Anthropic 的超级碗广告：针对 OpenAI 的蓝龟壳攻击？

我看到地平线上有一场大型 IPO。你被记者包围了。一夜成名，3000 双倍右。别错过信息。Courtney 宣布命令进场。我们开始吧。我们被将军包围了。守住你的位置。起来吧。相信专家。买代币。我们是专家。创始人模式。我看到地平线上有几个记者。待命。无人机上线。燃烧。双倍燃烧。三倍燃烧。双杀。错。警察请。团队死斗。我们是专家。三倍剑。我们开始吧。对。Marky 清理命令进场。来吧，起来。我们被记者包围了。守住你的位置。破一。击二。激活金毛模式。Marky 清理命令进场。5。我看到地平线上有几个记者。待命。创始人。你正在收看 TVPN。今天是 2026 年 2 月 5 日，星期四。我们正在 TVPN 超巨蛋现场直播。耶！科技殿堂，金融堡垒。让我给你介绍一下 ramp.com。时间就是金钱。两者都省。简单。使用公司卡、账单支付、会计等等，全部集中在一个地方。我们今天有一个相当棒的阵容。让我们调出线性阵容。认识一下现代软件开发系统。70%的企业在 Linear 的企业工作区使用智能体，今天我们为你准备了一个阵容。我们有来自 OpenAI 的 Sam Altman 加入 Sholto……如果你不熟悉，Sam Altman 是 OpenAI 的 CEO。他们是 ChatGPT 的创造者。

I see a large IPO on the horizon. You're surrounded by journalists. Overnight success the place of 3000 double right. Let's miss information. Courtney declaring order inbound. Let's just roll. We are surrounded by general. Hold your position. Come get up. Trust the experts. Buy token. We are experts. Founder mode. I see multiple journalists on the horizon. Stand by. UAV online. Blaze. Double blaze. Triple blaze. Double kill. Wrong. Cop is please. Team Deathmatch. We are experts. Triple blade. Let's just roll. Right. Marky clearing order inbound. Come on, get up. We are surrounded by journalists. Hold your position. Break one. Strike two. Activate golden retriever mode. Marky clearing order inbound. 5. I see multiple journalists on the horizon. Stand by. Founder. You're watching TVPN. Today is Thursday, February 5th, 2026. We are live from the TVPN Ultradome. Yahoo! The temple of technology, the fortress of finance. Let me tell you about ramp.com. Time is money. Save both. Easy. Use corporate cards, bill pay, accounting, and a whole lot more all in one place. We have quite the lineup today. Let's pull up the linear lineup. Meet the system for modern software development. 70% of enterprises on enterprise workspaces on linear are using agents and we have a lineup for you today. We got Sam Altman from OpenAI joining Sholto from... if you're not familiar, Sam Altman is the CEO of OpenAI. They are the makers of ChatGPT.

是的。我确实向某人解释过。我当时说，‘哦，你知道……’我在向一个陌生人解释这个节目。我说，‘哦，你知道 OpenAI 吗？’他们说，‘知道。’我说，‘哦，我们今天请了 Sam 上节目。’他们说，‘他是谁？’

Yes. I did actually explain to someone. I was like, 'Oh, do you know...' I was explaining the show to some random person. I was like, 'Oh, do you know OpenAI?' They're like, 'Yeah.' I was like, 'Oh, we're having Sam on the show today.' They're like, 'Are they like who?'

不，不，不，不。他们当时说，‘哦，那很酷。挺好的。我明白这个节目是什么了，对吧？’然后我们有来自 T1 Energy 的 Dan Barkello 谈论在美国建造太阳能电池板，当然还有我们的闪电轮，我对此非常兴奋。所以，我们一直在思考即将到来的超级碗。我们一直在思考广告、广告的回应、广告的来回交锋。Run 在这里发了一个好帖子。他说：‘戴上我的媒体观察员帽子。Anthropic 的广告相当聪明，因为它们在某种程度上是不诚实的，只会激怒 OpenAI 的高层和某些行业内部人士，但对其他人来说却有趣且引人注目。当你是一个看涨期权时。称他们为看涨期权。有点贬低。马里奥赛车的蓝龟壳。你熟悉吗？你玩过马里奥赛车吗？你明白吗？不太明白……我玩马里奥赛车是很久以前了，蓝龟壳当时还不存在。但我和孩子们一起玩，后来知道了蓝龟壳的重要性。蓝龟壳只针对第一名，只针对领先的人。这是一个很好的比喻，形容这里的情况。当你不是第一名时，你得到蓝龟壳。你可以向领先者开火，甚至不需要点名。所以，你只需要说这个类别很糟糕，每个人都会认为你在说你知道是谁。所以我觉得这很有趣。’

No, no, no, no. They were like, 'Oh, that's cool. Like, that's good. Like, I understand what the show is, right?' Then we have Dan Barkello from T1 Energy talking about building solar panels in America and of course our lightning round, which I'm very excited for. So, we've been thinking more about the Super Bowl that's coming up. We've been thinking more about ads, the response to the ads, the back and forth with the ads. Run had a good post here. He said, 'Putting on my media observer hat. Anthropic ads are pretty brilliant because they're dishonest in a way that's only going to rage bait OpenAI heads and certain industry insiders, but are funny and striking to everyone else. When you're a call option. Calling them a call option. Kind of a diss. Mario Kart Blue Shell. Are you familiar? Have you ever played Mario Kart? Do you understand? Not enough to get the... I played Mario Kart I think so long ago, blue shells didn't exist. But I've been playing with my kids and I've since learned the importance of the blue shell. The blue shell, it targets just the first player, just whoever's in first. It's a great metaphor for what's going on here. When you're not in first place, you get the blue shell. You can take a shot at the leader without even needing to call them out. So, you can just say the category is bad and everyone assumes you're talking about you know who. So I thought that was an interesting thing.'

Trey 说：‘Sam Altman，圆锥体收藏家。’

Trey says, 'Sam Altman, the cone collector.'

是的。是的。是的。就是那个。是的。内心深处。总之，昨天每个人都有看法。它引发的争议程度堪称完美。非常有趣。

Yes. Yes. Yes. That one. Yes. Deep down. Anyway, so everybody had a take on this yesterday. It was perfect in how much controversy it generated. It was wildly entertaining.

我想读一下我更新后的看法。我受到了一点反驳。我说他们玩得很脏。Signal 回复我说：‘一点也不脏。’所以，我想回应一下。

I wanted to read through my updated take. I got a little bit of pushback. I said they were playing dirty. Signal responded to me and said, 'Not dirty at all.' So, I wanted to address that.

好的，深入探讨。但首先，让我介绍一下 Console。Console 构建的 AI 智能体可以自动化 70%的 IT、人力资源、财务支持，为员工提供访问请求的即时解决方案。

Okay, dig in. But first, let me tell you about Console. Console builds AI agents that automate 70% of IT, HR, finance support, giving employees instant resolution for access requests.

总之，是的，我对此进行了更多思考。我们昨天显然看了一些广告。我们看了‘Get a Mac’活动。我们看了 Bud Light 的特别配送广告，讲的是 Bud Light 在城堡里。他们收到了一份玉米糖浆订单。他们说：‘我们不用玉米糖浆。那一定是给 Coors Light 和其他竞争对手的。’所以我当时在思考，区别在于那些广告是真实的，对吧？用过 PC 的人可能中过病毒，对吧？所以当 Mac 拿这个开玩笑时，那是真实的，对吧？

Anyway, so yeah, I was processing this more. We obviously watched some ads yesterday. We watched the 'Get a Mac' campaign. We watched the Bud Light special delivery one, which is about Bud Light in a castle. They get an order of corn syrup. They're like, 'We don't use corn syrup. That must be for Coors Light and other competitors.' And so like I was processing them and the difference there is that those advertisements are truthful, right? People who have had a PC have probably gotten a virus, right? So when Mac is riffing on that, it's truthful, right?

而且他们有一些数据支持。

And they had some data to back it up.

这不是欺骗。如果我在 2007 年或 2003 年戴上微软的帽子，我会说：‘嘿，我们确实有 Windows PC Defender。我们在对抗病毒。’Mac 上可能中病毒吗？可能。PC 上可能不中病毒吗？是的。而在微软方面，人们会说：‘是啊，没人会为你的电脑制作病毒，因为你卖得不多。投资回报率不高。’

It's not deceitful. I would say if I'm putting on my Microsoft hat in 2007 or 2003, I would say, 'Hey, we do have Windows PC Defender. We're fighting viruses.' And is it possible to get a virus on a Mac? Probably. Is it possible to not get a virus on a PC? Yes. And on Microsoft, people are like, 'Yeah, no one makes viruses for your computers because you don't sell very many. It's not very ROI positive.'

说得好。总之，Bud Light 的活动是真实的，尽管它很有攻击性，因为你可以查看竞争对手的配料表，看到他们确实使用了玉米糖浆，对吧？

That's a good point. Anyways, and then Bud Light's campaign was truthful even though it was aggressive in that you could look up the ingredient list of their competitors and see that they did in fact use corn syrup, right?

你可以自己决定是否喜欢那种成分，但他们只是引起了人们的注意。

And you can make your own decision on whether or not you like that ingredient, but they were just drawing awareness to it.

所以我的观点是，我认为 Anthropic 的广告更接近政治攻击广告，他们有点故意试图欺骗，对吧？他们没有违反任何法律。

And so my point is that I think that Anthropic's ads are closer to political attack ads and that they're sort of intentionally trying to be deceptive, right? They haven't broken any laws.

他们没有点名 ChatGPT。他们只是对整个类别泼脏水。

They don't name ChatGPT. They're just sort of throwing mud at the whole category.

我问 Claude，我说：‘Claude，你怎么定义玩得脏？’Claude 说：‘玩得脏通常意味着通过欺骗、不道德或违反特定情境下公认规则和规范的手段来实现目标，即使这些手段在技术上并不违法。这是你能做什么和应该做什么之间的差距。有几个维度。它涉及欺骗、误导他人关于你的意图、隐藏信息或制造虚假印象以获取优势。这些活动在制造虚假认知方面做得非常出色。’

I asked Claude, I said, 'Claude, how would you define playing dirty?' And Claude said, 'Playing dirty generally means achieving your goals through tactics that are deceptive, unethical, or that violate the understood rules and norms of a given context, even if they're not technically illegal. It's the gap between what you can do and what you should do. A few dimensions to it. It goes into deception, misleading others about your intentions, hiding information, or creating false impressions to gain an advantage. These campaigns do an amazing job creating a false perception.'

是的。

Yeah.

或者对 LLM 广告会是什么样子的印象。

Or impression of what ads in LLMs are going to be like.

我确实在想，我们之前好奇的广告反响如何？在 TBPN 之外，我们喜欢广告。广告没问题，它们不会搞什么奇怪的事。我们是坚定的支持者。公众会怎么接受？Claude 会因为广告效果这么好而飙升至榜首吗？普通消费者会相信聊天应用会因广告而变得奇怪这种说法吗？我昨晚完全随机地刷 Instagram Reels，并不是在找 Anthropic 的内容。我好像关注了 Claude 的账号，也许没有。它直接推送给我了。它给我看了一个竖屏版的广告。我觉得叫《欺骗》之类的。不对，是《侵犯》。《侵犯》才是那个。

I do think the response to the ads we were wondering about, like outside of the TBPN, we love ads. Ads are fine and they're not going to do anything weird. We're strong supporters. What will the public's reception be like? Will Claude skyrocket to the top of the charts because these ads are so effective? Will general consumers buy the line that the chat apps are going to get weird with the ads or not? I was scrolling on Instagram reels last night completely randomly. I was not looking for Anthropic content. I think I follow the Claude account. Maybe not. It just targeted me. It hits me with a vertical version of the ad. It's called Deception, I think. Something like that. No, Violation. Violation is the one.

不叫《欺骗》。那也太直白了。

It's not called Deception. That would be a little too on the nose.

我觉得有一个叫《欺骗》的。有一堆呢，名字都不一样。反正，这个叫《侵犯》。这儿有个截图。

I think there is one called Deception. There's a bunch. They all have different names. Anyway, it's called Violation. And there's a screenshot here.

哦，对。我记得。所以人们说 Anthropic 的《欺骗》广告是骗人的。

Oh, yeah. I remember. So, people were like the Anthropic Deception ad is deceptive.

对。所以《侵犯》弹出来，是这个 Claw AI，有将近 6000 个赞。就算它刚推给我，我的解读就是这招有效。这很受欢迎。不仅拍得漂亮，竖屏剪辑也很到位。

Yeah. So, Violation pops up and it's this claw AI and it has almost 6,000 likes. Even when it just got served to me, my interpretation was like this is working. This is popular. It's not just beautifully shot, it's well edited for vertical.

对。要么是引起了强烈共鸣，要么是投入了巨额资金，或者两者兼有。

Yeah. It's either really resonating or they're putting a massive amount of spend behind it, or both.

竖屏版是这样的。然后好玩的是。在这个广告里，你看到一个人在做引体向上很费劲。他问，那个应该是 LLM 吧？给我制定一个健身计划。然后健身机器人说：‘嘿，你知道还有什么能帮上忙吗？Height Max 的一英寸增高鞋垫之类的。’这看起来就是颜值最大化。非常搞笑。但那是一个一英寸的鞋垫，可以放进鞋里。我往上滑。Meta 给我推送的下一个广告是什么？一个三英寸鞋垫的广告。这个广告实际上有欺骗性。它说这个人可以从 5 英尺 9 英寸长到 6 英尺 1 英寸。那是 4 英寸。所以这是内置鞋垫的整双鞋。不知为何我收到了这个广告，尽管我并不需要鞋垫。算法就是知道我喜欢这些广告，因为它们很搞笑，而且非常符合颜值最大化的潮流。所以我不断收到这些广告。Meta 给我展示的全是这些增高鞋，因为我好像真的点进去过，还仔细研究了。

This is what it looked like vertically. And then this is the funny thing. So, in this ad, you see the guy struggling to do a pull-up. He asks, you know, what's supposed to be an LLM? Create a fitness plan for me. And then the fitness bot says, 'Hey, you know what else could help? One inch insoles from Height Max or something like that.' And it's like this looks maxing. It's very funny. But it's a 1-inch insert that would go in your shoes. I scroll up. What's the next ad that Meta serves me? An ad for 3-inch inserts. And the ad is actually deceptive. It says the guy can go from 5'9 to 6'1. That's 4 inches. And so these are full shoes that have the inserts built in. And for some reason I got it even though I'm not in the market for insoles. The algorithm just knows that I love these ads because they're very funny and they're very on trend with the looks maxing thing. And so I get served these ads constantly. This is all Meta shows me is these height enhancing shoes because I think I actually clicked on them and was digging in.

所以你买了，对吧？当然。寄给我的敌人。

So you bought them, right? Of course. To send to my enemies.

你想长到七英尺。

You're trying to get to seven feet.

那倒不错。Tyler，你有什么想说的吗？

That would be good. Tyler, do you have something on this?

对，我正想说我也在 Instagram 上看到了。我看到一些 Claude 的广告，评论里大家都在支持 Claude。

Yeah, I was just going to say I saw on Instagram as well. I saw some of the Claude ads and in the comments people were riding with Claude.

确实。

They are.

对。就像普通人完全赢得了氛围。他们一直在开发者中赢得氛围战，在 X 上也赢了，现在感觉他们即将在公共领域赢得氛围战。

Yeah. It was like normies totally winning the vibe. They've been winning the vibe war with developers and they've been winning the vibe war on X, and now it feels like they're about to win the vibe war in the public square.

这场活动可以同时成立两件事。它很出色、时机恰当，而且极具战略意义，原因我下面会列出来。同时，它旨在在数亿美国人的心中植入对 ChatGPT 即将推出的广告产品的错误印象。他们可以辩解说：‘哦，我们不是想那样做。’但你无法否认效果。Anthropic 实现了许多目标。这场活动娱乐了美国。它非常有趣，很搞笑。真的是完美的剧本。你能听到破折号、停顿。太棒了。代理公司叫 Mother。他们做得太棒了。他们也在 IPO 之前把自己推上了台面。我认为在某些方面，某些受众对 Anthropic 的了解比对 Claude 还多。即使你只是对投资 AI 感兴趣，你可能听到 Anthropic 比听到 Claude 更多。这在他们内部人士中建立了光环。如果我说他们在这项活动上花了 1 亿美元，而它只是帮助留住了一两个真正顶尖的研究员，那也值了。

Two things can be true about the campaign. It's brilliant, well-timed, and incredibly strategic for a few reasons I'll outline below. And it's designed to plant a false impression of ChatGPT's forthcoming ad product in the minds of hundreds of millions of Americans. They could argue, 'Oh, we're not trying to do that.' But you can't really argue with the effect. Anthropic accomplishes a lot. The campaign entertains America. It's wildly entertaining. It's hilarious. Truly the perfect script. You can hear the em dashes, the pauses. It's amazing. Mother is the name of the agency that did it. They absolutely crushed it. They're also putting themselves on the map ahead of the IPO. I think in some ways certain audiences would know more about Anthropic than Claude. Even if you're just generally interested in investing in AI, you're probably hearing about Anthropic more than you're hearing about Claude. It builds their aura with insiders. If I said if they spend $100 million on this campaign and all it does is help retain a couple of truly elite researchers, it's worth it.

我喜欢这个。Noah 说 Anthropic 在搞活动让自己被禁，就像虚假信息那样，走得太远引发反弹。我觉得这很好笑。我们可以回到那个话题。这多少延续了他们一贯的恐惧导向信息。这些文章有效地激怒了 OpenAI。他们完全上钩了。Sam 从全小写打字变成了‘我得用大写来回应’。很多回应。这增加了公众对广告推出的普遍审视，华盛顿也是一个因素。

I love it. Noah saying Anthropic is campaigning to get themselves banned just like with misinformation, going way too far into it, backlash. That's just funny to me. We can go back to that. It somewhat continues their fear-based messaging that they've been riding with in general. The essays effectively rage-baited OpenAI. They got fully baited. Sam switched out of his lowercase typing and was like 'I got to go into uppercase for this one.' Lots of responses. It increases the public's general scrutiny of the ads rollout, and then Washington too is another factor.

嗯，也许吧。我只是觉得这会被提出来。

Yeah, maybe. I just think it'll come up.

完全同意。你得问问马克·扎克伯格，你怎么赚钱？如今在华盛顿，什么都有可能。

Totally. You have to ask Mark Zuckerberg, how do you make money? Anything's on the table these days in Washington.

对。另一件事是它会广泛损害消费者对 LLM 的信任。有些人会想：‘等等，他们一直在我不知道的情况下从我身上赚钱？或者我怎么能相信每个输出都是真正的好建议，还是我被货币化了？’所以是的，这是你认为可能反噬他们的一点。其他五点可能也有风险，对吧？但这取决于他们的消费者策略。现在他们说自己不在乎消费者。他们说了很多次，但他们的行动在某些方面又有所不同。

Yeah. And then the other thing is it's going to broadly damage consumer trust in LLMs. Some people will just be like, 'Wait, they've been making money off me without me knowing? Or how can I trust every output as actually good advice or am I being monetized?' So yeah, and this is the one that you think could come back to bite them. The other five are potentially, right? But it depends what their consumer strategy is. Right now they're saying we don't care about consumer. They've said that a lot, but their actions kind of speak differently in some ways.

