GPT-5.4 评测、油价飙升、Anthropic 的 TGIF | 道格·德穆罗、文森佐·兰迪诺、马克斯·霍达克

是的。

Yeah.

如果有人来找你，说：‘我其实重新发明了数学。’

If somebody comes to you and they're like, I actually reinvented math Yeah.

为了完成这个任务。

In order to in order to accomplish this task.

你会说：‘什么？’

You're like, what?

根本不是这样。

It's like, no.

你只是希望有人能把工作做完。

You just want somebody to get the job done.

具体怎么做的并不重要。

Doesn't matter exactly how they do it.

我提到过一家无人机公司，他们没有使用Linux。

I talked about there was a drone company that wasn't using Linux.

他们说，是的。

They were like, yeah.

比如，我们别只是写自己的软件栈了，现在显然你应该使用 ImageNet 和现成的 SLAM 工具。

Like, let's let let like like, not not just like let's write our own software stack or like now, obviously, you should be using, like, ImageNet and, like, whatever off the shelf tools are available for SLAM.

Meta 有很多开源项目。

Meta has a bunch of open source stuff.

你应该把这些都整合起来，但没必要重写操作系统，只是为了获得完全的控制权。

You should be compiling all that together, but to rewrite, like, the the operating system, because, like, you want, you know, full control.

通常来说，如果有人想要一个网站，你不会说，我们从头开始吧。

Typically, you know, you if if somebody says a web wants a website, you're not gonna, like you're gonna be like, let's go.

如果需要，直接用 Shopify 搭建就行。

Maybe just set up a Shopify if you need that.

设置一个托管网站就可以了。

Set up a a a, you know, some sort of, like, you know, hosted site.

你不会去重新开发 NGINX 或其他作为网页服务器所需的组件。

You're not gonna rebuild NGINX or whatever else you need to actually use as a web server.

但这真是个很棒的演示。

But this is a cool demo.

它看起来功能相当完整。

It looks pretty fully functional.

我对游戏行业的繁荣感到非常兴奋。

I'm I'm I'm excited for the for the gaming boom.

我还在屏息等待，比如一些像音乐排行榜这样的事情。

I I think I'm I'm I'm still holding my breath to, like, still some things like charting.

我们在Spotify上看到过这种情况。

We saw that with with Spotify.

现在已经有AI音乐登上排行榜了，我也听过一些。

Like, there is AI music that's charting now, and I and I've listened to some of it.

还有一些混搭作品，比如用五十年代或六十年代的理发店四重唱风格演绎2000年代的说唱歌曲。

And some of the ones where, again, mashups, where it's like a fifties or sixties rendition barbershop quartet of an of a two thousands rap song.

这真的很酷，因为其中有一些真实参照，能让你保持根基和相关性。

That's really cool because it's like there's some ground truth there that, like, keeps you grounded and relevant.

很多图像混搭都特别酷。

A lot of the image mashups are really cool.

我猜很多游戏也都很棒。

And I imagine that a lot of the games are really cool.

比如一个以《使命召唤》为主题的模拟人生皮肤之类的。

Like, a Sims skin that's Call of Duty themed or something.

把这两种主题以一种有趣的方式结合起来，游戏引擎是熟悉的，内容也是熟悉的，但它们以前从未融合过，而在AI的世界里，它们现在可以了。

Like, taking these two themes, like, putting them together in an interesting way, whereas the way, like, the game engine's familiar, the content's familiar, but they never but they haven't crossed before and like they can in the world of AI.

西奥说，他已经用了5.4一周了。

Theo says, been using 5.4 for a week now.

这是一个极其出色的模型。

Absurdly good model.

我现在再也不想用别的东西了。

Don't really like using anything else anymore.

他真的很喜欢5.4。

He really likes he really likes 5.4.

他专门做了一场关于这个的直播。

He was doing a whole livestream about this.

Prime Agen，那个

What did the Prime Agen, the

其他主播有人回应Theo说：每次有新模型发布，你的说法就变了。

other streamers Somebody responded to Theo and said, Every time a new model release, your statement changes.

他说，那是因为新模型通常比旧模型更好。

And he says, That's because newer models tend to be better than older ones.

这太好笑了。

That's hilarious.

他总是对新模型有非常精彩的分析。

The the always has great great breakdowns of of new models.

是的。

Yeah.

我

I

会去他的账号看看。

would go over to his account.

他就说，是的。

And he's like yeah.

他处于一个很有趣的位置，因为他一直在直播并做评论，但同时也真的在开发东西。

He's he's in an interesting position because he's streaming a lot and and and and commentating, but he's also like building stuff.

还有那位Primegean说：希望大家都准备好迎接即将到来的5.4上釉浪潮。

As is the Primegean who says, I hope everyone is ready for the upcoming 5.4 glaze wave.

炒作周期将会非常盛大。

The hype cycle is going to be a big one.

嗯，

Well,

让我们为炒作周期鼓掌吧。

we will Let's give it up for hype cycles.

对。

Yes.

还有三重釉。

And and triple glazes.

三重釉。

Triple glaze.

三重釉。

Triple glaze.

三重釉。

Triple glaze.

怎么回事？

What's going on?

转到乔·魏森塔尔那里。

Heading over to Joe Wiesenthal.

在经济方面。

In the economy.

经济怎么样？

How's the economy going?

就业数据大幅不及预期。

We got a big miss in jobs.

美国二月份失去了9.2万个就业岗位。

US loses 92,000 jobs in February.

失业率上升至4.4%。

Unemployment rises to 4.4%.

经济学家原本预计新增5.5万个岗位，失业率将维持在4.3%不变。

Economists had expected 55,000 jobs to be added and for the unemployment rate to hold steady at 4.3.

好吧。

Okay.

