Tesla Limits Engineer AI Spending, Microsoft to Cut 4,800 Jobs, Is Claude Enabling a SaaSpocalypse?

欢迎大家收看《信息》频道的TITV。

Welcome everyone to the Information's TITV.

我是罗克特·德鲁。

My name is Rocket Drew.

今天是7月6日，星期一。

Today is Monday, July 6.

在节目开始前，我先给大家带来一条简短的新闻更新。

Before we start the show, I have a quick news update for you.

博通和苹果已同意将双方长期的技术合作延长至2031年。

Broadcom and Apple have agreed to extend their long standing tech collaboration through 2031.

该公司宣布，双方达成了新的多年协议，博通将为苹果开发并供应定制芯片，用于其产品。

The company's announced that they have struck new multi year agreements for Broadcom to develop and supply custom chips for Apple to use in its products.

本期节目，我们将深入了解一下，小公司如何用Claude代码开发自己的AI应用，逐步取代Salesforce的产品。

On today's show, we have an inside look at how small firms are replacing Salesforce products with their own custom AI apps made using Claude code.

此外，《信息》独家报道，特斯拉正在收紧员工在AI上的开支。

Plus, the information published exclusive reporting that Tesla is cracking down on employee AI spend.

我们会采访揭露这则新闻的记者。

We'll talk to the reporter who broke that story.

另一篇独家报道中，我们的微软记者拿到了一份内部备忘录，详细说明了公司打算如何全面改造它的AI应用。

And in another exclusive report, the information's Microsoft reporter got his hands on a memo detailing how the company plans to overhaul its AI app.

他将聊聊这则报道，还有微软今天在Xbox和销售部门宣布的裁员计划。

He'll talk about that story, plus the layoffs that Microsoft announced today in its Xbox and Sales units.

今天这期节目内容非常精彩，我们马上开始。

It's going be a great show, so let's get right into it.

随着AI在编程方面越来越强，许多软件投资者开始担心，一大批传统软件公司可能会在所谓的‘SaaS末日’中被淘汰。

As AI has gotten better at coding, it has spooked a lot of software investors into thinking that a lot of legacy software companies could become obsolete in a so called SaaSpocalypse.

我的同事劳拉·布拉顿，负责我们《应用AI》通讯，今天发表了一篇文章，指出小客户才是这场变革的先锋。

My colleague Laura Bratton, who authors our Applied AI newsletter, published a story today showing that smaller customers are the tip of the spear.

他们已经在用AI替代Salesforce了。

They're already using AI to replace Salesforce.

欢迎劳拉来到节目。

Welcome on the show, Laura.

嘿，火箭。

Hey, Rocket.

很高兴能来。

Good to be here.

当然。

Definitely.

我特别期待继续聊聊关于SaaSpocalypse的话题。

I'm really excited to continue our conversations about the SaaSpocalypse.

上次我们聊这个的时候，我还在想，五年后还会有软件公司吗？

I think the last time we talked about this, I was taking the position that, I don't know, will there even be software companies in five years?

你当时劝我别想太多，说软件公司还是有它的位置的。

And you were talking me off the ledge saying, well, software companies still have a place.

所以我很好奇，这个新故事有没有让你对SaaSpocalypse的全局看法有所改变？不过我们先不急着谈这个。

So I'm curious to see if this story has changed your bigger picture thinking about what's happening with the SaaS apocalypse, but we'll get to that.

我们这档节目经常聊全球顶尖科技公司，但你是怎么想到把焦点放在小企业身上的？

We talk about some of the biggest tech companies in the world on this show, but how did you get the idea to focus on small businesses for this story?

我在社交媒体和领英上看到很多创业公司和小企业高管在讨论，说他们已经这么做了。

I've been hearing a lot of chatter on social media LinkedIn, from founders and executives of smaller companies that have said they've done this.

此外，我在旧金山的各种活动中，还有在纽约工作时，也亲自遇到不少创始人，他们都表示自己在这么做。

And then I also have just met a lot of founders in person, at different events around San Francisco and when I was working from New York who said they'd been doing this.

所以我意识到这已经形成了一种趋势。

And so I knew that it was a trend happening.

但我还需要一些具体的案例，来把这件事说清楚。

I just needed some concrete examples to really drive the story home.

于是，这才有了这次的调查报道。

And so that's that's what led to this reporting.

明白了。

Got it.

那这些小企业是怎么在人工智能时代调整他们的软件使用方式的呢？

So what did you find with respect to how these smaller businesses are changing their software use in the AI era?

是的。

Yeah.

我了解到，很多创始人和高管正在使用像 Replit、Lovable、Anthropic 的 Claude Code 和 Cursor 这类新兴 AI 初创公司的软件，来开发内部应用，替代他们原本用的 Salesforce、HubSpot 以及其他小型客户管理工具。

I heard from founders and executives that they are using software from new AI startups like Replit and Lovable as well as Anthropics Claude Code and Cursor to create their own internal apps that replace their customer relationship management software from Salesforce, HubSpot, and other smaller apps.

这些应用每年的花费大概在四万美元左右。

And that those apps usually cost them around 40 ks annually.

所以，他们通常能省下这笔开销。

And so those are the cost savings they're typically seeing.

但对于二十到七十人的小公司来说，这笔节省相当可观。

But for small companies that are between twenty and seventy people, that's big saving.

你再详细说说这些节省是怎么来的。

Tell us more about the savings.

我猜 AI 的使用费用可能很高吧。

I guess AI bills can be very expensive.

现在我们经常听到这种说法。

That's something we're hearing a lot now.

