Tensor

@Tensor1992

Alumni @BayesBSchool. LLM Researcher, Senior Software Engineer, Ex-VCer/PM.

Joined November 2011

295Posts 53Followers 587Following

You might like

@zhanhr

@marilynjanehorn

Tensor reposted

Eason Mao☢

@KELMAND1

Nov 18

外交部介绍中日外交官会晤情况：中方再次向日方提出严正交涉 11月18日，外交部发言人毛宁主持例行记者会。日本NHK记者提问，据报道，日本外务省亚洲大洋洲局局长金井正彰与中国外交部亚洲司司长刘劲松举行了磋商，能否介绍磋商的具体内容？…

Eason Mao☢

@KELMAND1

Nov 18

名场面：“那年我双手插兜，不知道什么是对手” 11月18日，中国外交部亚洲司司长刘劲松与日本外务省亚洲大洋洲局局长金井正彰会面。下午约两点，金井正彰一行离开中国外交部。

李老师不是你老师

@whyyoutouzhele

Nov 17

11月17日，综合多家报道，台湾总统赖清德上午出席国家档案馆开幕典礼，媒体询问对于中日紧张局势的看法，赖清德表示，中国对日本进行复合式攻击，严重冲击印太和平稳定。他呼吁国际社会持续关注，也呼吁中国应该要克制，展现大国的风范，不要成为区域和平稳定的麻烦制造者，应该回到以规则为基础的国际…

whyyoutouzhele's tweet image. 11月17日，综合多家报道，台湾总统赖清德上午出席国家档案馆开幕典礼，媒体询问对于中日紧张局势的看法，赖清德表示，中国对日本进行复合式攻击，严重冲击印太和平稳定。他呼吁国际社会持续关注，也呼吁中国应该要克制，展现大国的风范，不要成为区域和平稳定的麻烦制造者，应该回到以规则为基础的国际…

Tensor

@Tensor1992

Oct 30

CURSOR 自研模型不仅仅在于他本身，如果cursor 自研模型超过Claude code，AI应用逻辑就能完美闭环了.，能打破了ai应用形成不了数据飞轮的枷锁。

歸藏(guizang.ai)

@op7418

Oct 29

Cursor 发布 2.0 重大升级，看来这次想一扫颓势了包含自己训练的编码模型 Composer，以及支持 Agent 并发的新界面两部分👇 首先是他们自己训练的代码模型 Composer，匿名内测时反响不错： - 速度比同类型模型快 4 倍 - 专为中低延迟的智能体编码而构建 - 大多数轮次可在 30 秒内完成 - 使用 MXFP8…

op7418's tweet image. Cursor 发布 2.0 重大升级，看来这次想一扫颓势了

包含自己训练的编码模型 Composer，以及支持 Agent 并发的新界面两部分👇

首先是他们自己训练的代码模型 Composer，匿名内测时反响不错：

- 速度比同类型模型快 4 倍
- 专为中低延迟的智能体编码而构建
- 大多数轮次可在 30 秒内完成
- 使用 MXFP8…

Tensor reposted

Cursor

@cursor_ai

Oct 29

Introducing Cursor 2.0. Our first coding model and the best way to code with agents.

Tensor reposted

Huge breakthrough from DeepMind! In their latest Nature paper, “Discovering state-of-the-art reinforcement learning algorithms,” they show that AI can autonomously discover better RL algorithms. "Enabling machines to discover learning algorithms for themselves is one of the…

jiqizhixin's tweet image. Huge breakthrough from DeepMind!

In their latest Nature paper, “Discovering state-of-the-art reinforcement learning algorithms,” they show that AI can autonomously discover better RL algorithms.

"Enabling machines to discover learning algorithms for themselves is one of the…

Tensor reposted

GitHubDaily

@GitHub_Daily

Oct 25

有位开发者，基于 DeepSeek-OCR 模型实现了一个简单实用的 OCR 前端应用：deepSeek_ocr_app，项目已开源。提供简洁直观的操作界面，只需要拖拽上传图片，即可开始处理，最大支持 100MB 文件。支持四种核心 OCR…

GitHub_Daily's tweet image. 有位开发者，基于 DeepSeek-OCR 模型实现了一个简单实用的 OCR 前端应用：deepSeek_ocr_app，项目已开源。

提供简洁直观的操作界面，只需要拖拽上传图片，即可开始处理，最大支持 100MB 文件。

支持四种核心 OCR…

Tensor

@Tensor1992

Oct 25

机器狗是那个背包的小朋友的，不是我的。一代人有一代人的信仰！过去，我妈：衣服手洗比洗衣机干净现在，传统程序员：代码手搓比AI写得好未来，铲屎官：真狗比电子宠物狗更有灵性.