一点都不好。

Good at all.

如果回过头看，从2025年4月起，经济就一直在裁员，如果考虑修正数据的话。

If I guess if you look back, the economy has just been a shedding job since April 2025 if you factor in Yeah.

修正数据下调了。

The revisions down.

是的

Yeah.

显然，医疗行业发生了罢工，因此医疗就业人数下降。

Apparently, there's a strike in the health care industry, so health care employment decreased.

此外，医疗、信息和联邦政府的就业持续下滑，可能与特朗普政府减少政府规模有关。

And then health care in information and federal government continue to trend down, probably something related to the Trump administration, you know, reducing the size of the government.

过去两个月的数据被向下修正了6.9万个。

Last two months were revised down by 69,000.

并非所有数据都很糟糕。

Not all of the data was terrible.

U6失业率下降，因为声称自己从事兼职但希望获得全职工作的人数实际上减少了。

U6 fell as the number of people who said they had part time jobs but wanted full time ones actually declined.

因此，股市期货下跌，但对这一数据本身的反应并不强烈。

So stock futures are lower, but not a ton of reaction to the number itself.

波尔托说，我注意到一些科技圈的朋友丢了工作。

Porto says, I've noticed some of my tech friends have lost jobs.

有些工作正变得极其简单。

Some jobs are becoming super easy.

有些工作之前有让人头疼的部分，现在这些部分变得非常轻松了。

Some jobs had parts that were a pain, and now those parts are super easy.

有些人积极拥抱人工智能以保持竞争力，但仍不确定它是否会完全取代他们。

Some people leaned into AI to stay relevant, still wonder if it will replace them entirely.

我们会继续关注这一点。

Well, we will continue to track it.

原油。

Crude oil.

接近90美元。

Close to $90.

上周日伊朗战争爆发时预测的每桶100美元目标，正越来越接近实现。

This prediction, the $100 ball the $100 barrel predicted on last Sunday as the Iran war broke out is is closer and closer to coming true.

他们现在预测

They're now predicting

是的

Yeah.

50

50.

原油今天上涨了11.5%。

Crude is up 11 and a half percent today.

是的

Yeah.

不太好。

Not good.

它如何影响经济或各个市场，这非常有趣。

It's it's very interesting how how it affects the economy or or or the various markets.

所以，纳斯达克下跌的幅度远小于道琼斯工业平均指数。

So, I mean, the Nasdaq slid a lot less than the Dow Jones Industrial Average.

道琼斯工业平均

The the Dow Jones Industrial

初创公司条款。

startup terms.

硬科技。

Hard tech.

如果你的

If your

你的AWS或GCP账单，你的Azure账单

a if your AWS or GCP bill, your Azure bill

天哪。

Oh, shit.

哦，是的。

Oh, that yeah.

上涨了100%。是的。

That went up a 100% Yeah.

一周。

Week.

所以但是

So But

这关系到整个经济。

it's for the entire economy.

纳斯达克下跌了0.3%，但工业板块下跌了1.6%。

The Nasdaq slid point 3%, but the industrial slipped 1.6%.

显然，因为工业板块对石油更敏感，它们制造东西、建造东西，所以任何对石油敏感的行业都受影响。

Obviously, because industrials are more sensitive to oil because they make things, they build things, and so anything that's sensitive to oil.

有趣的问题是，美国的石油钻探公司会如何应对？

The interesting question is how will the American oil drillers react?

至少在本周早些时候的期刊中，它们在反思：如果这是一次短暂的事件，它们可能不想立即增加大量新产能。

They were, at least in the journal earlier this week, sort of reflecting on the fact that if this is a short lived engagement, then they might not want to spin up a bunch of new capacity.

但如果这是持续性的，且油价继续上涨，那么显然会形成增加钻探的经济动力。

But if it's ongoing and the price of oil is continuing to rise, then obviously that creates an economic incentive to drill more.

因此，我们将看到产出会如何变化，储备会在哪里被开采，等等。

And so we will see where, like, the outputs go, where the reserves get tapped, all of that.

关于你昨天和丹的讨论，有个后续内容。

There's a follow-up to a discussion you were having with Dan yesterday.

首先，让我跟你介绍一下 Console。

First, let me tell you about Console.

Console 构建了人工智能代理，可自动化 70% 的 IT、人力资源和财务支持工作，为员工提供即时解决访问请求和密码重置的服务。

Console builds AI agents that automate 70% of IT, HR, and finance support, giving employees instant resolution for access requests and password resets.

另外，让我也跟你介绍一下 CrowdStrike。

And let me also tell you about CrowdStrike.

你的业务是人工智能。

Your business is AI.

他们的业务是保障人工智能的安全。

Their business is securing it.

CrowdStrike 保障人工智能安全并阻止入侵事件。

CrowdStrike secures AI and stops breaches.

所以你之前问丹·普里马克的问题，大概是这样的：伊朗战争是否会影响海湾国家邻国投资团体对美国风险投资基金的投资？

So the question you had for Dan Primack was something along the lines of, will the Iran war affect investment from Gulf State neighboring Gulf State investment groups into American venture capital funds?

随着风险投资的蓬勃发展，许多公司的融资额已达到数百亿甚至上千亿美元，中东一直是风险投资股权的强劲有限合伙人需求来源。

We've seen as venture capital has boomed and the fundraisers have gotten into the tens of billions, the high billions for many firms, The Middle East has been a source of strong LP demand for venture capital stakes.

因此，随着伊朗局势不稳，人们可能会改变关注重点。

And so with instability in Iran, you might just see folks change focus.

由于对国防投资和军事建设的新需求，来自海湾国家流入美国风险投资公司的资金可能会减少。

With new demands for defense investing, military buildup investing, you might see less dollars flow out of The Gulf into American venture capital firms.

