钢钢钢蛋

嘿嘿😬

[Nassim Nicholas Taleb @nntaleb]

I declined an invitation to lecture in Morocco. I will wait for a regime change before setting foot in that country.

[钢钢钢蛋 @cying388]

@lianda_edu 教授目前说的没问题。
你有言论自由，但要付出代价。
心理学数据无法复现，操作p值的手法已经是公开化了，有什么奇怪的？

[Jensen Huang @JensenHuang]

Attackers have frontier AI. Defenders need a frontier AI ecosystem—the best open and closed models, force-multiplied by a global community.

During the Hugging Face incident, closed AI blocked essential forensics. An open-weight frontier model helped contain the intrusion.

That’s why we created the Open Secure AI Alliance.

[钢钢钢蛋 @cying388]

传说，你他妈的也是欠杀了。

[George Galloway @georgegalloway]

@ScottAnd67 It doesn’t exist. It is urban myth.

[钢钢钢蛋 @cying388]

人类的多样性

[Valeriy M., PhD, MBA, CQF @predict_addict]

One can get Field’s medal, but not erase the fact that she wasn’t the smartest person in her class.

It would be interesting to see what her classmates achieved, not everyone wants civil servant position in France.

It is very likely many of her classmates work in finance and AI in China and have achieved much more than writing something on a piece of paper.

[Hedgie @HedgieMarkets]

🦔2.2% of US households pay for an AI subscription as of April. The median spend is $20 a month. More Americans pay for sports betting apps (5%) than pay for AI. ChatGPT has 900 million weekly users but only 5% convert to paid. 37% of consumers say none of AI's uses are helpful. 30% say they're less likely to buy a product marketed as "AI-powered." Companies are spending $600 billion on AI infrastructure this year to serve a market where 97.8% of households don't pay for the product.

My Take
2.2% penetration nearly four years after ChatGPT launched. Netflix hit 25% household penetration in a similar timeframe. Spotify hit 15%. The growth is fast in percentage terms but off a base so small that doubling it still leaves you in single digits. The capex projections, the bond issuances, the IPO valuations all assume consumer adoption eventually catches up to the spending. Nearly four years in, it hasn't.

The pushback I'll get is that enterprise revenue is what counts and consumer subscriptions are a sideshow. Alphabet's cloud grew 82%. Anthropic reportedly hit $4.5 billion in ARR. But a lot of that enterprise spending is funded by VC money and hyperscaler capex, which makes it circular. Consumer willingness to pay is the closest thing to a genuine market signal, and right now that signal says $20 a month from 2.2% of households. The $600 billion in spending this year is a bet that the other 97.8% eventually show up. I'm not seeing it.

Hedgie🤗

[Jensen Huang @JensenHuang]

For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.

AI will transform every industry, power every company, and be built by every country.

Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.

The world needs both frontier closed models and frontier open models.

https://t.co/AUKzoQ5Ikb

[钢钢钢蛋 @cying388]

笑死，现在三明治时代爽不爽？

[朱韵和 @zhu0588]

龙应台：在二零零五年二月一日台湾所通过的所谓“广电三法”。一个法案三读通过了，多么枯燥的新闻，然而，知道迈向民主法治的路有多崎岖、多昏暗，就会知道这是多么划时代、惊天动地的三个法：广播电视法，有线广播电视法，卫星广播电视法的三读通过，意味着党、政、军三种国家权力彻底退出新闻媒体。

[Hedgie @HedgieMarkets]

🦔South Korea's stock market crashed 30% from its June peak in under a month. 1.2 million accounts received margin calls. 360,000 were liquidated to zero. The trigger wasn't an earnings miss or a recession. It was fear that AI spending might slow down. Samsung and SK Hynix make up over half the index and when both dropped, margin calls cascaded into forced selling which drove prices lower which triggered more margin calls. As of today the KOSPI is still swinging 5-6% per session and remains 25% below its June peak.