我说过 Anthropic 一直告诉市场他们不在乎消费者，但我不确定。

I said Anthropic has consistently told the market they don't care about consumer, but I'm not sure.

支持广告的理由是，它能让那些付不起订阅费的人免费使用大语言模型，但 Anthropic 已经输掉了服务数十亿人的竞赛，对吧？我不认为——你看看 Gemini 的势头、OpenAI 的 ChatGPT 的势头，争夺 30 亿月活用户的竞赛似乎已经结束了。我不相信 Claude 能后来居上并达到那个规模。而且不靠广告他们做不到这一点，对吧？因为你可以算算账——就像马里奥赛车里的蓝龟壳一样，后来者很难超越领先者。

The argument for ads is that they'll make LLMs free for people that can't afford to pay a subscription, but Anthropic has already lost the race to serve billions of people, right? I don't think that they're—when you look at Gemini's traction, OpenAI's ChatGPT traction, it seems like the race to get to three billion monthly active users is kind of over. I don't believe Claude is going to come behind and get there. And they wouldn't be able to do that without doing ads, right? Because you can kind of run the numbers—the blue shell take out in front.

所以我想引出的问题是：他们能否向数亿规模的小众群体提供一款奢侈品——就像 iPhone 的销量那样？全球大约有 15 亿部活跃的 iPhone。那些人本可以买更便宜的安卓手机和更便宜的设备，但他们愿意为 iPhone 支付溢价，因为他们有能力，而且对很多人来说，iPhone 提供了更好的体验。

So the question that I was taking this to is: can they deliver a luxury product to a smaller cohort in the hundreds of millions—to kind of iPhone numbers? There are roughly one and a half billion iPhones active in the world. Those people could all buy a cheaper Android and cheaper devices, but they've paid a premium for the iPhone because they can, and for many people it delivers a better experience.

iPhone 不是第一款智能手机。Claude 也不是第一款面向消费者的 LLM。iPhone 早期确实在规格上实现了差异化，这和模型卡没什么不同，但苹果最终转向了更感性的理由，告诉你为什么应该让别人看到你用 iPhone。它告诉人们你关心环境，你的应用商店里没有泛滥的成人应用，你重视隐私。这些策略取得了不同程度的成功。每家科技公司都能讲一个 ESG 故事。我无法想象今天有苹果高管会说出“色情”这个词，尽管史蒂夫·乔布斯在 2010 年非常尖锐地提到了这一点。他说：“你知道，安卓上有一个色情商店。你可以下载色情内容。你的孩子也可以下载。那是一个我们不想去的地方，所以我们不会去那里。”直接点名竞争对手并抛出这种话——

The iPhone was not the first smartphone. Claude was not the first consumer LLM. The iPhone did differentiate on specs early, not unlike a model card, but Apple did eventually pivot to more emotional arguments for why you should be seen with an iPhone. It tells people you care about the environment, that you don't have adult apps flooding your app store, and that you take privacy seriously. These have had varying levels of success. Every tech company was able to tell an ESG story. And I can't imagine an Apple exec even saying the word porn today, even though Steve Jobs was very pointed about it back in 2010. He said, "You know, there's a porn store for Android. You can download porn. Your kids can download porn. That's a place we don't want to go, so we're not going to go there." Calling out the competition by name and dropping that is like—

是的。关键在于他说的是事实，是真的。

Yeah. And the key thing here is that it was factual. It was true.

没有欺骗性。

Like it wasn't deceptive.

对，没有。

No. No.

所以我不觉得那很激进，但他没有玩阴的。

And so I don't think that was edgy, but he wasn't playing dirty.

是的，是的。

Yeah. Yeah.

但苹果确实围绕隐私制定了一套坚实的传播策略，并借此向广告商施压。市场做出了调整，广告行业显然存活了下来，但普通美国人仍然觉得 Facebook 在监听他们的对话以投放定向广告。消费者有权选择。如果他们想为无广告版本付费，那很好。但大多数人不会。顺便说一句，在欧洲，Facebook 只有 1%的用户选择付费。这个数据是：他们都有选择为无广告 Facebook 付费的选项，但只有 1%的人这么做。所以消费者有权选择，但他们不应该被误导，以为广告平台是如何运作的。安卓为世界创造了巨大价值。谷歌整体也是如此。让消费者选择，但要让他们明智地选择。

Apple did however carve out a solid messaging strategy around privacy and allowed them to put the screws to advertisers. The market adjusted and the ad industry obviously survived, but average Americans still feel that Facebook is listening to their conversations to target ads. Consumers deserve choice. It's great if they want to pay for ad-free tiers. Most don't. 1% in Europe for Facebook, by the way. That's the stat. They all have the option to pay for ad-free Facebook and only 1% do. So consumers deserve choice, but they should not be misled about how ad platforms work. Android has generated an immense amount of value for the world. So has Google broadly. Let consumers choose, but let them choose intelligently.

是的，我有反驳意见，但我要先跟你聊聊 Lambda。Lambda 是超级智能云，构建用于训练和推理的 AI 超级计算机，规模从一块 GPU 到数十万块。所以我的反驳——我的最佳论证——是——

Yes, I have a rebuttal, but I'm going to tell you about Lambda. Lambda is the super intelligence cloud building AI super computers for training and inference that scale from one GPU to hundreds of thousands. So my rebuttal—my steel man—is that—

需要明确的是，我不是说 Anthropic 不应该这么做。我只是说这有点脏。

And to be clear, I'm not saying Anthropic shouldn't have done this. I'm just saying that it was a little dirty.

是的，他们在泥里打滚。这很好。他们身处低谷。我们喜欢低谷。我们活在低谷。我们为低谷而生。我的最佳论证是，他们没有越界。他们没有玩阴的，因为他们没有直接点名 ChatGPT。好吧，你怎么理解都行。

Yeah, they're rolling around in the mud. It's good. They're in the trough. We love the trough. We live in the trough. We live for the trough. My steel man is that they didn't cross the line. They didn't play dirty because they didn't call out ChatGPT directly. Okay, you can take that whatever you want.

另一件事是，他们在向上挑战。他们在挑战强者。是的。

The other thing is they are punching up. They're punching up. Yes.

但第三点，存在一个世界，这样的事情会发生。存在一个世界，广告会以如此无缝的方式被整合。如果你看看 Google 的 10 条蓝色链接的演变，最初是 10 条蓝色链接，没有广告。然后是一个非常清晰的黄色框，标着“广告”，非常清楚那是广告。随着时间的推移，UI 的演变使得告知你这是广告的方式变得不那么咄咄逼人。而 Meta 平台上的广告有时确实会让人毛骨悚然。你谈论某件事，然后你就看到了广告，也许那只是确认偏误或某种认知——你只注意到那些奇怪的。所以它们都感觉很奇怪。你看到很多你没在谈论的东西。那不会触发什么。但当你看到你刚刚和朋友谈论的东西——我刚刚和你谈论毛衣，然后我就看到了毛衣的广告。我就想，它怎么知道的？实际上，它知道是因为你刚刚上了 Facebook，找到了那件毛衣，买了它，它知道我们是朋友，我们在私信，我们在聊天，我们在平台上确实是朋友。所以，如果 Jordy 喜欢这个，他们整天混在一起，互相发私信，为什么不直接向 John 展示我刚卖给 Jordy 的东西呢？这完全合理。这可以用 Facebook 和 Meta 内部的核心 AI 的普通机器学习来实现，他们做得非常有效，但有时会让人感到毛骨悚然，有些人被吓到，他们会谈论这件事。所以，这种互动可能发生并不是完全的科幻小说。这是可能的。所以他们是在警告：嘿，如果你想确保这种事永远不会发生，我们承诺不会——那甚至不在考虑范围内。

But third, there is a world where something like this will happen. There is a world where the ads do get integrated in such a seamless way. If you look at the evolution of Google's 10 blue links, it started with 10 blue links, no ads. Then it was a very clear yellow box with "ad" and it was very clear that it was an ad. And over time the UI evolved to be a lot less aggressive about telling you that it's an ad. And the ads on Meta platforms do get creepy sometimes. You talk about something and then you see the ad and maybe that's just confirmation bias or some sort of cognitive—you only notice the ones that are weird. So they all feel weird. You see a lot of stuff that you weren't talking about. That doesn't trigger anything. But when you see the thing that you were just talking to your friend about—I was just talking to you about sweaters and I see an ad for a sweater. I'm like, how did it know? And realistically, it knows because you just went on Facebook, you found that sweater, you bought it, it knows that we're friends, we're DMing, we're talking, we're literally friends on the platform. And so it's like, well, if Jordy likes this and they're hanging out all the time, sending each other DMs, why don't I just show John what I just sold to Jordy? That makes perfect sense. That's something that can be done with just stock vanilla machine learning at core AI inside Facebook and Meta and they do that very effectively but it can feel sort of creepy sometimes and some people get creeped out by it and they talk about it. And so the idea that an interaction like that might happen is not complete science fiction. It is possible. And so they are sort of warning that hey, if you want to make sure that this never happens, it's our pledge that that won't—that's not even on the table.

是的。

Yeah.

现在的大问题是，Anthropic 什么时候推出广告？

Now the big question is when's Anthropic launching ads?

我们得让他们推出广告。

We got to get them to launch ads.

嗯，我觉得他们现在还不能。

Well, I don't think they can now.

不，他们必须这么做。没关系。我给他们许可。我要说，如果你们——每个人都会吐槽，哦，你食言了。不，我会是你最坚强的战士。我会是你最坚强的战士，我会说这是正确的事。在 Claude Code 里放广告。在 Claude 里放广告。在你写的代码注释里放广告。这样如果有人审查代码，他们就会看到广告。这就是我想要的。这就是我愿意为之而活的未来。总之，Turbo Puffer 无服务器向量和全文搜索，从第一性原理和对象存储构建。快速，便宜 10 倍，而且极其可扩展。哦，我还得告诉你更多关于广告的事。Vibe，让 DTC 品牌、B2B 初创公司和 AI 公司在流媒体电视上做广告。选择频道，定位受众，衡量销售。就像在 Meta 上一样。我们这里热爱广告，我们喜欢为广告做广告。

No, they have to. It's okay. I give them permission. I will say if you—everyone's going to be dunking, oh, you went back on your promise. No, I will be your strongest soldier. I will be your strongest soldier and I will say it's the right thing to do. Put ads in Claude Code. Put ads in Claude. Put ads in the comments of the code that you write. So if someone reviews the code, they're seeing an ad. That's what I want. That's the future I want to live for. Anyway, Turbo Puffer serverless vector and full text search built from first principles and object storage. Fast, 10x cheaper, and extremely scalable. Oh, and you know I got to tell you more about ads. Vibe where DTC brands, B2B startups, and AI companies advertise on streaming TV. Pick channels, target audiences, and measure sales. Just like on Meta. We love ads here, and we love doing ads for ads.

Zack Kukov 说：“每次我看到 Anthropic 和 OpenAI 试图通过传播营销来区分自己，我就意识到我们正在重演 90 年代的 PC 战争。”

Zack Kukov says, "Every time I see Anthropic and OpenAI try to distinguish themselves with comms marketing, I realize how much we are replaying the PC wars from the '90s."

Anthropic 有品味、优雅、有主见、专业、面向企业；OpenAI 则大众化、广泛吸引人、低消费者、低消费者加典型企业。

Anthropic, tasteful, elegant, opinionated, proumer, expert, enterprise, open AI, populist, broadly appealing, low consumer, low consumer plus typical enterprise.

对。所以 Anthropic 是苹果，OpenAI 是微软，而且 OpenAI 还与微软结盟或部分由微软拥有。

Yeah. So, Anthropic is the Apple and OpenAI is the Microsoft and they're also aligned with Microsoft or owned in part by Microsoft.

是啊。我是说，听 Daria 纯粹讲一个小时广告和 AI 的风险会很有趣。

Yeah. I mean, it would be interesting to hear Daria just talk for an hour purely about just the risks of advertising and AI.

对。没错。因为那会很有力。当然不会像把 OpenAI 拖到数亿人面前那样有效。

Yeah. Right. Because that would be powerful. Certainly wouldn't have been as effective as dragging OpenAI in front of hundreds of millions of people.

是啊。但 Anthropic 不是有播客吗？Anthropic 播客。我觉得他们有。

Yeah. But doesn't Anthropic have a podcast? Anthropic podcast. I think they do.

TJ 在帮 Sam 做公关。他说如果 Anthropic 有任何消费者在使用他们的产品，他们可能会更认真地考虑加广告。

TJ was helping Sam with some comms. He said Anthropic might think more seriously about adding ads if they had any consumers using their product.

只是对情况做出非常得体的回应。就像说，‘兄弟，如果你加大力度呢？’

Just taking a very tactful response to a situation. Just being like, 'What if you amped it up, brother?'

是啊。我是说，如果你只是抛出一大堆废话，那你就已经输了。

Yeah. I mean, you've already lost if you're just dropping a massive word salad.

也许吧。我不知道。我觉得这里有很多细微差别，这很好。我确实认为重要的是不要被激怒，不要生气，不要让别人让你的皮质醇飙升。

Maybe. I don't know. I think there's a lot of nuance here and it's good. I do think it's important to not mold, to not be angry and mad and let people spike your cortisol.

完全正确。就像他们会在超级碗上做手势，你不能让它影响你。你得专注于今天发布的新模型。有一堆新东西。所以 GPT 5.3 编解码器今天发布了，还有一个非常酷的 Frontier 更新，这是一个构建 AI 同事的产品，我很兴奋能和 Sam 聊聊，因为这感觉像是我们昨天讨论的编排产品的第一次亮相。Gastown 一直在起飞，让我们给 John 一些赞誉。

Exactly. Like they're going to be gesture maxing at the Super Bowl and you can't let it affect you. You got to focus on the new models which have launched today. There's a bunch of new stuff. So GPT 5.3 Codex launched today and there's a very cool update to Frontier, a product for building AI co-workers that I'm very excited to talk to Sam about because it feels like the first glimpse of an orchestration product which we talked about yesterday. Gastown has been taking off, let's give some credit to John.

是啊。不，可能看起来 John 对今天的 OpenAI 发布有先见之明。他没有。

Yeah. No, it might seem like John had early insight on today's OpenAI launch. He did not.

我们没有。不，我们确实没有。

We didn't. No, we actually didn't.

但事后看来，你预测得完美。

But in hindsight, you called it perfectly.

水晶球。

Crystal ball.

精英。阿尔法。一天的阿尔法。

Elite. Alpha. One day of alpha.

一天的阿尔法。是啊。如果你昨天推出了一个面向企业的编排平台。

One day of alpha. Yeah. If you had launched an enterprise focused orchestration platform yesterday.

是的。你可以融资，进行二次出售，然后在 OpenAI 立即推出一个非常……之前。

Yes. You could have raised money, got sold secondary, all right before OpenAI immediately launched a very...

你要从 OpenAI 那里偷吗？他们甚至没在想这个。他们第二天就发布了。

Are you going to steal from OpenAI? They're not even thinking about this. They launch it the next day.

谁知道呢？不管怎样，我们继续聊 Shopify。Shopify 是一个电商平台，它随着你的业务增长而增长，让你能在网上、店内、移动端、社交媒体、市场以及现在通过 AI 智能体在几秒钟内销售。

Who knows? Anyway, let's move on to Shopify. Shopify is the commerce platform that lets you grow that grows with your business and lets you sell in seconds online, in store, on mobile, on social, on marketplaces, and now with AI agents.

Matt Turk 说：‘普通创业公司。我们签了几个客户，发布了一些新功能。这周还不错。’Anthropic。我们用超级碗广告摧毁了主要竞争对手，并通过宣布一些插件让整个软件类别在公开市场暴跌。这周还不错。

Matt Turk says, 'Regular startup. We closed a few customers and shipped some new features. Good week so far.' Anthropic. We destroyed our main rival with our Super Bowl ads and tanked the entire software category in public markets by announcing some plugins. Good week so far.

哦对，我们几乎没谈这个，但 Anthropic 推出了一个口袋里的律师。他们发布了一个法律工具，或者只是宣布了。我不知道。它真的在应用里可用了吗？因为这感觉可能和 Harvey 竞争。我还没在云端应用里看到它。

Oh yeah, we barely even talked about this, but Anthropic launched a lawyer in your pocket. They launched a legal tool or they announced it. I don't know. Is it actually available in the app yet? Because this feels like maybe it competes with Harvey. I'm not seeing it yet in the cloud app.

我不认为它真的和 Harvey 竞争。

I don't really think it competes with Harvey.

不，不，不。所以，我不认为它竞争，因为他们直接面向消费者销售。至少我相信是这样。但这似乎是一个很棒的产品。似乎对它的需求会非常惊人。

No, no, no. So, I don't think it does because they're selling it direct to consumer. I believe at least. But it seems like an amazing product. Like it seems the demand for this would be incredible.

它会和 Legal Zoom 竞争。比如 Legal Zoom 下跌了 15%。自从这个公告以来，它现在是一家 13.8 亿美元的公司。

It will compete with Legal Zoom. Like Legal Zoom is down 15%. Since this announcement, it's now a 1.38 billion company.

是啊。有点像 CHEG CE 模式。而且不仅仅如此。我的意思是 Legal Zoom 有点不同，因为他们实际上可以提交公司注册文件，而且他们长期以来一直面临 Stripe Atlas 在 LLC 等方面的压力。但说实话，如果你找到工作，雇主给你发录用通知书，拿给律师看可能非常贵。如果是你的第一份工作，你可能不会去审查。很多人可能只是把它复制粘贴到 Claude 里，让父母看看。但能够直接转发邮件或集成你的 Gmail，然后说：‘嘿，我收到了这份录用通知书，有什么看起来奇怪的吗？有什么我应该问的吗？我没有太多筹码，但我想理解这份文件。’Claude 应该能做到，这很有道理。我预计 OpenAI 会尽快推出这个产品。

Yeah. Sort of in CHEG CE mode. And not just for that. I mean Legal Zoom is a little bit different because they can actually file incorporation documents and they've faced pressure from Stripe Atlas for a long time on LLCs and whatnot. But truly, if you're getting a job and your employer gives you an offer letter, taking that to a lawyer can be really expensive. If it's your first job, you're probably not going to review it. A lot of people are probably just copy-pasting it into Claude, asking their parents to take a look. But just being able to forward the email or integrate your Gmail and say, 'Hey, I got this offer letter, does anything look weird? Is there anything I should ask about? I don't have a ton of leverage but I want to understand this document.' Claude should be able to do that and it makes a lot of sense. I expect OpenAI to launch this product ASAP anyway.