但你提到，传统 SaaS 软件的实际拥有成本，往往远高于标价。

But you wrote that the true cost of ownership for more traditional SaaS can be a lot higher than the sticker price.

那人们具体能省下多少钱呢？

So what kind of savings are people seeing?

是的。

Yeah.

我举个例子吧，西雅图海狼队。

I guess one example I can give is the Seattle Seawolves.

这是西雅图的一支职业橄榄球队。

It's a pro rugby team, in Seattle, of course.

他们大概有七十名全职员工，花了四个月时间，让四个工程师开发了一个自己的应用，取代了原来的票务销售系统和 Salesforce 的客户关系管理软件。

And, they have about 70 full time staffers, and they used four engineers over the course of four months to create their own app that replaced their ticketing sales system and Salesforce's CRM.

对于一家七十人的小公司来说，创始人告诉我，他们根本不需要 CRM 和 AXS 那么复杂的全套功能——他们之前得把两个系统的数据手动合并到表格里，再用这些表格里的数据来决定怎么调整观众的座位安排，或者怎么向他们推荐更高级的观赛体验，这就是他们原来的做法。

And for a small company with 70 people, I mean, what the founder said is that they didn't really need all the bells and whistles of a CRM and AXS because what they were having to do is merge data from these two separate apps into spreadsheets and then use data from their spreadsheets to inform, you know, how they changed their fans experiences by placing them at different points in the arena and, know, how they could upsell them certain fan experiences is the way they described it to me.

现在有了自己的应用，他们能实时看到观众的购票数据，所有客户管理的功能都整合在这个新开发的系统里了。

And now that they have their own app, they can see sort of data about those fans and which tickets that are buying in real time and have all the functionalities of a CRM in this, you know, new app that they created.

所以在这个案例里，他们的收入在增长，软件开支也降了差不多一半。

So I think that in this example, in particular, they're seeing revenue go up, and they're also seeing their software budget, their software costs decrease by about 50%.

我觉得这个例子特别好。

And so I thought that that was a really good example.

他没告诉我他们每年的软件预算具体是多少。

He didn't give me an exact annual software budget.

另外，我听其他客户说，他们通过自己开发解决方案，每年在Salesforce上的支出省了大约四万美元，而这些自建系统的年维护成本只有几百美元。

And then, you know, from other customers that I heard, they were saving around 40 k annually in payments to Salesforce by creating their own solutions that cost in the hundreds of dollars per year to maintain.

确实很有趣，自己开发不仅更便宜，还更好，因为你可以按需打造真正需要的功能。

Yeah, it's super interesting that building your own can be not just cheaper but also better because you can build whatever features, the exact features that you're looking for in an app.

这个例子真的很有意思。

So that's a really interesting example.

你还能再讲一个让你印象特别深的公司案例吗？

Was there another example of a company that you want to walk us through, one that was especially interesting to you from this story?

是的。

Yeah.

我想提一下，不只是Salesforce这一个例子。

I want to call out because it wasn't just Salesforce.

有一家初创公司叫 Arcade Dev，特别有意思。

So there's a startup arcade.

他们和 Anthropic 合作，参与了我们都知道的 MCP 协议的开发。

Dev, and they're really interesting because they helped Anthropic create the they collaborated with Anthropic to create part of the MCP protocol that we all know.

最近他们拿到了数千万美元的投资。

And they recently had a big raise in the tens of millions.

所以他们已经是 AI 领域里不可忽视的一员了。

And so they're they're really part of the AI conversation.

他们告诉我，换掉 HubSpot 并不是因为价格，而是实在受不了那个用户界面。

And they told me that they replaced HubSpot not necessarily because of the cost, but because they were just tired of the user interface.

他们想要一个高度定制的系统，用 AI 代理直接抓取数据，不用每次都登录网站。

And they wanted to have something super customized where they use AI agents to pull data rather than having to log into a site.

我觉得很明显，这些软件巨头现在都希望你用 AI 代理来和数据互动。

And I think obviously the argument can be made that all of these software giants, they want you to use AI agents to interact with data now.

他们都在数据上面加了一层 AI 接口。

They've created AI layers on top of data.

我们看到Salesforce推出了无头360，还说用户界面已经过时了。

We saw Salesforce launch headless three sixty and declare that, you know, they were like, the user interface is dead.

他们要给你一种不用登录就能直接操作数据的方式。

We're going to give you, you know, a way to interact with our data without having to log in.

但我觉得，刚才那个arcade的例子说明，这些软件巨头正不得不重新思考如何改造自己的用户界面。

But I think that the exam the arcade example shows how software giants are having to reckon with how they change their user interface.

这不光是成本的问题。

It's not only about cost.

挺有意思的是，不知道这种趋势会不会也影响硅谷以外的其他公司。

And, you know, it'll be interesting to see whether that applies to all the other companies outside of Silicon Valley.

他们是想用AI代理来操作数据，还是依然愿意用我们原来的界面？

Like if they want to use AI agents to interact with their data or if they still want to use our interface.

我觉得也有公司确实更愿意用我们原来的界面。

I think that an argument can be made that some companies do want to use our interface.

因为他们技术没那么强。

They're less technical.

所以这些多样化的例子真有意思，它们反映出小公司们都在想些什么。

So really interesting how this broad range of examples kind of shows what smaller companies are thinking about.

对，我觉得定制化的用户界面其实是SaaS危机中被严重低估的一个方面。

Yeah, I think just sort of customized UIs are a really underappreciated aspect of the SaaS apocalypse.

