https://blog.gusibi.site/tags/ai/Recent content in AI on 古思乱想Hugozh-cnMon, 03 Aug 2026 00:27:59 +0800https://blog.gusibi.site/post/agent-runtime-objects/Mon, 03 Aug 2026 00:27:59 +0800https://blog.gusibi.site/post/agent-runtime-objects/拆分 Session、Run、ToolCall、Artifact 等 Agent runtime 对象，建立清晰边界。https://blog.gusibi.site/post/agent-run-object/Mon, 27 Jul 2026 22:21:51 +0800https://blog.gusibi.site/post/agent-run-object/用 Run 对象把一次 Agent 执行过程建模清楚，避免 Agent 只是 Chatbot 套壳。https://blog.gusibi.site/post/quickdash-english/Fri, 24 Jul 2026 00:18:10 +0800https://blog.gusibi.site/post/quickdash-english/一个利用编码等待间隙练习英语的 macOS 工具 BestLearn。https://blog.gusibi.site/post/chatbot-to-agent-service/Sun, 19 Jul 2026 17:10:36 +0800https://blog.gusibi.site/post/chatbot-to-agent-service/从 Chatbot、Tool-using Chatbot 到 Agent Service 和 Agent System，理解差别为何在运行时。https://blog.gusibi.site/post/agent-service-vs-demo/Fri, 10 Jul 2026 11:27:12 +0800https://blog.gusibi.site/post/agent-service-vs-demo/从 Chatbot 到 Agent Service，再到 Agent System 的工程边界与课程地图。https://blog.gusibi.site/post/loop-engineering/Tue, 23 Jun 2026 23:51:02 +0800https://blog.gusibi.site/post/loop-engineering/Loop Engineering 的核心不是继续提示 Agent，而是设计一个能持续运行的工程循环。https://blog.gusibi.site/post/agent-service-overview/Tue, 23 Jun 2026 23:13:07 +0800https://blog.gusibi.site/post/agent-service-overview/从真实系统拆解 Agent 服务从 Demo 走向上线所需要的运行时、工具、状态和权限边界。https://blog.gusibi.site/post/matt-pocock-skills-workflow/Fri, 19 Jun 2026 18:34:23 +0800https://blog.gusibi.site/post/matt-pocock-skills-workflow/从 Matt Pocock 的 skills v1 出发，拆解一套从想法到交付的 AI 工程工作流。https://blog.gusibi.site/post/anthropic-sandbox-runtime-agent/Sat, 13 Jun 2026 14:28:55 +0800https://blog.gusibi.site/post/anthropic-sandbox-runtime-agent/把 Anthropic sandbox-runtime 接入 Agent 后，对执行边界、审批恢复与多渠道运行时的实践思考。https://blog.gusibi.site/post/ai-agent-sandbox-safety/Fri, 12 Jun 2026 15:41:22 +0800https://blog.gusibi.site/post/ai-agent-sandbox-safety/AI Agent 具备执行命令的能力后，如何通过沙箱建立清晰的安全边界。https://blog.gusibi.site/post/claude-code-dev-tips-en/Fri, 12 Jun 2026 15:41:22 +0800https://blog.gusibi.site/post/claude-code-dev-tips-en/I'm Boris and I created Claude Code. I wanted to quickly share a few tips for using Claude Code, sourced directly from the Claude Code team.https://blog.gusibi.site/post/agent-sandbox-product/Fri, 12 Jun 2026 15:41:22 +0800https://blog.gusibi.site/post/agent-sandbox-product/在真实 Agent 产品中落地 Sandbox、权限审批与安全配置的全过程。https://blog.gusibi.site/post/agent-sandbox-runtime/Mon, 11 May 2026 12:13:27 +0800https://blog.gusibi.site/post/agent-sandbox-runtime/深度拆解主流 Agent 沙箱方案与 Anthropic Sandbox Runtime。https://blog.gusibi.site/post/llm-training-overview/Mon, 27 Apr 2026 11:04:22 +0800https://blog.gusibi.site/post/llm-training-overview/***https://blog.gusibi.site/post/obsidian-llm-wiki-plugin/Tue, 07 Apr 2026 21:05:00 +0800https://blog.gusibi.site/post/obsidian-llm-wiki-plugin/前几天，Karpathy 发了一条推。https://blog.gusibi.site/post/ai-glossary/Mon, 23 Mar 2026 00:00:00 +0800https://blog.gusibi.site/post/ai-glossary/你最近肯定听过很多 AI 词汇。你大概知道它们是什么意思，但可能又没那么确定。https://blog.gusibi.site/post/agent-tool-call-engineering/Mon, 09 Mar 2026 14:30:00 +0800https://blog.gusibi.site/post/agent-tool-call-engineering/在使用大模型的过程中，你肯定发现目前的大语言模型（LLM）在逻辑推理、代码编写和文本生成方面表现优异。但是，它们都有一个共同的局限：无法直接干预外部世界。https://blog.gusibi.site/post/minimal-agent-impl/Fri, 06 Mar 2026 10:19:31 +0800https://blog.gusibi.site/post/minimal-agent-impl/这两年大家一提 Agent，脑子里很容易浮现出一个“会思考、会规划、会调用工具、还会自我修复”的高级智能体。听起来很玄，但如果把那些花哨概念都剥掉，一个能跑起来的极简 Agent，其实没有那么复杂。https://blog.gusibi.site/post/coding-planning-pricing/Wed, 25 Feb 2026 09:11:23 +0800https://blog.gusibi.site/post/coding-planning-pricing/| 厂商 | 计划名称 | API类型/功能 | 定价模式https://blog.gusibi.site/post/cloudflare-worker-gemini-api-proxy/Thu, 15 Feb 2024 16:51:26 +0800https://blog.gusibi.site/post/cloudflare-worker-gemini-api-proxy/一、在 Cloudflare 中创建一个 worker gemini-api-proxy，保存并部署https://blog.gusibi.site/post/coze-gpt4-api/Thu, 15 Feb 2024 16:26:07 +0800https://blog.gusibi.site/post/coze-gpt4-api/1. 打开 [discord开发者平台](https://discord.com/developers/applications) 。