silence

[锐汉街评 @ruihanmeimei01]

这女的肯定认为这个男的不会还手呀，结果失算了！所以别太相信自己的魅力，女孩子在外面保护好自己~

[Amto @XAMTO_AI]

想保持英文阅读习惯，订杂志的费用一年下来不是小数目。GitHub 上有一个每周自动更新的外刊资源库，把几本主流英语杂志的下载链接整理在一起。

收录范围包括《经济学人》、《纽约客》《连线》《大西洋月刊》这些，每期同时提供 epub、mobi、pdf 三种格式，Kindle、手机、平板都能适配。

每周五经济学人更新，周六纽约客到，基本是跟着发行节奏走的。我用了大半年，每周更新很准时，基本没断过。

不需要订阅，也不需要到处找资源，点开就能下。对于想保持英文阅读习惯又不想花太多钱的人来说，算是一个很实在的资源。

🔗 ：https://t.co/g2MxUFBzJC

[0xkevin (🖤 , 💙) @0xKevin00]

看完汗毛直立，提问和回答的水平都是超一流的

王煜全，海银资本创始人，他是一位长期研究全球科技创新的投资人，投资圈里少有的直言派和前瞻派

视频里非常敢说，2029年 AI 泡沫将破灭，2030年遍地是黄金....

[伊利丹 @YLDLZN]

感觉张桂梅现在真的开始塌房了......

[Kevin Kelly @kevin2kelly]

Lately I’ve been asking myself: what might artificial intelligence be good for besides answering questions and writing code? My answer is the latent spaces within AIs themselves will become a new medium for creativity. https://t.co/qsxBMmK4En

[deep Manifold @BetaTomorrow]

https://t.co/nqsx8Tr6zu

[爱吃折耳根的Ace @_zheergen]

🚨卧槽！高盛开源了他们的量化工具库 gs-quant

Goldman Sachs 把内部 quants 使用的 Python 量化工具库开源了，目前 GitHub 已获得 11.3K Star。

它提供了丰富的量化分析工具，适合做结构化产品、风险建模和策略研究等工作。虽然完整 API 需要高盛机构客户权限，但核心库已经开源，普通用户也可以安装使用（pip install gs-quant）。

核心亮点：

> 机构级量化工具：支持衍生品定价、风险管理、回测、交易策略开发等全流程
> 生产级代码质量：直接来自高盛交易台的真实工具栈
> 支持 Claude Skills：官方已集成 Claude 技能，可直接在 Claude 中使用
> 覆盖广泛资产类别：股票、期权、期货、外汇等衍生品分析能力极强
> 持续维护更新：最新版本 2.1.1 刚刚发布
> Apache-2.0 开源协议，可直接用于研究和开发

对想学习机构级量化建模、做衍生品定价与风险管理、或者用 AI 辅助量化研究的开发者来说，这是目前开源里含金量最高的量化工具库之一。

[大漂亮| C Labs @giantcutie666]

世界银行昨天出了一份中国经济简报，内容很劲爆😅

因为内需疲软，中国的房价还会继续下跌，且GDP增速会持续放缓

世行说内需疲软的原因是政府乱花钱😂

政府把43％的预算拿去投资，远高于发达国家13.5％的水平

而民生投入仅占28％，远低于同行60％以上的水平

且这些民生投入，主要还是去的体制内🤣

[前低级经理X @happypaidaxing]

中国经济每次都是选择短期路径最优
但是长期就走成死路
到底是谁的"亲自指挥"?

[Michael Pettis @michaelxpettis]

Noah Smith gets much right in his argument as to how China managed the huge property-bubble deflation in a "painless" way, and why it's a little silly to say that Beijing managed to deflate a massive housing bubble at no cost. The key point, as he notes, is that whenever China's faces an economic shock, Beijing "uses its direct control over the banking system to push banks to lend more."

What's more, contrary to some of the more dizzy comments to which his essay is a response, China is not the first country to have figured out how to manage adverse economic shocks by setting of compensating investment bubbles. Many investment-driven economies have done the same, most notoriously the US in the mid 1920s, Latin America in the early and mid 1970s (in response to the oil shocks), and Japan in the mid 1980s.

What this means in practice is that the system is not allowed to adjust. Each economic shock is countered by setting off a new investment bubble. For example the post-2009 collapse in China's current account surplus (from 10%of GDP to 3% of GDP) was balanced by inflating an infrastructure bubble.

