记录 AI 模型实测、Agent 用法、编程与工具折腾的个人博客。https://ohmyself.com/zh-CNhttps://ohmyself.com/blog/how-to-eval-two-models/https://ohmyself.com/blog/how-to-eval-two-models/榜单跑分和你的活不是一回事。这是我给 Opus 5 和 Fable 5 出 8 道题的完整方法：判分怎么机械化、主观题怎么盲评、以及我把题出砸了的那部分。Sun, 26 Jul 2026 11:00:00 GMTAI模型评测方法论https://ohmyself.com/blog/blog-seo-audit/https://ohmyself.com/blog/blog-seo-audit/建站几天后回头看，最该修的不是关键词，是分享出去没有图、和 12MB 的封面。附一份可以照抄的检查清单，和两个我自己踩到的坑。Sun, 26 Jul 2026 10:30:00 GMTAISEOAstro踩坑https://ohmyself.com/blog/ai-agent-beginner-guide/https://ohmyself.com/blog/ai-agent-beginner-guide/聊天是你动手，Agent 是它动手。八件今天就能让它替你做完的事，外加五个让它真正靠谱的技巧。Sun, 26 Jul 2026 10:00:00 GMTAIAgent入门https://ohmyself.com/blog/claude-opus5-vs-fable5/https://ohmyself.com/blog/claude-opus5-vs-fable5/用完全相同的提示词让 Opus 5 和 Fable 5 跑 8 项测试：数学、逻辑、算法、找 bug、指令遵循、信息提取、决策备忘录、Agentic ETL。总分 52.63 vs 53.25，差距比价格差小得多。Sun, 26 Jul 2026 09:20:00 GMTAIClaude模型评测https://ohmyself.com/blog/gpt-5-6-juice-values/https://ohmyself.com/blog/gpt-5-6-juice-values/GPT-5.6 Sol、Terra、Luna 各推理档位的 Juice 值测试记录。Thu, 23 Jul 2026 02:00:00 GMTAIGPT-5.6模型评测https://ohmyself.com/blog/codex-install-china/https://ohmyself.com/blog/codex-install-china/官网 403、下载超时、安装器报错后，我用第三方安装器装上了 Codex App。Wed, 22 Jul 2026 07:00:00 GMTAI踩坑教程https://ohmyself.com/blog/what-is-agents-md/https://ohmyself.com/blog/what-is-agents-md/一个 Markdown 文件，让 AI 助手从「通用聊天机器人」变成「懂你项目的队友」。Wed, 22 Jul 2026 06:00:00 GMTAI工程实践效率https://ohmyself.com/blog/qoderwork-guide/https://ohmyself.com/blog/qoderwork-guide/QoderWork 是什么、能干什么、怎么跟它说话效率最高。附一批可以直接抄的提示词。Wed, 22 Jul 2026 03:00:00 GMTAI工具效率https://ohmyself.com/blog/ai-one-sentence-blog/https://ohmyself.com/blog/ai-one-sentence-blog/从一句建站需求到网站上线，这次我几乎没碰配置。Wed, 22 Jul 2026 02:30:00 GMTAI技术cloudflarehttps://ohmyself.com/blog/hello-world/https://ohmyself.com/blog/hello-world/The obligatory first post. Why I started this blog and what to expect.Wed, 22 Jul 2026 01:00:00 GMTmeta