丹的观点似乎是，这基本不会受到影响，因为实际在海湾国家投资基金工作、向美国风险投资公司拨款的人并不住在中东。

Dan's point, it seemed like, was this should mostly be unaffected because the folks who actually work at the Gulf State Investment Funds, who write the checks into American venture capital firms, they're they don't live in The Middle East.

他们住在纽约。

They live in New York.

他们与纽约的其他风险投资投资者关系人士会面，并说：嘿。

And they meet with other venture capital investor relations folks in New York, and they say, hey.

你们的基金表现如何？

How's your fund performance?

好的。

Okay.

我们正在分配这笔资金。

We're allocating this.

然后丹的第二个观点是，他们之所以投资美国科技、风险资本和美国的前沿技术，主要是为了从海湾地区分散风险。

And then Dan's second point was, well, a big part of why they're investing in American technology and venture capital and frontier technology in America specifically is because they wanna diversify away from The Gulf.

他们想摆脱地缘政治风险，也想摆脱对石油的依赖。

They wanna diversify away from the geopolitical risk, and they wanna diversify away from oil.

没错。

Exactly.

是的。

Yeah.

话虽如此，我之所以问这个问题，是因为想想如果你有一份工作。

That said, the reason I asked is think about if you have a job.

假设你有一份大科技公司的工作。

Let's say you have a big tech job.

你过得还不错。

You're doing quite well.

是的

Yep.

你正在做一些天使投资，然后你失去了工作。

You're doing some angel investing, and then you lose your job.

嗯

Mhmm.

然后你不得不找一份差得多的工作。

And then you have to get a much worse job.

你还会继续以同样的节奏进行天使投资吗？

Are you gonna keep angel investing at the same pace?

你可能会想，嘿，我感觉有点脱节了，现金流变少了。

You might be like, hey, I'm kind of have this kind of disconnect, less cash flow.

是的

Yep.

我现在不想再投入那么多现金了。

I don't wanna be outlaying as much cash now.

我认为，归根结底，如果现金流的来源受到干扰——比如石油生产的话，

And I think that that's like ultimately, you know, if the source of the cash flow gets disrupted, in this case being oil production

是的。

Yep.

那么你就更不可能继续以同样的速度进行投资分配。

Then you're much less likely to continue to allocate at the same pace.

《金融时报》昨晚晚些时候发布了一篇报道。

So the Financial Times has a story this late yesterday.

海湾国家可能重新评估海外投资，以缓解因伊朗战争引发的财政压力。

Gulf States could review overseas investments to ease financial strains caused by Iran war.

随着美国和以色列对德黑兰的行动持续，三个主要中东经济体正在考虑应对方案。

Three, leading Middle East economies consider options as US Israeli campaign against Tehran continues.

海湾国家预算面临的压力可能促使它们重新评估海外投资和未来承诺，以寻找缓解战争带来的财政压力的方案。

Pressure on the Gulf States budget could cause them to review their overseas investments and future commitments as they consider options to ease the financial strain caused by the war.

一位海湾官员表示，这可能影响从对外国政府或企业的投资承诺，到体育赞助、与企业或投资者的交易，以及资产出售等各个方面。

A Gulf official said it could impact could have an impact on anything from investment pledges to foreign states or companies, sports sponsorships, with businesses and investors or sales of holdings.

这位官员表示，四大海湾经济体——沙特阿拉伯、阿联酋、科威特和卡塔尔——已共同讨论了它们预算和经济所面临的压力，但并未点名具体国家。

The official said three of the four big Gulf economies Saudi Arabia, UAE, Kuwait and Qatar had jointly discussed the strains being put on their budget and economies, but they declined to name the states.

一些海湾国家已开始进行内部审查，以确定是否可以在现有合同中援引不可抗力条款，同时重新评估当前和未来的投资承诺，以缓解当前战争带来的预期经济压力，尤其是如果战争及其相关开支继续以当前速度持续的话。

A number of Gulf countries have begun an internal review to determine whether force majeure clauses can be invoked in current contracts, while also reviewing current and future investment commitments in order to alleviate some of the anticipated economic strain from the current war, especially if the war and related expenses continue at the same pace.

他们补充说，这一举措是一种预防性措施，源于这些国家因能源收入减少（由于产量下降或运输受阻）以及旅游和航空业受损，再加上国防支出增加而面临的预算压力。

They added that the move was a precautionary measure that was the result of the budget strains these countries are facing due to reduced income from energy due to the slowdown in output or the inability to ship and from the tourism and aviation sectors in addition to the increase in defense spending.

多种不同因素的叠加影响。

A bunch of different factors combining.

一位海湾政府顾问表示，这些富裕国家即将进行投资审查的前景引起了白宫的关注。

An advisor to a Gulf government said the prospect of an investment review by the wealthy states had caught the White House's attention.

去年，沙特阿拉伯、阿联酋和卡塔尔管理着全球规模最大、最活跃的主权财富基金，并在特朗普总统访问该地区后承诺向美国投资数千亿美元。

They managed some of the world's largest and most active sovereign wealth funds in Saudi Arabia, The UAE and Qatar last year pledged to invest hundreds of billions of dollars in The US after President Donald Trump visited the region.

它们也是全球各类体育赛事的重要支持者，并且一直在国内大力投资，以推动国家发展和经济多元化。

They are also big backers of sporting events around the world and all and have all been investing heavily domestically to develop their nations and diversify their economies.

任何影响对美或其他西方国家投资的举措，都可能增加特朗普寻求外交策略以结束战争的压力。

Any move that affects investments in The US or other Western states may raise the pressure on Trump to seek a diplomatic strategy to bring the war to an end.