At the time this was hailed as a brilliant move by Chinese and (especially) foreign economists, although today many Chinese economists will tell you (mostly in private, but occasionally even in public) that this was a mistake, and was the beginning of China's debt problem.

The reason is that if China had been severely underinvested in infrastructure, balancing the contraction in trade with an expansion in infrastructure investment would have been both clever and sustainable, but by 2009 China already had almost as much infrastructure as it could productively absorb, and so the surge in infrastructure investment was decided on not because China needed more infrastructure but because a slowdown would have been more disruptive than China could absorb.

Consider what happened next. As China tried to bring infrastructure investment levels down in the mid-2010s, it could only do so "without pain" by setting off what turned out to be the last stage of the property bubble. It did this in 2015-16 by eliminating many property purchase restrictions, lowering minimum payments, reducing mortgage rates, and forcing banks to expand mortgage lending (often in spite of growing concerns among banks).

These policies, not surprisingly, set off a new wave of surging home prices and expanding property development, but this couldn't last long, and when the property sector, gorged on almost unimaginable levels of debt, began its sharp contraction in 2021-22, Beijing once again responded by setting off an investment bubble in another sector. This time, as Smith notes, it set if off in the manufacturing sector, with the decline in property investment matched almost dollar for dollar with an increase in manufacturing investment.

Once again, if China had been severely underinvested in manufacturing, this would have been a great move, but because at the time it was already over-reliant on a very large manufacturing sector (even larger than the property sector at its height) and suffering from excess manufacturing capacity, the result within 2-3 years was massive increases in excess capacity, severe involution, and an uncontrollable surge in the trade surplus.

"So what?" some of the less historically-literate might wonder. If every time an investment bubble contracts, and the authorities can negate the economic impact by setting off an investment bubble in another sector, shouldn't they keep doing it forever?

Yes, if they can but, for what should be obvious reasons, they cannot, and the longer they postpone the adjustment, the more costly it is likely to be. The problem, of course, is that each bubble has to be bigger than the last to keep economic activity from slowing. We can see this most obviously in the acceleration in the growth of the country's debt burden, already the second highest in the world (after Japan's) and the fastest-growing in history.

To the extent that debt funds productive investment, of course, rising debt should not translate into a rising debt-to-GDP ratio, but in an economy in which nearly all debt funds investment, and very little funds consumption or transfers, many years of a rapidly-rising debt-to-GDP ratio is pretty strong evidence that the investment is increasingly unproductive.

Among other things this means that a bubble-replacement strategy can look very clever in the short term, but it can only run on as long as growth in the debt burden can accelerate.

[JinYu金鱼 @JinYu762]

字节股神 @leto_bao 又推荐了一本好书

对过山车行情下怎么保护资产感兴趣的，推荐看《避风港》，普通人特别实用

我顺手把这本 + 他之前推荐的另外两本也准备好了，网盘自取：https://t.co/ZXy7nUNQfO

仅做书单分享，不构成投资建议

[RJ @RJDAIGOGO]

12分钟讲清芯片行业上下游，从晶圆讲到芯片设计和封装，以及如何识别一家公司在半导体产业链中的角色。文科生也能看懂。

原创者“净发大财”，原视频链接见评论区。

[大洋弥生mio&ocean @hiroe_yoshida]

这篇关于日元下跌的文章太牛了，感谢@e_satellite 推荐！

作者认为，日元持续贬值的根源不是利率，而是日本经济增长模式失灵和全球资本对日本长期前景失去信心，因此日元的长期贬值趋势尚未结束。

我去年写过一篇文章，《日本不希望让你有钱》我的观点就是日本放弃了经济增长，只想维持现状

这种制度背后其实反映的是日本对“社会动荡”的深刻恐惧：
高龄化社会不能承受剧烈失业；
失控的通胀或资本波动可能重演80年代灾难；
民众太有钱、太有能力，也意味着可能有更多挑战体制的能力。
所以，日本选择了一个看似“保守”的路径，其实是维稳型国家资本主义的高级形式。

日本不是民主国家，是社会主义国家，大政府管制和分配财富，只会越来越穷！这就是为什么税越来越高！
我的观点是因，这篇文章的观点是果！

[梭哈.AI @SUOHA_AI]

