🧩 DeepSeek Harness v0.1 is now available in Developer Preview!

🔹 We’re opening it up to developers building agent harnesses worldwide and open-sourcing the codebase in MIT license.
🔹 Powered by the Cordis meta-framework, DeepSeek Harness is an agent

ds harness 来啦来啦

最近我让 Kimi K3 用 go 语言重构了我的博客系统 Rables,原先是用 Rails 写的,内存占用至少 500 MB 起,重构后,只有 16 MB 左右,不要太爽了

https://github.com/versun/rables

tweet-2085639284552540500-a8f53fd0.pngtweet-2085639284552540500-2ab4ad16.png

完成重构了,效果不错,比我预想中的要快很多,总共消耗了5个 swarm+goal+max,累计90轮左右,累计消耗 1.3B token

tweet-2087819397251690746-b885285e.pngtweet-2087819397251690746-0be19fcb.png

Old news actually from a bunch of days ago, but crossed that 15M. Enjoy a nice reset everyone. Landing in the next hour or so, go /fast.

看来昨晚的 ds 和 grok 还是给到 tipo一些压力的😜

Kimi 最近是不是卡到货了,明显稳定耐用了很多

我一个早上9点20开始的 swarm+goal+max 会话,到现在竟然还没断,已经进行到第11轮了

在以前,我还没遇到过能超过8轮的,而且中午也没有被429打断

tweet-2087457668457984009-cb2003d6.jpg

Mojo 🔥 1.0 is here.

Since we open-sourced the standard library, nearly 200 contributors have landed more than 1+ pull requests, changing 200k+ lines of code, and more than 1k+ others have filed issues that shaped the language.

To every developer w

想不到 Mojo 还活着,简单回顾了下:

- 2023.5 出道,放话比 Python 快 3.5 万倍

- 2025 年从"语言"转向"推理基础设施",推出 MAX 推理框架

- 2026.6 公司被高通 39 亿美元收购了。。

- 磨了三年到 1.0,承诺秋天开源编译器

四年半,3.8 亿融资进,39 亿收购出,也算大成功了

We need to rethink APIs for agents rather than humans.

@insforge has been experimenting with a better agent API.

Key changes:

- Status states

- Detailed context on error (and success too)

- Clear next steps

tweet-2077422545142534275-0b04cea4.jpg

API for agent,有意思

Here’s how the 4 new harnesses scored:

- Pi Agent passed 20 of 30 tasks.

- Deep Agents (by LangChain) passed 16 of 30 tasks.

- Hermes Agent passed 15 of 30 tasks.

- Prime Agent passed 15 of its 24 valid runs (6 runs were excluded; more on that la

tweet-2086814491920974069-7228c7aa.jpg

Pi 打败了Oh My Pi 🤣

赞美 @orca_build 的一天!

我之前就在想,如果能用 Fable5 来做需求的拆解、技术架构的设计,再用 GLM 或者 Kimi 负责执行,让 Codex 去测试 debug,最后由 Fable5 验收,那该有多好,质量与速度都能兼顾。

其实我早就发现了 ORCA 的 orchestration 能力,但一直苦于没有时间尝试。最近糊了一个 orchestration 的 viewer 来展示编排完成后的任务,终于能把这个想法落地了。

感谢 ORCA 拯救了我的 Fable5 额度

这个有意思,就是安装麻烦了些,建议简化下安装方式