我可以想象，用不了多久，我手机里的每个应用都会根据我个人喜欢的风格来定制界面。

Like, I imagine a day not too long from now where every app on my phone has a customized UI just based on the look and feel that I prefer.

有意思的是，对你提到的这些公司来说，用户界面已经是个关键因素了。

So it's interesting that the UI is already a factor for the companies that you're talking to.

更有趣的是，这次是HubSpot，而不是Salesforce。

It's interesting also that it was HubSpot and not Salesforce.

我们每次聊这些话题，好像总是绕不开Salesforce。

I feel like it's always Salesforce that we're talking about in these conversations.

Salesforce有什么特别之处吗？

Is there something special about Salesforce?

它是不是其他SaaS公司的风向标？

Is it a leading indicator for other SaaS companies?

为什么我们总在聊它呢？

Why do we end up talking about it so much?

我觉得主要是因为它本质上是一款客户关系管理软件。

I think because it's a customer relationship management software primarily.

它最出名的就是CRM功能。

I mean, it's known for its CRM.

虽然它现在有好多不同的应用，但人们一提到Salesforce，首先想到的还是CRM。

That's what it obviously has a lot of different apps, but that's what people think of when they think of Salesforce.

很多小公司当初也是从Salesforce买的这个功能。

And that's what a lot of small companies bought from Salesforce.

所以我觉得，正是因为Salesforce在这个领域里是标杆，我们才总听到它。

So I think that because Salesforce is the leading example in this particular category, that's why we hear about it so much.

而且它本身就是个软件巨头。

And it's just a massive software giant.

你知道吗？它的CEO马克·贝尼奥夫早在2000年初就率先提出了软件即服务这个概念。

You know, its CEO, Marc Benioff, pioneered SaaS in general, in the early 2000s, the idea of software as a service.

所以我觉得，这就是为什么我们总能听到它被当作例子。

So I think that that's why it's always an example that we hear about.

其实我听一些顾问说，像 Salesforce 这样的软件，并不是企业换系统时的首选。

I think, actually, what I heard from consultants is that ripping out software like Salesforce isn't like Salesforce's CRM isn't necessarily the first choice.

企业真正想做的，是在现有系统基础上搭建 AI，把它当作核心数据源。

What companies really wanna do is build AI on top of the software as like a system of record.

所以，我这次报道并不能证明 SaaS 正在被大规模取代。

So I think that my reporting for this story doesn't necessarily prove that the SaaS populace is happening.

我觉得它更说明了一点：很多小公司其实并不需要这些系统，却花了不少钱，现在想把它们换掉。

I think it shows that there is interest in small companies ripping out software systems that arguably they didn't need in the first place, and that they spent a lot of money on.

不过，我接触的顾问都提到，虽然换系统能省不少钱，但如果你能基于现有软件开发定制化的 AI 助手，那才是更好的选择。

But, you know, consultants I spoke to said that while there can be a lot of cost savings, it's really a better alternative if you have custom AI agents that you can use on top of your software.

因为终止这些合作关系、解除合同的成本其实相当高。

Because the cost of ending, you know, those relationships and ending those contracts is actually pretty high.

而且如果系统里存着大量历史数据，那就更值得把它当作核心数据源来用。

And if you have historical data in them, it makes sense to use them as kind of a system of record.

是的

Yeah.

这说得通。

That makes sense.

我想说的是，如果Salesforce真的出问题了，这在多大程度上说明其他软件公司也都面临危机？

The thing I'm trying to get at is if Salesforce is in trouble, how much does that suggest that all other software companies are in trouble?

听起来，一方面，Salesforce可能是最先被放弃的，因为你为一堆用不上的功能付了钱。

It sounds like on the one hand, Salesforce might be first on the chopping block because you're paying for all these features you don't want.

这价格确实不低。

It's kind of expensive.

但另一方面，要把Salesforce彻底移除并替换掉，过程也挺折腾的。

On the other hand, it sounds like it's kind of disruptive to rip out Salesforce and replace it.

所以如果人们愿意花这么大代价去换，那也许其他软件公司也真不安全了。

So if people are willing to go to those lengths, maybe other software companies are in trouble too.

对。

Yeah.

我觉得现在还不能说他们真的出问题了，因为据我了解，我在跟一些为大小公司做咨询的人聊天时，他们提到大公司基本都没在替换自己最主要的软件供应商。

I don't think we can say that we're they're in trouble yet just because if anything, you know, when I was exploring these conversations chatting with consultants that work with both small and large companies, they said that large companies are largely not replacing their biggest software providers.

但我认为，这已经是一个早期信号，表明大公司可能会开始考虑这类替代方案。

But I think that this is an early signal that larger companies may begin to explore these kinds of options.

而且我已经在其他报道中听到，一些大公司正在把部分工作流程从他们的核心系统中移出来，不管是Salesforce、SAP还是ServiceNow，他们正把任务交给AI代理来自动化，以前这些事都是靠买大软件公司的服务来做的。

And we're already hearing from bigger companies for other stories I've reported that they're moving some workflows out of their systems of record, whether it's Salesforce, SAP, ServiceNow, they're moving workloads out of those big software providers, and using AI agents to automate some tasks that they used to pay large software providers to do.

所以我觉得，软件供应商的定价能力正面临压力。

So I think we're seeing pricing power from software providers is under pressure.

而且小公司也在积极寻找替代方案。

And, know, smaller companies are exploring alternatives.

至于这会不会导致SaaS行业的崩盘，现在还不好说。

Whether that will lead to the SaaS apocalypse or not, I think is to be seen.

