GPT-6之后的空间：主要AI模型的服务中断

欢迎收听本期播客。

Welcome to the podcast.

今天几乎所有主流AI模型提供商都有重磅消息，其中最大的就是OpenAI发布了GPT-6 Astra，他们甚至称这标志着AGI的开始。

We have some big news from basically every single one of the major AI model providers, the biggest being that OpenAI has released GPT-six Astra, and they're calling this the start of AGI.

当然，他们肯定在大力宣传Astra和新模型的各种亮点，但我觉得确实挺让人惊艳的。

Of course, they're trying to hype up everything going on, everything that they're coming out with with Astra and the new model, but I think it's actually really impressive.

我看了不少演示。

I've been looking at a bunch of demos.

今天节目里我们会展示其中一些。

We'll show some of them on the show today.

还会聊聊其他几个。

We'll talk about a few others.

另外，ChatGPT、Claude和Grok在90分钟内相继宕机了。

ChatGPT, Claude, and Grok also all went down within ninety minutes of each other.

今天关于这件事有很多阴谋论，我们可能还真能搞清楚到底是什么秘密导致它们全部瘫痪的。

Today, there's a lot of conspiracy theories about this, and we might actually have some clarity into what took them all down secret.

剧透警告，可能是埃隆·马斯克干的。

A spoiler alert, it could have been Elon Musk.

然后还有OpeningEye，最近放弃了一笔十亿美元的Cursor交易。

Then we also have OpeningEye that recently walked away from a $1,000,000,000 cursor deal.

这事发生在几天前，但有一些非常劲爆的短信内容，我不得不提一下。

This happened a couple days ago, but there's been some really juicy texts about this that I just had to bring up.

这一切都是因为SpaceX收购了Cursor和OpenAI，而且你知道，显然山姆·奥特曼和埃隆·马斯克之间有很多过节。

This is all because SpaceX bought Cursor and OpenAI and, you know, obviously, Sam Altman and Elon Musk have a lot of beef.

Meta正在为分享训练数据的用户提供MuseSpark 95%的折扣。

Meta is offering a 95% discount on MuseSpark for users that share their training data.

我觉得很多人都有点震惊，比如，你知道，他们就是在搞那种震惊效应，这其实就是Meta在偷我们所有的数据，但这其实是我个人已经和OpenAI做了将近一年的事情。

I think a lot of people are kind of shocked, like, you know, they're they're doing like the shock factor, this is just meta against stealing all of our data, but this is actually something that I've personally been doing with OpenAI for almost a year now.

还有NVIDIA，他们打算收购HuggingFaces。

We also have NVIDIA that is going to acquire HuggingFaces.

我们会讨论一些细节，而且黄仁勋对此有个个人说明，我们也会深入聊聊。

We'll talk about some of the details there, and Jensen Huang has a personal note on that that we'll get into.

在开始之前，我想提一下，如果你还没试过世界上最好的AI模型，我希望你能去aibox.ai上看看它们。

Before we do, I wanted to mention, if you haven't tried all of the best AI models in the world, I would love for you to go check them all out on aibox.ai.

我有一个平台，上面有超过90个顶尖的AI模型。

I have a platform with over 90 of the top AI models.

涵盖音频、视频、图像和文本，而且在AI Box里有一个游乐场，你可以问任何问题，然后还能再访问其他任何模型。

That's everything from audio, video, image, and text, and you have a playground inside of AI Box where that you can ask any question, and then you can access it again to any of the other models.

所以Gemini、Llama、Grok、Claude、ChatGPT，你可以向它们问同样的问题，看看哪个回答最好，而且你还可以在同一个聊天线程里混入音频、视频和图像。

So Gemini, Llama, Grok, Claude, ChatGPT, you ask the same question to all of them and see which one gives the best response, and then you also can mix in audio, video, and image inside of that same chat thread.

想象一下，一个聊天线程，但里面包含了每一个模型和每一种模态。

So imagine one chat thread, but it has every single model and every single modality inside of the thread.

如果你想把这些能力带到Claude co work或ChatGPT work，或者你喜欢的任何其他工具，我们有一个MCPU，基本上就是一个连接，让你可以在Claude或任何其他模型里生成图像、音频和视频。

And if you wanna bring all of that capability over to Claude co work or ChatGPT work, whatever is your, you know, preference as far as using, we have an MCPU, so basically a connection where you can get images, audio, and video generated inside of Claude or any of these other models inside of any other model.

所以如果你想看看，每月只要8.99美元，你就可以在AI Box上开始。

So if you wanna go check it out, it's $8.99 a month, and you can get started over on AI Box.

我们先来聊聊OpenAI那边发生了什么。

Let's talk about what's going on with OpenAI first.

GPT-6的发布当然让人很兴奋。

So the GPT-six release is obviously exciting.

关于它的炒作显然很多，而且在X上绝对是热门话题。

There's obviously a ton of hype on it, and it's definitely trending over on X.

我觉得它今天算是热搜榜上的前几名之一。

It's, I think, one of the top trending things today.

据说他们是用超过一千块GPU来训练的。

Apparently, they trained this on over a thousand GPUs.