是啊。我是说，所以我认为实际上为云端法律功能推出的只是一个 Claude 协作中的插件，这意味着它基本上就像 MD 文件，对吧？就像技能。

Yeah. I mean, so I think all that was actually launched for the cloud legal thing was just a plugin in Claude co-work, which means it's basically just like MD files, right? It's like skills.

但这感觉超越了技能，因为他们可能不得不自己做一些法律工作，以确保他们不是在提供法律建议，而是恰当地措辞。我认为这也是为什么 ChatGPT Health 在一个不同的领域，因为他们不想冒任何在预训练中用到你的睾酮水平的风险。然后我进入下一个版本的 ChatGPT，我说：‘Tyler 的睾酮水平是多少？’它就知道，因为它从聊天中学到了。当然，除了睾酮水平，还有一大堆私人信息。

But it feels like it goes beyond skills because they probably had to do some legal work of their own to make sure they're not giving legal advice and they're couching things properly. I think that's also why ChatGPT Health is in a different area because they don't want any risk of pre-training on your testosterone levels. And then I go into the next version of ChatGPT and I say, 'What are Tyler's testosterone levels?' And it just knows it because it learned it from the chats. There's a whole bunch of private information beyond testosterone levels, of course.

不管怎样，Railway 是一体化智能云提供商。使用你最喜欢的智能体来部署 Web 应用、服务器、数据库等，而 Railway 自动处理 Scaling（规模扩张）、监控和安全。Key 在 X 上说：‘说实话，Anthropic 的广告很好，但我认为它们是个坏主意。普通人不会想，哇，这就是 ChatGPT 的样子。我最好订阅 claude.com。’他们会想：‘哇，这就是 AI 的样子。’

Anyway, Railway is the all-in-one intelligent cloud provider. Use your favorite agents to deploy web apps, servers, databases, and more while Railway takes care of scaling, monitoring, and security automatically. Key over on X said, 'Tbh, the Anthropic ads are good, but I think they're a bad idea. Normies are not going to think, wow, this is what ChatGPT is going to be like. I better subscribe to claude.com.' They're going to think, 'Wow, this is what AI is going to be like.'

哦，是啊。

Oh, yeah.

Rune 同意。他一点也不偏颇。他说：‘自杀式炸弹策略。对他们不好，但对 OpenAI 更糟。你几乎不得不尊重它。’

Rune agrees. He's not biased at all. He says, 'Suicide bombing strategy. It's bad for them, but worse for OpenAI. You almost have to respect it.'

我发给你的短信是什么？我说，引用：‘我不知道怎么上 Anthropic。’

What was the text I sent you? I said, quote, 'I don't know how to get on Anthropic.'

是啊。John 问了一个科技圈外的朋友，他们问：‘Anthropic 是什么？我怎么上它？我不知道怎么上。’

Yeah. John was asking a friend outside of tech and they were like 'What's Anthropic? How do I get on it? I don't know how.'

是啊。我们来看看 Eric Seford。他说 Anthropic 的广告简直令人反感。然后我们继续其他故事。

Yeah. Let's go to Eric Seford. He said the Anthropic ad is simply obnoxious. And then we'll move on to other stories.

Anthropic 的大部分收入来自企业业务，因此它可以负担得起不最大化消费者收入机会。但这种道貌岸然的说教把广告描绘成一种愤世嫉俗的商业模式选择。这既不是愤世嫉俗，也不是选择。优质的数字广告支持模式是唯一反复被证明能让消费科技产品大规模触达人类的途径。有些人付不起每月 20 美元。OpenAI 的广告收入将使其能够为免费用户提供更大模型的访问权限。如果 Anthropic 认为由前沿模型驱动的聊天机器人对社会有价值，它就应该努力让尽可能多的人使用。现在，他们可能会说，‘嘿，我们是高端产品。我们确实向免费用户提供前沿模型驱动的聊天机器人，只是他们只能使用一定量，然后就得升级。’但 Eric 的观点仍然成立。广告能做到这一点，而其他消费科技商业模式都无法证明能做到。这种对数字广告居高临下的态度是经济沙文主义，代表着一种技术守门的欲望。认为广告必然会左右聊天机器人内容的说法被夸大了，近乎不合时宜。不合时宜是因为现在不再那样做了。你会等到，嘿，这个人正在买鞋垫或增高垫，同时他们在问关于罗马帝国的问题，这正是完美时机，因为他们正悠闲地读着罗马帝国，心里想，‘哦，好吧，我确实需要买那个。’然后他们切换话题。你不需要把广告放在相关内容的旁边。那只是对在线广告的过时看法。它忽略了功能性……再说一次，我认为 Sam Altman 在 2024 年 10 月确实说过，‘我把广告视为我们商业模式的最后手段。’所以，他这次得食言了。

Anthropic generates most of its revenue from enterprise business, so it can afford not to maximize its consumer revenue opportunity. But the sanctimonious moralizing here presents advertising as a cynical business model choice. It's neither cynical nor a choice. The premium digital advertising-supported model is the only repeatedly proven pathway for a consumer technology product to reach humanity at scale. There are folks who can't pay $20 a month. OpenAI's advertising revenue will enable it to offer access to larger models for free-tier users. To the extent Anthropic thinks chatbots powered by frontier models are valuable to society, it should aspire to provide access to as many people as possible. Now, they'd probably say, 'Hey, we're premium. We do give access to chatbots powered by frontier models to free users. They just only get a certain amount and then they have to upgrade.' But Eric's point still holds here for sure. Ads allow for that to an extent that no other consumer technology business model demonstrably can. This kind of condescending attitude toward digital advertising is economic chauvinism and it represents a desire for technological gatekeeping. The idea that ads will necessarily influence chatbot content is overstated, borderline anachronistic. Anachronistic because you don't do it that way anymore. You wait until, hey, this person's buying insoles or lifts and they're asking about the Roman Empire and this is the perfect time because they're just chilling reading about the Roman Empire and they're thinking, 'Oh, okay. Yeah, I did need to buy that.' And then they switch gears. You don't need to put the ad right next to the content that relates to it. That's just an antiquated way of thinking about online advertising. It ignores functional... Again, I think Sam Altman did say in October of 2024, 'I kind of think of ads as a last resort for us as a business model.' So again, he's going to have to eat his words on this one.

总之，在我们继续之前，CrowdStrike，你的业务是 AI，他们的业务是保护 AI。CrowdStrike 保护 AI 并阻止入侵。

Anyway, before we move on, CrowdStrike, your business is AI, their business is securing it. CrowdStrike secures AI and stops breaches.

Claude 还宣布了 Anthropic 的新模型，推出了 Claude Opus 4.6。我们将在 12:30 邀请 Sholto 来节目讨论。他们说这是他们最聪明的模型，并且得到了升级。Opus 4.6 规划更仔细，能更长时间地维持智能体式任务，在大型代码库中可靠运行，并能发现自己的错误。这也是他们第一个拥有 100 万 token 上下文窗口（测试版）的 Opus 级模型。非常令人兴奋。他们整理了模型卡。表现特别好。Tyler，哪个基准测试让你印象深刻？

Claude also announced a new model from Anthropic, introducing Claude Opus 4.6. We have Sholto coming on the show at 12:30 to discuss that. They say it's our smartest model and it got an upgrade. Opus 4.6 plans more carefully, sustains agentic tasks for longer, operates reliably in massive code bases, and catches its own mistakes. It's also our first Opus-class model with a 1 million token context window in beta. So very exciting. They put the model card together. They did particularly well. What was the benchmark that stuck out to you, Tyler?

是的，我认为是 ARC AGI V2。现在大约是 69%。我记得之前是……我相信是 5.2，大概在 55% 左右。所以是一个相当大的升级。是的。但这些东西总是很难说。现在真的只是质上的差异，你得用上一两个小时才能看出区别。是的。而且在新场景中，比如简单的提问，它已经能给出很好的答案好几年了。但它之前无法去拉取一堆财务数据，全部核实，然后放进 Excel 表格。这正是他们现在推动的，也是 Cluso Investments 如此兴奋的原因。哦，Cluso 说。Anthropic 更新 AI 模型以处理复杂的金融研究。所以 Opus 4.6 旨在进行金融研究和其他工作相关功能。公司扩展到新领域，包括法律服务，这震动了华尔街，并引发了关于哪些公司和服务将被 AI 颠覆的担忧。SaaS 末日来临了。我们将问 Sam Altman，软件死了还是我们回来了？我们回来了？谁知道呢？

Yeah, I think it was ARC AGI V2. It's now at like 69%. I think previous was... I believe it was 5.2 which was at I want to say 55 around. So like a pretty sizable upgrade. Yeah. But I mean with these things it's always so hard to tell. It's really just like qualitative differences at this point where the models are like you have to use it for like an hour or two and then you can kind of tell what the differences are. Yeah. And also in new contexts like in terms of just the vanilla go and ask a question. It's been able to get you a pretty good answer for like years now. But it hasn't been able to go and pull a bunch of financial data together, fact-check it all, put it in an Excel sheet. This is what they're pushing and this is what Cluso Investments is so excited about. Ooh, says Cluso. Anthropic updates AI model to handle complex financial research. So Opus 4.6 is designed to carry out financial research and other work-related functions. The company's expansion into new areas, including legal service, has rattled Wall Street and sparked concerns about which companies and services will be disrupted by AI. The SaaS apocalypse is upon us. We will be asking Sam Altman, is software dead or are we back? Are we back? Who knows?

Claude 4.6 Opus 仍然是最好的。它在所有模型中仍然拥有最好的 SVG 结果。品味极高，Lissan Algib 说。这些是一些非常漂亮的图片。你熟悉 SVG 编程吗？你基本上要用代码画出每个方块和线条。所以你必须说，我想要一个从这像素到这像素到这像素到这像素的方块。然后你把它们叠加起来。手工做会非常耗时。但 Claude 可以一次搞定，而且看起来非常符合品牌风格。这是一个有趣的基准测试，因为这并不是你通常要求的东西。这些不是扩散意义上的生成图像。这不是 Midjourney 模型。这不是 Sora 用的东西。这是另一回事，要求模型基本上编写生成图像的代码。所以它必须真正理解它在构建什么之间的来回关系。很酷的基准测试。我喜欢。我很享受。

Claude 4.6 Opus is still best. It still has the best SVG results out of all the models. Just incredibly high taste, says Lissan Algib. And these are some pretty beautiful pictures. Are you familiar with SVG programming? So you basically draw each square and line in code. So you have to say, I want a square that goes from this pixel to this pixel to this pixel to this pixel. And you add and layer all those up. It can be extremely time-consuming if you do it by hand. But Claude can just sort of one-shot it and it looks pretty on brand. It's an interesting benchmark because this isn't something you necessarily ask. These are not generative images in the diffusion sense. This isn't the Midjourney model. This isn't what Sora uses. This is a different thing asking the model to basically write code that generates an image. So it really has to understand the back and forth between what it's building. So cool benchmark. I like that. I enjoyed that.

我还喜欢纽约证券交易所。你想改变世界吗？在纽约证券交易所筹集资金。就这么简单。总之，Anthropic 的账号被一个只发他最喜欢的 AI 安全形式——Wordle 的人持有。他只发 Wordle。哦，而且他正好有 @Anthropic。Paula Rambles 说这很有趣。所以这是 X 上的一个奇怪现象。他们对账号有严格的政策，比如 @Ramp 并不归 Ramp 所有。@Anthropic 也不归 Anthropic 所有。当你看到这些账号时，你会想，相比之下，这真的是对这样一个账号的最佳利用吗？在 Anthropic 工作的 Paul Jangura 明确表示他不是 AI 公司。俄亥俄人，自由派，书虫，新闻迷，独身。如果他是书虫，他可能很喜欢持有 Anthropic 账号，因为 Anthropic 一直在像虫子一样啃书，嚼书。是的，我们还没谈到那篇文章。那篇文章很有趣。你得喂 Claude。没关系。书得喂，你知道，他们必须……是的，可能比水的事情还大。就是那么直观。人们喜欢书，喜欢看着它们被毁。是的，如果你没看过被毁的书，我同意。但人们不喜欢毁书。他们对那很敏感。

I also like the New York Stock Exchange. Do you want to change the world? Raise capital at the New York Stock Exchange. Then it's as easy as that. Anyway, the Anthropic handle being owned by a guy who only posts his favorite form of AI safety: Wordles. He only posts Wordles. Oh, and he just has @Anthropic. That's very funny from Paula Rambles. So this is a strange thing with X. They have policies around their handles that they really stand by in that @Ramp is not owned by Ramp. @Anthropic is not owned by Anthropic. And when you look at the accounts, you think is this really the best use of a handle like this in comparison. Paul Jangura who is at Anthropic's is emphatically not an AI company. Ohioan, liberal, bookworm, news hound, celile. If he's a bookworm, he probably loves sitting on the Anthropic handle given that Anthropic has been worming their way through some books, chewing through some books, much like a worm. Yes, we didn't get to that article. That was a funny article. You got to feed Claude. It's okay. The books got to feed, you know, and they have to... Yeah, potentially bigger than the water thing. Like, it's just so visceral. People love books and watching them be destroyed. Yeah, if you didn't see destroyed book, yes, I agree. But people don't like destroying books. They're sensitive about that.

非常有趣的是，保罗在推特（X）上已经迁移到 Blue Sky 做很多事情，但他没有拿到 Anthropic 的账号。他在那边是 Anthropic42。所以如果你在闲逛找 Wordle 进度，可以关注他。

The very funny thing is that Paul on Twitter on X has migrated to Blue Sky for many things, but he didn't get the Anthropic handle on Blue Sky. He's Anthropic42 over there. So shoot him a follow if you're hanging out looking for Wordle progress.

人们在玩名字决定论。是的。

People are playing around with the nominative determinism. Yes.

比如谷歌的 CEO 叫 Pichai，他的使命就是推销 AI。

Say Google's CEO is named Pichai. His purpose is to pitch AI.

Dario 的姓字面意思就是 AI 模型。

Dario's last name is literally AI model.

也许我应该改名叫 pitch MongoDB。John pitch MongoDB。选择为灵活性和扩展性而构建的数据库，配备一流的嵌入模型和重排序器。MongoDB 拥有你构建所需的一切。接下来是什么？

Maybe I should change my name to pitch MongoDB. John pitch MongoDB. Choose a database built for flexibility and scale with best-in-class embedding models and rerankers. MongoDB has what you need to build. What's next?

Dylan Patel 分享了一些数据。他说目前 GitHub 上 4% 的公开提交是由 Claude Code 编写的。按照目前的趋势，我们相信到 2026 年底，Claude Code 将占每日提交量的 20%。

Dylan Patel is sharing some data. He says 4% of GitHub public commits are being authored by Claude Code right now. At the current trajectory, we believe that Claude Code will be 20% of all daily commits by the end of 2026.

我觉得这个数字低了。我本来预期是 99%。抱歉。

That feels low to me. I was expecting 99%. Sorry.

嗯，我觉得你可以这样理解：因为当你用 Claude Code 提交时，它会自动添加署名。所以有很多人用了 Claude Code 但没有归因到 Claude。

Well, I think the way you can see this is because whenever you actually do a commit with Claude Code, it adds itself. So there are a ton of people that are using Claude Code but then not attributing back to Claude.

是的。

Yes.

我们还有一些消息。来自 Analysis 的 Doug Olaflin 明天将加入节目，分析 Claude Code。我们可能得把他从精神病中拉出来。他对 Claude Code 上瘾了。他一直在运行多个实例，并且在 Semi Analysis Slack 中更新了他的头像，以表达他对 Shog 的忠诚，我想。

We also have some news. Doug Olaflin from Analysis will be joining the show tomorrow to break down Claude Code. We might have to pull him out of his psychosis. He is addicted to Claude Code. He's running multiple instances at all times and he has updated his profile picture in the Semi Analysis Slack to reflect his devotion to the Shog, I suppose.

OpenAI 发布了 Frontier，一个用于构建 AI 同事的产品。这出现在《华尔街日报》上，OpenAI 也发布了相关消息。这个新平台在市场对 AI 颠覆软件的担忧中推出，旨在帮助企业开发与人类协作的 AI 智能体。所以这里有一些有趣的问题……

OpenAI unveils Frontier, a product for building AI co-workers. This is in the Wall Street Journal and OpenAI also posted about it. The new platform launched amid market fears over AI's disruption to software is aimed at helping businesses develop AI agents that work alongside humans. So there are some interesting questions here about...

把这些看作是在积极试图抢你工作的 AI 同事。

Think of these as like AI co-workers that are actively trying to take your job.

就像他们在努力帮忙，但同时也想要你的头衔。他们想要你的薪酬。是的。

Like they're trying to help, but they also want your title. They want your comp. Yes.

他们想了解让你变得优秀的一切，他们想要……

They want to learn everything about what makes you great and they want...

也许，或者也许他们只是想赋能你，Jordy。也许他们只是想让你变得更好。我不知道。也许他们只是过来讲笑话，试图分散你的注意力。

Maybe or maybe they're just trying to empower you, Jordy. Maybe they're just trying to make you a better you. I don't know. Maybe they're just coming over cracking jokes trying to distract you.

是的。

Yeah.

同时在学习，这样他们就能……

While learning so that they can...

他们确实可以两者兼得。他们确实可以两者兼得。

They can literally do both. They can literally do both.

在我们深入讨论之前，让我介绍一下 Reream。一个直播流，30 多个目的地。如果你想多平台直播，请访问 reream.com。

Before we dive into this, let me tell you about Reream. One live stream, 30 plus destinations. If you want to multiream, go to reream.com.

Frontier 与 OpenAI 之前宣布的 AI 智能体构建工具配合使用，使企业更容易整合智能体执行任务所需的数据源。这家 AI 公司表示，智能体将能够处理来自各种来源的信息，并完成诸如处理文件和运行代码等任务。OpenAI 表示，不再需要将所有内容复制粘贴到 ChatGPT 企业版中。它应该能访问你的网络，接入你所有不同的系统。我想你将能够编写 API 绑定，可能还会有一些来自 OpenAI 的前沿部署工程师或助理帮助你完全接入所承诺的智能体工作流。是的，他们正在招聘多少顾问来帮助推向市场？

Frontier works with OpenAI's previously announced AI agent building tools and makes it easier for businesses to combine sources of data that agents need to perform tasks. The AI company said the agents will be able to process information from various sources and complete tasks like working with files and running code. OpenAI said no more copy-paste everything into your ChatGPT Enterprise Edition. It should have access to your network, plug into all your different systems. You'll be able to write API bindings I imagine and there might be some forward deployed engineers or associates from OpenAI that are helping you actually onboard fully to the agentic workflows that have been promised. Yeah, they're hiring how many consultants to help with this go to market.

他们正在招聘数百名 AI 顾问以促进企业销售。这是信息。

They are hiring hundreds of AI consultants to boost enterprise sales. This is information.