你也可以从另一个角度看，如果错过了在这些小公司早期就部署软件的机会，将来它们可能成长为像OpenAI或Anthropic这样的巨头，那可就真是错失了良机。

And you can also make the argument that, you know, it is a missed opportunity to have your software installed at these smaller companies that could later arguably become the OpenAI and Anthropics of the world.

所以，这会不会是未来收入增长的一个潜在机会呢？

So is there a missed opportunity for future revenue growth?

是的

Yeah.

这确实有道理。

That's totally fair.

但目前来看，Salesforce失去一些小客户合同，还不至于造成太大影响。

But for now, it's not too devastating for Salesforce to be losing some of these smaller contracts.

不过，这些小公司的做法，能不能告诉我们，那些Salesforce更在意的大客户接下来会怎么走？

But does the experience of these smaller companies, does it tell us anything about what's going to happen next with the larger customers that say Salesforce cares more about?

是的，我觉得这确实说明了一些问题。我采访过Replit，这是一家帮企业开发定制应用的初创公司。

Yeah, I think that it does show, you know, I spoke to Replit, which is, you know, a startup that helps companies build custom applications.

他们说，尤其是那些有替代方案的小公司，绝大多数都在直接淘汰旧软件，改用全新的定制解决方案。

And they said that, you know, they're seeing, in particular, small companies that have software replacements, the large majority of smaller software companies are just ripping out their older software and replacing them with new custom solutions.

他们认为，这种趋势最终也会蔓延到大公司身上。

And they think that this will eventually lead to larger companies doing the same thing.

当然，不同咨询公司和帮助企业规划软件支出的机构，对这一点看法不一。

That's obviously a point of contention among different consultancies and companies that help Fortune 500 firms and other small companies with their software spend.

所以，我觉得这到底是不是一个更广泛的趋势，现在还不好说。

So yeah, I think it remains to be seen whether this is a broader trend or not.

但我的报道显示，至少在小范围内，这种做法是完全可以实现的。

But I think that my reporting is an early signal that it's possible to be done at a small scale.

而且我觉得，这种模式完全可以推广到更大的规模。

And I think it's, know, it could be replicated at a larger scale.

比如，我采访了一家法国生物制药公司赛诺菲，他们说正计划把80%的工作负载从ServiceNow和其他软件系统中移出来，转而使用一套他们自己开发的定制化工作流程——这套流程基于他们内部打造的AI助手平台，名叫Concierge，是用Cursor和Claude代码构建的。

For example, I spoke to a French biopharmaceutical company, Sanofi, and they said that they're aiming to take 80% of their workloads out of ServiceNow and other software apps and move them into kind of a more customized workflow that consists of AI agents they built, they built this custom in house platform for employees to use called Concierge, and they built it with Cursor and Claude code.

这些AI助手还会和一家叫Elementum的初创公司提供的AI助手，以及Snowflake里的数据进行交互。

And those AI agents then talk to AI agents from a startup called Elementum and their data in Snowflake.

同时，他们也接入了SAP提供的AI助手，虽然这整个体系看起来错综复杂，但正是靠这些工具，他们把过去依赖大型软件系统的大量流程，逐步迁移到了更灵活的定制化方案中。

And AI agents that they're provided by SAP, which I know is just this tangled web of AI agents, but they're using that to take a lot of workflows they used to have in larger software systems and move them into a more custom process.

这可是个巨大的威胁。

And that's a massive threat.

我的意思是，这背后涉及的年均支出至少上千万美元。

I mean, it represents 10s of million, at least 10s of millions of dollars annually.

每年一千万欧元。

€10,000,000 annually.

不过呢，

But Yeah,

没错。

Absolutely.

听到你从一线看到的这些进展，真是让人着迷。

Well, it's fascinating to hear what you're seeing on the front lines of this development.

下次我们得好好聊聊那80%里具体包含什么，他们决定从Salesforce里去掉哪些功能，又留下哪些。

Next time we'll have to talk about what is in that 80% and what are they choosing to remove from Salesforce and what gets to stay for now.

再次感谢你做客节目，劳拉。

So thanks again for coming on the show, Laura.

特斯拉现在开始限制员工使用AI。

Tesla is now capping employees' AI usage.

尽管这家公司的一部分战略仍然依赖于AI技术。

Even as the company is banking part of its strategy on technology that leverages AI.

我的同事格蕾丝·凯负责报道埃隆·马斯克的商业帝国，她挖到了这个消息的内部细节。

My colleague, Grace Kay, who covers the Elon Musk Business Empire, scooped inside details of that story.

我现在请她来分享她了解到的情况。

I wanna bring her on now to share what she knows.

欢迎来到节目，格蕾丝。

Welcome on the show, Grace.

你好。

Hi.

是的。

Yeah.

上个月初，特斯拉的IT部门通知员工，将对AI使用支出设为每周200美元的上限，这是公司首次推出此类限制。

So we saw that, earlier last month, Tesla's IT department notified employees that they were introducing a $200 weekly cap on AI spending, which is the first cap they've introduced at the company so far.

对。

Yeah.

我觉得现在看到这种情况真的很不寻常。

I think that's pretty remarkable to see right now.

这确实是时代变化的信号。

Definitely a sign of the times.

这项新政策背后的理由是什么？

What's the rationale behind this new policy?

这明显是为了控制开支。

It's definitely targeted towards cutting spending.

我接触过一些人，听说工程团队，尤其是软件工程师，每周在AI令牌上的花费高达好几千美元。

I've talked to some people who saw that a lot of people on the engineering side, software engineers in particular, were spending upwards of several thousand dollars a week on AI tokens.

