Tday: All-in-One Harness Agent Terminal unbug.github.io/tday/ ; CODELF (Github star 14k); MIHTool (Mentioned in Google I/O'13)
Joined January 2008
- Tweets 3,147
- Following 715
- Followers 1,424
- Likes 220
332 Photos and videos
我不会下载也不会测，我打赌肯定拉，你们测完就会扔，快没什么阮用，Nvidia Nemetron3 系列最快，谁在用？
Jun 10
DiffusionGemma is an open, experimental model that brings our text diffusion research to Gemma 4. It’s a racehorse 🏇achieving up to 4x faster inference by generating entire blocks of text simultaneously vs predicting token-by-token (word-by-word) output!
75
要想你的 Agent 变得更好，就要舍得给他们花钱😂
一分钟读论文：《思想的经济：Agent经济交互中的多智能体智能涌现》 | Micropaper unbug.github.io/one-minute-r…
57
MTP on MoE is not stable, I am counting on Dflash github.com/ggml-org/llama.cp…
130
果然是好东西！上幼儿园中班的娃对着 Scratch Copilot 口喷，Vibe Coding 开发各种游戏自娱自乐！
有了 AI，中学学历的家长也能带娃一起学编程，编程是成本最低最有效的逻辑思维和计算思维启蒙工具。我三岁多带娃学到四岁，每月搭积木 编程，娃变化肉眼可见，不仅理解数学和规律，拼积木卡住会说“对齐”，逛超市找不到路会说“先规划”，平日沟通很顺，5岁放手他自会自学东西。
unbug.github.io/Scratch-Copi…
246
好个大厂割小厂韭菜的一个愿打一个愿挨时代。
AI 范式雷达：《裁员换 AI，为何成本长期上升、利润反而负增长》 | Micropaper unbug.github.io/paradigm-rad…
117
“We’ve gone from ‘move fast and break things’ to ‘move fast and ship things nobody asked for.’”
AI 范式雷达：《AI Coding 时代，功能膨胀如何杀死好产品》 | Micropaper unbug.github.io/paradigm-rad…
80
Holy wild! OpenClaw Qwen3.6-35B-a3b harness without my engagement for 3 weeks. Release 160 version and 300 features. If my wife not turned off my laptop it still keeps going through. So never underestimate a local model! See the project yourself unbug.github.io/monday/
本地大模型有多能打？你可以持续关注这个让你可以在浏览器里一键运行小模型的开源项目: 全 Harness 自主迭代，0人工干预。一流的设计、流畅的功能、适配移动端。一周已迭代到 v0.4，按 Roadmap 估计一个月到 v1.0。(@lmstudio @openclaw @Alibaba_Qwen Qwen3.6-35b-a3b) unbug.github.io/monday/
260
AI 范式雷达：《AI 正在让开源协议失去约束力》 | Micropaper unbug.github.io/paradigm-rad…
127
有了 AI，中学学历的家长也能带娃一起学编程，编程是成本最低最有效的逻辑思维和计算思维启蒙工具。我三岁多带娃学到四岁，每月搭积木 编程，娃变化肉眼可见，不仅理解数学和规律，拼积木卡住会说“对齐”，逛超市找不到路会说“先规划”，平日沟通很顺，5岁放手他自会自学东西。
unbug.github.io/Scratch-Copi…
1
1
1
603
I’ve been thinking about a breaking engagement approach with LLM: UILess. So much performance potential we’ll benefit from it, because agents work together by carrying less context which supposed to explain for humans, that is key matters.
Ok i think I found a terminal rendering system for the ds4-agent that is a bit different than the normal route, from the POV of 1. simplicity. 2. not taking control of the terminal, so when the session ends the history is still all there. This is how it works (thread):
1
1
437
虽然已经有很多开源的一流 Coding Agent 支持DeepSeek，但是我还是打心里期待官方的版本，甚至不期待有什么质的飞跃。在 AI 时代赢家通吃就是彻头彻尾的永动机。
May 20
🚀 We’re hiring! DeepSeek is forming a new Harness team to build Code Harness from the ground up—may be you can call it DeepSeek Code or something like this hhh🤣🤣🤣
📍 Based in Beijing. Two roles open:
🧠 Harness Product Manager → app.mokahr.com/social-recrui…
👨💻 Harness R&D Engineer → app.mokahr.com/social-recrui…
Research meets product—let's build it together. Hit the links and apply directly! 🔥
#DeepSeek #CodeHarness #AI #Hiring #ProductManager #Engineering #Beijing #Referral
1
432