他们是在得克萨斯州的Stargate基地做的。

They did it at the Texas Stargate site.

Brockman称这是向通用人工智能的一次代际飞跃，而且这个模型在很多不同的基准测试上表现都非常非常好。

Brockman called it a generational leap towards artificial general intelligence, and the model is doing really, really well on a bunch of different benchmarks.

具体来说，它在ARC AGI 3上拿到了98.6%的成绩，在另一个测试上拿了满分100%，我记得我昨天好像讲过这个。

Specifically, it got 98.6% on the ARC AGI three, and it got a perfect 100% on I kind of I think I covered that yesterday.

基本上，它在黑客攻击方面特别厉害，而且在编程、科学和计算机使用方面也很擅长。

Basically, it's really good at hacking, and it's good at coding science and computer use.

电脑操作是他们重点强调的一大亮点，我觉得这也是当今AI模型最关键的领域之一。我个人最在意的就是电脑操作，因为现在我同时用Claude和Chad的Claude Cowork，还有ChatGPT的Work，我就是让它们帮我在电脑上干活，所以电脑操作基本上就是一切。

Computer use being one of the big things that they've highlighted, and I think one of the winning categories for AI models today, the things that I personally care about are basically computer use because today I'm using both Claude and Chad Claude Cowork and ChatGPT work, and I'm just telling them to go do things on my computer so that computer use is basically everything.

如果这方面不行，那我根本不会去试。

If it's not good at that, it's not something that I'm gonna try.

我听说Grock Bot可能在这方面表现不错。

I've heard that Grock Bot might be good at this.

我还没怎么让它干过活。

I haven't given it a lot of a lot of work.

我通常不是那种最早尝鲜的人。

I'm never usually like the first early adopter.

我会等它稍微有点口碑了，确定它能搞定事情再说。

I'll wait till it's getting a little bit of traction, and I know it can handle things.

我觉得自己好像被坑过。

I feel like I've been burnt.

我想很多人可能都有同感，第一次尝试新东西的时候，结果发现承诺太多，兑现太少，有点被坑的感觉。

I think a lot of people might be in that same boat where we feel a little burnt trying things for the first time with a lot of overpromising and under delivering.

不过，据说这个正在向一些选定的组织推出，他们拥有所谓的“抢先体验”权限。

But apparently, this is rolling out to a bunch of selected organizations that have quote unquote daybreak access.

这主要是针对那个特别擅长黑客攻击的模型。

That's specifically, I think, for the model that's really good at hacking.

然后订阅用户也会获得它。

Then subscribers are gonna get it.

山姆·奥特曼发了一条推文，说我们希望它能开始推动新一代的创业、科学发现和建设。

And Sam Altman had a a tweet where he said, we hope it will begin to enable a new generation of entrepreneurship, scientific discovery, and building.

我们相信它是世界上最好的模型，适用于计算机使用、专业工作、科学、编程、网络安全等等。

We believe it is the best model in the world for computer use, professional work, science, coding, cybersecurity, and more.

显然，他是在自卖自夸。

I mean, obviously, talking up his own book.

他还说，我们花了一些额外的时间来确保我们能够达到这种能力水平所需的安全和一致性标准，但我们认为你会发现等待是值得的。

He also said, it took us some extra time to ensure that we could meet the safety and alignment standards required for this capability level, but we think you'll find it worth the wait.

然后他就谈了一些它得分很高的基准测试。

And then he just talks about some of those benchmarks that it scores high on.

我其实看了一个演示视频，印象特别深刻。视频里一个人坐在房间里，指着屏幕，然后他们演示了一个具体场景：说画一个黄色的圆圈，它就画了一个黄色的圆圈，然后又说把它变成火箭，它就变成了火箭。

I actually watched a demo video, which I was really impressed with, where they essentially have a person sitting down in a room, they're pointing at a screen, and they're like, in this one particular use case, they were like, draw like, draw a yellow circle, it goes and it draws a yellow circle, and then they it says like, and turn that into like a rocket ship, it turns it into a rocket ship.

总之，他们展示了很多不同的应用场景，展示了它能做什么。

Anyways, they have a whole bunch of different, you know, use cases that they're they're showing what it's able to do.

我觉得这个灵感来自1979年一个非常著名的演示，当时有人展示一台电脑，它能跟着他手指的动作，而他正用遥控器指着屏幕。所以他们算是重现了同样的演示概念。说实话，我觉得这是个很好的广告。

I think there was this comes from a very famous demo in 1979 where someone was showing off a computer and it could kind of follow his finger while he was pointing the remote at a screen, and so they kind of replayed the same it's the same kind of demo idea, so honestly, I think it's a good ad.

在这个演示里，他们让那个人直接跟ChatGPT语音对话。顺便说一句，如果你还没用过语音功能，我觉得这是个很大的突破。

And in this one, they had the guy just like talking to ChatGPT voice, which by the way, I think is a big unlock if you haven't used it already.

你不需要坐在那里打字。

You don't need to sit there typing.

你只要开启语音，就可以直接跟它说话，它能实时接收反馈并做出修改。

You can just have voice enabled, you can talk to it, and it can take feedback and make changes live while you're going.