我认为截至今天早上，他们表示我们对停火不感兴趣，并且已经准备好进行地面入侵。

I believe as of this morning, said, we're not interested in a cease fire and we're ready for a ground invasion.

所以基本上是在说

So basically calling calling

我之前没考虑到这会对旅游业造成什么影响。

I hadn't considered the the impact that this would have on tourism.

我的意思是，我也没意识到旅游业在中东有多重要。

Like, I I mean, I I also didn't understand how big of a deal tourism is in The Middle East.

我刚查了一下。

Just looked it up.

据称，阿联酋、迪拜，特别是迪拜，旅游业占其GDP的12%。

Apparently, The UAE, Dubai, you know, specifically, 12% of their GDP is tourism.

沙特阿拉伯是7%到9%。

Saudi Arabia is 7% to 9%.

卡塔尔是8%到10%。

Qatar is 8% to 10%.

当然，他们在世界杯期间有过一次大规模的推广。

Of course, they had that big push during the World Cup.

科威特的占比低得多，只有2.2%到3%。

Kuwait's much lower at 2.2 to 3%.

但这仍然很重要。

But this is still significant.

你得想象一下，旅游业会断崖式下跌。

You have to imagine that this the tourism will fall off of a cliff.

此外，还可能有一些外籍人士离开，比如你因为……搬到了迪拜。

And then there also might be some expatriation where if you have, you know, moved to Dubai because

哦，已经有关于人们把资金从迪拜转移到新加坡的报道了。

Oh, there's already reports of people moving money out of Dubai towards In Singapore.

是的。

Yeah.

新加坡。

Singapore.

没错。

Exactly.

所以所有这些都可能产生连锁反应。

So all of that could have knock on effects.

当我第一次看到这篇文章，你跟我提到这个时，我觉得这可能是为了给美国施加商业压力，意思是：嘿。

When I first saw this this article and you said this to me, was like, this might just be a way to, you know, put business pressure on The United States saying like, hey.

我们赶紧了结吧。

Let's wrap this up.

别让这里产生经济影响。

Let's not have an economic impact here.

这是一些我们打算撤出对美国投资的举措。

This is the the the this is something that, you know, we're gonna pull out of investments in The United States.

因此，这给美国创造了动力，促使它尽快干净利落地结束战争，并寻求和平解决。

And so that creates an incentive for The United States to to to, you know, bring a clean close to the war as fast as possible and and seek a, you know, peaceful resolution.

是的。

Yeah.

海湾国家曾

The Gulf States were

表示赞赏。

Appreciation.

你明白的，他们鼓励政府不要卷入冲突，而是寻求外交解决方案。

You know, encouraging the administration not to enter into a into a conflict and and pursue a diplomatic solution.

阿联酋知名商人卡拉夫·阿尔霍达克反映了海湾国家的挫败感。

Kalaf Al Hodak, a prominent Emirati businessman reflected Gulf frustrations.

他昨天在X上表示：直接问一句，是谁赋予你权力，将整个地区拖入与伊朗的战争？你做出这个危险决定的依据是什么？

He went on X yesterday and said, A direct question, who gave you the authority to drag a region into a war with Iran and on what basis did you make this dangerous decision?

你在按下按钮之前，有没有计算过附带损害？

Did you calculate the collateral damage before pulling the trigger?

总之，另外有个明显令人沮丧的消息：乔·维森塔尔正在研究德克萨斯州的工人营地，这些营地配备了从定制餐饮到高尔夫模拟器等各种设施，旨在吸引工人前往建设数据中心。

And so anyways, obvious frustration in other news, Joe Wiesenthal is studying Texas man camps with everything from catered food to golf simulators that are springing up in order to lure workers out to construct data centers.

哦，是吗？

Oh, yeah?

他们在建造工人营地。

They're building man camps.

现在正是工人营地的牛市。

It is a bull market in man camps.

AI工人营地提供免费高尔夫等福利，以吸引德克萨斯州的工人。

AI man camps offer golf free stakes to lure workers in Texas.

这个泰勒可能会支持我们。

This Tyler could get on board with with We

我们应该在这里建个工人营地。

should build a man camp here.

我们应该。

We should.

我们其实已经有点在做了。

We kind of we kind of do.

我的意思是，我们之前讨论过要搞赛车模拟器，但最后还是放弃了，因为你在办公室的时候，大概还是应该工作。

I mean, we talked about we talked about getting racing simulators and then we ultimately backed off just because when you're at the office, you should probably be working.

我以前试过。

I've I've I've tried it before.

而且，办公室里玩游戏总是有点奇怪，因为总会有一些小互动，让我觉得我在等你做点什么。

And, like, games in the office, it's always just weird because, like, there's all these little interactions where I'm like, oh, I'm sort of waiting on you to do something.

某件事。

Something.

你在等我做点什么。

You're waiting on me to do something.

如果你看到我正在那里删东西，你可能会想：你不能，你知道的，把这个修好吗？

If you just see me, like, erasing there, you're gonna be like, can't you, like, you know, fix this?

或者总还有别的事要做。

Or there's always something else to do.

所以，还不如好好组织一次正式的外出活动，那就是

And so it's much better just to do, like, proper off sites That is

现在非常失望。

extremely disappointed right now.

我的意思是，我不确定。

I mean, I don't know.

还有，有多少

Also How much

我的投票在这里算数吗？

does my vote count here?

你支持赛车模拟器吗？

You're you're in favor of racing simulators?

我觉得

I think

它确实是，我见过它们被建起来，然后又没人用，因为人们要么想离开回家独处，要么即使在那里，也只是说，好吧。

it is I I I've just seen them get built and then seen them not get used because people either wanna leave and go be at home, in private, or if they're if they're there, it's just like, okay.