所以确实是在大幅削减。

So definitely cutting back.

这在行业内也是一个更大的趋势。

It's also a larger trend in the industry.

我这里的同事乔迪·曼一直在深入报道Meta公司这方面的情况。

My colleague here, Jody Mann, has been reporting on this very extensively with Meta.

现在整个行业都在推动员工更多地使用AI。

There's been this kind of trend of all the companies are pushing people to use more AI.

后来他们发现，由于大家把AI令牌用到极限，还像玩游戏一样刷用量，成本实在太高了。

And then, you know, they've realized that with token maxing and people kind of gamifying the system that it's very expensive, you know, these AI tokens.

所以我们得更明智地使用这些工具。

And, perhaps we need to be a little bit more wise about how we use them.

在特斯拉，这看起来就像是给AI使用设了个上限。

At Tesla, this looks like an AI cap.

明白了。

Got it.

现在每周上限是200美元。

So $200 per week now.

以前的上限是多少？

What was the cap before?

之前有上限吗？

Was there one?

据我所知，以前并没有设过上限。

There wasn't a cap from my understanding.

那时候简直谁都能随便用。

It was kind of a free for all.

有个排行榜，大家都能看到各自的令牌使用量。

There's this leaderboard where people could see token usage.

甚至还能看到你们部门里谁用得最多。

You could even see who the top token users were in your department.

这样一来，我觉得他们其实是在鼓励大家多用令牌。

And then that way, I think they were kind of encouraging people to use more tokens.

这个排行榜现在员工还能看到，但效果好像反过来了，倒像是个‘尴尬榜’，让人开始想：我这么用，公司得花多少钱？能不能省着点？

That leaderboard is still accessible to employees, and I think it's kind of having the reverse effect now of like maybe a board of shame or, you know, maybe there's this idea that like, okay, here is how much I'm costing the company and, what ways can I cut back?

要是我是个工程师，肯定一头雾水。

If I was an engineer, I'd be so confused.

这变化也太快了。

It's like whiplash.

昨天还让我多用，今天又让我少用。

It's like one day you want me to use it more, one day you want me to use it less.

我肯定会一头雾水。

I'd be so confused.

你觉得工程师们会满足于每周只给200美元的令牌额度吗？毕竟以前他们花的可多得多。

Do you think engineers will be satisfied with just $200 of tokens per week if they were sounds like spending a lot more before?

是的。

Yeah.

我觉得这得看情况。

I think it really depends.

很明显，对那些每周花几千美元的人来说，这是一次大幅削减。

I mean, obviously, for those people who are spending thousands of dollars a week, this is a substantial cutback.

但特斯拉也给员工开了些特例。

But Tesla has also allowed employees, you know, some exemptions.

也就是说，他们可以申请超过每周200美元的额度，但必须得到上级批准。

So, you know, they can get more than that $200 a week, but it has to be approved by higher ups.

这样一来，特斯拉能更好地掌控局面，而不是像以前那样放任自流，想花多少就花多少，还几乎没什么约束。

So this way I think it gives Tesla a little bit more control versus kind of having a free for all where, you know, people could spend as much as they wanted and, you know, there didn't seem to be as many consequences.

现在你真得好好想想自己花的钱，还有是怎么用AI的。

Now, you know, you really have to think about what you're spending and like how how you're using AI.

你觉得申请豁免会很难吗？还是他们只是想在流程里加点障碍，让你多走一步？

Do you think it will be hard to get that exemption or do you think they just wanna add a little bit of friction in the process so there's one more hoop you have to jump through?

是的。

Yeah.

这个不好说。

It's it's hard to say.

我觉得不会太难，毕竟特斯拉一直在推动大家多用AI。

I I don't think it'll be that difficult because, obviously, Tesla has been pushing them to use more AI.

不过确实，这让他们对员工的控制更强了。

But yeah, it definitely gives them more control over employees.

明白了。

Got it.

我猜员工们花掉大量令牌的一个方式，就是用来买Claude这类代码助手。

I imagine one way that employees are burning a lot of these tokens is paying for like Claude code and these kinds of coding agents.

我也在想，Grok 在这里面占什么位置。

I also wonder where Grok fits into this.

这个政策下，Grok 会不会有特殊待遇？

Like, does it get any special treatment under this policy?

是的。

Yeah.

对。

Yeah.

这确实挺有意思的一点。

That was one interesting thing.

特斯拉的一个特殊之处在于，因为你跟埃隆·马斯克的其他公司联系紧密，特斯拉员工能提前试用 Grok 的产品。

So a unique thing about working at Tesla is because you're so connected to Elon Musk's other companies, people at Tesla get to beta test early versions of Grok products.

今年早些时候，他们就在测试 Grok Build，那是个功能更强的开发者平台。

Earlier this year, they were testing Grok Build, which is like this larger developer platform.

现在他们还在测试其他 Grok 产品。

They're still testing different Grok products.

而且他们还直接和X AI的产品负责人安德鲁·米利奇合作，他经常跟特斯拉的员工沟通，收集反馈，甚至帮忙解决技术问题。

And actually working directly with X AI's product lead, Andrew Millich, who is talking to people, you know, at Tesla, getting feedback, even troubleshooting issues.

所以Grok的测试版产品是免费的，不计入那200美元的限额内。

So with Grok beta versions of products, those are free, so they're not included in that, $200 limit.

明白了。

Got it.

但仅限于测试版。

But only the beta version.

一旦产品正式上线，就连特斯拉的员工也得付费了，大概是这样。

So once the product is released, then even people at Tesla would have to pay for it presumably.

