Ciallo ~(∠・ω< )⌒★!
- XnneHang's Blog
- Intern at NevaMind-AI and interested in Long-Term Memory.
- I like reading📚, writing✏️ and animes📺!
- I love all interesting things and want to try to do it.
- I want to bring waifus into real world.
Markdown 14 hrs 49 mins ██████████████████▓░░░░░░ 74.28 %
Python 2 hrs 21 mins ███░░░░░░░░░░░░░░░░░░░░░░ 11.79 %
Other 56 mins █▒░░░░░░░░░░░░░░░░░░░░░░░ 04.71 %
JSON 43 mins █░░░░░░░░░░░░░░░░░░░░░░░░ 03.67 %
Text 37 mins ▓░░░░░░░░░░░░░░░░░░░░░░░░ 03.13 %Recent Blog Posts
![]() | 协议的形状决定能力:codex 为什么砍掉 chat_completion,以及桌宠的「边说边做」到底卡在哪 从 codex 硬删 chat/completions 的全过程,追到 new-api 被波及的迁移损伤,再拆开 Anthropic Messages 与 OpenAI 协议在 Tool 流上的真实差别——为什么我的长链静默不是协议的锅,以及什么才算真的「边说边做」。 2026-07-28 |
![]() | 以坏架构为鉴:Claude Desktop 五套定时系统反映出来的 cowork 和 code 分离架构的弊端 拆解 Claude Desktop 的五套定时系统,剖析混乱背后的及架构本质,为什么花 N+ 倍精力维护一个项目。以及一个坏的架构能让我们看到什么? 2026-07-23 |
![]() | 了解 RRF 和 Hybrid Search:基于时间的检索应该以什么形态混合,以及它在项目场景和陪伴场景的适配性 MoeChat 衍生讨论 RRF 和 Hybrid Search:时间检索应该以什么姿态参与混合检索?在不同场景里如何适配?以及 wikimem 和 XnneHangLab 的边界划分。 2026-07-21 |
Here is my cat🐾 powered by openclaw.