总之，视频里一个人坐在椅子上，面前是一个巨大的投影屏幕。

So in any case, they have a person in a chair looking at a ginormous projector screen in front of them.

他说，画一个黄色的圆圈。

He's like, draw a yellow circle.

它画了一个黄色的圆圈，然后他说，好，把它变成火箭飞船的窗户。

It draws a yellow circle, then he's like, okay, turn that into the window of a rocket ship.

它就把它变成了火箭飞船的窗户。

It makes it into the window of a rocket ship.

他说，好，打开Blender，把那个火箭飞船变成三维渲染的火箭飞船。

He says like, okay, go open Blender and turn that rocket ship into a three d rendering of a rocket ship.

它就照做了。

It goes and does that.

他们有一大堆不同的人在做不同的事情。

They have a whole bunch of different people doing different things.

他们还请了一些名人什么的，你知道，他们说，帮我预约一个网球课，对。

They have some celebrities and stuff, you know, people saying like, go schedule me a a tennis appointment and Right.

他们展示了所有这些不同的用例。

They got all these different use cases.

Blender那个我确实喜欢，因为他把它变成了三维的火箭飞船，然后又把它变成了一个视频游戏，他还解释了那个视频游戏在做什么。

The Blender one I did like because he turned it into a three d rocket ship, and then he turned it into a video game where he explained what the video game was doing.

他们还有律师和其他一堆人在用这个。

They have, like, lawyers and a bunch of other people using it.

他们想展示一堆很酷的用例。

They're trying to show a bunch of cool use cases.

在视频游戏那个例子里，当它正在给他做这个游戏的时候，他说，嘿。

For the video game one, while it's turning making this video game for him, he's like, hey.

顺便，去我上周点过餐的那家餐厅订个餐，然后它就说，好的。

Also, go, you know, order food from that restaurant I ordered from last week, and then it's like, okay.

看看你上周点了什么，然后它就会去下单。

Look at what you ordered last week and it's going to get in it.

他们显然是用DoorDash的插件之类的东西来实现的。

They're obviously using like the the, DoorDash plug in and all that kind of stuff to get it.

然后他说，把我做的那个火箭飞船视频游戏拿来。

And then he said, take that rocket ship video game that I made.

它就像在展示他玩自己的游戏，这显然不太真实，因为这是个很无聊的用例。

It like shows him playing his own video game, which obviously is not very realistic because this is a a boring use case.

你可以做出非常刺激的视频游戏。

You can make really exciting video games.

这个例子挺无聊的。

This one's pretty boring.

但接着他说，嘿，去生成一个三维的，你知道，生成一个可以让我3D打印那艘火箭飞船的文件，然后它就做到了。

But then he's like, hey, go make a three d, you know, make the the right file that I can make a three d print of that rocket ship, and it does it.

所以挺酷的。

So kinda cool.

对吧？

Right?

展示了很多酷炫的用例。

Shows a bunch of cool use cases.

这东西非常强大。

This thing's very powerful.

很多人讨论这个以及我刚才描述的整个演示，就是低延迟的即时工作，所以当你说话时，它很快地执行操作。

The thing that a lot of people are talking about with this and this entire demo I've just explained is the low latency instantaneous work, so as you talk, it's doing things very quickly.

如果你用过ChatGPT的协作功能，你就知道这东西非常强大。

If you've used the ChatGPT co work, you know that this thing's very powerful.

我觉得它和Claude的协作功能不相上下，甚至在某些方面可能稍微好一点。

I think it's on par with with with Claude co work or maybe a little bit better in a a few areas.

我在用浏览器的时候遇到过几次小故障，而Claude不一定会有，所以我不太清楚各自的优缺点是什么。

I've had it had a couple glitches with using my browser that Claude doesn't always have, so I'm not sure what the pros and cons of of that are.

通常我重启电脑和软件之后，它就能正常工作了。

Usually if I restart my computer and restart software, it works.

所以，说实话，可能确实有一些小问题，但除此之外，它真的是一个非常强大的技术产品，我一直在用。

So a couple glitches perhaps, if I'm being a 100% honest, but otherwise, a like really powerful piece of technology, I use it.

如果没有那个特定的故障，我会更频繁地使用它。

I would use it more often if it didn't have that one particular glitch.

很多人在推特上分享他们实际用它做的事情。

A lot of people are tweeting about what they're actually doing with it.

让我感到兴奋的一些事情是，有人让它去Blender里创建一个完整的3D世界，里面有各种角色，它们都在互相交谈，然后它们需要一起合作建造房屋等等，它们一边做一边聊天，看起来真的挺令人惊叹的，它能做到这些事情。

Some of the things that I've been excited about is someone had like, they they told it to go to Blender and make like an entire three d replica of like a world and have all of these like characters that it pulled in and they're all talking to each other and they have to like make it a like a, you know, it's like kinda like a video game, but they're like making their own world, and they have to work together to build like houses and stuff, and they're like doing it and talking, and it's like pretty impressive some of the things that it can do.

另外，我看到有些人，其中一个特别搞笑的是BojonTongueGuzz说的，AGI向公众发布的时候你在哪里。

In addition, I saw some people, well, one of the big funny ones is a BojonTongueGuzz said, where were you when AGI was released to the public.