是的。

Yes.

对。

Yes.

我的理解就是这样。

That's my understanding.

你觉得人们会用吗？

Do you think people mhmm.

去用吗？

Go for

其实挺有意思的，虽然beta版本是免费的，也可能有人鼓励特斯拉员工多用Grok，但我接触过的那些特斯拉工程师，其实并不怎么喜欢Grok。

Oh, I was gonna say it's it's interesting too because, like, even though beta versions are free and, you know, there might be, you know, a push to use more Grok products at Tesla, the Tesla engineers from who the ones I've talked to at least are not big Grok fans.

对。

Mhmm.

看起来Claude Code和Codex反而更受欢迎一些。

And it seems like, you know, Claude Code and Codex are kind of the more popular products.

明白了。

Got it.

好的。

Okay.

这么说来，他们可能没指望上特斯拉员工提供足够的beta测试反馈。

So it sounds like they're maybe not getting all of the beta testing that they're hoping for from Tesla employees.

是的。

Yeah.

看起来，尽管可能有某种激励，因为beta产品没有使用限制，但特斯拉的工程师们并不一定喜欢Grok。

It seems like, you know, even though maybe there is an incentive because there's not that limit on the beta products that, you know, Tesla engineers just aren't a fan necessarily.

这里顺便问一下，过去特斯拉在AI使用上是否更偏向埃隆旗下公司的产品？

For context here, has Tesla's AI usage favored products from Elon's companies in the past?

据我所知，并没有。

Not from my understanding.

他们确实合作得很密切。

So, you know, they've collaborated a lot.

特斯拉的车里已经集成了Grok，而且他们还在共同推进一个叫MacroHard的大型办公AI代理项目。

Tesla has Grok in their cars, you know, and they are working together on MacroHard, which is this larger white collar AI agent project.

他们合作很多，但我没听说有工程师被强制使用Grok产品之类的。

So they've they've worked together a lot, but like, I haven't seen any engineers like saying that they are being forced to use Grock products or anything like that.

完全是自愿的。

It's totally voluntary.

明白了。

Got it.

我觉得这个政策变化挺惊人的，因为在我印象里，埃隆简直就是AI的代言人。

I mean, I think this policy change is pretty remarkable because I think of Elon as such a poster child for AI.

他就是那种典型的AI派人物。

He's like one of the AI guys.

我本来以为他会说，AI让人的效率提升这么多，我们愿意砸钱，因为回报实在太大了。

I expected that he would be saying AI is making people so productive that we're happy to throw money at it because they're getting so much value out of it.

那这个政策变化该怎么解释呢？

How do I square this with the policy change?

他说，每周超过200美元，对大多数人来说就不值了？

He's saying, Past $200 a week, it's not worth it for most people?

是的。

Yeah.

对。

Yeah.

这也很有意思，因为我们都看到马斯克一直在发公司全员邮件，谈AI代理。

That's an interesting thing too because we saw that Elon Musk has actually been emailing, sending company wide emails about AI agents.

他还发了链接，教大家怎么使用AI代理。

He sent them a link about how to use AI agents.

他鼓励大家用Composer，也就是Cursor的编程模型。

He's he sent them a link kind of encouraging them to use Composer, which is Cursor's coding model.

而我们知道，Cursor和XAI正在合并，Cursor要加入SpaceX了。

And, you know, as we know, Cursor and X AI, you know, Cursor's joining SpaceX.

所以很明显，他一直在推动AI的使用。

So he's definitely been, like, encouraging AI use.

但我认为，特斯拉一向非常注重成本，这是我的理解。

But I think, you know, Tesla is always very cost conscious is my understanding.

我觉得这个限额，其实就是他们想稍微多掌控一点的方式。

And, you know, I think this limit is just one way for them to have a little bit more control.

明白了。

Got it.

这确实是个挺有意思的进展。

Well, it's a fascinating development.

谢谢你能来参加节目，Grace，为我们详细解读了这些情况。

So thanks for coming on, Grace, and walking us through it.

微软正在裁员4800人，主要集中在Xbox和销售部门。

Microsoft is laying off 4,800 employees, primarily in its Xbox and sales units.

我们请微软的记者亚伦·霍姆斯来聊聊，这次调整反映出公司怎样的战略方向。

Let's bring in our Microsoft reporter, Aaron Holmes, to talk about what this move indicates about the company's strategies.

欢迎回来，亚伦。

Welcome back, Aaron.

你好。

Hi.

很高兴能来。

Happy to be here.

好的。

Alright.

我们早就知道这些裁员要来了。

So we knew these layoffs were coming.

最让你意外的是，裁员主要集中在哪些部门？

Does anything surprise you about where most of the cuts were made?

正如你所说，这些裁员其实早有预兆，大家都知道Xbox会是重灾区，但结果并没有。

So, yeah, like you said, the these cuts have been, you know, previewed for a while, and, we knew that they were going to hit Xbox heavily, which they didn't.

另一个大头是销售部门。

The other big area that we've seen is sales.

今天发给员工的公告里，微软在措辞上还挺谨慎的。

And, you know, in in the announcement that went out to staff today, Microsoft was, you know, a little bit careful about how it framed this.

它说公司正在把业务重心转向人工智能，这没错，但同时也特意说明，被裁的岗位并不是直接被AI取代的。

It said that, you know, it's reorienting its business around AI, which is true, but it also, you know, was careful to say that the roles that are being cut are not actually being directly replaced by AI.

更准确地说，这是在重新调整业务结构，让AI销售成为微软业务中更重要的部分。

It's more about just sort of reorganizing the business to to make AI sales a bigger part of what Microsoft does.

