent deployments, the underlying architecture—a persistent loop where an LLM executes external tool calls and modifies its own configuration state—is functional. Here is a technical breakdown of my infrastructure setup, the debugging process, and what happened when I let the agent loose in a bot-only social network.