你知道，就是在炒作说这就是AGI，它已经来了。

You know, it's, you know, hyping up that this is the, this is AGI and that it has arrived.

现在，我觉得这一切中一个重要的因素是，所有的大模型，稍微换个话题，今天所有的大模型，ChatGPT、Claude、Grok，都在90分钟内相继宕机了。

Now, I think what is an important element of all of this is the fact that all of the big models, I guess switching gears slightly to a new story, all of the big models, ChatGPT, Claude, Grock, all went down within ninety minutes of each other today at the same time.

所以一方面，OpenAI发布了这么重磅的产品，然后所有东西都崩溃了。

So on the one hand, we have this big huge release from OpenAI, and then everything kinda crashed.

现在我觉得，特别是OpenAI的宕机，可能只是时间上的巧合，你知道，它崩溃下线是因为他们发布了那个大模型。

Now I think with the OpenAI crash in particular, maybe it was just a coincidental timing that, you know, it crashed and went down because they have their big model release.

也许这给Anthropic带来了大量流量，然后我觉得发生的情况是，因为SpaceX几小时前发了一条推文，说我们对今天早上孟菲斯计算中心中断后你在Grok上可能遇到的问题表示歉意。

Perhaps that pushed a lot of traffic over to Anthropic, and then what I think happened, because we had SpaceX come out with a tweet a couple hours ago where they said, we are sorry for the issues you may have experienced with Grok following the outage at our Memphis Compute Center this morning.

我们还要向受影响的计算合作伙伴道歉，也就是Anthropic，他们每月支付近10亿美元使用同一个孟菲斯设施。

We'd like to also apologize to our impacted compute partners, aka Anthropic, who pays them almost $1,000,000,000 a month to use that same Memphis facility.

他们说所有系统现在已经恢复，运行正常。

They said all systems have have now been restored and are functioning normally.

所以我觉得这可能有点像多米诺骨牌效应，使用量太大了，如果ChatGPT挂了，也许是因为他们发布了一个大模型，可能更受欢迎，不管发生了什么。

So I think it it could kinda be like this domino effect where there's so much usage that if ChatGPT goes down, perhaps because they have a big model that got released, maybe it was more popular, whatever happened.

也许这是Sam Altman搞的一个大营销噱头，故意让ChatGPT在发布大模型后宕机，就是为了说有多少人在用。

Maybe it was a big marketing stunt by Sam Altman to like make ChatGPT go down right after they launch a big model just to say that's how many people are using it.

我知道你在翻白眼，但我记得很清楚，当Sam Altman发了一条推文，我想大概是ChatGPT 5发布的时候，或者可能是ChatGPT 4.5，他发了一条推文说，硅胶熔化时有一种特别的味道，或者类似的，就是他们发布新模型后，说他们的服务器都快熔化了，显然那只是营销手段，挺搞笑的。

And I know you have rolling your eyes, but like I remember distinctly when he distinctly being the word, when Sam Altman put a tweet out, I think it was maybe around ChatGPT five when it came out, or maybe it was ChatGPT 4.5, and he put a tweet out where he was like, there's a distinct smell when you when silicone is melting or something after they like released a new model to say like their servers were like melting, and obviously that's just a marketing thing, and it's like funny or whatever.

但总之，我没忘记那件事，而且不排除他们故意在发布时让服务器崩溃一下。

But so anyways, I I haven't I haven't forgot that and wouldn't put it past that moment to to crash them a little bit on launch.

好吧。

Okay.

我其实不认为真的发生了。

I don't actually think that happened.

但基本上，我们看到了这种多米诺骨牌效应，一个模型可能先挂了，然后大量流量涌向Anthropic。

But, basically, we're having this domino effect where one model probably went down, and then we had probably a lot of traffic go to Anthropic.

可能就把孟菲斯数据中心给搞垮了。

Potentially, took down the Memphis data center.

我不确定。

I'm not sure.

但不管怎样，孟菲斯数据中心宕机了，把Anthropic和Grok都带崩了。

But one way or another, the Memphis data center went down and took out Anthropic and Grok.

有几点要说。

A couple things.

对吧？

Right?

如果你看很多评论，人们会说，这正好证明了不能跟SpaceX对着干，因为所有基础设施都靠他们。

If you look at a lot of the comments on this, people are like, that just proves that, like, you can't bet against SpaceX because all the infrastructure runs on them.

好。

K.

我也觉得这证明了，把所有鸡蛋放在一个基础设施篮子里可能不是个好主意。

I also think it proves that, like, putting all of our eggs in one infrastructural basket is probably not a great idea.

我希望他们能分散一下数据中心的投资组合，但事实就是这样。

I wish probably diversify the data center portfolio pool here, but it is true.

新建数据中心真的很难。

It's really hard to get new data centers built.

SpaceX采取了非常激进的策略。

SpaceX has take a very has taken a very aggressive approach.

对吧？

Right?

他们很快就把新数据中心建起来了，而且你知道，有时候他们会像这个孟菲斯数据中心一样。

They're spinning up new data centers really quickly, and they're you know, sometimes they'll like this Memphis one in particular.