字节炒美股财富自由的老哥推荐了两本书

投资怎么入门？

花两周看完《持续买进》《生命周期投资法》，已下载电子书，分享给大家

网盘自取：https://t.co/hDvAnW4EuI

[RJ @RJDAIGOGO]

疑似华强北大佬举重若轻，有实物讲解光模块中的新事物“玻璃桥”，五分钟讲解感觉信息量堪比我上午看的半小时动画讲解。最后结论……出乎意料。😅

[Kay @justsosooknow]

每次有人问我“视频用啥下载”，

我都懒得再解释了，

直接一次性整理好，

你自己存👇

1、X / 推特视频下载
https://t.co/hMR6RrU1wJ

2、 抖音 / TikTok 去水印
https://t.co/fck4ugyt2C

3、Instagram / Facebook 视频
https://t.co/DxVCxBeRB4

4、全平台通用（国内平台）
https://t.co/mcANK1zfWs

5、视频去水印
https://t.co/nnU7tmn3ZT

6、图片去水印
https://t.co/tVQdc8Tx6a

7、YouTube 专业下载（开源）
https://t.co/FSbQi7i87A

8、懒人桌面工具
https://t.co/Lh0AZRXeg2

全部免费，

够用了。

别再来问我用啥下载视频了。

[Line @0xLinehigher]

非常建议每一个选择计算机系的大学生，在大学时期将cs336啃完，不开中文字幕，只开英文字幕。啃完之后，你对LLM的理解和英语能力至少在国内前百分之1%。

这门课超过国内任何一所大学里面计算机的课程。

《Stanford CS336: Language Modeling from Scratch》 是一门斯坦福大学计算机科学系的课程，这门课每一年春季都会有，每一年都会在油管免费开放。

里面的内容非常前沿而且实用，目前油管上的版本是25年的版本，第一节课提到的模型已经是GPT-4。而与此同时，国内的计算机课程还在用着至少5年前的PPT。

依稀记得，我的第一节C语言课程，PPT的里面的演示系统是XP...

啃完CS336，你会收获LLM的全栈技术栈，从数据收集，清洗，模型训练，优化，评估到部署，Transforemer架构，注意力机制，各种目前最新的机器学习训练方法

这门课的目标就是从头让你手搓一个LLM，亲手实现完整的pipeline。

两位主要讲师：
1、Percy Liang
MIT 本科 + MEng，UC Berkeley 博士（2011）。曾在 Google 做 post-doc，后加入 Stanford 任教。

2、Tatsu
Harvard 本科（统计与数学），MIT 博士（co-advised by Tommi Jaakkola 和 David Gifford），之后在 Stanford 做 Percy Liang 和 John Duchi 的 post-doc

耐心啃完这个，所有计算机课程几乎都不需要上了...因为在过程中必然会遇到很多不懂的点，边学边查，然后顺带就把学校关于机器学习和深度学习的专业课给学完了。

如果刚上大一，可以先看cs50，然后再到cs336，先稍微打点计算机的基础。

与此同时，一定要把线性代数学好，将高数丢了都要把线性代数学好，线性代数是机器学习的底层DNA。

如果线性代数没有学好，会比较难理解神经网络，相当于没学好加减乘除就想去求导。

[Ralph Sueppel @macro_synergy]

"Vanilla Options: A Unified Mathematical Framework for Pricing, Hedging, and Volatility Modeling" https://t.co/m8T79kURFW

[Kithe @kuk47377341]

我来学习下

老师们都是如何把Codex用到极致？

[Smartpig @Smartpigai]

每次有人问我“视频用啥下载”，我都懒得再解释了，直接一次性整理好，你自己存👇

1、X / 推特视频下载
https://t.co/kpPSahQuqI

2、 抖音 / TikTok 去水印
https://t.co/6R5ukv4Zex

3、Instagram / Facebook 视频
https://t.co/xIssJnIgoO

4、全平台通用（国内平台）
https://t.co/Qvp7h92Ntr

5、视频去水印
https://t.co/FR0i3ahHtc

6、图片去水印
https://t.co/G8GzXL6QoY

7、YouTube 专业下载（开源）
https://t.co/LFmsZd57RK

8、懒人桌面工具
https://t.co/n4VKGXW3jw

全部免费，够用了。别再来问我用啥下载视频了。