创造就业。

Job creation.

是的，这是一份好工作。如果你在这个市场，我强烈建议抓住这个机会。

Yeah, this is a great gig. I would highly recommend jumping on this if you're in this market.

OpenAI 正在招聘数百名新员工，以扩大一个技术咨询团队，帮助大型企业开发定制 AI 应用和智能体来自动化员工任务。据一位知情人士透露，这一招聘努力可能有助于它击败主要竞争对手 Anthropic，后者也在服务企业方面加大了力度。与此同时，OpenAI 正准备推出一项新的企业产品，以统一企业使用 AI 的努力。ChatGPT 制造商正在增加其技术顾问（也称为前沿部署工程师）的数量，他们可以使用客户自己的数据定制 OpenAI 的模型。一位人士表示，这些工程师可以例如帮助 T-Mobile 开发 AI 来响应客户服务请求，或帮助 Intuit 为其客户提供税务准备服务。所以你有了所有这些数据，你想在上面进行长上下文强化学习。长上下文强化学习在编程领域非常成功，因为 Git 拥有每一行代码的完整历史，以及每条注释说明原因。你拥有构建软件时发生的一切的完美记录，并且有大量的开源仓库。你基本上可以下载整个 GitHub，看看软件是如何开发的，然后训练模型。嗯，你无法从企业那里获得同样水平的免费数据。很多数据都被锁定了，而且很多是特定于某个业务流程的。

OpenAI is hiring hundreds of new staffers to expand a technical consulting team that helps large corporations develop custom AI applications and agents to automate employee tasks. According to a person knowledgeable with the company's plans, the hiring effort could help it beat back competition from arch rival Anthropic, which has also upped its game in catering to enterprises. It comes as OpenAI prepares to launch a new enterprise offering that would unify businesses' efforts to use AI. The ChatGPT maker is expanding its number of technical consultants, also known as forward deployed engineers, who can customize OpenAI's model using a client's own data. A person said these engineers can for example help T-Mobile develop AI to respond to customer service requests or help Intuit provide its customers with tax preparation services. So you have all this data, you want to do long context reinforcement learning on it. Long context reinforcement learning has been very successful in the coding world because Git has a complete history of every line of code that's been written, every comment why it happened. You have this perfect record of everything that happened when you built a piece of software and there's a ton of open source repositories. You can download all of GitHub basically and see how software is developed so you can train the model on that. Well, you can't really get the same level of free data with enterprises. A lot of this data is locked up and a lot of it's specific to a specific business process.

是的。

Yeah.

让我们调出一张图片。这是他们从 OpenAI 的 Frontier openai.com/index/slash introducing 制作的图表。

Let's pull up an image. There's a little graphic here that they made from OpenAI's Frontier openai.com/index/slash introducing.

是的。

Yeah.

如果你能稍微放大底部，你会看到你的记录系统，以及业务上下文、智能体执行、评估和优化。你的智能体、OpenAI 智能体、第三方智能体。这感觉很重要，对吧？他们想成为编排者，对吧？如果你想引入其他人来帮忙，很好。至少目前是这样。

And if you can zoom in a little bit at the bottom you have your system of record and you have business context, agent execution, evaluation, and optimization. Your agents, OpenAI agents, third-party agents. That feels significant, right? They want to be the orchestrator, right? And if you want to bring in some other folks to help out, great. At least for now.

然后他们有界面，即 ChatGPT 企业版、OpenAI Atlas 和其他商业应用。

And then they have interfaces which are ChatGPT Enterprise, OpenAI Atlas, and other business applications.

我有点失望他们没有采用《疯狂的麦克斯》主题，比如 Gas Town。我喜欢 pole cats、市长、执事。我觉得那是个有趣的比喻。他们选择了更企业化的东西，但我认为 Frontier 是个好名字。我不知道。它当然很顺口。它不是 Frontier 5.2 high，所以我们得庆幸。但我认为 Frontier 不错，你知道吗？你正在做一个 Frontier 集成项目吗？哦，那些来自 OpenAI 的顾问，他们正在让我们接入 Frontier。听起来不错。Tyler 为什么在笑？

I'm a little upset that they didn't go with a Mad Max theme like Gas Town. I like the pole cats. I like the mayor. I like the deacon. I thought that was a fun metaphor. They went with something a little more enterprisey, but I think Frontier is a good name. I don't know. It flows, of course. It's not Frontier 5.2 high, so we got to count our blessings here. But I think Frontier works, you know. Are you working on a Frontier integration project? Oh, those consultants from OpenAI, they're getting us onboarded to Frontier. It sounds good. Why is Tyler laughing?

Tyler 在笑。

Tyler's laughing.

是的，我在评论里被喷了。Tyler 怎么了？

Yeah, I'm getting flamed in the comments. What's up with Tyler?

是的。发生了什么？你用了洗发水还是护发素？

Yeah. What happened? Did you use shampoo or conditioner?

我需要理发。也许你需要一顶帽子。也许你可以拿那边的一顶 TDP 帽子。

I need a haircut. Maybe you need a hat. Maybe you can grab one of those TDP hats over there.

关于 5.2 的最后一点，最后一个有趣的基准是：GPT-5.2 在高等（而非特高等）推理努力下，在我们扩展的软件任务套件上，50%的时间跨度约为 6.6 小时。这是我们迄今为止报告的时间跨度测量的最高估计值，并且正好以每 128 天翻倍的速度增长。这取决于你在寻找什么，但其含义是越来越多的任务适合这些模型。不过，我认为我们正在经历一个编排方面的转变，可能需要一个新的基准，因为如果单个模型 GPT-5.2 太高，基本上可以完成一个六小时的任务。但你可以部署五个相互通信的模型，组合起来就能完成 20 小时的工作。图表上会出现一个跳跃，但这并非模型本身的技术进步。所以，大概需要一个新的基准。

So the last thing on 5.2, the last benchmark that's interesting is GPT-5.2 with high not extra high reasoning effort has a 50% time horizon of around 6.6 hours on our expanded suite of software tasks. This is the highest estimate for time horizon measurement we have reported to date and it's right on track doubling every 128 days. It depends on what you're looking for but the implications of this are that more and more tasks are suitable for these models. Although I do think we're going through a shift with the orchestration thing where we might need a new benchmark because what happens if the individual model GPT-5.2 too high can do one six-hour task basically. But you can deploy five that talk to each other and when you combine those it adds up to 20 hours of work. You just get a jump in the graph but it's not technically the model. So you need sort of a new benchmark supposedly.

Jessica Lesson 喜欢 Sam Altman 的帖子。她说这是 Sam Altman 的精彩对比。一个令人瞠目的数据，表述简单易懂。但我讨厌这种新兴的倾向，即把商业模式辩论框定为精英主义与反精英主义，而聊天机器人一开始就采用了订阅商业模式。我曾以为这将是消费科技与广告和订阅相结合的伟大新时代，就像我们在 Netflix 和 Spotify 上看到的那样。如果 OpenAI 与 Anthropic 的竞争抹杀了这一点，让人觉得订阅只是精英的领域，那将是一种遗憾。许多其他消费企业证明了并非如此，而且它们因平衡而受益。我喜欢这一点。Signal 说要么是你，要么是我，已经失去了处理现实的能力。

Jessica Lesson likes Sam Altman's post. She says this is excellent comps from Sam Altman. An eyepopping stat in an easy to understand claim. And I hate though this budding movement to frame the business model debate as elitism versus anti-elitism when chatbots launched with subscription business models. I thought it would be a great new era of consumer tech blended with ads and subs much as we've seen with Netflix and Spotify. It would be a shame if the OpenAI versus Anthropic rivalry wipes that away and makes it seem like subscriptions are only the domain of the elite. So many other consumer businesses show otherwise and they are better off for the balance. I like that. Signal says either you or I have lost the ability to process reality.

是的，我的意思是不是回应这个。

Yeah, I mean not in response to this.

回应那个。好的。是的，很多人根本不认为这是好的沟通。我的意思是，我看到的大多数帖子都是这样的：'嘿，这感觉像是你被钓鱼了。'

In response to that. Okay. Yeah, a lot of people didn't think this was great comms at all. Like I mean I would say most of the posts I saw were like, 'Hey, this feels like you got baited.'

是的。因为你可以选择不回应，我猜。

Yeah. Because you could have just not responded, I guess.

是的。或者你可以在当天晚些时候说：'我很高兴地分享，Codex 已有 100 万用户。'

Yeah. Or you could have said he said later in the day, 'I'm excited to share that Codex has 1 million users.'

是的。他们这周推出的产品，对吧？

Yeah. Which product they launched this week, right?

是的。也许只关注自己吧。我不知道。

Yeah. Just focus on yourself maybe. I don't know.

与此同时，在硅谷，Greg 说 GB200 确实让我们能够做一些了不起的事情。

Meanwhile over in Silicon, Greg says GB200 has really been enabling us to do some amazing things.

是的，值得注意的是，推出速度相对较慢。显然，它已经非常快了，但从 Nvidia 宣布这些芯片到实际制造、上架、布线、准备训练新模型，确实需要时间。所以呢？是的，我的意思是，我认为 Dean Ball 今天有一篇新文章，他说，你知道，公司甚至还没有 Blackwells。所以我们还没有看到任何在 Blackwells 上训练的模型。

Yeah, it is remarkable how relatively slow the rollout has been. Obviously it's been very quick but it does take time from the time that Nvidia announces these chips to actually fabbing them, racking them, wiring them up, getting ready to train new models. And so anything? Yeah, I mean I think Dean Ball had a new post today and he said like, you know, companies don't even have Blackwells yet. Like so we haven't seen there's no models that have been trained on Blackwells.

是的，相当。所以准备好看到一些 AI 进步吧，但 Mike Isaac 在聊天中说：'沟通不好。'

Yeah, pretty. So it's going to get ready to see some AI progress, but we got Mike Isaac in the chat. He says not good comms.

不好。

Not goods.

我同意 Mike 的看法。

I'm with Mike on that one.

让我告诉你关于谷歌的财报，因为我们提到过这一点，但在上下文中很重要。Alphabet 的销售额创下纪录，支出翻倍。他们全力投入 AI。谷歌母公司 Alphabet 报告第四季度收入增长 18%，由数字广告增长推动。在我们深入之前，Mike 昨天发了一条精彩推文。他说：'从我看到的 OpenAI 员工发推文谈论 Anthropic 超级碗广告的数量来看，Anthropic 应该向 OpenAI 支付赚取的媒体费用。'

Let me tell you about Google earnings because we touched on this but it's important in the context. So Alphabet sales hit record spending to double. They're going all in on AI. Google parent Alphabet reported an 18% jump in fourth quarter revenue driven by growth in digital advertising. Before we go into this, Mike had a banger yesterday. He said, 'By the number of OpenAI employees I see tweeting about Anthropic Super Bowl ad, Anthropic should be paying OpenAI earned media fees.'

是的，那是个好观点。

Yeah, that was a good take.

那是个好观点。销售额达到近 1140 亿美元，超出分析师预期。净利润 345 亿美元，同比增长 30%。公司报告 2025 年销售额创纪录的 430 亿美元。利润 1320 亿美元。还不错。Alphabet 的股价波动很大。《华尔街日报》报道称，盘后交易下跌 1%，但有很多因素。谷歌和其他科技公司一样，计划花费数百亿美元开发 AI 模型，并建设训练和运行它们所需的数据中心。公司表示，预计 2026 年资本支出在 1750 亿至 1850 亿美元之间，而 2025 年为 910 亿至 930 亿美元。所以，他们都在翻倍，这是指数级增长，宝贝。准备好迎接一些 AI 进步吧。

That was a good take. Sales reached nearly $114 billion ahead of analyst expectations. Net income $34.5 billion, a 30% increase compared with the period a year earlier. The company reported a record $43 billion in sales for 2025. Profit $132 billion. Not too bad. Alphabet shares were sort of all over the place. The Wall Street Journal printed that they went down 1% in after hours trading, but there's a lot going on. Google, like other technology companies, plans to spend tens of billions of dollars to develop AI models and build the data centers needed to train and run them. The company said it expected to spend between $175 billion and $185 billion in capex in 2026, up from $91 billion to $93 billion in 2025. So, they're like all doubling, which is exponential growth, baby. Get ready for some AI progress.

是的。有人说，2026 年的预计资本支出将超过谷歌截至 2021 年的历史资本支出总和。所以，在一年之内，他们将超过这个数字，这太疯狂了。

Yeah. Somebody was saying that the 2026 projected capex will be more than the lifetime capex for Google up to 2021. So, in a single year, they're going to eclipse that, which is just insane.

Buco 有一个好观点。他说：'谷歌故意加大资本支出。在 IPO 之前告诉市场，这就是击败我们所需要的代价。'

Buco had a good take. He said, 'Google capex on purpose. Tell the market this is what it will take to defeat us before IPOs hit.'

是的，如果你与他们竞争，当然令人紧张，但在与 Anthropic 和 OpenAI 的资本战中，他们在资本方面有优势，但竞争仍然真实存在。

Yeah, certainly nerve-wracking if you're competing with them, but they have the edge on the capital side in the capital war over Anthropic and OpenAI, but the race is still real.

《华尔街日报》提供了更多关于谷歌的背景信息。'谷歌大力押注其 AI 冬季地位。'是 AI 冬季地位，不是 AI 冬天。'广告和云增长加速证明了 Alphabet 股价近期飙升的合理性，但惊人的资本支出预测仍然令人窒息。今天人工智能竞赛的口号应该是：如果你有，就花掉。这是 Meta Platforms 上周在第四季度报告中牢记的信息，这家 Facebook 和 Instagram 的母公司宣布计划在资本支出上花费高达 1350 亿美元，而去年约为 720 亿美元。谷歌周三加码，计划今年花费高达 1850 亿美元，大约是去年支出的两倍。谷歌的年收入现已超过 4000 亿美元，大约是 Meta 的两倍。尽管如此，新的支出目标，即使对于一家最近全速运转的公司来说，也令人窒息。股价在第四季度报告和电话会议后盘后交易中下跌。谷歌拥有政治和金融资本来下这样的赌注。该公司的 Gemini 3 模型已使其在 AI 模型性能上名列前茅。'

The Wall Street Journal had some more context on Google. 'Google leans hard into its AI winter status.' AI winter status, not AI winter. 'Ad and cloud growth acceleration justify the recent surge in Alphabet stock, but blowout capex forecast still takes one's breath away. The motto for the artificial intelligence race today should be if you've got it, spend it. That's a message that Meta Platforms took to heart during its fourth quarter report last week when the Facebook and Instagram parent announced plans to spend up to $135 billion on capex compared to about $72 billion last year. Google managed to up the ante Wednesday with its own plan to spend as much as $185 billion this year, which would be about double last year's outlay. Google's annual revenue has now topped $400 billion, about twice as large as Meta's. Still, that new spending target, even for a company that has been firing all cylinders lately, takes one's breath away. The stock price slipped in after hours trading after its fourth quarter report and conference call. Google has both the political and financial capital to lay such a bet. The company's Gemini 3 model has put it on top of a heap of performance for AI models.'

尽管其搜索引擎和 Gmail 等产品无与伦比的覆盖范围迅速推动了采用，谷歌周三表示，仅 Gemini 应用就有超过 7.5 亿月活跃用户，而这只是 Gemini 实际用户的一部分，因为它被整合到了所有不同的产品中。聊天室里发生了什么？

While the unmatched distribution of its search engine and products like Gmail have quickly driven adoption, Google said Wednesday that it has more than 750 million monthly active users just on its Gemini app, which only represents a portion of Gemini's actual users because it's vended into all the different products. What's going on in the chat?

40 军团进入了聊天室。有意思。他们用 #ke keep40 刷屏。好吧。他们想让 Sam Altman 听到。我们……不是 keep 40。40 不是已经被弃用了吗？所以他们不应该升级标签，改成“bring back 40”吗？复活 40。复活 40。呃，实际上它要到 2 月 13 日才下线。哦，好吧。所以还有时间撤销决定。我很想知道维持 40 的运营成本，因为感觉如果它运行良好，显然是人们愿意付费的东西。是啊。能不能就把它放在角落里让它跑着？你知道，这可能更多是一个用户行为决策，而不是财务决策。呃，因为运行旧模型似乎会变得非常昂贵。

40 Army has entered the chat. Interesting. They're hitting the chat with #ke keep40. Okay. They want to be heard by Sam Altman. We'll be... It's not keep 40. Hasn't 40 been deprecated? And so, shouldn't they upgrade the update the hashtag to say like bring back 40? Revive 40. Revive 40. Uh, it's actually not leaving till February 13th. Oh, okay. So, there is time to reverse the decision. I am interested to know about the operating cost of keeping 40 alive because it feels like if it rolls clearly something people were willing to pay for. Yeah. Can you just kind of leave it in a corner and just kind of run it? It was, you know, maybe more of a user behavior decision than a financial one. Uh, because it seems like running legacy models gets really expensive.

Gemini 的强劲反响，加上谷歌在联邦政府拆分公司的诉讼中获胜，让投资者在科技和 AI 情绪低迷时感到振奋。Alphabet 股价在过去三个月上涨了约 20%。同期，英伟达、微软、亚马逊和博通都出现了下跌。但资本也很重要。而谷歌的商业模式在这里最能体现其价值。该公司的广告业务长期以来一直是利润丰厚的现金牛，目前仍保持着两位数的增长率。增长实际上还在加速，第四季度广告收入同比增长 14%，而上一季度为 13%。这和我们之前在 Meta 看到的情况一模一样。广告越来越好了，宝贝。该公司的云计算部门更令人印象深刻，收入同比增长 48%，达到 178 亿美元。自 2021 年初以来，谷歌云还没有出现过这样的增长，当时该业务规模还不到现在的三分之一。谷歌云上一季度实现了创纪录的 53 亿美元营业利润，比华尔街目标高出 45%。该公司蓬勃发展的业务在 2025 年产生了近 1650 亿美元的经营现金流，是标普 500 指数中最高的。这些强劲的业绩将帮助投资者消化最新的投资计划。但支出可能占到年收入的 40%，用于 AI 芯片和相关基础设施，这仍然是一场豪赌。此类投资将大幅提高折旧费用，上一季度折旧使净利润减少了 18%。谷歌已不再是不到一年前那种因拆分担忧导致股价低于预期收益 16 倍时的便宜货了。我的问题是，AI 建设何时停止？因为如果你不断投入更多，现金流从哪里来？作为分析师，你会看到，他们花费收入的 50% 或 40% 用于 AI 建设和数据中心，但他们需要这样做多久？如果他们必须永远这样做，那么你的业务就会永久性地变差，因为你只是在不断购买硬件。

The strong reception of Gemini along with Google's victory over the federal government's efforts to break up the company have cheered investors when sentiment on technology and AI is faltering. Alphabet stock price has jumped around 20% in the last three months. Nvidia, Microsoft, Amazon.com, and Broadcom have all lost ground during that time. But capital matters, too. And here's where Google's business model pays off the most. The company's advertising arm has long been a lucrative cash cow that is still managing double-digit growth rates. Growth has actually accelerated with ad revenue up 14% in the fourth quarter compared to 13% in the previous one. It's exactly what we saw at Meta. The ads are getting better, baby. The company's cloud computing division was even more impressive with revenue growth jumping 48% year-over-year to hit 17.8 billion. Google Cloud hasn't seen growth like that since early 2021 when the business was still less than a third of its current size. Google Cloud turned in a record 5.3 billion in operating profit in the last quarter, a figure that was 45% higher than Wall Street's targets. The company's booming business produced nearly $165 billion in operating cash flow in 2025, the highest in the S&P 500. These strong results will help investors digest the latest investment plan. But spending could be what, 40% of annual revenue on AI chips and related infrastructure is still a sizable gamble. Such investments will sharply elevate depreciation charges which in the latest quarter reduce net income by 18%. Google isn't the bargain it was less than a year ago when breakup fears had the stock trading at less than 16 times projected earnings. My question is like when does the AI buildout stop? Because if you're constantly investing more and more, where does the cash flow come from? Like as an analyst, you look at this and you say, okay, they're spending, you know, 50% of their revenue or 40% of their revenue on the AI buildout on data centers, but how long will they need to do this? Like if they have to do this forever, then you just permanently have a worse business because you're just constantly buying hardware.