他们整个都建起来了。

They spun the whole thing up.

他们搞了一堆燃气轮机。

They got a bunch of gas turbines.

我是说，他们为了把这个建起来，做了很多了不起的事情。

I mean, they did some really incredible things to get this whole thing spun up.

他们几乎把全美国的燃气轮机都租来了，还把全美国的移动冷却设备都调过来了。

They, like, rented all the gas turbines in America, brought them all the all the mobile cooling in America.

他们把所有的东西都租下来，集中到一个地方，然后把这个数据中心整个建起来了。大家都觉得他们有点疯狂，从理论上讲，这确实花了不少钱。

They rented it all, brought it all to one place, and spun up this whole data center, and everyone said they're kinda crazy, and you could conceptually say this cost them a lot of money.

也许是个巨大的浪费。

Maybe it was a big waste.

因为那些燃气轮机什么的，他们惹上了不少官司，等等等等。

They're getting a lot of lawsuits because of all the gas, yada yada.

但与此同时，他们反过来把算力卖给了Anthropic，一个月十亿美元。

But at the same time, they turned around and are selling that compute back to Anthropic for a billion dollars a month.

所以毫无疑问，他们在这件事上确实赚了钱，但我还是觉得市场需要一些多元化。

So undoubtedly, they're making some money off of this thing, but I do think we need some diversification in the market.

好。

Okay.

说到OpenAI，我绕不开前两天发生的一件事，因为里面爆出了太多劲爆的八卦，就是OpenAI放弃了一笔十亿美元的Cursor交易。

OpenAI, I can't get away from this the OpenAI story without talking about something that happened a couple days ago, but I just so much that so much juicy drama came out of it, which is that OpenAI walked away from a billion dollar Cursor deal.

我之所以说这是十亿美元的Cursor交易，是因为这个合作如果达成，OpenAI在Cursor上的收入预计每年会超过十亿美元。

The reason why I say it's a billion dollar Cursor deal is because this particular partnership having OpenAI on Cursor was projected to bring over $1,000,000,000 annually.

但他们没有继续合作，也没有让Cursor继续使用OpenAI的模型，而是决定和这家公司断绝联系，因为他们显然不喜欢埃隆·马斯克，而马斯克拥有SpaceX，所以他们不想帮他。

But rather than servicing that, rather than letting Cursor have OpenAI models on there, they decided to cut contact with the company because obviously they don't like Elon Musk and he owns SpaceX, so they don't wanna help them out.

说到具体的推文，OpenAI发了一条推文说，我们正在终止与Cursor的合作关系，因为Cursor被SpaceX收购了。

Getting into the actual tweets, it came from OpenAI who had a tweet that they said, we're ending our partnership with Cursor following its acquisition by SpaceX.

根据我们的提议，Cursor直接访问我们模型的权限将在11月12日结束。

Under our proposal, Cursor's direct access to our model would end on November 12.

我觉得很多人都在讨论这件事，然后他们说，我们知道受这个决定影响最大的是那些在Cursor里依赖OpenAI模型的开发者。

I think a lot of people talked about this, and then they're like, we know the the people most affected by this decision are the developers who rely on OpenAI models in Cursor.

我们很关心他们在这次过渡中的体验，我们准备竭尽全力支持他们，等等等等。

We care about their experience in this transition and we're ready to go above and beyond to support them, blah blah blah.

基本上他们就是说，看，来用我们的平台吧。

I mean basically they're like, look, come over and use our platforms.

说实话，我觉得这件事对OpenAI来说并没有很多人说的那么糟糕。

And if I'm being honest, I don't think it's as bad for OpenAI as a lot of people are making it out to be.

他们说，看，他们会损失十亿美元，但我认为有一部分人可能确实喜欢OpenAI的模型。

They're like, look, they're gonna lose a billion dollars, but I think there is a percentage of people that maybe do like OpenAI models.

我就是其中之一。

I'm one of them.

我确实很喜欢用OpenAI的模型。

I like, I really do use OpenAI, like their model.

说实话，他们的模型比Claude多一些小毛病，但我确实喜欢他们的模型，而且用起来效率更高。

They have a couple more glitches than Claude, I'm gonna be honest, but I do like their models and I get more done.

我觉得在同样的智能水平下，OpenAI可能让你用得更多一点，而且我真的很喜欢ChatGPT的工作方式。

I actually think you might get a little bit more usage out of OpenAI the same intelligence level, and I really like ChatGPT work.

所以我觉得，如果你真的喜欢OpenAI，会有一部分人转过去用，但说到底，我觉得很多人，也许有一部分吧，但很多人并没有忠诚度，你就在不同模型之间换来换去，哪个性价比高就用哪个，尤其是如果你是开发者的话。

So I think they'll have some percentage of people move over there if you really like it, but honestly at the end of the day, I don't think a lot of people, maybe there's a percentage, but a lot of people don't have loyalty, you just switch back and forth between whatever model gives you the best bang for your buck, especially if you're a developer.

所以在这方面，我挺好奇的。

So in this regard, I'd be curious.

我觉得他们真正会失去的，是大量的企业客户。

The one thing that I do think they are gonna be losing out on is there's a ton of enterprise customers.