我觉得这种说法挺有意思，原因很多，可能也和大家原先的预期不太一样。

And and I think that that, framing is interesting for many reasons and maybe a little different than what people were expecting.

是的。

Yeah.

说得通。

Fair enough.

微软真正的问题出在哪里？

Where have the real issues at Microsoft been?

今年到目前为止，它的股价在大型科技股中表现垫底。

Its stock has been among the worst performers in the mega cap tech stock so far this year.

对。

Yeah.

我觉得有几个原因。

And I think that's for a few reasons.

一方面，投资者担心微软可能面临来自OpenAI和Anthropic的潜在生存威胁，它们正试图自动化微软核心办公软件中的工作内容。

You know, one is that investors are nervous that Microsoft could face these potentially existential threats from OpenAI and Anthropic, which are trying to essentially, automate the type of work that goes on in Microsoft's flagship office software.

与此同时，微软还在持续大量投资建设AI数据中心，但推广自家Copilot产品的进展却时断时续。

And at the same time, Microsoft has continued to invest heavily in new CapEx for AI data centers, and it its efforts to sort of sell its own Copilot product have been a little bit stop and go.

今年早些时候，Copilot的销售终于开始有了一些起色。

They they've started to show some momentum for Copilot sales earlier this year for the first time.

但与此同时，我觉得这些数字还没强到足以让华尔街安心，毕竟微软在AI上投了这么多钱，回报却没达到预期。

But at the same time, I don't think that those numbers have quite been strong enough to assuage the concerns on Wall Street that Microsoft isn't getting, you know, a a sizable enough return from the huge amounts that it's invested into AI so far.

是的。

Yeah.

现在大家都把希望寄托在Copilot上了，我觉得是这样。

There's a a lot of hopes riding on Copilot right now, I feel like.

微软裁员时说，被裁的岗位并不会被AI取代，这挺有意思的。

It was interesting with these job cuts that Microsoft said the roles are not being replaced by AI.

我本来没想到他们会这么说。

I was surprised they said that.

说实话，现在大家总爱把裁员归咎于AI，这都成潮流了。

Frankly, I feel like it's in fashion to always blame AI for these job cuts.

你怎么看这一点？

So what did you make of that?

你说得对，这点很有道理。

You know, it it's a good point.

我觉得在更广泛的科技圈里，我们现在正看到两种现象。

I think that we we are kind of seeing two things in the broader tech world.

一方面，正如你所说，公司都希望展示自己如何用AI实现回报和成本节约，尤其是通过减少人员编制。

One, to your point, you know, companies do wanna show that they're using AI to to get returns and to get cost savings, especially in the form of potentially headcount reductions.

但与此同时，最大的AI供应商们也开始表现出谨慎态度。

But at the same time, we're now seeing this sort of cautiousness from the biggest sellers of AI.

比如微软，还有OpenAI、Anthropic和谷歌，它们都不希望AI被看作是导致大规模失业或取代工作的罪魁祸首。

So Microsoft, but also OpenAI and Anthropic and Google, who don't really want AI to be seen as this villain that is going to lead to widespread unemployment or replace jobs.

说白了，我觉得微软在这一点上有很多公关和政策上的考量，所以才强调AI并没有取代人。

I you know, reading between the lines, I think that Microsoft has a lot of sort of PR and policy reasons to to say that AI is not replacing people.

我们已经看到这家公司很刻意地宣传，比如在建设AI数据中心的地区投入社区发展，推动AI技能培训等等。

And we've seen the company be pretty sappy about, you know, its efforts to say that it's going to, for example, invest in communities where it's building AI data centers and that it's gonna do AI job skilling.

总的来说，微软并不想散布AI会导致员工失业——无论是它自己公司内部，还是整个社会——这样的说法。

And so across the board, you know, Microsoft doesn't wanna go around spreading this idea that AI is going to lead to, like, people losing their jobs either at Microsoft or more broadly.

所以我觉得，今天的这些说法其实和这一点是一致的。

And so I think some of the messaging today aligns with that.

对。

Yeah.

没错。

Yeah.

这完全说得通。

That totally makes sense.

那说到Xbox，微软是不是准备把Xbox业务拆分出去？

So in terms of s Xbox, is Microsoft gearing up to spin off the Xbox division?

如果是的话，会是谁来收购呢？

And if so, who would buy it?

是的。

Yeah.

我一个月前报道过，微软并没有排除将Xbox拆分的可能性。

So I reported about a month ago that Microsoft had not ruled out spinning out Xbox.

据我目前了解，微软并没有马上要把Xbox卖掉的计划。

My current understanding is that there's not any immediate plans to put Xbox up for sale.

我觉得现在Xbox的首席执行官阿莎·夏尔马，主要精力还是放在如何让现有业务变得健康起来。

And I think right now, Asha Sharma, the CEO of Xbox, is focused on just trying to make the business healthy in its current form.

她今天提到，这些裁员是为了让业务走向更稳健的状态。

You know, she said today that some of these job cuts are about trying to get the business to a healthier place.

Xbox目前已经出售了旗下四个工作室。

And what Xbox has done is it sold four of its studios that it owned.

今天他们宣布，其中两个工作室是卖回给了原管理团队。

It announced today, you know, two of them were sold back to management.

听说售价很低，而且还给了这些团队一笔启动资金，好让他们能独立运营。

I understand for a very low price and and also while giving those studios cash so that they can stand alone on their own.

另外两个工作室则卖给了其他买家。

And then two other studios have been sold to additional buyers.