Tensor reposted

Nathan Lambert

@natolambert

Oct 16

The first fantastic paper on scaling RL with LLMs just dropped. I strongly recommend taking a look and will be sharing more thoughts on the blog soon. The Art of Scaling Reinforcement Learning Compute for LLMs Khatri & Madaan et al.

natolambert's tweet image. The first fantastic paper on scaling RL with LLMs just dropped. I strongly recommend taking a look and will be sharing more thoughts on the blog soon.

The Art of Scaling Reinforcement Learning Compute for LLMs
Khatri &amp; Madaan et al.

Tensor reposted

Robert Youssef

@rryssf_

Oct 15

Holy shit... Tencent researchers just killed fine-tuning AND reinforcement learning in one shot 😳 They call it Training-Free GRPO (Group Relative Policy Optimization). Instead of updating weights, the model literally learns from 'its own experiences' like an evolving memory…

rryssf_'s tweet image. Holy shit... Tencent researchers just killed fine-tuning AND reinforcement learning in one shot 😳

They call it Training-Free GRPO (Group Relative Policy Optimization).

Instead of updating weights, the model literally learns from 'its own experiences' like an evolving memory…

Tensor reposted

ginobefun

@hongming731

Oct 14

打算再细看一遍回头再总结视频地址：youtube.com/watch?v=6_BcCt… BestBlogs AI 转录文字稿：bestblogs.dev/video/087a1f3 PPT 地址：docs.google.com/presentation/d… drive.google.com/file/d/1QGJ-Br…

Tensor reposted

Alex Prompter

@alex_prompter

Oct 13

Holy shit. MIT just built an AI that can rewrite its own code to get smarter 🤯 It’s called SEAL (Self-Adapting Language Models). Instead of humans fine-tuning it, SEAL reads new info, rewrites it in its own words, and runs gradient updates on itself literally performing…

alex_prompter's tweet image. Holy shit. MIT just built an AI that can rewrite its own code to get smarter 🤯

It’s called SEAL (Self-Adapting Language Models).

Instead of humans fine-tuning it, SEAL reads new info, rewrites it in its own words, and runs gradient updates on itself literally performing…

Tensor reposted

Rohan Paul

@rohanpaul_ai

Oct 11

New paper from @Google is a major memory breakthrough for AI agents. ReasoningBank helps an AI agent improve during use by learning from its wins and mistakes. To succeed in real-world settings, LLM agents must stop making the same mistakes. ReasoningBank memory framework…

rohanpaul_ai's tweet image. New paper from @Google is a major memory breakthrough for AI agents.

ReasoningBank helps an AI agent improve during use by learning from its wins and mistakes.

To succeed in real-world settings, LLM agents must stop making the same mistakes. ReasoningBank memory framework…

Tensor

@Tensor1992

Oct 10

double standards

The Wall Street Journal

@WSJ

Oct 10

China’s newest restrictions on rare-earth materials would mark a nearly unprecedented export control that stands to disrupt the global economy on.wsj.com/42xove5

WSJ's tweet card. President Trump says the administration is weighing a U.S. response to Beijing’s newest restrictions.

China’s Rare-Earth Escalation Threatens Trade Talks—and the Global Economy

Source: wsj.com

Tensor reposted

Robert Youssef

@rryssf_

Oct 9

RIP fine-tuning ☠️ This new Stanford paper just killed it. It’s called 'Agentic Context Engineering (ACE)' and it proves you can make models smarter without touching a single weight. Instead of retraining, ACE evolves the context itself. The model writes, reflects, and edits…

rryssf_'s tweet image. RIP fine-tuning ☠️

This new Stanford paper just killed it.

It’s called 'Agentic Context Engineering (ACE)' and it proves you can make models smarter without touching a single weight.

Instead of retraining, ACE evolves the context itself.

The model writes, reflects, and edits…

Tensor reposted

Kevin Ma

@kevinma_dev_zh

Sep 24

推荐两个曾经颠覆了我的英语学习认知的视频，我的英语学习理论就是来自于这里。我现在听力和阅读基本上没有障碍，我的女儿们不到两岁，也能听懂很多单词和短语。我敢说这是我在 B 站上发现的最宝藏的视频，强烈推荐你多看几遍。链接放评论区 👇

Tensor

@Tensor1992

Sep 17

有一说一，说的很中肯。

Commentary: Trump Truth Social Posts On X

@TrumpTruthOnX

Sep 16

Donald J. Trump ReTruthed Donald J. Trump: > When Foreign Companies who are building extremely complex products, machines, and various other “things,” come into the United States with massive Investments, I want them to bring their people of expertise for a period of time to…