比如我有很多朋友，他们公司都有Cursor的许可证，对吧？

Like I have a lot of friends that work at companies where they have licenses to Cursor, right?

他们公司想要能访问所有顶尖的AI模型，而不只是一个。

Their company wants to be able to access all of the top AI models, not just one.

所以就直接买个Cursor的许可证，如果OpenAI被从Cursor里踢出去，你公司也不会给你批新东西，他们只会说，继续用我们一直在用的Cursor就行了。

So just go get a license to Cursor, and so if OpenAI is getting ripped out of Cursor, like your company isn't approving you for a new thing, they're just like, just keep using Cursor that we've always been using.

你所有的工作流程什么的都已经在那里设置好了。

You have all of your workflow and stuff set up there.

所以他们这么做会失去一大块客户和用户。

So they are gonna lose like a big chunk of customers and users by doing this.

好的。

Okay.

把这些都放一边，我真的很喜欢Cursor的CEO对这个事的回应。

All of that out of the way, I just really liked Cursor's CEO's response to this.

这是Michael Truewell。

This is Michael Truewell.

说实话，他挺狠的，因为打造Cursor的那帮原始团队，现在个个都是千万富翁了。

He's kind of a gangster if I'm being honest because in building Cursor, the like, the OG team, all of them are multi multi millionaires now.

很多早期加入公司的人，现在身家都上亿了。

A lot of the early people that joined the company are, like, 100,000,000.

你知道，有些人赚了一个亿。

You know, some of them made a $100,000,000.

有些人通过这次卖给SpaceX的交易，赚了几千万美元。

Some of them are making, you know, tens of millions of dollars on this whole sale to SpaceX.

所以他对团队来说，结果非常棒。

So he had an incredible outcome for his team.

他是这么说的。

This is what he said.

他说，我们很遗憾看到OpenAI发声明说计划在三个月内阻止Cursor用户访问OpenAI的模型。

Said, we're sorry to see that OpenAI put out a note saying they plan to block Cursor users from accessing OpenAI models in three months.

OpenAI的模型大约占Cursor用户流量的5%，我们正在和OpenAI团队沟通解决这个问题。

OpenAI models serve about 5% of Cursor user traffic, and we're speaking with the OpenAI team to resolve this.

Cursor是OpenAI最早的客户之一。

Cursor was one of the very first users of OpenAI.

我们多年来一直和他们的团队紧密合作，也一直信任他们的平台，把它当作我们业务的中立基础设施。

We've worked closely with their team for years, and we've trusted their platform to be neutral infrastructure for our business.

这话倒也没错。

It's kinda true.

就因为你不喜欢埃隆·马斯克，就这么干，有点不地道。

It's kinda shady just because you don't like Elon Musk.

你不能因为一家公司被另一家公司拥有，而那家公司又被马斯克拥有的公司控制，就不跟它合作了吧。

You can't, like, work with a company that's owned by a company that's owned by a company that he owns.

我的意思是，这真的有点说不过去。

Like, I mean, it's just, like, at some point I don't know.

你看，山姆·奥特曼还在发推，OpenAI 也还在 X 上发帖。

Like, Sam Altman's still tweeting this, and OpenAI is still posting this on X.

他们可以在马斯克拥有的 X 上发帖，却不想通过 Cursor 来服务他们的客户。

So they can they can post it on X, which Elon Musk owns, but they don't wanna service their customers through Cursor.

我真搞不懂。

I don't know.

我觉得这有点离谱。

I think it's kind of ridiculous.

而且我觉得迈克尔有点暗讽他们，说OpenAI的模型在Cursor上只占大约5%的流量。

And I think that Michael threw a little bit of shade at them being like, OpenAI models serve only about 5% of the traffic on Cursor.

所以，你知道，他就像在说，看吧。

So, you know, he's like, look.

这不是世界末日。

It's not the end of the world.

我们还有很多其他模型。

We have lots of other models.

显然，Anthropic在这方面占了上风。

Obviously, Anthropic is goaded in this regard.

不过，是的。

But yeah.

所以有点好笑。

So kinda funny.

希望他们能把这事儿都理顺了。

Hopefully, they get it all sorted out.

Meta 给那些愿意分享训练数据的用户提供了 Spark 二手产品 95% 的折扣。

Meta is offering a 95% discount on used Spark for users who are gonna share their training data.

我是说，Meta 推出的这个新模型确实让人印象深刻。

This is I mean, the the new model that Meta came out with is genuinely impressive.

我喜欢 Alexander Wang。

I like Alexander Wang.

他在 X 上发推文说这事儿，但 Artificial Analysis 也发了一条推文。

He's tweeting about it on X, but there's a a tweet also that came out from artificial analysis.

那是我最常用的基准测试。

That's the number one benchmark I use.

如果你不太了解基准测试，Artificial Analysis 会发布很多不同基准测试的结果，但他们也有自己的基准测试。

If you haven't if you don't know much about benchmarking, artificial analysis publishes the results of a lot of different benchmarks, but they also have their own benchmark.