是啊。我的意思是，你会获得巨大的能力提升，大量劳动力转移到数据中心，最终收入激增。是啊。你能设想一个场景，继续增加资本支出是合理的，因为收入增长更快。是啊。嗯，值得注意的是，即使财报更新后，英伟达今天仍在下跌，你会以为人们会看到，嘿，你知道，黄仁勋对未来几年给出了一些相当疯狂的预测，你必须……这里的更新让……在 TPU 上花了很多钱，但他们肯定还会很快购买大量英伟达芯片。让我告诉你关于 Plaid 的事。Plaid 为你用来消费、储蓄、借贷和投资的应用提供支持，通过连接银行账户来转移资金、打击欺诈、改善贷款，现在还有 AI。Richard 说谷歌是一家不炒作的公司。他们从 900 亿增加到 1800 亿的资本支出，可能是长期投资者能看到的最乐观的事情，因为它显示了未来收入增长的规模。我很惊讶，在这个阶段大多数人仍然不明白这一点。是啊。对此非常兴奋。呃，Doug 在 Fabricated Knowledge 上说了什么？他说谷歌去年 CFO 达到 1630 亿，现在计划明年资本支出 1800 亿。现在，聋子认为加速即将到来，但哇，我宁愿破产也不愿输掉比赛。引用 Sergey 的话，过去 12 个月的自由现金流令人震惊。嗯，今年将产生 2000 亿现金流。目前仍低于资本支出。还将有 800 亿净现金，可能还有 1000 亿的各种少数非流动性股权。所以是的，他们拥有 SpaceX 和一堆 Waymo。呃，以目前的增长率，他们可以在 2030 年筹集 3000 亿，甚至不用负债。完全同意，但这对我来说感觉非常接近全力以赴，Doug 说。

Yeah. I mean, you get a massive capability increase, lots of labor moves into data centers and eventually the revenue spike. Yeah. Can you enter a scenario where it makes sense to continue to increase capex because revenue is accelerating even faster. Yeah. Yeah. Um, not notable that Nvidia is still trading down today even after that update on earnings like you would expect people to look and see like hey, you know, Jensen had given some pretty kind of wild projections right looking out over the next couple years and you have to... updating here makes it a lot more... spend a lot on TPU but still they will be buying a lot of Nvidia chips for sure really quickly. Let me tell you about Plaid. Plaid powers the apps you use to spend, save, borrow, and invest securely by connecting bank accounts to move money, fight fraud, and improve lending now with AI. Richard says Google is a company that doesn't do hype. For them to go and increase capex from 90 billion to 180 billion is probably the most bullish thing long-term investors can see as it shows the scale of future revenue growth. I'm shocked that at this stage most still don't understand this. Yeah. So excited about it. Uh, what did Doug say over at Fabricated Knowledge? He said Google made 163 billion in CFO last year and is now planning to spend 180 billion in capex next year. Now, deaf think acceleration is coming but wowza, I'd rather go bankrupt than lose the race. Quoting Sergey, the trailing 12 months of free cash flow is staggering over there. Um, going to make 200 billion in cash flow this year. Still less than capex for now. Also gonna have 80 billion in net cash and probably a hundred billion in myriad minority illiquid equity stakes. So yeah, they own like SpaceX and a bunch of Waymo. Uh, with current rates of growth, they can raise 300 billion in 2030 without even going into debt. Totally agree, but this feels pretty close to the pedal to the metal to me, says Doug.

所以，Joe Weisenthal 说：“地球上平均每人每天看 25 个 YouTube Shorts。YouTube Shorts 日均观看量达 2000 亿次。”这太多了。简直疯狂。这太多了。我的意思是，你看一个 YouTube Shorts 大概需要多久？平均 5 秒，对吧？因为你跳过一些，看一个 5 秒，一个 15 秒。所以，呃，25 个 YouTube Shorts 大概相当于浏览 5 到 10 分钟，但仍然相当惊人。是啊，我还没有迁移，我还没有刷 YouTube Shorts。我发现如果我搜索某个东西，比如我在查一辆车，如果我没时间看完整评测，看个 60 秒的短片也不错，但我不会坐在那里刷。是啊，它显然仍然服务很多人，但我很少刷，而在 Instagram 上我会刷 Reels 的 feed。呃，是啊。继续内容领域，Khi 分享说，仅 YouTube 去年就产生了超过 600 亿收入，超过了 Netflix。Polymouse 说：“YouTube 正在用这种非常巧妙的内容策略击败 Netflix，他们的创作者制作人们想看的内容，然后通过观看量和金钱获得回报。”这简单概括了 YouTube 讨厌这个一个奇怪的简单技巧。

So, Joe Weisenthal says, 'The average person on Earth is watching 25 YouTube shorts every single day. YouTube shorts average 200 billion daily views.' That's a lot. That's just insane. That's a lot. I mean, you can watch a YouTube short in like what? 5 seconds on average, right? Because you skip one, you watch one for 5 seconds, one for 15 seconds. So, uh, I mean 25 YouTube shorts that feels like maybe five or 10 minutes of browsing, but still pretty staggering. Yeah, I haven't migrated any, like I don't scroll on YouTube shorts yet. I find if I search for something like let's say I'm looking up a car, it's nice to get a 60-second if I don't have time to watch a review, but I'm not like sitting there scrolling. Yeah, it still serves clearly a lot of people, but I rarely scroll through whereas on Instagram I will scroll the feed of reels. Uh, yeah. Staying in the content world, Khi shared that just in YouTube generated over 60 billion last year, more than Netflix. And Polymouse says, 'YouTube is beating Netflix with this really sneaky content strategy in which their creators make stuff people want to see and are then rewarded for it with views and money.' And it is a simple encapsulation of the YouTube hates this one one weird simple trick.

UGC 是一个非常大的生意。谁能想到呢？但质量的规模……我希望我们很快能和 MrBeast 聊聊他对 YouTube 的看法以及它如何变化，因为他把内容带到了一个令人难以置信的方向，质量堪比 Netflix。我是说，他把内容卖给 Amazon Prime，对吧？所以他肯定做到了。司法部的反垄断部门今天发布了一个更新。DOJ 反垄断部门提交了通知，将对谷歌非法垄断互联网搜索和搜索广告案的补救措施决定提出交叉上诉。所以他们还在解决这个问题。让我告诉你关于 Labelbox：强化学习环境、语音、机器人、评估和专家人类数据。Labelbox 是全球领先 AI 团队背后的数据工厂。所以大型科技公司正在向印度投入大量资金，以寻求 AI 霸权。我们昨天稍微讨论了这个故事。我们没能深入了解。

UGC is a big, big business. Who would have thought? But the scale of the quality... I'm hoping we're going to talk to MrBeast about his outlook on YouTube and how it's changing soon, because he's taken it in such an incredible direction where the content is Netflix quality. I mean, he sells it to Amazon Prime, right? So he's certainly there. In the antitrust division of the Department of Justice, they posted an update today. The DOJ antitrust division filed notice that it will cross-appeal from the remedies decision in its case against Google's unlawful monopolization of internet search and search advertising. So they're still hashing that out. Let me tell you about Labelbox: reinforcement learning environments, voice, robotics, evals, and expert human data. Labelbox is the data factory behind the world's leading AI teams. So big tech is throwing cash into India in its quest for AI supremacy. We talked about this story a little bit yesterday. We couldn't get to the bottom of it.

彭博社有一篇报道说，谷歌在印度的办公面积大幅扩张。问题是，他们为什么要扩张呢？他们在资本支出、服务器和数字工人上投入了这么多，但当然他们也需要人类。所以美国科技巨头正在争夺人工智能的领导地位，不仅在美国，而且在全球。印度张开双臂欢迎他们，《华尔街日报》说。印度正成为全球最热门的市场之一，美国 AI 巨头希望迎合这个国家庞大且精通数字技术的人口。为了吸引更多科技投资，印度政府上周末宣布计划，给予科技公司基于印度的全球数据服务所获得的海外收入 20 年的税收减免。所以你去设立一个数据中心或科技业务，你从整个数据中心或你销售的软件中赚钱，而 20 年内不必向印度缴税。那是很长很长的时间。Techno Chief 昨天指出了这一点。

Bloomberg had a story that effectively Google is scaling square footage in India pretty dramatically. And the question was, why are they scaling it up at all? They're investing so much in capex and servers and digital workers, but of course they need humans as well. So American tech behemoths are racing to establish leadership in artificial intelligence, not just in the US, but also around the world. India is welcoming them with open arms, says the Wall Street Journal. India is becoming one of the hottest markets globally for US AI titans looking to cater to the country's massive and digitally savvy population. Looking to attract more tech investments, the Indian government announced plans over the weekend to give tech firms a 20-year tax break on overseas revenue gleaned from global data services based in India. So you go and set up a data center or a tech business, you're making money off of that whole data center or whatever you're selling in terms of software, and you don't have to pay taxes to India for 20 years. That's a long, long time. Techno Chief called this out yesterday.

此举是印度政府推动该国成为主要 AI 服务提供商的一部分，包括解决本地问题的低成本工具，同时将尖端创新留给美国和中国的资金雄厚的公司。印度科技部长表示，这将使印度有机会成为主要的 AI 中心。过去几个月，美国科技公司宣布了在印度数据中心数百亿美元的投资，因为他们竞相在全球建设 AI 基础设施。10 月，谷歌宣布在印度东南部投资 150 亿美元建设数据中心和海底电缆，该公司称这是其在美国以外最大的单一 AI 中心。12 月，微软宣布了其在亚洲有史以来最大的投资，承诺投入 175 亿美元发展该国的云和 AI 基础设施。亚马逊也承诺到 2030 年在印度运营中投入 350 亿美元。被称为超大规模企业的美国大型数据中心公司被这个拥有 14 亿消费者的国家所吸引，这些消费者是数据和 AI 聊天机器人最活跃的用户之一。非常有趣。

The move is part of the Indian government's push to make the country a major provider of AI services, including low-cost tools to solve local problems while leaving cutting-edge innovation to deep-pocketed firms in the US and China. This will give India the opportunity to become a major AI hub, said India's technology minister. In the past few months, US tech companies have unveiled tens of billions of dollars in investment in Indian data centers as they race to build AI infrastructure around the world. In October, Google announced a $15 billion investment in data centers in southeastern India as well as undersea cable links in what the company described as its largest single AI hub outside the United States. In December, Microsoft unveiled its largest ever investment in Asia with a $17.5 billion pledge to develop the country's cloud and AI infrastructure. Amazon also pledged $35 billion across its operations in India up until 2030. The big US data center firms known as hyperscalers are drawn to a country whose 1.4 billion consumers are some of the most prolific users of data and AI chatbots. Very interesting.

《信息》杂志有更多关于英伟达的消息。英伟达因内存短缺推迟了一款新的游戏芯片。这是你不希望跨越的界限。游戏玩家会起义，他们会冲击数据中心。他们会打电话给他们的代表，给他们的参议员打电话。Pause AI 有了新的群体，如果你不给我新的游戏显卡来玩最新的游戏。

There's more news from Nvidia in The Information. Nvidia is delaying a new gaming chip due to memory shortage. This is the line you don't want to cross. The gamers are going to rise up and they're going to storm the data centers. They'll be calling their representatives, calling their senators. Pause AI has a new cohort if you don't give me a new gaming graphics card to play the latest games.

这是 30 年来第一次在日历年内不发布一款。每年都有 1080 Ti、2080、3080……这些是很棒的显卡，现在不再了，只是推迟。我想你可以用云游戏之类的。但确实感觉很多视频游戏在标准显卡上已经达到了极限，所以每年升级的需求没那么大。你怎么看？

It's the first time in 30 years it won't release one in the calendar year. Every year the 1080 Ti, 2080, 3080... These were great graphics cards and now no longer, just delaying it. I guess you can use cloud gaming or something. But it does feel like a lot of video games have sort of maxed out what they can do with stock graphics cards, so there isn't as much need to upgrade every year. What do you think?

而且有了 Genie 3，我们在游戏方面有了新的进步。所以他们应该对此感到高兴。你将一无所有，但你会快乐。Tyler，你昨天玩了多久 Genie 3？很有趣，对吧？很有趣。所以你玩了。

Also with Genie 3, we're getting new advancements in gaming. So they should be happy about this. You'll own nothing and be happy. How much Genie 3 did you play yesterday, Tyler? It's fun, right? It's fun. So you play it.

我想过玩它。哦，好吧。你想过玩它，但机制不到位，对吧？你需要 Genie 4。我要打败 DA。那是目前的目标。我想要机制。我想要在 Genie 3 汽车模拟器、赛车模拟器中驾驶纽博格林赛道时换轮胎、加油。

I thought about playing it. Oh, okay. You thought about playing it, but the mechanics weren't there, right? You need Genie 4. I'm going to beat DA. That's the goalpost currently. I want mechanics. I want tire changes, gas refill as I'm driving the Nurburgring in my Genie 3 car simulator, race simulator.

内存芯片是 GPU 的关键组件。非常期待问 Sam Altman 关于 AI 瓶颈的各种问题，因为我们也应该问他关于世界模型的问题。那会很有趣。他不是那种不发布东西的人。他发布了很多东西。Ben Thompson 一直持有这种观点，即 AI 生成的视频就是元宇宙，你在 Meta Vibes 中看到的就是这样。你在 Sora 中看到的也是这样。这就是事情的发展方向。英伟达还因内存短缺削减了其当前游戏芯片系列 GeForce RTX50 GPU 的产量。一位人士表示，由于过去一年的稀缺性，英伟达最新游戏 GPU 的价格已经在零售店和网站上上涨。我想知道现在建造一台现代游戏 PC 到底要花多少钱，因为已经有一段时间了。过去几千美元就能买到顶级配置。我想知道现在所有短缺的情况下是什么样子。《信息》杂志称，如果市场改善，英伟达高管仍可能改变主意并发布游戏芯片，因为该公司以灵活和行动迅速而闻名。由于 AI 热潮，对计算机内存芯片的需求激增，因为训练和运行机器学习模型需要大量内存芯片。内存芯片充当存储数据的仓库。内存芯片短缺预计将导致消费电子产品价格上涨。

Memory chips are a key component of GPUs. Very excited to ask Sam Altman about the various pieces of the AI bottleneck since we should ask him about world models, too. That would be interesting. He's not one to not launch. He's launched a lot of stuff. Ben Thompson has had this take for a while that AI-generated video is the metaverse and that's what you're seeing with Meta Vibes. That's what you're seeing with Sora. This is sort of where things go. Nvidia is also slashing production of its current line of gaming chips, the GeForce RTX50 GPUs, because of the memory shortage. One of the people said prices of Nvidia's latest gaming GPUs have already risen at retail stores and websites due to their scarcity over the past year. I wonder how much it costs to actually build a modern gaming PC these days because it's been a while. It used to be like a few thousand dollars would get you a top-tier setup. And I wonder what it's looking like now with all the shortages. It's possible that Nvidia executives could still change their mind and release a gaming chip if the market improves, as the company is known for being flexible and moving quickly, says The Information. Demand for computer memory chips has skyrocketed due to the AI boom, as they are needed in large quantities to train and operate machine learning models. Memory chips act as warehouses for storing data. The memory chip shortage is expected to lead to higher prices in consumer electronics.

上周，苹果 CEO 蒂姆·库克表示，内存芯片价格上涨将影响公司三月份的季度利润率。他还提到在台积电获得产线时间方面有些困难。因此，整个科技市场都存在许多瓶颈。他暗示未来影响会更大，并指出：“我们确实看到内存市场价格持续大幅上涨。一如既往，我们会考虑各种应对方案。”游戏和 AI 芯片使用不同类型的内存，但两者都由来自三星电子、SK 海力士和美光科技这三大主要供应商的相同原材料制成。

Last week, Apple CEO Tim Cook said the rising prices of memory chips would have an impact on the company's March quarter margins. He also mentioned having trouble getting line time at TSMC to some extent. So there are many bottlenecks working their way through all the technology markets. He hinted that the impact would be greater in the future, noting, "We do continue to see market pricing for memory increasing significantly. As always, we'll look at a range of options to deal with that." Gaming and AI chips use different types of memory, but both are made of the same raw materials coming from one of three main suppliers: Samsung Electronics, SK Hynix, and Micron Technologies.

那么，话不多说，我们节目的第一位嘉宾，OpenAI 的 CEO 山姆·奥特曼。他正在休息室等候。欢迎来到节目，山姆。你好吗？

Well, without further ado, our first guest of the show, Sam Altman, the CEO of OpenAI. He's in the green room waiting. Welcome to the show, Sam. How are you doing?

欢迎回来。

Welcome back.

很好。谢谢你们再次邀请我。

Good. Thank you guys for having me back.

非常感谢。重要的一天。

Thanks so much. Big day.

重要的一天。

Big day.

开始吧。我们从哪里开始？从模型开始还是从 Frontier 开始？

Kick us off. Where should we start? Should we start with the model or Frontier?

我们能从模型开始吗？因为我玩得很开心。

Can we start with the model? Just because I'm having fun with it.

当然可以。详细说说。你们发布了什么？

Yeah, absolutely. Break it down. What did you launch?