你知道的，你在这儿干什么呢？

Like, you know, what what are you doing here?

也许这是一种共享的东西，但拥有它们的成本通常比租用外部场地高得多。

May maybe it's, like, some sort of, like, shared thing, but the cost is usually much higher to own as opposed to just, like, rent for an off-site.

我不知道。

I don't know.

我还没见过。

I I I haven't seen it.

我还没看到足够的正面案例来认为它很好。

I haven't seen enough, like, positive case studies to to think that it's good.

让我跟你聊聊Plaid。

Let me tell you about Plaid.

Plaid为那些用于消费、储蓄、借贷和投资的应用提供支持，安全地连接银行账户，实现资金流转、防范欺诈并改进贷款服务，现在还加入了AI功能。

Plaid powers the apps you use to spend, save, borrow, and invest, securely connecting bank accounts to move money, fight fraud, and improve lending now with AI.

让我跟你聊聊Figma。

Tell you about Figma.

无论你的想法从哪里开始——无论是Figma、代码、Codex还是手绘草图，Figma画布都是让想法交汇、产品成形的地方。

No matter where your idea starts, Figma may quad code, codex, or a sketch, the Figma canvas is where ideas connect and products take shape.

用Figma朝着正确的方向构建。

Build in the right direction with Figma.

OTP说，我同意办公室里玩游戏是个坏主意，但养个宠物动物园呢？

OTP says, I agree games in the office are a bad idea, but what about a petting zoo?

哦。

Oh.

宠物动物园不错。

A petting zoo is good.

对。

Yeah.

我明白你想说什么。

Like where you're going with this.

宠物动物园跟兄弟会养狗差不多危险，而那从来都不会有好结果。

A petting zoo is dangerously close to frat dog though, which never works out.

你不能搞一个集体养的狗。

You can't you can't get a collective dog.

狗是男人最好的朋友，而不是一群男人最好的朋友。

Like a dog is man's best friend, not men's best friend.

需要有一个专属的主人。

Like needs an individual owner.

一个照顾者。

Someone a caretaker.

责任必须落在某个人身上，你知道的。

The buck has to stop with someone, you know?

所以我认为这就是办法。

And so I think that's the way.

公司正在为招聘工人建设数据中心而竞争，却发现带慢速WiFi的汽车旅馆房间毫无吸引力。

Companies are competing for workers to build data centers and are finding that a motel room with sluggish WiFi isn't much of a draw.

试试免费的飞镖和高尔夫模拟器。

Try free stakes and golf simulators.

我喜欢这个点子。

I like that.

随着人工智能的兴起，数据中心的建设急剧扩张，对水资源和电力供应的竞争正推动建设向缺乏住房与基础设施的偏远地区延伸，而这些地区往往无法容纳建设巨型服务器仓库所需的数百甚至数千名临时工人。

As data center development has exploded with the rise of AI, competition for water and power supplies is pushing construction further into rural areas that often lack the housing and infrastructure to support the hundreds or thousands of temporary workers needed to build hulking warehouses of computer servers.

这迫使开发商越来越多地依赖一种在2010年代页岩油热潮期间流行起来的权宜之计——即所谓的‘工人营地’。

That's forcing developers to increasingly lean on a stop gap solution that was popularized during the shale oil boom of the 2010, sprawling temporary villages known as man camps.

这些临时居住区可以是木结构的两层公寓楼，也可以是集装箱式模块化住宅或配备电力和水源的拖车公园。

These temporary housing villages can vary from wood frame two story apartment buildings to containerized modular homes or trailer parks supplied with electricity and water.

但为了吸引急需的电工、焊工和管道工，开发商们不惜额外投入，提供游戏室、牛排和往返工作地点的班车服务。

But to lure in demand electricians, welders and pipe fitters, developers are going the extra mile and offering game rooms, rib eyes and shuttle rides to work.

这为Target Hospitality和Civio等专注于移动住房的公司创造了有利可图的细分市场。

That's fueling a lucrative niche for companies including Target Hospitality and Civio that specialize in mobile housing.

这实际上是一种间接参与投资的方式，据彭博情报估计，全美有7000亿美元的项目正在规划中，另有1600亿美元的项目已启动。

It's effectively a backdoor play for a share of what Bloomberg Intelligence estimates is 700,000,000,000 of projects in the planning stage and another 160,000,000,000 already underway throughout The US.

这是我见过的最大、最切实可行的项目管线。

It's the largest, most actionable pipeline I've seen.

这些营地谈不上奢华，但绝对比住廉价酒店、睡车里、把每日津贴全花在吃饭上强多了。

The sites aren't the epitome of luxury, but they sure beat what shrink calls hotel F250, sleeping in a vehicle and spending a per diem allowance on food.

通常，与Target Hospitality合作的建筑公司会为员工免费提供住宿、餐饮和配套设施，只要他们还在工作期间。

Typically, the construction companies that work with Target Hospitality offer housing, meals and amenities to their employees for free as long as they're working.

这对已经享受显著加薪的技工来说又是一个吸引力，有些数据中心电工的年收入超过15万美元。

It's an additional lure for tradespeople already enjoying significant pay hikes with some data center electricians making more than 150,000 a year.

很不错。

Not bad at all.

我们是不是该转到泰勒的文章《美国国有化简史》？

Should we kick it over to Tyler's piece, Short History of Nationalization in The United States?

泰勒·科斯格罗夫今天在TBPN通讯的每日观点栏目中撰文。

Tyler Cosgrove took the daily op ed in the TBPN newsletter today.

谢谢，泰勒。

Thank you, Tyler.

他带我们快速回顾了国有化的历史。

And he took us on a whirlwind tour of the history of nationalization.

