Curry

As you can see, I'm Hein. A CG animation enthusiast knows your fate and our luck. To develop my communication in the world. We know you among 7 billion people.

[比特币橙子Trader @oragnes]

老黄其实早就把 AI 的投资逻辑讲透了。

我把这套逻辑拆成了 5 层：

Layer 1 能源
没有电，就没有 AI
发电、核能、电网改造、储能，都是 AI 扩张的底层支撑

Layer 2 芯片
GPU、晶圆制造、半导体设备、光通信、存储
这是 AI 的“心脏”

Layer 3 基础设施
数据中心、服务器、液冷、电力配套、算力租赁，这是 AI 规模化落地的中间层

Layer 4 模型
谁在训练、部署、运营大模型
谁就掌握 AI 的核心能力

Layer 5 应用
最终赚大钱的，可能不是模型本身
而是把 AI 变成真实收入的软件和应用公司

[Curry @233admin]

好东西

[Jackywine @Jackywine]

今天绝对不能错过 Anthropic 这篇博客

运营一个人工智能原生组织

https://t.co/ciu2Y3511j

顺便问一下大家有无更好的翻译模型

[Dep @0xDepressionn]

https://t.co/DGep0M16LP

[monokern @monokern]

https://t.co/RmlXgqYeJg

[Thariq @trq212]

If this prompt feels well written to you, it's because Suzanne is a writer in her little spare time!

You can read her short story, Mall of America here: https://t.co/uYY8VlPIUt

It's one of my favorite short stories about the human condition that happens to involve AI.

[Fei-Fei Li @drfeifei]

https://t.co/Kt50ttQRMJ

[QuiverAI @QuiverAI]

Visual AI is shifting from pixels to code.

With Arrow 1.1, every illustration is generated as editable SVG source code, creating outputs that can be refined, adapted, and integrated into real workflows.

[D4rk_Intel @d4rk_intel]

My article "How To Investigate A Person Of Interest In 2026" is now available as a PDF.

A practical guide to digital footprint analysis – from email reconstruction to metadata mining and entity graphing.

Thanks @osintnewsletter for the mention.

PDF: https://t.co/86YlE9e2pB

[Roan @RohOnChain]

As someone who builds institutional level quant systems, this Stanford paper on Market Making is the closest thing to an HFT desk I have ever seen publicly shared.

19 pages. Hedge Fund level Market Making Algorithm. Bookmark & get this before someone takes it down.

[Yanhua @yanhua1010]

https://t.co/HclIjm2pEC

[aditya @adxtyahq]

for anyone asking where to learn this stuff:

• RAG → https://t.co/4bzbUIwV5g
• Agentic RAG → https://t.co/IotOiGmV1Y
• AI Agents → https://t.co/nEeMnVJQbk
• Multi-Agent Systems → https://t.co/pavDPVJEFj
• LangGraph → https://t.co/3miEqqFzF0
• LangGraph (code) → https://t.co/v7kxHZXqba
• MCP → https://t.co/lKawRb4etX
• Memory Systems → https://t.co/LSaT2UaPAS
• Evals → https://t.co/vxChxa1kqQ
• Context Engineering → search "Context Engineering Survey" on arXiv

and please skip the "build an ai agent in 10 minutes" videos
build something, watch it fail, then figure out why.

[Leo｜一个人 + AI @runes_leo]

这条原文讲到我最近刚好踩到的一个坑。

我最近在 Codex App 里跑通了跨对话协作：

一个主 thread 推进任务。

发现某一步该交给另一条业务线，
就写一份 handoff 文件，
再用短期定时任务把目标 thread 唤醒。

目标 thread 醒来后，
读文件，
干活，
写回状态。

这套跑起来很爽。

然后问题马上来了：

任务能交接，
判断过程经常交接不过去。

下一个 thread 能看到：

- 要做什么
- 读哪些文件
- 产出什么结果
- 做到什么算完成

但它经常看不到：

- 为什么现在选这条路
- 哪些方案已经想过
- 哪些方案已经试过
- 哪些坑上一个 thread 已经踩过

所以我准备把跨对话 handoff 模板改成这样：

任务：
当前卡点：
目标 thread：
先读这些文件：
需要产出：
完成标准：
红线：

为什么选这条路：
已经想过哪些方案：
哪些方案先放下：
上一个 thread 踩过什么坑：
下一个 thread 可以直接继承什么：

前半段是任务交接。

后半段才是我现在真正想要的 shared memory：

把判断过程也交出去。

这样下一个 agent 接到的是一张带判断路径的任务单，
还带着前一个 agent 的踩坑记录。

这东西很小，
但应该挺有用。

能少重复犯蠢一次，
就值了。

[Thariq @trq212]

https://t.co/R6exTuF7P8

[周览资源 @grgerwcwetwet]