我们发布了 5.3 Codex。我认为它是世界上最好的编程模型。我们收集了人们对 5.2 Codex 的大量反馈，并将其整合到一个模型中。它在编程方面更聪明，但也更快。你可以中途与它交互。我觉得它的个性更好了。它非常擅长计算机使用。所以感觉像是向前迈出了一大步。有趣的是，今天早上我们部署它的时候，几个非常精通这些模型的专家用户注意到了，并说：“天哪，Codex 真的不一样了。”他们在部署过程中就发现了。所以我认为你很快就能感受到。

We launched 5.3 Codex. It is, I think, the best coding model in the world. We took a lot of the feedback that people had about 5.2 Codex and got it into one model. It is much smarter at programming, but it's also way faster. You can interact with it mid-turn. I think it's got a much better personality. It's really good at computer use. So it feels like a very big step forward. It was funny as we were deploying it this morning, a couple of extremely expert users of these models noticed and said, "Man, something's really different with Codex." And they caught it mid-deploy. So I think you can really feel it quickly.

太棒了。你是说 OpenAI 之外的人，日常使用？

That's great. Oh, you're saying people outside of OpenAI, just everyday use?

是的，就像我们发布后的一小时内，在我们……

Yeah, like the start of an hour that we put it out before we...

谈谈中途交互。这是如何工作的？为什么重要？它解锁了什么？

Talk about interacting mid-turn. How does that work? Why is that important? What does it unlock?

人们开始一次性使用这些工具处理非常长的工作，比如多小时的任務。有时你指定得不对，对吧？有时某些东西设置不对。有些东西就是搞砸了。沿途引导它们的能力至关重要。它们可以在没有引导的情况下做出惊人的事情，但如果你沿途引导它们，它们可以做出更惊人的事情。所以这是我觉得这个模型最新颖的地方之一。

So people are starting to use these tools for very long pieces of work at one time, you know, multi-hour tasks. And sometimes you don't specify it correctly, right? Sometimes something's not set up right. Something just screws up. The ability to steer them along the way is crucial. They can do amazing things with no steering, but they can do much more amazing things if you steer them along the way. So this has been one of the things I felt most new about this model.

那么谈谈编排，这如何融入 Frontier，因为我想象……

So talk about orchestration, how this fits into Frontier, because I imagine...

等一下。值得注意的是：如果你看到同事犯错而不打断他们，那很粗鲁，对吧？这非常低效。

One second. It's notable: if you see a coworker making a mistake and you don't interrupt them, that's rude, right? It's deeply inefficient.

这些模型在没有反馈的情况下能做的事情令人难以置信。就像如果你想到一个新同事，尤其是，你训练他们，早期给他们很多反馈，他们学习工作，你纠正他们，他们得到练习。模型很快也会做到这一点，但现在它们还做不到。所以我们只能依赖要么它们一次做对，要么我们沿途纠正它们。

It is incredible what these models can do without any feedback. Like if you think about a new coworker, especially, you train them and give them a lot of feedback early on, they learn the job, you correct them, they get practice. The models will soon do that, but right now they don't. So we just rely on either they get it right one shot or we correct them along the way.

是的。所以有很多人在运行多个智能体和多个标签。他们开始考虑编排。Frontier 似乎是其中的一部分。但如果你与一个中途运行的模型交互，对于开发者来说，5.3 的用户体验如何改变？然后在 Frontier 世界中会是什么样子？

Yeah. So there are a lot of people running multiple agents and multiple tabs. They're starting to think about orchestration. Frontier feels like a piece of that. But if you're interacting with a model that's running mid-turn, how does the user experience change for developers with 5.3? And then what will it look like in the Frontier world?

我认为我们将走向一个工作流程，很多人感觉就像在管理一个智能体团队。随着智能体变得更好，它们将在越来越高的抽象层次上运作。至少从目前观察到的情况来看，人们会很好地实现这一跳跃。模型现在非常出色。能力过剩如此严重，以至于构建更好的工具让人们做到这一点——我们周一发布的 Codex 应用在这方面迈出了一大步——将非常重要。但你将管理非常复杂的工作流程。智能体会不断改进，所以你将一直以你的管理带宽或认知能力的上限工作，以跟踪所有事情。而让这变得容易的工具，我认为在一段时间内会比智能更重要，因为智能已经过剩了。

I think we will be heading towards a workflow where a lot of people just feel like they're managing a team of agents. As the agents get better, they'll keep operating at a higher and higher level of abstraction. At least watching what's happening so far, that is a jump that people are going to make pretty well. The models are so good now. There's such a capability overhang that building better tools to let people do that—which the Codex app we launched on Monday was a great step forward for—will be very important. But you will be managing very complex workflows. The agents will keep getting better, so you'll keep working at the maximum of your management bandwidth or cognitive ability to keep track of all the stuff. And the tools to make that easy to do will matter, I think, more than intelligence for a little while, because there's such an intelligence overhang already.

到今年年底，OpenAI 的前沿部署工程师的角色是什么？能力过剩对前沿部署工程师来说就像生肉。他们就是解决那个问题的，对吧？

What's the role of a forward deployed engineer today at OpenAI towards the end of the year? Capability overhang feels like raw meat for a forward deployed engineer. It's like they solve that problem, right?

是的。是的，我的意思是，看，最终模型会变得非常好，它们会帮助公司自我部署，而前沿部署工程师将再次在更高的抽象层次上工作。但就目前而言，你进入一家非 AI 原生公司，说：“好吧，你说你想部署 AI。”他们真的不确定该怎么做：如何将它连接到我的系统？我需要对代码库进行微调吗？如何考虑编排智能体和使用不同公司的产品？最重要的是，或者至少我们最常听到的是，如何考虑数据安全，以及如何知道这些 AI 协作智能体不会访问大量信息并以不应该的方式共享，或者遭受上下文利用之类的事情。所以前沿部署工程师利用这项令人难以置信的新技术和像 Frontier 这样的平台，说：“我们将把你的公司连接到一个 AI 平台，这样你就可以使用所有这些智能体、工作流程以及你想要的任何其他东西。”

Yeah. Yeah, I mean, look, eventually the models will get so good that they'll help companies deploy themselves, and the forward deployed engineers will again get to work at a higher level of abstraction. But for now, you go into a company that is not AI native and say, "Okay, you've said you want to deploy AI." They really are not sure what to do: how do I hook this up to my systems? Do I need to fine-tune a model on my codebase? How do I think about orchestrating agents and using things from different companies? Most of all, or at least what we hear most frequently, is how do I think about security of my data and how do I know that these AI coworking agents are not going to go access a bunch of information and share it in ways they shouldn't, or get a context exploit or something like that. So the forward engineers take this incredible new technology and a platform like Frontier and say, "We will connect your company to an AI platform so that you can use all these agents and workflows and everything else you want."

这些比喻有多重要，或者它们有多临时？我对读到关于 Gas Town 的内容非常感兴趣，你们有这些 pole cats，整个疯狂的麦克斯世界。这感觉可能只是一个暂时的异常，你为特定任务设置智能体，但这也可能在向大公司解释他们将如何在整个组织中整合 AI 时非常有价值。

How important are these metaphors or how temporary are they? I was very interested in reading about Gas Town and you have these pole cats and it's this whole Mad Max world. That feels like maybe just a temporary aberration where you're setting up agents for specific tasks, but also that could be incredibly valuable in explaining to a large corporation how they're going to integrate AI across the whole organization.

是的，我怀疑就像在一个快速发展的行业中发生的一切一样，所有这些在足够长的时间尺度上都是暂时的。随着这些模型变得更有能力，这些智能体在很长的时间范围内运作，能够自己解决问题，我们的信任和它们的稳健性不断提高，那么也许你就不需要我们今天需要的很多抽象了。

Yeah, I suspect like everything else that's happening when an industry is moving so fast, all of this is somewhat temporary on a long enough time scale. As these models become more capable and these agents are operating on very long time horizons with the ability to just kind of figure it out, and our trust and their robustness keeps going up, then maybe you don't need a lot of the abstractions we need today.

你知道，也许你公司里只有一个 AI 机器人，你可以说，‘嘿，我想推出这个新产品’，然后它就会完成一个雄心勃勃的人会做的一切。但我们现在还做不到。所以今天我们必须多花点功夫把各个部分拼凑起来。

You know, maybe you just have a single AI bot that runs at your company and you can say, 'Hey, I want to launch this new product' and it does everything an ambitious person would do. But that's not where we are today. So today we have to put in a little more work to get the pieces put together.

是的。你怎么看待长任务时间跨度的基准测试？你们在排行榜上名列前茅。同时，如果我们谈论智能体集群，感觉可能需要一个新的排行榜，因为它们能处理持续数周的任务，但会细分工作。推理模型内部有一些细分，但并没有真正的并行化，至少据我所知是这样。那么，在一个你调用一个模型，但它却在底层启动一大堆不同模型的世界里，会是什么样子？

Yeah. How are you thinking about the meter benchmark for long task horizon? You're at the top of the charts. At the same time, it feels like we might need a new chart if we're talking about agent swarms because they'll be able to do things that go for weeks, but they will subdivide the work. There's some subdivision that happens within a reasoning model, but it doesn't truly parallelize, at least that I'm aware of. So what does it look like in a world where you go to a model, but now it's spinning up a whole bunch of different models underneath?

我认为这个问题包含了整个领域的两个关键见解。第一，AI 领域的任何排行榜都不会持续超过几年，这个暗示是对的。至于这个排行榜，我们看看它还能有用多久。

I think there are two key insights of the whole field in this question. Number one, the implication that no chart in AI lasts more than a few years is right. And this one, we'll see how much longer it's really useful.

第二点是很多人认为我们需要一个超长任务和超长任务时间跨度，所以我们需要超长上下文。但人们已经在编码智能体上看到，通过智能体分解工作、良好编排、分派给子智能体，即使有当前技术的限制，我们也能做到一些事情——这并不令人惊讶，因为这与人类做事并完成大量工作的方式类似。所以看着很酷。我认为这将继续下去。

The second is a lot of people thought that we're going to need a super long task and a super long task horizon, so we need a super long context. And definitely what people have already seen with coding agents is by agents breaking up work, orchestrating it well, farming it off to sub agents, even with the current limitations of the technology, we can do something which should not be surprising because it's similar to how people do things and get amazing amounts of work done. So that's been cool to watch. I think that will keep going.

OpenAI 的一些人开玩笑说，很快重要的排行榜就只是 GDP 影响了。

A joke that some people at OpenAI make is that soon the chart that matters is just going to be GDP impact.

是的。

Yeah.

然后问题是什么是 GDP 之后的下一个？但其他所有代理指标，模型所做的工作已经产生了巨大的经济价值。

And then the question is what's the one that comes after that? But everything else, these proxy metrics, there's now so much economic value in what the models are doing.

你认为之后会是什么？

What do you think could come after?

我不知道。你有想法吗？

I have no idea. Do you have an opinion?

幸福感。

Happiness.

我不知道。

I don't know.

当你回顾 10 年前的一些博客文章时，你的预测通常相当准确。

When you look back at some of your blog posts from 10 years ago, your predictions were usually pretty on point.

是的。

Yeah.

也许很难预测。

Maybe it's hard to predict.

谢谢。谢谢。

Thank you. Thank you.

特别是合并，基本上有很多好的预测。比如现在我们收到成千上万条关于 40 的聊天消息，而你在 2016 年就预测人们会对聊天机器人产生强烈依恋。

Specifically the merge, basically there's a bunch that are good. Like right now we're getting thousands of messages in the chat about 40 and you predicted in 2016 that people would become very attached to a chatbot.

是的。我正在写一篇大的预测博客文章，未来 10 年似乎太远了，但未来 5 年可以。不过因为我很确定很多预测会是错的，但尝试仍然很有趣。

Yeah. I'm working on a big prediction blog post for the next 10 years seems too far, but the next five. But because it's like I'm sure a lot of it'll be wrong and it's still fun to try.

与聊天机器人的关系显然是我们现在需要更多担忧的事情，不再是一个抽象概念。甚至 GDP 之后是什么这个问题，我认为有趣的一个原因是，我们现在衡量 GDP 的方式可能会开始下降，即使生活质量大幅上升。

The sort of relationships with chatbots clearly that's something now we got to worry about more and is no longer an abstract concept. Even the question of what comes after GDP, one reason I think that's interesting is the way we measure GDP now could start going down even though quality of life goes way up.

当然。

Sure.

而且我们对此类事情没有太多经验。但大规模通缩……

And we don't have a lot of practice with things like that. But the massively deflationary...

不只是欧洲，我们也希望生活质量上升，而不仅仅是下降，你知道的。

Not just Europe, we want quality of life going up too, not just going down, you know.

是的。

Yeah.

换个话题，你怎么看待 Neol boom 以及整个硅谷正在进行的研究工作？感觉大家都承认需要研究突破，每个人都在尝试不同的方向。你认为这些公司会找到突破然后加入一个实验室，还是推出自己的产品？

Switching gears, what do you think about the work that the Neol boom, the research efforts that are happening all over Silicon Valley? It feels like there's an acknowledgment that there's research breakthroughs that need to happen and everyone's taking different shots at those. Do you think that those companies will just find a breakthrough and join a lab or launch their own products?

我认为这太棒了。首先，我们创办 OpenAI 时想做的元事情之一是，曾经有一段时间，特别是硅谷的科技行业在创建新研究实验室或进行行业新研究方面非常出色，然后它有点崩溃了，好一阵子没有好的实验室。我们希望展示的一部分——而且不只是我们，很多对新研究实验室感到兴奋的人——是工业界可以再次做研究。

I think it's fantastic. First of all, one of the meta things that we wanted to do when we started OpenAI is that there had been a period where the technology industry in Silicon Valley in particular was amazing at new research labs or just doing new research in industry in general, and then it kind of fell apart and there hadn't been a good one in a while. And part of what we hoped to show, and this was not only us, like a lot of people were excited about new research labs, is that industry could do research again.

嗯。

Mhm.

所以看到现在这在所有新实验室中变得流行，我觉得非常棒。有些会成功，有些会失败，有些会转向其他努力，但让工业界以创业风格支持研究，我认为这很美妙。

So seeing that now become fashionable in all of these new labs, I think it's totally awesome. Some will succeed, some will fail, some will go into some other effort, but having industry support research in startup style, I think it's wonderful.

未来两年，你预计会收购更多独立产品公司还是更多研究实验室？

Over the next two years, would you expect to acquire more individual product companies or more research labs?

好问题。我没有强烈的意见。我打赌最优秀的那些往往看起来是两者的混合。我现在想到的一个例子就非常像两者的混合。所以也许未来的形态就是：真正卓越的产品工作将包含越来越多的研究成分，它会变得更混合。

Good question. I don't have a strong opinion there. I would bet the very best ones will often look like a mixture of both. The one that I have in mind right now is something that very much looks like a mixture of both. So maybe that's the shape of things to come: the truly extraordinary product work will have more and more research component, and it'll be kind of a more hybrid thing.

数据是新的石油吗？有价值吗？几天前我们开玩笑说，他们有一堆数据，但他们不懂 AI。他们不知道如何变现。

Is data the new oil? Is there value? We were joking a couple days ago that they have a bunch of data, but they don't understand AI. They don't know how to monetize.

嗯，那个短语十年前实际上就被浪费了。

Well, and that phrase was effectively wasted a decade ago.

是的。

Yeah.

所以现在说它听起来很傻，但感觉现在可能比以往任何时候都更真实。

And so to say it now sounds really silly, but it feels like it could be more true now than ever.

是的。他们十年前确实浪费了它。我只是在想那些曾经这么说的人，然后他们从中走了出来。

Yeah. They really did waste it a decade ago. I was just thinking of the kind of people that used to say that and came out of it.

TED 演讲，你不应该点名批评他们。

TED talks, you're not supposed to call them out.

TED 的人很好。

They're nice people over at TED.

你知道，过去八年的神奇关系，不管你怎么称呼它，就是我们投入越来越多的资源、算力、数据、新想法等来创造一个产物，然后它的对数会变得更好。

You know, definitely the magic relationship of this last eight years, whatever you want to call it, has been that we can put in more and more resources, compute, data, new ideas, whatever into creating an artifact and it gets like the log of it gets better.

嗯。

Mhm.

所以你可以投入，这就是为什么我们有这种巨大的指数级资源增长，但我们不断得到越来越好的模型。尽管人们担心它会达到顶峰或放缓等等，但没有人说对过。我的意思是，有时看起来他们对了几个月，当我们消化一个新模型或进入新形态时，但过去六八年里这非常平稳。那些资源是什么。

So you can throw in, that's why we have this huge exponential increase in resources, but we keep getting better and better models. And for all of the concern people have about it's going to top out or it's slowing down or whatever, no one's been right about that. I mean, sometimes it looked like they were for a couple of months as we digested a new model or came to a new form factor, but it has been an incredibly smooth last six or eight years of this. What those resources are.

这中间会有一些权衡，你知道，有时候把钱花在更好的数据上更划算，有时候花在更多算力上，有时候是别的东西。总的来说，'算力是新石油'这句话对我来说最接近真相，但也会有其他方面。

There can be some trade-off between, you know, sometimes it's better to spend your money on better data, sometimes on more compute, sometimes something else. On the whole, compute power is the new oil is the statement that feels closest to true to me, but there will be other parts too.

有道理。软件死了吗？

Makes sense. Is software dead?

情况不同了。它绝对没有死。但软件——比如你如何创建它、如何使用它、每次需要时有多少代码会为你自动生成，以及你希望有多少一致的用户体验——这一切都将改变。过去几年，随着这些模型的推出，SaaS 股票出现了多次大规模抛售。我预计还会有更多。我也预计软件会出现大繁荣。我认为它只会在一段时间内波动。

It's different. It's definitely not dead. But what software like how you create it, how you're going to use it, how much you're going to have written for you each time you need it versus how much you'll want sort of a consistent UX. That's all going to change. There have been a number of these like big sell-offs of SaaS stocks over the last few years as these models have rolled out. I expect there will continue to be more. I expect there will be big booms in software. I think it's just going to be volatile for a while.

嗯。随着人们弄清楚这到底是什么样子，过去几周有句话一直印在我脑海里，有人对我说：现在每家公司都是 API 公司，不管他们愿不愿意。

Mhm. As people figure out what this looks like, the statement that someone said to me that has stuck in my mind most these last couple of weeks is that every company is an API company now whether they want to be or not.

哦，是的。

Oh yeah.

因为智能体将能够编写 CS。

Because agents are just going to be able to write a CS.

是的。昨天我们请来了 Uber 的 D，他的态度相当令人耳目一新。我们问起将智能体与 Uber 集成的问题，他承认，如果你能在 ChatGPT 里叫 Uber，广告业务确实可能受到威胁，但他基本上说，你必须考虑消费者。消费者想通过他们偏好的智能体叫 Uber，你就应该让他们这么做，否则你会遇到其他问题。

Yeah. We had D from Uber on yesterday and he had a pretty refreshing kind of approach. We were asking about integrating agents with Uber and he recognized that yeah, the ad business could potentially be threatened if you can order an Uber in ChatGPT, but he basically said like you have to think of the consumer. The consumer wants to order an Uber via their preferred agent, you should let them, otherwise you're going to have other problems.

是的，这肯定是正确的看法。至少我是这么认为的。

Yeah, that is the right take for sure. And I don't or I think so at least.