还有一个工作室目前正在考虑出售。

There's a fifth studio that is also exploring a sale.

所以目前来看，Xbox短期内不太可能从微软分拆出去。

So right now, it doesn't look like Xbox itself is being spun out from Microsoft anytime soon.

我觉得他们的计划是先看看未来六到十二个月内，能把业务改善到什么程度。

I think the plan is to essentially see how much they can shore up its business in the next six months to a year.

不过我也听说，微软最终还是不排除对Xbox进行重组或分拆的可能性。

But I I've also been told, you know, eventually restructuring or spinning out Xbox is not off the table for Microsoft.

明白了。

Okay.

我想换个话题，确保我们有时间聊一下你上周写的另一个故事。

I want to switch gears to make sure we have time to get to another story that you wrote last week.

你看到一份备忘录，里面说明了微软打算把面向消费者和企业的Copilot应用合并。

You viewed a memo outlining how Microsoft is merging its consumer and enterprise Copilot apps.

能说说这份备忘录的内容吗？

Tell me about this memo.

它具体说了些什么？

What did it say?

是的

Yeah.

简单来说，微软正在对Copilot进行一场重大转型，目的是像这份备忘录的作者所说那样，向客户证明Copilot确实有存在的价值。

So, basically, this is, you know, a transformation that Microsoft is making to Copilot in an effort to basically, in in the words of this executive who wrote this memo, to prove that Copilot has a right to exist to customers.

这位名叫雅各布·安德鲁的高管刚上任不久，过去六个月里他负责Copilot的用户体验，现在正试图彻底重构这款应用。

And so, you know, this executive, Jacob Andrew, who just started, to oversee Copilot's user experience in the past six months is essentially trying to overhaul the app entirely.

其中一部分工作就是把原本分散在消费者端、企业端和代码助手等功能模块全部整合进一个应用里，据我所知，目标是下个月八月上线。

And so, you know, part of that involves combining its different consumer enterprise coding agent features all into one app, which I'm told is, you know, aiming to ship by August, so next month.

同时，安德鲁还在砍掉很多用户不喜欢或者觉得烦人的功能。

And then at the same time, Andrew has been trying to cut a lot of features from Copilot that people didn't like or that they found annoying.

我还听说，他一直在和微软的一些大客户紧密合作，就是为了搞清楚他们到底喜欢或不喜欢Copilot的哪些地方。

I also hear he's been working very closely with some of Microsoft's biggest customers just to really understand, you know, what they like and don't like about Copilot.

更广泛地说，这背后是一场全面的冲刺，想让Copilot更快地卖出去，真正证明它能和ChatGPT、Claude这些在企业市场更占优势的AI聊天机器人正面竞争。

And and sort of the broader effort there is an all out push to get Copilot to to sell more quickly and, to essentially, you know, prove that it can compete directly with ChatGPT and Claude and and some of the other AI chatbots that have been more so dominating in the in the enterprise space.

对

Yeah.

这种‘存在权’的说法我觉得特别有意思，也反映出Copilot现在所处的困境。

That right to exist framing I think is really interesting and kind of telling about the tough position that Copilot is in.

换句话说，微软得自己证明，Copilot现在凭什么存在。

Like, the burden of proof is on Microsoft to show why Copilot should even exist right now.

他们还得慢慢赢得客户的信任。

And they need to sort of earn customers' trust about that.

这次公告之后，Copilot的实际功能有什么变化吗？

Is anything actually changing about what Copilot is with this announcement?

到底发生了什么样的改变？

What is the change that's happening here?

是的。

Yeah.

我听说，下个月客户最明显能感受到的变化是，以前在Office里有个叫365 Copilot的工具。

So the biggest change that will show up for customers, I'm told, next month is that previously there was this three sixty five Copilot that lived in office.

还有GitHub Copilot，就是那个AI编程辅助工具。

There was GitHub Copilot, which was like the AI coding tools.

还有一个通用的 Copilot 聊天机器人。

And then there's just kind of the generic Copilot chatbot.

现在这些功能都要整合进一个应用里，你以后不管用哪个，入口都是一样的。

Essentially, of those things are being combined into a single app, that, you know, you will basically have the same point of entry for all of them.

然后你自己决定哪些功能值得花钱去用。

And then from there, you know, choose what you wanna spend extra to use.

另外，我觉得 Copilot 会越来越少出现在用户明确表示不喜欢的地方。

Additionally, you know, I think that Copilot is going to show up less in places where people have told Microsoft that they don't like it.

以前 Windows 里好多地方都会弹出 Copilot 窗口，但微软后来分析数据发现，这些弹窗要么让人烦，要么一不小心点中就立刻关掉。

So, you know, there was multiple places in Windows where you would see Copilot pop ups that essentially Microsoft looked at data and found that those pop ups were usually annoying to people or were immediately closed out if they accidentally clicked on them.

现在微软正在逐步减少这些弹窗。

And, you know, Microsoft is trying to pair those back now.

不过总的来说，我听说安德鲁的风格是，他之前在 Snap 工作，去年加入微软，特别看重数据分析——只要用户不用某个功能，或者数据表明它不受欢迎，就该立刻调整或优化。

But in general, I you know, what I'm told about Andrew's approach is that he he comes from Snap before he joined Microsoft last year, and he has this very kind of, you know, analytics forward approach where if if a user is not using a feature or if the data shows that it's not immediately beloved, it should just be caught, or, you know, improved.

我觉得这正好契合了他所说的，要证明 Copilot 值得存在。

And I think that feeds into this sort of effort to to prove Copilot's right to exist in his words.