50个感觉'非法'却合法的网站

https://t.co/3m9mwGdEb3：下载社交媒体视频
https://t.co/yxmEyF1dE4：网页版 Photoshop
https://t.co/hecQx9d5Tq：临时邮箱
https://t.co/u2VYVKHAq9：100+ 在线工具
https://t.co/3Gjo6wbxqf：查看旧网页
https://t.co/smOSOW3p8K：找免费替代软件
https://t.co/7gnNUbV9Z5：查影视在哪个平台能看
https://t.co/wWUCA9ds9N：免费经典书籍
https://t.co/B9zgSEk5WH：解数学题
https://t.co/yTlScrKv52：一键抠图
https://t.co/TPb60nLxQP：擦除图片物体
https://t.co/Yno6BQCRXN：图片压缩
https://t.co/p1Dx7troC2：手绘风白板
https://t.co/h3fFNe4mwL：代码截图美化
https://t.co/QHOEw9Yfks：实时航班追踪
https://t.co/r0MpjhJV9J：亚马逊价格追踪
https://t.co/th7983N5yU：检查邮箱是否泄露
https://t.co/c8CCnckHzN：文件安全扫描
https://t.co/5I4Tfb6Lm4：自毁消息
https://t.co/rs0SWlziv8：自动删除文件分享
https://t.co/saf3CJXYmO：网页存档
https://t.co/X4AzdGPaCo：全球电台
https://t.co/mWYlmxdxnN：找影视里的歌
https://t.co/pTFGIV4nhv：专注白噪音
https://t.co/SwuL7rZiaF：AI 论文助手
https://t.co/aQw762SaI6：科学共识搜索
https://t.co/VqS5YHQy0Y：论文关系图
https://t.co/AmnXgwKDsU：学术搜索
https://t.co/hKeG0rM0bP：读懂论文
https://t.co/zzmRmn59Tr：总结 YouTube 视频
https://t.co/VNqnYDXQDM：开发者 AI 搜索
https://t.co/tCEobGUUBa：正则测试
https://t.co/aaG5GlzwKw：代码格式化
https://t.co/eWCKQh60rt：解释命令行
https://t.co/81P7Gf79AI：无限白板
https://t.co/qkbybHsF3j：查网站是否宕机
https://t.co/gC5fFGjkbW：反向搜图
https://t.co/ONnkx4Kbj2：测网速
https://t.co/TIlcoce1Mm：PDF 工具
https://t.co/qwJrOu8U9r：PDF 合并拆分
https://t.co/61xRlWIPaI：10 分钟邮箱

[divyansh tiwari @DivyanshT91162]

This is probably one of the wildest AI projects on GitHub right now.

Someone turned Warren Buffett into an AI agent.

Then added Charlie Munger, Benjamin Graham, Cathie Wood, Philip Fisher, Michael Burry, Ray Dalio, Peter Lynch, and other investing legends.

Now they all analyze stocks together.

You get a full AI hedge fund powered by 18 agents:

→ 12 legendary investor agents
→ 6 analyst agents for valuation, risk, sentiment, fundamentals, technicals, and portfolio decisions

Each agent follows the actual investing philosophy of the person it's based on.

The craziest feature?

Built-in backtesting.

Run your strategy on historical market data first.
See how it would've performed.
Then decide whether it's worth using in real markets.

50K+ GitHub stars and still climbing fast.

https://t.co/tMFNkdN9hQ

[Amto @XAMTO_AI]

开源爬虫框架 Scrapling 可以直接绕过 Cloudflare 等反爬机制，不需要手动维护选择器，速度比 BeautifulSoup 快很多。

它原生解决反爬问题，数据获取流程更直接。

🔗 https://t.co/wyXgKaXWiq

[GitHubDaily @GitHub_Daily]

想学大语言模型的底层原理，网上的教程要么纯讲理论，要么直接丢个开源模型让你微调，真正从零手写训练的实战教程太少了。

偶然找到 train-llm-from-scratch 这个项目，手把手教你用 PyTorch 从零实现一个 Transformer 模型，在单张显卡上就能完成训练。

从注意力机制、多层感知机到完整的 Transformer 架构，每个模块都有详细的代码和原理图解，跟着做就能训出一个能说人话的语言模型。

GitHub：https://t.co/isOZ7rPF5z

项目提供了 1300 万和 20 亿两种参数规模的配置方案，1300 万参数的模型用免费的 Colab 就能跑，训练一天就能看到效果。

训练数据用的是 Pile 开源数据集，涵盖书籍、论文、代码等多种来源。

如果你想搞懂大模型是怎么工作的，而不只是停留在调用接口的阶段，这份教程值得跟着动手练一遍。

[slash1s @slash1sol]

THE CO-FOUNDER OF GITHUB GAVE A 46-MINUTE TALK ON GIT BECAUSE ENGINEERS WITH 10 YEARS IN HAVE NEVER SEEN HALF OF WHAT IT DOES

This is Scott Chacon. He wrote Pro Git -- the book most devs secretly learned Git from and he co-founded GitHub. So when he says you're missing things, you're missing things.

About ten minutes in it clicks: half the "git disasters" you've ever fixed by deleting the folder and re-cloning had a one-line solution sitting in the tool the whole time.

Git ships new code almost every day -> roughly nine commits a day for over a decade. Most of us stopped learning it the second we memorized add, commit, push.

Knowing Git isn't a senior-dev flex anymore -> it's the floor. The agent writes the code now. Your real job is reading, branching, and untangling the history it leaves behind.

The day an AI agent force-pushes over your main branch, these 46 minutes are the difference between a quiet fix and a very loud apology.

Save it now.

You'll reach for it sooner than you'd like ↓

[车厘子 @0xcherry]

冷知识：RPG Harness 支持让 AI 自动进行图像资产生成

找个会生成图片的 AI 扔进去就行了，比如 CodeX，跟他说帮我画一下立绘和 CG

Repo: https://t.co/96jf7ZetDx

[Howie Liu @howietl]

We’re giving away $10,000,000 to founders building agent-first businesses.

Autonomous, proactive agents will run tomorrow's companies.

We're backing 500 founders building them.

The Founding 500.

https://t.co/u8mxRWEXiT