是的。

Yeah.

而且你知道，我们以前也经历过这样的平台转变，我的意思是，没有 iPhone，Uber 就不会存在。直到 iPhone 出现，你才能在外出时合理地叫一辆 Uber 到你所在的地方。所以我认为会有全新的事情发生。其他东西你会以新的方式使用。但自从我开始使用 Codex，我对让智能体去为我做事、同时仍然使用其他服务、支付其他服务感到兴奋。我相信我们需要找出新的商业模式和收入分配方式，但这会发生。

And you know, we've been through platform shifts like this before where you I mean Uber wouldn't have existed without one. It wasn't until the iPhone where you could like have it make sense to order an Uber to ride where you were as you were out in the world. So I think there will be totally new things that happen. Other things you'll use in new ways. But definitely as I've started using Codex, my excitement about having agents go off and do things for me and still use other services, pay other services. I'm sure we'll need to figure out new business models and how revenue gets shared around, but that will happen.

是的。再多谈谈 Codex Desktop。

Yeah. Talk more about Codex Desktop.

再问一个关于 SaaS 的问题。有没有公开市场的 SaaS 公司试图与 OpenAI 达成软着陆？你认为这有价值吗？

One more question on SaaS. Have any public market SaaS companies tried to get a soft landing with OpenAI? And do you think there's any value?

据我所知，没有公开的 SaaS 公司尝试过与 OpenAI 这样做。你看，其中一些公司肯定会是持久的，现在正在打折，可能只需要新的能量和全新的方法，也许 OpenAI 可以提供。

Not that I'm no one has no public SaaS companies that I'm aware of have tried that with OpenAI. Look, some of them will certainly be durable and are on sale right now and potentially just need new energy and need kind of a new entirely new approach and maybe OpenAI could provide that.

是的。

Yeah.

是的。我认为有些会非常有价值。有些现在确实感觉像一层更薄的层。但我不确定，我最近和一群 SaaS 公司聊过，他们并不感到不兴奋，他们说我们要经历一场大转型，你知道，是的，其他人现在可以即时编写软件，但我们也可以，而且我们有一个很棒的记录系统，这似乎合理。当然，有些会失败。

Yeah. I think some will be incredibly valuable. Some do feel like a thinner layer now. But I don't know like I was talking recently to a bunch of SaaS companies and they do not feel unexcited like they're like we're going to go through a big transformation here and you know yeah sure other people can instantly write software now but so can we and we got a great system of record and seems reasonable. Some won't make it of course.

是的。是的。再多谈谈 Codex Desktop 的发布。感觉下载量很成功，但对那些可能有点技术背景但没时间设置 IDE 和配置环境来真正开始写软件的人来说，这是一个关键转变。我想知道整合到手机的计划。我认为对很多人来说，Claude bot、ChatGPT bot、开放的东西是一个重大时刻，就像哦，我可以发个短信，它就会去写代码，这很有价值，解锁了一种新的智能体式体验。你认为 Codex Desktop 生态系统会走向何方？

Yeah. Yeah. Talk more about the Codex Desktop rollout. It feels like a successful amount of downloads but a key shift for people who are maybe lightly technical but don't have time to set up an IDE and configure an environment to actually start writing software. I want to know about plans to integrate to the phone. That was a big moment I think for a lot of people with the Claude bot, the ChatGPT bot, the open thing was like oh I can text something and it will go and write code and that's valuable and that unlocks a new agentic experience. Where do you see the Codex Desktop ecosystem going?

Codex Desktop 让我有点意外，人们如此喜欢它，包括我自己也非常喜欢。你知道，我认为这是一个很好的例子，说明对使用这些模型的体验进行 10%的打磨，尤其是在能力过剩的情况下，对你能够构建什么以及如何与这些东西互动有极大的帮助。当然，我们应该有能力从手机端启动新任务，我们会做到的。我的意思是，你真正想要的是你的单一 AI，它在统一的后端为你工作，访问你所有的数据、想法、东西、记忆和所有上下文，并且能够在多个界面上工作，而你通常会在桌面端。你通常会在手机上，只想添加一些东西。但这对我自己的工作流程来说是一个相当深刻的转变，不仅仅是编码任务，还有更通用的任务。如果你至少不是相当技术性的话，它仍然有点难用。但显然，我们会找到这个产品的一个版本，它可以做其他知识工作任务，控制你的电脑等等，你不需要具备技术背景，它将把构建东西的魔力带给很多人，因为即使你从不看代码，你也能够构建相当复杂的东西。我在玩新的 Codex 应用时构建的东西之一，就是我一直想要的这个神奇的自动完成待办事项列表。

I am so Codex Desktop has been a somewhat of a surprise to me in terms of how much people love it including how much I love it myself. You know, I think it's a great example of 10% of polish of the experience of using these models, especially when there's so much capability overhang goes an extremely long way to what you can build and how you interact with this stuff. Of course, we should have an ability to kick off new tasks from mobile and we'll do that. I mean really what you want is like your single AI that's working for you on a unified back end, access to all of your data and your ideas and your stuff and your memory and your all the context and the ability to work across a lot of surfaces and often you'll be at your desktop. Often you'll be on your phone and you just want to add something in. But it is a pretty profound shift in my own workflow, not just for coding tasks, but more general purpose tasks. It's still kind of hard to use if you're not at least reasonably technical. But obviously, we'll find a version of this product that can do other knowledge work tasks and control your computer and things like that where you don't have to be and we'll and it'll bring the magic of building stuff really to a lot of people because even if you never look at code, you'll be able to build something reasonably sophisticated. One of the things that I have built when I was playing around with the new Codex app is this thing I had always wanted just like this magic autocompleting to-do list.

哦，是的。我真的使用待办事项列表，这个想法是，我只需把任务放进去，它就会尝试去完成它们。如果它能完成，它就完成。如果需要问题，它会问我问题。如果需要，你知道，如果我必须做某事，我仍然可以用老方法做。但这样的界面，你知道，所有你想做的事情，你只需向电脑或你的 AI 解释，它就会尝试去完成它们。当然，如果你在手机上，你只需在手机上添加一个任务，或者，你知道，如果你想轻松地从邮件中导入一些东西，你也会这样做。感觉真的很好。所以，我对所有它将成为一个通用知识工作智能体的方式感到兴奋。

Oh yeah. I really work with to-do lists and this idea that I could just put tasks in and it would try to go do them. If it could complete them, it could complete them. If it needed questions, it would ask me questions. If it needed, you know, if I had to do something, I could still do it the old-fashioned way. But an interface like that where, you know, all the stuff you want to do, you just sort of explain to a computer or your AI and it tries to go off and do them. And sure, if you're on your phone, you're going to just add a task on your phone or, you know, if you want to easily import something from email, you're going to do that. It felt really good. So, I'm excited about all of the ways that this will just become a general knowledge work agent.

是的。

Yeah.

看到像 OpenClaw 这样的产品来自开源，你感到意外吗？因为那种用户体验，我想你早就知道会出现，但 OpenClaw 的魔力在于，大型科技公司很难做到。Peter 没怎么给超大规模云服务商打电话说‘嘿，我要集成你的 API’，它就这么成了。回想 Sky 收购案时，这种体验可能正是你心中所想，也是你们内部在努力的方向。

Were you unsurprised to see a product like OpenClaw come from open source? Because that kind of user experience is something I imagine you knew would be a thing, and yet part of the magic of OpenClaw is that it would be very difficult for a large tech company to make. Peter didn't make many phone calls to hyperscalers to say, 'Hey, I'm going to be integrating your API.' It just went. And when I think back on the Sky acquisition, this kind of experience was probably top of mind and things you were working towards internally.

我喜欢 OpenClaw 的一切精神。

I love the spirit of everything about OpenClaw.

是啊。

Yeah.

你说得完全对，一个单人开源项目做这种事，比一个害怕诉讼、数据隐私等问题的公司要容易得多。我认为这就是创新的方式。这样的东西开始出现，显然很了不起。会有办法做出那个产品的大众市场版本，但让创造者去创造，那种家酿计算机俱乐部的精神，非常重要。

And you are totally right that it's much easier to imagine a one-person open source project doing something like that than a company that is going to be afraid of lawsuits, data privacy, and everything else. I think this is kind of how innovation works. Something like that starts. It's clearly amazing. There will be a way to make a mass market version of that product, but letting the builders build, the equivalent of the Homebrew Computer Club spirit, is so important.

完全同意。

Yeah, totally.

我们能聊聊社交吗？我觉得如果我在谷歌上搜 Sam Altman 社交，出来的全是 AI 和 Sora，还有关于人形社交网络的需求或预测。你大体上认为社交会走向何方？你希望如何将 AI 融入其中并赋能未来？Moldbook 那件事是一个很有趣的社交实验，我认为它指向了智能体在某种社交空间中互动，希望至少在某种程度上代表人类，这可能相当有趣。我觉得我们还没想好怎么做，但感觉社交将发生很大变化。我对这样的社交体验很感兴趣：当你的智能体和我的智能体对话并产生新东西时。显然，在现有社交平台上放大量 AI 机器人只会让所有人抓狂，并不好玩。所以那不是正确答案。

Can we switch to social? I feel like if I Google Sam Altman social, I get pure AI and Sora, and also demand or predictions about a humanoid social network. Where do you see social going broadly? How do you want to integrate it with AI and power it in the future? The Moldbook thing was a very interesting social experiment to watch, and I think it points to agents interacting in some sort of social space, hopefully on behalf of people at least to some degree, which could be quite interesting. I don't think we know what to do there yet, but it feels like social is going to change a lot. And I am interested in the space of what a social experience can look like when your agent is talking to my agent and coming up with new stuff. Clearly, putting a lot of AI bots on existing social platforms is just making everyone crazy and not that fun. So that's not the right answer.

但我认为我们可以为这项技术的能力设计一些新的东西，让它感觉良好且有用。

But I think we can design something new for what this technology is capable of that will feel good and useful.

是啊。有没有解决机器人问题的办法，就是所有实验室都与其他平台整合？即使你无法检测出是 AI 生成的，你也可以直接说，‘我们刚刚生成了那些 token，那些确切的 token 就在我们的数据库里。’

Yeah. Is there a solution to the bot problem that's just all the labs sort of integrating with all the other platforms? And even if you can't detect it's AI generated, you can literally say, 'We just generated those tokens, those exact tokens are in our database.'

做不到，因为开源模型现在已经足够好了，能写出来。

Can't do that because open source models are good enough to write at this point.

是啊。我对于某种‘人性声明’感到兴奋，而不是试图检测 AI。

Yeah. I am excited about sort of assertion of humanity instead of trying to detect AI as a thing here.

我不知道社交平台是否有兴趣解决这个问题。因为至少在短期内，它带来了大量参与度和使用量的增长。所以我相信如果他们想解决，是可以做到的。但我不确定这符合他们的利益。实际上我甚至不——我不喜欢这样，但有些人似乎确实乐在其中。

I don't know if the social platforms have an interest in solving this. It is because it's creating, at least in the short term, a lot of engagement and increased usage. So I believe they could solve it if they wanted to. I'm not sure it's in their interest. I'm actually not even—I don't like it, but some people do seem to enjoy it.

是啊。你能多谈谈 Sora 作为视频生成模型的未来方向吗？感觉工具使用可能被低估了。加入推理，不仅仅是扩散模型。它赋予这些模型进行线性剪辑和叠加动态图形的能力。当我刷 Instagram Reels 时，我看到氛围短片有剪影、负片翻转，还有调色。这些你或许都能用扩散生成，但教模型使用 After Effects 或任何动态图形套件也很酷。这是一个有趣的解锁吗？你怎么看？

Yeah. Can you talk a little bit more about where Sora as a video generation model is going? It feels like tool use is maybe underdiscussed. Adding reasoning, it's not just the diffusion model. It's giving these models the ability to make linear cuts and overlay motion graphics. And when I scroll through Instagram reels, I see vibe reels with cutouts, it flips negative, and it's all color graded. It's stuff that you could probably diffuse it all, but it's pretty cool just to teach a model to also use After Effects or whatever video motion graphics suite you want to use. Is that an interesting unlock? What do you see going on?

所以所有这些都会发生，我同意你的看法。模型会变得非常擅长做这些。人们喜欢生成视频。但我要说，我们还没有找到让人们真正喜欢看别人视频的方式。这对很多其他 AI 来说也是如此——人们喜欢和 ChatGPT 之类的东西聊天，但对大多数人来说，看别人生成的东西并没有那么吸引人。所以我认为这里面有点问题。

So all of that stuff will happen and I agree with you. The models will get really great at doing that. People love generating videos. I would say people, we have not yet found a way that people really love watching other people's videos. This is true for a lot of other AI—like they love talking to ChatGPT or whatever, but it's not that compelling for most people to read other people's generations. So I think there is something.

但这对所有写作和所有视频不都一样吗？

But isn't that for all writing and all video?

是啊，在我看来这种情况比一般情况更明显，但也许你是对的。

Yeah, it seems stronger to me in this case than the general case, but maybe you're right.

是啊，但如果有人说，‘嘿，我生成了一个 15 分钟的视频，我很兴奋想让你看’，你看了前 10 秒并不觉得吸引人，那我才不管它是不是人做的。

Yeah, but if somebody says, 'Hey, I generated a 15-minute video, I'm really excited for you to watch it,' and you watch the first 10 seconds and you're not that captured by it, I don't care that it was human.

是啊，也许你是对的，这并非特例。

Yeah, maybe you're right and this is not a special case.

你在 Sora 下载数据中看到这一点了吗？因为我注意到，我会在 Sora 上生成东西，下载下来分享到群聊，然后这成了我和其他五个人之间的小玩笑。我们看到家庭群聊里说‘哦，这是我们的狗和我们的孩子’，但并没有那种‘好吧，这是个生意，大家都喜欢这个’的感觉。

Do you see that in the data with Sora downloads? Because I've noticed that I'll generate stuff on Sora, download it, and share it to a group chat, and then it's this little in-joke that me and five other people get. And we see this like family group chats of 'Oh, it's our dog and our kids,' but there's not really like, 'Okay, yeah, this is a business, everyone likes this.'

绝对是的。我会说最常见的用例就是那样。你知道，群聊里的表情包算是一种杀手级用例。

Absolutely. I would say that the most common use case is something like that. You know, memes on group chats is a real killer use case of sorts.

迪士尼的推广进展如何？我对此非常兴奋。Jordy 从商业策略角度非常看好它。

How is the Disney rollout going? I was super excited about it. Jordy was extremely bullish on it from a business strategy perspective.

是啊。当你回顾历史上图像模型如何推动各种大语言模型发展时，现在你将拥有一个图像和视频模型，能做其他大语言模型做不到的事，至少在法律上如此。

Yeah. When you look at how image models have grown various LLMs historically, and now you're going to have an image and video model that can do something that no other LLM can do, at least legally.

是啊。Bob Iger 要加入 OpenAI 吗？Bob，我喜欢这个。不邀功。

Yeah. And is Bob Iger joining OpenAI? Bob, I love it. No credit for that.

他马上要找工作了，是个自由人。签下他。

He's going to be looking for a job and a free agent here. He's a free agent. Pick him up.

替我们联系他。你知道，做点招聘。那太好了。

Hit him up for us. You know, do some recruiting. That'd be great.

我认为在图像和视频中生成角色对人们来说会非常重要，他们真的很喜欢。否则，就像我们说的，我不认为很多人想看我和某个星球大战角色一起做点什么，但我可能会觉得这很酷。

I think that generating characters in images and videos is going to be very important to people and they really like that. Otherwise, as we were saying, I don't think many people want to watch me and some Star Wars character doing something together, but I might think it's cool.

现在 ChatGPT 上有一个真正的趋势：‘根据你对我的所有了解，给我和我的工作画一幅漫画。’那种东西人们确实喜欢看别人的。

There's a real trend going on right now with ChatGPT where it's 'Make a caricature of me and my job based off everything you know about me.' And those kinds of things people actually do like looking at other people's media.

是啊。

Yeah.

这几乎就像个面部滤镜之类的。里面仍然有足够多的人性，所以你的和我的不一样。它仍然是个性化的。

It's almost like a face filter or something. There's enough of the human still in it that it's not—yours is not the same as mine. So it's still personalized.

它很个性化，能体现你的特点。之前图像生成领域很多爆款，我觉得就是能让人看起来比现实中更吸引人或更酷。

It's personalized and it says something about you. And a lot of these things that have gone viral before with image gen, I think it's like if you can make people look a little bit more attractive or cool than they are in real life.

对。

Yeah.

不用特意要求。

Without having to ask for that.

对。

Yeah.

你如何看待实际推广策略？我们之前讨论过，是全面放开，允许生成任何迪士尼角色，还是像蜘蛛侠周那样，大家发蜘蛛侠，然后米老鼠周再来一波病毒式传播。

How are you thinking about the actual rollout? We were debating between like open the floodgates, you can generate any Disney property versus like it's Spider-Man week and everyone's posting Spider-Man and then it's Mickey Mouse week and there's another viral moment.

我不确定团队的具体计划。我知道迪士尼对合作方式有一些不同意见，我们努力做好合作伙伴，但个人而言，我很期待全面放开。

I'm not sure what the team is planning there. I know Disney's had some different opinions about what they want to do and we try to be a good partner there, but I'd be excited to open the floodgates personally.

哦，那会很有趣。酷。聊聊你的第一个演讲视频，第一个超级碗广告。感觉不像生成的，有很多动态图形，黑点汇聚在一起。那个广告的目标是什么？你想对谁说话？它不像直接回应的二维码，下载应用。使命是什么？

Oh, that'd be fun. Cool. Talk about your first speaking video, your first Super Bowl ad. It felt like not generated, lots of motion graphics, the black dots coming together. What was the goal with that ad? Who were you trying to speak to? It didn't feel like a direct response QR code, download the app. What was the mission?

我喜欢那个广告。我觉得它很酷。它显然不是面向大众市场或直接回应的广告。

I love that ad. I think it was such a cool one. It was clearly not meant to be a mass market or direct response ad.

对。

Yeah.

但它是在对这场革命的核心人物说话，试图庆祝过去的一切和未来的一切。

But it was speaking to the people who are at the center of this revolution and just trying to celebrate everything that has come before and everything that will come after.

对。

Yeah.

普通 ChatGPT 用户对此反响不大，但领域内的研究者反响强烈，产生了共鸣。它绝对不是生成的，是用传统方式制作的。很多人喜欢，很多人讨厌，中间很多人不理解，我觉得这没问题。

We didn't hear a lot about it from average users of ChatGPT, but we heard a lot from researchers in the field and there was a lot of resonance there. It was definitely not generated. It was done the old fashioned way. And a lot of people loved it, a lot of people hated it, and many in the middle didn't get it, and I felt okay about that.

总结得很好。

It's a great encapsulate.

我喜欢我们周日的广告。

I like our ad for Sunday.