当然，这件事最近上了新闻，因为Anthropic与国防部之间的对峙仍在继续。

Of course, is in the news as the showdown between Anthropic and the Department of War continues.

我们当然希望未来能形成统一战线，持续展开多方互动，或许近期会有一些让步，但随着事态发展，我们拭目以待。

We're, of course, hoping for a unified front moving forward and lots of back and forth, potentially some capitulation recently, but we will see as the story keeps evolving.

但我们想知道这段历史，以便预测未来。

But we wanna know the history so that we can predict the future.

所以，Tyler，给我们讲讲吧。

So Tyler, take us through it.

关于美国国有化的历史，你学到了什么？

What did you learn about the history of nationalization in The United States?

是的。

Yeah.

这非常有趣。

So it's very interesting.

直接说重点。

Just immediately.

就是这样。

There we go.

完美。

Nailed it.

好的。

Okay.

所以我认为是的。

So I think yeah.

所以不管怎样，比如Anthropic和五角大楼的合作，很明显，这不会是政府交易的最后一次。

So so whichever way, like, the Anthropic Pentagon stuff, like, like, I I think it's very clear that, this is not, the last time government deals will be made.

对吧？

Right?

比如，OpenAI与政府的交易不会是OpenAI与政府的最后一次合作。

Like, the OpenAI deal is not the last deal that OpenAI will make with the government.

是的。

Yeah.

而且之前就已经有讨论了，比如后备方案。

And there was already discussions, like, the backstop.

当后备方案的门槛出现时，大家就产生了一个大问题，就是说，好吧。

And when the backstop gate happened, there was a big question about, like, okay.

这看起来像是看跌，但同时也该有个兜底机制吗？

It seems like, you know, bearish, but also, like, should there be a backstop?

比如，如果这是一种重要的技术的话。

Like, maybe if it's an important technology.

我们以前也这么做过。

Like, we've done this before.

所以给我讲讲一些

So take us through some

我觉得所有实验室的负责人都一直在反复讨论这个问题。

And I I think, like, all the lab leaders are also they've been talking about this again and again.

对吧？

Right?

你也注意到，我在这儿也说过，他们总是对此说得非常模糊。

You've seen I also say in this that, you know, they're they're always, like, extremely vague about it.

对吧？

Right?

他们说，你知道的，我们都得想清楚，得讨论一下该怎么应对这个问题。

They say, you know, we all gotta figure out We gotta talk about figure out what we're gonna do about this.

我们得跟政府沟通，而政府又说，得去跟AI实验室的负责人谈，但没人真正表态。

We gotta talk to the government, and the government are is like, we gotta talk to the AI lab leaders, no and one really says

同意我们应该多交流，这正是TBPN看涨的原因，因为我们一直在谈。

agrees that we should be talking, which is why bullish for TBPN because all we do is talk.

是的。

Yeah.

但但但，基本上，嗯。

But but but, basically, like Yeah.

如果你想想未来的监管，对吧，你就会看到一个光谱，一边是完全什么都不做，这更接近我们现在的情况，是的。

If you think about, like, future regulation, right, you have this this spectrum of, like, we'll do literally nothing, which is, like, closer to to kind of what we have now Yeah.

或者你进行彻底的国有化。

Or you do, like, full scale nationalization.

是的。

Yeah.

所以这里有一个这样的范围。

So there's kind of this spectrum.

所以我很好奇，过去发生了什么？

So so I was curious, like, okay, in the past, like, what happened?

核武器的类比是最极端的例子，就像我可以去建一条铁路。

Is nuclear weapon analogy, like, the most extreme example where it is like like, I can go build a railroad.

铁路，正如你们告诉我们的，以前曾被国有化过，但我法律上完全不能成立一家制造核武器的公司。

Railroads, as you'll tell us, have been nationalized before, but I I legally cannot start a company that builds nuclear weapons at all.

所以这是最彻底的国有化例子。

And so that's like the most nationalized

是的。

Yeah.

我认为还有许多不同的分包商提供最终组成核武器的零部件。

I think also there's a bunch of different subcontractors that supply the components that ultimately

对。

Yes.

但我的意思是，我们最终会陷入人工智能就是核武器的阵营，因为你根本不能训练基础模型，这本质上就像研发核弹头。

But I'm saying is that like, we end up in the in the AI is nuclear weapons camp, like, you cannot train a foundation model because that's, like, the nuclear warhead development, essentially.

这无疑是最极端的例子。

That's, like, the most extreme.

是的。

Yeah.

我认为总体上，这确实是最极端的例子。

I I think broadly that that's that's the most extreme example.

这也就是利奥波德所提到的。

That's also what, you know, Leopold.

对。

Yeah.

在态势感知方面，他有一个项目。

In situational awareness, he has the project.

对吧？

Right?

这简直就是曼哈顿计划。

That that is kind of the the Manhattan Yeah.

项目。

Project Yeah.

你懂的，AI的版本。

You know, version of AI.

对吧？

Right?

是的。

Yeah.

而且AI在2027年也预测了自然化。

And AI 2027 also sort of predicted naturalization.

是的。

Yeah.

是的。

Yeah.

它和它们有时间表。

It and they have the timeline.

它们有，我觉得是Open Brain，还有政府。

They have I think it's Open Brain, and there's also the government.

有这两种竞争性的实验室。

There's kind of these two competing, like, labs.

对吧？

Right?

所以也许它们会保持分离。

So may maybe they kind of stay separate.

也许它们会变成同一件事。

Maybe they come become the same thing.

好的。

Okay.

是的。

Yeah.

但，是的，我可以列出一些在美国历史上有趣的事件，你知道的，

But, yeah, I I can just list out a few, like, interesting, you know, events in in US history that's

至少是这样。

like Yeah.

至少。

At least.