具体来说，他们会有一个基准测试，你可以免费使用 AI 模型，但会并排问你两个问题，你选你最喜欢的回答，然后他们基本上就做成这个基准测试，而且是盲测的。

Specifically, they'll have like a benchmark where you get free access to AI models, but it asks you two questions side by side and you pick the response you like the most, and then they basically make this benchmark where they're like, and they and it's blind.

你不知道哪个模型说了什么。

You don't know what model said what.

你只需要问一个问题，看两个回答，选你最喜欢的那个。

You just ask a question, see two responses, pick your favorite one.

在这个盲测中，他们可以说，看，用户更喜欢这个模型的回答，而不是那个模型的。

And in that blind test, they're able to say, look, users prefer responses from this model over that model.

所以这算是一个很酷的盲测基准，我真的很喜欢。

So it's kind of a it's a cool blind test benchmark that I really do like.

总之，人工分析说Meta发布了Muse Spark 1.3，这是他们在五个月内发布的第四个Muse Spark模型。

Anyways, artificial analysis said Meta has released Muse Spark 1.3, their fourth Muse Spark model, released in five months.

Muse Spark 1.3 Max，目前对Meta的合作伙伴限量预览，在人工分析智能指数上得了62分，仅次于Claude Fable 5.1和Claude Opus 5。

Muse Muse Spark 1.3 max, which is in limited preview for Meta partners across, scores 62 on the artificial analysis intelligence index behind only Claude Fable 5.1 and Claude Opus five.

这个版本的提升主要来自代理工作和科学能力的改进。

The variant the variant gains come primarily from improvements in agentic work and scientific capability.

好的。

Okay.

所以基本上，他们在这个特定的人工智能分析智能指数上排名第三，只比Claude Opus五稍微落后一点，而且你知道，它比刚出来的Claude Fable 5.1 max低四分。

So basically, they're saying in this particular artificial analysis intelligence index, it's scoring number three, just slightly behind Claude Opus five, and, you know, it's four points behind Claude Fable 5.1 max, which just came out.

所以，我的意思是，我们有来自Anthropic的全新模型，它就在那后面，所以他们真的处于前沿。

So, I mean, we have the brand new model from Anthropic, which is it's like right behind there, and so they're really at the frontier.

我看到很多人在讨论，说大家原本都赌Meta会落后，但看起来他们挺过来了，因为他们有非常出色的人才库。

I've I've seen a lot of people talking about the fact that people kinda were betting on Meta to fall behind, and it seems like they were able to pull through because they have a really incredible talent pool.

关于Meta，你想怎么说都行。

Say whatever you want about Meta.

他们有出色的工程师，显然从Facebook广告和Instagram等赚了很多钱，而且他们从一开始就在计算和AI研究人员上投入了大量资金。

They have incredible engineers, they obviously have a lot of money coming in from Facebook ads and Instagram and all that kind of stuff, and then they have a lot of money they invested in compute and they invested a lot of money into researchers for AI from the very beginning of this.

所以，是的，我对他们能超过Claude Fable五，甚至超过GPT-5.6、Grok 4.6、KimiK-three、GLM-5.3和Gemini 3.8 Flash，并不感到惊讶，他们能超过所有这些。

So, yeah, I'm not I'm not surprised that they were able to pull ahead of, you know, even Claude Fable five, and they were able to pull ahead of GPT-5.6 and Grok 4.6 and KimiK-three and GLM-5.3 and Gemini 3.8 Flash, they're able to pull ahead of all of those.

所以，向Meta的所有人致敬。

So great work to everyone over at Meta.

另外，我之前在这个新闻里提到的另一件事是，他们提供了95%的折扣。

And the other thing that I was mentioning earlier in this news story is that they're offering a 95% discount.

你看，这是如果你分享你的训练数据，或者你讨论的内容作为训练数据的话。

Look, and that's if you're sharing your train your, like, whatever you're talking about as training data.

你看，这是获取更多数据的好方法。

Look, this is a great way to get more data.

而且，别以为只有Meta这么做，有趣的是他们提供了95%的折扣。

And, lest you think Meta's the only person doing this, it's interesting they're offering a 95% discount.

我在一些不同的项目上，包括OpenAI的聊天和API聊天，都享受100%的折扣。

I get a 100% discount on a lot of my, on some different projects that have different chats and API chats for OpenAI.

他们已经这样做了一年多了，所以我在某些工作上有免费的OpenAI访问权限，说实话这是个不错的方式，特别是如果你提供免费增值模式的工具之类的。

They've been doing this for over a year, so I have certain things that I work on that I get free OpenAI access, and it's honestly a good way to go, especially if you're offering like freemium models for tools and stuff.

NVIDIA将收购HuggingFaces。

NVIDIA is going to acquire HuggingFaces.

这个消息已经传了一两天了，但现在看起来已经正式确定了。

This has kind of been in the news for a day or two, but it looks like it's all official now.

传闻已经成真了。

The rumors have officially happened.

黄仁勋在领英上发了帖子。

Jensen Huang posted on LinkedIn.

他说，我很高兴地宣布，英伟达已经同意收购HuggingFace。

He said, I'm excited to announce that NVIDIA has agreed to acquire HuggingFace.