My Take
I think most US investors saw the Korea crash and assumed it couldn't happen here because our market is more diversified. It is, by percentage. But US margin debt is $1.5 trillion, 37 times Korea's $40 billion peak. Hedge funds are borrowing $7.2 trillion. Leveraged semiconductor ETF assets went from $163 billion to $60 billion in weeks, meaning $100 billion in speculative money evaporated and the broader market barely flinched.
That means the leverage is deep enough to absorb a $100 billion hit without repricing. When the hit is big enough that it can't be absorbed, the unwind will be proportional. And there has never been this much leverage.

All it takes is one disappointing earnings call or one capex cut from a hyperscaler. The selling wouldn't start with fundamental investors reading a 10-Q. It would start with leveraged ETFs rebalancing at the close, pushing prices down, triggering margin calls overnight, forcing more selling at the open. That's exactly what happened in Korea. The difference is Korea's forced selling ran through $40 billion in margin. The US version of that cascade would run through $1.5 trillion.

Hedgie🤗

[钢钢钢蛋 @cying388]

推特奇葩大赏之

给不起钱屁话多。

人家是想要工作，多半都是现编的想法，迎合你个傻逼又嫌弃没行动力。

有想法有行动力的人，需要给你个傻逼打工？

逗死了

[钢钢钢蛋 @cying388]

再过段时间加入对抗叙事就可乐了。

[钢钢钢蛋 @cying388]

@OverwatchBrasil 像坨屎

[钢钢钢蛋 @cying388]

@Jackywine 你放个鸡毛心，目前模型厂商招人什么门槛？轮得到你这个简历连ai筛都过不了的人来用宏大叙事蹭热度？
a÷起个头，中美弱智都跟着对抗叙事走。
想想就可乐。

[钢钢钢蛋 @cying388]

笑死人了，谁是我们，谁是他们，而且谁会跟着你成为我们，行为招笑。

[Jackywine @Jackywine]

一种朴素的情感，或许没有人像我们一样希望国产大模型能够百花齐放更好的发展，从我最开始连什么是 API 接入都不太懂的就给智谱充值的时候，到现在每天主力模型就是 DeepSeek 再到现在 Kimi ，看着这几个国产大模型一点点成长真的特别开心🥹

[钢钢钢蛋 @cying388]

笑死我了哈哈哈哈哈哈

[evan loves worf @esjesjesj]

Have you met your boss

[钢钢钢蛋 @cying388]

这个帖子一股上古味道。

还不如ai生成的信息流。

[钢钢钢蛋 @cying388]

姜萍这件事非常反应老中人特征。
就一17岁的小女孩，家徒四壁，被老师拿来当个道具，能一直被开盒至今。
但想一想这群人同时支持恢复肉刑和咬人猫狗活该虐杀，也就释然了

[钢钢钢蛋 @cying388]

笑死我了，在这儿忽悠前两年没关注过光伏板的傻逼

[Journeyman @JourneymanChina]

刷到一个主播分享的新疆光伏相关细节，觉得挺有意思的。

国家规定大型光伏项目必须拿出2%-3%的股份由村集体代持，每个村每年能分到几十万分红；
项目的建设阶段和后续的维护阶段有务工需求时，也强制优先聘用本地工人，大量新疆当地群众因此受益。

光伏项目既带动了新疆本地的基础设施建设、提升了居民收入，也让东部地区获得了充足的廉价电力，
是实打实的多赢局面，也进一步推动了内地与新疆的经济融合和人员交流。

再加上国家现在也在加强当地的普通话推广，语言沟通也越来越顺畅了。

[钢钢钢蛋 @cying388]

是的

[Nassim Nicholas Taleb @nntaleb]

Any help with LLMs for writing will make you sound like a journalist, stripped of all idiosyncracy, which, for many, is not a bad thing.
This is structural, the way statistical machines operate. LLMs are bland, dull, & devoid of personality.

[钢钢钢蛋 @cying388]

讨厌阿里的同时，a畜也不是什么好东西

[Susan Zhang @suchenzang]

instead of indulging in all this fear-mongering frontier-stealing narrative slop, @Alibaba_Qwen might as well release all "29 million Claude exchanges" for the benefit of open-source research and call it a day

let the public decide for themselves how "dangerously critical" these tokens actually are, relative to the other trillions upon trillions being used to train these models