拿你那个最接近国有化的例子吧。

Take your least nationalization adjacent.

你为什么在装死？

Why are you doing crickets?

好吧。

Okay.

所以，那个，是的。

So so the Yeah.

第一个是，比如，联邦银行。

First one is is the, like, the the net like, the federal bank.

对吧？

Right?

所以，是的。

So Yeah.

美国第一银行和第二银行。

First and second bank of the of the United States.

我不确定人们是否真的认为这是国有化。

I I don't know if people really consider this nationalization.

对吧？

Right?

这非常像公共政策，但它是一种公私合营的

It's very much like public policy, but it is like a a public private

是的。

Yeah.

一种政府拥有实际股权的机构。

You know, institution where the government owns, like, actual equity.

是的。

Yeah.

好的。

Okay.

美国成立于1776年。

So America founded 1776.

三十年内，他们就说：我们得把一些东西国有化。

Within thirty years, they're like, we gotta nationalize some stuff.

1791年，美国第一银行被国有化了。

'19 1791, the first bank of The United States is is nationalized.

它最初是一家私营公司，然后政府说：嘿。

So it was it was started as a private company, and then the government said, hey.

我们想持有股份。

We want a stake in this.

是这样吗？

Is that right?

我认为对于这一点，我将国有化定义为：他们并没有真正接管一家私营公司，而是做了通常由私营公司负责的事情。

I I think for this, I I define nationalization as just like maybe they didn't they're not actually taking over a private company, but they're doing something that is usually regulated to private companies.

明白了。

Got it.

好的。

Okay.

对。

Right.

所以即使政府是作为一项计划来发起它

So even even if the government is, like, founding it as a project

是的。

Yeah.

比如曼哈顿计划，那就是一种国有化。

So so, like, Manhattan Project is, like, nationalization.

我认为大多数人都是这么描述的，但当时并没有一家原子弹公司被他们接管。

I think most people describe it as that, but there wasn't an atomic bomb company that they took over.

哦，这倒是真的。

Oh, that's true.

还有像NASA和登月任务这样的例子。

And same thing with, like, the the like, NASA and, like, the moon mission.

阿波罗计划并不是某个私营实体被国有化的案例。

Like, the Apollo program was not some wasn't a private entity that got nationalized.

所以，这可以说是完全由政府项目从零开始的最极端例子。

So so that's sort of the most extreme, like, de novo starts as a government project.

好的。

Okay.

是的。

Yes.

这说得通。

That makes sense.

对。

Yeah.

所以你会看到，在战争期间，这种情况很常见。

So so then you see stuff like during war, you'll see this as a a common occurrence.

比如在内战期间，铁路和电报就出现了。

So during the civil war, there was the railroads and the telegraphs.

是的。

Yep.

这些基本上都被国有化了。

These were essentially nationalized.

对吧？

Right?

1862年，内战时期的铁路和电报法案赋予总统在公共安全需要时接管所有电报线路和铁路的权力。

1862, Civil War Railroad and Telegraphed Act gave the president authority to take possession of any and all telegraph lines and railroads when public safety required it.

有意思。

Interesting.

是的。

Yeah.

然后你就看到，是的。

And then you see Yep.

1914年，出现了阿拉斯加铁路。

1914, there's the Alaskan Railroad.

所以是宏大的阿拉斯加梦想。

So big big Alaska Dream

对我来说。

for me.

是的。

Yeah.

那里发生了什么？

What happened there?

是的。

Yeah.

所以基本上，根本没有真正的私人投资进入阿拉斯加。

So so there basically was not any, like, real real, like, private investment going into Alaska.

而且

And

然后是的。

then Yeah.

政府说，好吧。

The government says, okay.

我们要介入。

We're gonna step in.

是的。

Yeah.

我们要资助它。

We're gonna fund it.

但后来他们实际上一直耗尽了资金，我想直到1985年。

But then we actually they they they ran out until, I think, 1985.

明白了。

Got it.

然后，我认为它又回到了阿拉斯加州。

And then it went back, I I believe, to the the state of Alaska.

哦，明白了。

Oh, okay.

是的。

Yeah.

所以，有点像政府提供资金来启动某项事业，然后又撤出。

So sort of, like, government funding to get something off the ground that then leaves.

也就是说，另一种人们没有考虑到的有趣情况是某种临时的国有化。

And that, I mean, that that is another interesting scenario that I don't think people are are considering is, like, some sort of, like, temporary nationalization.

是的。

Yeah.

我的意思是，是的。

I mean Yeah.

这是安全的。

It's safe.

我们没问题，而且是的。

We're good and, like, yeah.

我们准备好了，可以发布这个了。

We're ready to, like, release this.

所以，我们再把它私有化吧。

So, like, let's privatize it again.

是的。

Yeah.

这种事经常发生。

That happens all the time.

有一个有趣的方面，比如政府拨款。

There's, an interesting aspect where, like, like, grants.

如果你获得了政府的拨款。

If if you get a grant from the government

是的。

Yeah.

这不算国有化。

That's not nationalization.

对吧？

Right?

只是给你钱而已。

Just giving you money.

这是资金支持。

It's funding.

但如果政府以提供资金为条件获得股权，是的。

But if the government takes equity in return for that Yeah.

好的。

Okay.

也许你正在稍微接近真相了。

Maybe you're like getting a little bit closer.

当然。

Sure.

然后在某个时刻，如果他们拥有了全部，那就确实是国有化了。

And then at some point if they own all of it, that's very much nationalization.

对吧？

Right?

所以这其实是一个非常模糊的灰色地带谱系。

So it's this very kind of like, know, blurry gray area spectrum.

嗯。

Mhmm.

但再者，第一次世界大战期间，你看到铁路和电报电话系统被再次国有化了。

But then again, during World War one, you you see the railroads and the telephone telegraphs Yep.