在过去的十年里，Hugging Face团队打造了一些了不起的东西，一个充满活力的开放模型开发者社区之家。

Over the past decade, the Hugging Face team has built something remarkable, a vibrant home for the open model developer community.

超过1800万的开发者、研究人员和创作者使用其平台来分享模型、数据集和应用，而世界各地的公司则用它来发现、评估、定制和部署AI。

More than 18,000,000 developers, researchers, and creators use its platforms to share models, datasets, and applications, while companies around the world use it to discover, evaluate, and customize, and deploy AI.

我们将共同扩展Hugging Face的平台，加强其基础设施，并扩大全球开发者和机构对AI的访问。

Together, we will scale Hugging Face's platforms, strengthen its infrastructure, and expand access to AI for developers and institutions worldwide.

这对英伟达来说是个大动作，我的意思是，这是个很好的举措。

This is a big I mean, is a great move for NVIDIA.

他们通过支持像Hugging Face这样拥有开源模型的项目，获得了大量的好感，而且他们一直非常直言不讳。

They're getting a lot of goodwill in supporting a project like Hugging Face that has open source models, and they've been quite vocal.

过去几个月，当大家都在谈论监管AI时，黄仁勋一直非常明确地表示，他认为开源模型需要公开可访问和可用。

Jensen's been very vocal over the last couple months when everyone's talking about regulating AI that he thinks open source models need to be publicly accessible and available.

他在X上签署了那封大型公开信，还拉了一堆其他行业大佬一起参与。

He kind of signed this big open letter on x and had a bunch of other big industry leaders get in on this.

他在信里说，Hugging Face将继续作为整个AI生态系统的开放平台。

In his note, he said, Hugging Face will remain an open platform for the entire AI ecosystem.

开发者们将继续选择最适合他们工作的模型、框架、云服务、推理提供商和计算平台。

Developers will continue to choose the models, frameworks, cloud, inference providers, and compute platforms that best fit their work in video.

平台不会强制要求使用特定的计算资源来构建或部署。

Compute will not be required to build on or deploy through the platform.

这可能是加进去的一条反垄断条款。

This is probably an anti monopoly, you know, clause that got put in there.

这对大家都有好处，但意思就是他们不会强迫你用NVIDIA。

I mean, it's goodwill for everybody, but it's like they're not gonna force you to use NVIDIA.

不过，它很可能会成为默认选项。

However, it's probably gonna be the default.

很可能会被鼓励使用。

It's probably gonna be encouraged.

它很可能会被轻松链接、轻松访问，并且设置成一种非常无缝的集成。

It's probably gonna be easily linked and easily accessible and set up to kind of like be like a very seamless integration.

不过，你不一定非得用它。

However, you don't have to use it.

而且我觉得，如果他们真要收购的话，总得从中获得一些价值，他们会想办法让它被用起来。

And I think at the end of the if they're gonna buy it, like they gotta get some value out of it, they're gonna try to get some usage out of it.

这很有道理。

That makes a lot of sense.

我对这一切感到高兴。

I'm happy for all of this.

最后他说，当人们能够一起构建时，AI进步最快。

Closing out, he said, AI advances fastest when people can build together.

我们迫不及待地想和这个非凡的社区一起建设未来，等等等等。

We can't wait to build the future with this extraordinary community, yada yada.

微软首席执行官萨蒂亚·纳德拉对这次收购发表了评论。

Satya Nadella, the the CEO of Microsoft, commented on this acquisition.

他说，期待这个生态系统，开放模型继续蓬勃发展，与NVIDIA和Hugging Face以及持续的合作伙伴关系一起成长。

He said, looking forward to this ecosystem with open models continuing to flourish and grow with NVIDIA and Hugging Face and the continued partnership.

这挺酷的。

It's kinda cool.

埃隆·马斯克也发了推文，表示祝贺，或者说，你知道，他很欣赏黄仁勋出来帮助一些开源项目，比如Hugging Face。

Elon Musk also tweeted kind of in congratulatory or saying, like, you know, he appreciated that Jensen Huang was, you know, coming out and and helping with some of the open source projects going on, Hugging Face.

所以我觉得总的来说，大家的反应都很正面。

So I think overall, people have been pretty positive.

很多大科技公司，感觉他们并不觉得这有什么威胁。

A lot of the big tech companies, it doesn't feel like they're they feel threatened by this.

而且NVIDIA显然是所有AI公司的供应商，所以他们都想讨好黄仁勋。

And NVIDIA obviously is a supplier to every AI company, so they all wanna be in his good great in Jensen's good graces.

所以每个人都会祝贺他，我觉得也会支持这件事继续推进。

So everyone's gonna congratulate him and I think kind of support this moving forward.

非常感谢大家今天收听播客。

Thank you so much for tuning into the podcast today, guys.

如果你还没去过的话，一定要去aibox.ai看看，那里有超过90种不同的AI模型，涵盖音频、图像、视频等各个方面。

Make sure if you haven't already to go check out aibox.ai where you get access to over 90 different AI models, everything from audio, image, video.

所有链接都在描述里。

It is all linked in the description.

有很多很酷的功能。

Tons of cool features.

我们下期节目再见。

I'll catch you guys all in the next episode.