又被国有化了。

Get nationalized again.

是的。

Yep.

这只不过是想把一切整理清楚，确保铁路系统确实非常繁忙。

This is just like, know, just to clean everything up to make sure like, okay, the railroads, they're they're very busy.

有很多事情在同时发生。

There's a lot of, like, stuff going on.

我们需要主要优化军事、建筑这类方面的投入。

We need to basically just optimize just for, military, you know, construction and stuff like this.

是的。

Yeah.

很有趣。

Interesting.

到了1933年，出现了田纳西河流域管理局。

So then 1933, there's the Tennessee Valley Authority.

好的。

Okay.

这基本上是第一个联邦公用事业公司。

So this is basically one of the first, like, you know, federal utility companies.

是的。

Yeah.

这个想法是，我们必须为美国许多农村地区实现电气化。

The the the idea is, like, we gotta electrify a lot of, like, rural parts of The US.

好的。

Okay.

他们从田纳西河流域开始。

They start in, you know, Tennessee Valley.

对吧？

Right?

是的。

Yep.

然后他们随着时间推移逐步扩展。

And then they expand over time.

但通常这还是由私营公司来完成的。

But it it's still, like, this is a usually a private company doing this.

我觉得田纳西河流域管理局最近还在新闻里出现，是因为那个数据中心的扩建吧？

I feel like the Tennessee Valley Authority is still in the news because of like, did the data center build out?

我觉得他们确实运营着至少部分数据中心所依赖的电网之一。

Like, I'm pretty sure that they're they they operate one of the grids that is relevant to at least some of the data centers.

我在新闻里见过他们

Seen them in the

我的意思是，我们可以跳过1942年曼哈顿计划这段。

I mean, we we can skip over the there's a see, 1942, you have Manhattan Project.

这可以说是国有化的典型代表。

This is kind of like the epitome of of nationalization.

每个人回溯时都会提到这个例子。

This is the example everyone goes back to.

是的。

Yep.

你看，二战期间有很多这样的例子。

You see I mean, a lot of stuff during World War two.

对吧？

Right?

比如，他们接管了肉类加工厂和石油工业设施。

There's the seizure of, like, meat packing facilities, petroleum industry facilities.

这些很多都是因为工会与罗斯福政府之间的分歧。

A lot of these were were because of basically union disagreements with the with FDR.

然后他干脆说：好吧，我们就直接接管这些了。

And then he basically said like, okay, we're just gonna seize this.

是的。

Yes.

其中一些很有趣，因为它们发生在战争结束后。

Some of these were interesting because they happened sort of after the war.

比如，肉类加工厂的接管发生在1946年。

Like, the seizure of meat packing facilities happened in 1946.

那肯定是战后的事了。

So that must have been postwar.

需求仍然存在各种波动。

There's still, like, gyrations in demand.

在政府从唯一买家或主要买家角色退出的过程中，仍然存在一些异常现象。

There's still there's still, like, anomalies as you're transitioning out from the government being the only buyer or, like, the major buyer.

工会正试图获得更多控制权，重新为业务注资，并转型为更正常的公司。

The labor unions are trying to, like, get more control and and recapitalize the business and become more of, a normal corporation.

因此，政府介入来适度调整这一过程，这一点很有意思。

And so the government stepping in to sort of, like, massage that is sort of interesting.

我真的很想深入研究一下这个话题。

I I would love to dig into that more.

是的。

Yeah.

然后到了1970年，你就看到了美铁。

Then 1970, you see Amtrak.

对。

Yeah.

所以这些案例中，铁路问题反复出现，这一点也很明显。

So so a lot of these are also, like I mean, you you see rail come up over and over.

是的。

Yeah.

但这是Conrail，你知道的，同一时期，只是稍晚一点。

But this was this and Conrail, which is, you know, same decade a little bit later.

是的。

Yeah.

许多铁路或轨道在某些地区基本上不再被频繁使用。

A lot of the railroads or or, like, rails were basically just, like in some areas, they're they're they're not used as much.

嗯。

Mhmm.

所以它们逐渐开始变得越来越差。

So then they kind of start, like, you know, becoming worse in in quality.

人们不再维护它们。

People aren't keeping them.

它们没有得到保养。

They're they're not maintaining them.

于是政府介入并说，好吧。

So then the government steps in and said, okay.

我们要来运营这个。

We're gonna run this.

1979年，你看到发生了第二次石油危机，

1979, you see you have the, like, kind of second oil shock,

对，

right,

源于伊朗革命。

from the Iranian revolution.

因此克莱斯勒最终被

So Chrysler ends up being

通过贷款担保获得救助。

Rescued via loan guarantees.

是的。

Yeah.

所以不是股权，而是对债务提供保障，并允许他们

So not equity, but a backstop on the debt to And allow them

再说一遍，这难道真的是国有化吗？

again, I mean, this is like, okay, is that really nationalization?

嗯。

Mhmm.

但我

But I

我觉得这仍然是这样。

I think it is still like Yeah.

好的。

Okay.

这是

It's the

政府介入了私人领域。

government stepping into, you know, the the private sphere.

2001年，运输安全管理局，我认为这是国有化的一个好例子。

2001, TSA, I think this is a good example of nationalization.

对吧？

Right?

在那之前，基本上都是私人承包商负责安检工作。

You see before this, you basically have private contractors doing, like, security screening.

2001年，你知道的，9/11事件发生了。

2001, you know, 09/11 happens.

运输安全管理局接管了。

TSA comes in.

我们要全面接管这些工作。

We're gonna take all this over.

到了2008年，发生了一系列事情。

You have in in 2008, there's bunch of stuff.

对吧？

Right?