好的。是关于 Codex，不意外。但谈谈广告的演变，如何更清晰地传达实际用例和价值。你现在对视频的广告策略想表达什么？

Okay. It's about Codex, no surprise. But talk about the evolution of the advertising to be more clear about the actual use case, the value. What are you trying to say with your advertising strategy now as it relates to video?

我最希望我们传达的，也是鉴于模型现状的新挑战，是教会人们能用 AI 做什么。我认为 AI 现在在大多数领域已经非常强大，但人们还只是像在 ChatGPT 上问基本问题。现在每个人都能构建令人惊叹的东西，每个人都能做各种工作。科学家将做出新发现。如果我们的广告能在一定程度上教会人们如何使用它，那就太棒了。

The thing I would most like us to say, and I think this is a new challenge given where the models are, is to teach people what they can go do with AI. I think AI is now unbelievably capable in most of the world. It's still like asking basic questions on ChatGPT. Everyone can go build amazing things now. Everyone can go do all kinds of work. Scientists are going to make new discoveries. And I'd like to, to the degree that advertising we do can teach people how to use this, I think that'd be awesome.

所以关键绩效指标是广泛减少能力过剩。

So the KPI is like reduce the capability overhang broadly.

我认为这应该是我们的总体关键绩效指标，不仅仅是广告。

I think that should be a general KPI for us, not just of our ads.

我们构建的产品，以及如何教人们使用这些产品，这感觉非常重要。

The products that we build, how we teach people to use those products, that feels very important.

对。

Yeah.

Anthropic 也在超级碗投放了大量广告。你认为他们对聊天应用广告推广方式的描述哪里错了？

Anthropic also had a bunch of ads in the Super Bowl. Seems like they ran a ton. What do you think they got wrong about their characterization of how ads will roll out in chat apps?

嗯，就是错了。我认为主要问题是，我们不傻。我们尊重用户。我们明白，如果我们做了那些广告所描绘的事情，用户会理所当然地停止使用产品。没有人，我们的首要原则是不把广告塞进大语言模型的输出流中。那会显得疯狂反乌托邦，像糟糕的科幻电影。所以这些广告的主要问题是用欺骗性广告来批评欺骗性广告。我不知道，我对此感到不舒服。我问 Claude 什么是玩脏的，它说：‘误导他人关于你的意图，隐藏信息，或制造虚假印象。’

Well, it's just wrong. The main thing I think is we are not stupid. We respect our users. We understand that if we did something like what those ads depict, people would rightfully stop using the product. No one, our first principle with ads is that we're not going to put stuff into the LLM stream. That would feel crazy dystopic, like a bad sci-fi movie. So the main thing that's wrong with the ads is using a deceptive ad to criticize deceptive ads. I don't know, something doesn't sit right with me about that. I asked Claude what the definition of playing dirty is, and it said, 'Misleading others about your intentions, hiding information, or creating false impressions.'

对。

Yeah.

我觉得有点脏。我觉得它玩得漂亮，确实漂亮。而且是个有趣的广告。关于 ChatGPT 个性中最让我恼火的部分，我们很快会修复，我觉得他们在广告中抓住了这一点。所以那部分很有趣。

I thought it was a little dirty. I thought it was well played, but it was well played for sure. And it was a funny ad. And the stuff about the ChatGPT personality that most annoys me, which we'll fix very soon, I thought they nailed in the ad. So that part was funny.

对。

Yeah.

但我不确定，我也觉得他们不做广告挺好的。我们的业务形态不同。我注意到他们在声明中说‘我们可能稍后修订这一决定，并会解释原因。’所以，那篇博客文章确实很好地安抚了支持广告的人，并给自己留了后路。

But I don't know, I also think it's great for them not to do ads. We have a different shaped business. I did notice that they said in their thing like 'we may later revise this decision and we'll explain why.' So, yeah, the blog post kind of did a good job of disarming the pro-ad people and gave themselves an out in the future.

你觉得他们在意吗？

Do you think they care?

我在意。我觉得这只是个插曲。人们热衷于公司间的争斗，但这些模型的惊人能力、产品、围绕 Codex 的兴奋浪潮，对我来说重要得多。

I do matter. Like I think it's a sideshow. People are excited for a food fight between companies, but the amazing capabilities of these models, the product, the groundswell of excitement around Codex, that feels way more important to me.

如何解决语音模式中的停顿问题？需要新硬件还是模型能力问题？

How do you stop the pausing that happens in voice mode? Do you need new hardware for that or is it a model capability thing?

我们需要一个新模型。可能也需要一些新硬件，但主要是新模型。我认为今年年底前我们会有一个很棒的语音模式。

We need a new model. We may need some new hardware too, but mostly we just need a new model. I think we'll have a great voice mode by the end of this year.

更大的瓶颈是什么？能源还是芯片？

What's the bigger bottleneck? Energy or chips?

两者交替。现在又是芯片。

It goes back and forth. Right now again it's chips.

芯片。社会或美国是否应该更积极地增加晶圆厂供应？

Chips. Is there anything society or America should be doing more aggressively to increase the supply of fabs?

我认为它可能会自行解决。正常的资本主义可能解决。但我认为，如果社会决定增加全球晶圆产能并为此提供资金，获得所需的整个供应链和人才，那将是一件非常好的事情。

I think it may get solved on its own. Normal capitalism may solve it. But I think somehow deciding as a society that we are going to increase the wafer capacity of the world and fund that, and get the whole supply chain and talented people we need to make that happen, would be a very good thing to do.

你认为模型智商有上限吗？目前的竞赛是，你很聪明，但聪明不了几天，只能聪明几小时。能走得更远，变得更聪明吗？

Do you think there's an upper bound on model IQ? The race right now is you're smart, but you're not smart for days. You're smart for hours. Can you go much farther and get much smarter?

似乎必然。上限，我不知道。我还不知道如何思考这个问题。

It seems certain. Upper bound, I don't know. I don't know how to think about that question yet.

我甚至无法推理 2000 智商是什么样子。我不知道那意味着什么。

I can't even reason about what 2,000 IQ looks like. I don't know what that means.

你这么说很有趣。我的意思是，我无法推理思考一个问题一万人类年意味着什么。

It's funny you say. I mean, I can't reason about what it means to think about a problem for like 10,000 human years.

这又是一个好问题。是啊，太疯狂了。但也许智商（IQ）会感觉更奇怪。我不知道。我总觉得这不会像听起来那么奇怪。原因有很多，我们太关注别人了，太关注自己的生活了。我们太以人类为中心了。所以，好吧，这个东西真的很聪明。它为我们发明新科学，运营公司，做所有这些事情。这听起来应该会非常奇怪，我想它确实会非常奇怪。

That's another good one. Yeah, that's crazy. But maybe IQ is going to feel even weirder. I don't know. I feel like somehow this isn't going to feel as strange as it sounds. For a bunch of reasons, we're so focused on other people, our own lives. We have such a human-centric nature. So okay, this thing is really smart. It's inventing new science for us, running companies, doing all this stuff. That sounds like it should be impossibly weird, and I think it'll just be very weird.

你认为太空数据中心会在未来两三年、五年内为 OpenAI 提供有意义的算力吗？

Do you think space data centers will provide a meaningful amount of compute for OpenAI in the next two to three years, five years?

不会。

No.

十年呢？

10 years?

你继续问吧。一万年。祝你好运。

You just keep going. 10,000 years. I wish you luck.

关于广告的整个争论，有趣的一点是，在我们这个领域，批评是你没有尽早推出广告。有没有一个世界你希望自己更早推出？实际推广进展如何？广告商满意吗？你有一个很长的路线图，还是你认为你会更快地赶上广告的前沿？

The funny thing about the whole back and forth about ads is that in our world, the criticism is that you didn't launch ads early enough. Is there a world where you wish you launched earlier? How is the actual rollout going? Are advertisers happy? Do you have a really long roadmap, or do you think you'll be faster at catching up to what's frontier in ads?

我们甚至还没有开始测试。我们很快就要开始测试了。但我们需要几次迭代才能找到合适的广告单元和正确的方式。我是否希望我们更早开始？我们从三年前三个月前还不是一家公司，只是一个研究实验室，到现在是一家拥有很多产品的大公司。所以我希望很多事情能做得更快。我认为我们在权衡需要做的事情上是正确的。我希望我们今天早上推出了这个很酷的企业平台。也希望更早做，但必须应对 ChatGPT 和 Codex 等产品的巨大增长。这是好问题。

We haven't even started to test yet. We start to test soon. But it's going to take us some number of iterations to figure out the right ad unit, the right way this all works. Do I wish we had started earlier? We have gone from not a company, three years and three months ago we were a research lab, and now we are a pretty big company with a lot of products. So there are many things I wish we had done faster. I think we were correct on the trade-off here of how we balance things that we need to do. I wish we had launched this very cool enterprise platform this morning. Wish we had done that earlier too, but had to deal with the monstrous growth of ChatGPT and Codex and all that other stuff. Good problems to have.

最后一个问题。那个你用来写文章的内部写作模型后来怎么样了？感觉那真的很酷，但我们从未真正看到它面世。

Last question for me. What happened to that internal writing model that you used to write the essay? That feels like something that was really cool, but we never really saw the light of day.

我们会把那种精神大量融入未来的模型中。再说一次，有太多事情在发生。我们必须做出这些艰难的优先级决策。我很想要一个酷的写作模型，但不如我想要一个酷的编码模型那么强烈。而现在编码和科学领域是可能的。那是我最兴奋的事情。加速各种研究，包括 AI 和其他领域，真正加速经济。我认为那是我们在新能力方面最应该优先考虑的事情。但你也想要一个能写出优美文字的模型，因为那意味着它能非常清晰地思考并表达出来。这在日常工作中很有用。

We're going to get a lot of that spirit into a future model. Again, it's like there's so much stuff happening. We have to make these hard prioritization decisions. I would love a cool writing model, not as much as I would love a cool coding model. And it's what is possible now for coding, for science. That's the thing I'm most excited about. For accelerating all kinds of research, AI and otherwise, for really accelerating the economy. I think that's the right thing for us to most prioritize in terms of new capabilities. But you want a model that can write beautifully because it means it can think very clearly and express that very clearly. That's just useful in normal work.

最后一个问题。与 OpenAI 整个领导团队的对话如何？你们处于这样一种位置：在任何情况下说的任何一个词或句子都可能立即被曲解成头条，然后你们不得不进行危机公关，纠正说法。但当然，原始信息往往比更正更广泛地传播。这似乎是一个有趣的挑战。

Last question for me. How have conversations been with the broad OpenAI leadership team? You guys are in a position where any single word or sentence you say in any situation can be spun into a headline immediately, and then you have to go on damage control correcting the narrative. But of course the original message is often seen more broadly than the correction. It seems like an interesting challenge.

这是一种奇怪的生活方式。我不知道有哪家私营公司曾如此备受新闻关注和显微镜审视。在某种程度上，这令人沮丧。我们完全处于每个人的焦虑和每个试图打倒我们的竞争对手的视线中，每个人都在想 AI 会对他们的业务或生活产生什么影响。有很多等离子体在寻找不稳定性来坍缩。但从另一个意义上说，主观体验是我们忙于太多令人兴奋的事情，以至于常常感觉周围有一场疯狂的飓风，而我们坐下来时却相当平静。媒体或 Twitter 某天对某件事发疯，他们谈论着疯狂的崩溃。我们觉得那太疯狂了。人们整天谈论它，后来发现是错的，看起来浪费了很多精力。但我们只是觉得我们有一个很棒的新模型即将推出。人们在建造不可思议的东西。公司在转型。我们试图弄清楚如何获得更多算力并应对算力紧张。我们只是继续前进，很忙。然后如果我们打开 Twitter，抬起头看新闻，就会觉得哇，那是一件完全脱离现实或 99% 脱离现实的疯狂事情。有人会纠正它，但然后我们回去工作，人们又发疯了。当我们向外看时，这很奇怪，但内部比你在媒体报告中想象的要平静得多。

It is a strange way to live. I don't know of any private company that has ever been so in the news and under a microscope. At some level it's frustrating. We're squarely in the sights of everybody's anxieties and every competitor trying to take us down, and everybody wondering what will happen with AI to their part of the business or their own lives. There's a lot of plasma looking for an instability to collapse on. In some other sense though, the subjective experience is we are so busy on so much exciting stuff that it often feels like there's this crazy hurricane turning around us, and when we sit it's fairly calm. The media or Twitter goes insane about something one day, they're talking about a crazy meltdown. We're like that is insane. People talk about it all day and then later find out it's wrong, and it seemed like a lot of wasted energy. But we're just like we have this great new model coming. People are building incredible stuff. Companies are transforming. We're trying to figure out how to get more compute and deal with this compute crunch. And we just keep going and we're busy. Then if we open Twitter, pop up our heads and look at the news, it's like wow, that is an insane crazy thing happening completely divorced from reality or 99% divorced from reality. Someone will correct it, but then we get back to work and people flip out again. It is weird to watch when we look outside, but it is less chaotic internally than you would imagine from reading the media reports.

是的，有道理。非常感谢你抽出时间来和我们聊天。恭喜……我很期待看到 Codex 广告。我也是。请试试它。这个应用和 53 我觉得是我们一段时间以来做的最酷的事情。是的。用一个提示，我把 tbpn.com 的主页重建得看起来像伯克希尔·哈撒韦，而且立竿见影。非常有趣。有趣的选择。纯文本。非常简单。一次性搞定。没有真正挑战它的极限，但我玩得很开心。非常感谢你上节目。我们很快再聊。很高兴叙旧。再见。干杯。

Yep, that makes sense. Well, thank you so much for taking the time to come chat with us. Congrats on the... I'm excited to see the Codex ad. Me, too. Please try it. The app and 53 have been like, I think the coolest thing we've done in a while. Yes. With one prompt, I rebuilt the tbpn.com homepage to look exactly like Berkshire Hathaway and it was just immediate. It was very fun. Interesting choice. Plain text. It was very easy. Immediately one-shot it. Did not really push it to its limits, but I'm having fun. So, thank you so much for coming on the show. We'll talk to you soon. Great to catch up. Goodbye. Cheers.

让我介绍一下 11 Labs。构建智能实时对话代理。用 11 Labs 重新想象人机交互。我还要介绍一下 Sentry。Sentry 向开发者展示出了什么问题，并帮助他们快速修复。这就是为什么 15 万个组织用它来保持应用运行。闲话少说，我们有来自 Anthropic 的 Sholto Douglas。他是技术团队成员，现在在电视上。Sholto，你好吗？

Let me tell you about 11 Labs. Build intelligent real-time conversational agents. Reimagine human technology interaction with 11 Labs. And I'm also going to tell you about Sentry. Sentry shows developers what's broken and helps them fix it fast. That's why 150,000 organizations use it to keep their apps working. And without further ado, we have Sholto Douglas from Anthropic. He's a member of technical staff and he's here in the TV. Sholto, how are you doing?

嘿，大家好，你们好吗？

Hey guys, how you doing?

我很好。看起来你更好。告诉我们新闻。你们发布了什么？发生了什么？

I am doing great. Seems like you're doing better. Tell us the news. What did you release? What happened?

好的。嗯，又一次，我们有了一个新模型。真的很棒。

All right. Well, once again, we've got a new model. It's really fantastic.

呃，摄像头有点卡，有点奇怪，但应该没问题。这个模型真的非常非常棒。我觉得人们一直在比较上一代 OpenAI 和 Anthropic 的模型，他们注意到了一些差异，对吧？OpenAI 的模型在解决难题时更努力一些，而 Anthropic 的模型则快得多，等等。所以，他们专注于速度，而我们则致力于让模型在真正困难的问题上表现更好。这就是这个模型的闪光点：它能够思考很长时间，能够消耗大量的测试时算力来深入思考问题。这也是我们一直在开发的‘同事’（coworker）方向的一大步。我记得上次提到过，我们看到模型在知识工作的其他方面正逐渐变得和编程一样强大。我对此感到非常兴奋。你可以看到它在计算机使用方面好多了。我们有 Claude for PowerPoint，据说在 Excel 上也变得相当不错了。虽然还不完美，但已经是巨大的进步了。

Uh the camera's a little bit laggy, a bit strange, but um you know, should be good. This model is really, really, really fantastic. So, I think people have been comparing the previous generation of OpenAI and Anthropic models, and they've noticed there's some differences, right? The OpenAI models were a bit better at trying really, really, really hard on tough problems, but the Anthropic models were much faster and so forth. And so, they worked on speed while we worked on making the models much much better at really, really tough problems. And so that's where the model really shines: it's able to think for a really really long time. It's able to expend a lot of test time compute on thinking hard about problems. It's also a massive step towards the coworker that we've been working on. I think I mentioned last time that we're seeing this continuous progression towards models that are as capable at the rest of knowledge work as they are at coding. And I think I'm really really excited about that as well. You can see it's a lot better at computer use. We've got Claude for PowerPoint, claiming to become actually quite capable at Excel. It's not perfect there yet, but huge steps, huge progress.

你们现在有像 PowerPoint 这样的内部基准测试吗？

Do you have internal benchmarks for things like PowerPoint now?

我们做的每件事都有基准测试。

We have benchmarks for everything that we work on.

但哪些会进入模型卡（model card）呢？

But what makes it to the model card?

哪些会进入模型卡？哦，模型卡通常只包含公开发布的基准测试。

What makes it to the model card? Oh, the model card is typically only publicly released benchmarks.

因为对于内部基准测试，我们通常保留用于内部测试。

Because for our internal benchmarks, we usually hold them out for internal testing.

哦，当然，当然。

Oh, sure, sure, sure.

给 PowerPoint 做基准测试的流程是怎样的？因为它不是……

What's the process for benchmarking a PowerPoint? Because it's not...

我不能过多谈论 PowerPoint 的基准测试，但呃，你知道，我不确定你是否知道几年前我在麦肯锡工作过，所以有时候也许他们只需要……

I can't talk too much about benchmarking a PowerPoint, but um you know, I'm not sure if you know back a couple of years ago I did work at McKinsey, so you know sometimes maybe they just need...

所以他们给你看，然后你说好或不好？

So they show them to you and you say it's good or...

没错，好或不好。

Exactly, good or bad.

是的，可能有很多人参与。跟我们聊聊 4.5 和 4.6 之间的区别意味着什么。人们可能熟悉预训练，以及改变模型的各种流行词。在过去的几个月里，Anthropic 在技术上做了什么来实现这一点？

Yeah, probably a lot of people involved. Yeah, talk to us about what a difference between 4.5 and 4.6 means. People might be familiar with pre-training, the different buzzwords that go into changing a model. What technically happened at Anthropic over the last few months to enable this?

是的。实际上，我们将模型发布的版本号与任何具体的技术细节解耦。它更多是关于我们兴奋的整体能力水平。实际上，很多人问我为什么我们不直接把 4.5 命名为 5，因为它是一个巨大的进步。我想也许我们低估了。我的意思是，上次我在这里的时候，我甚至没有完全意识到从 4 到 4.5 的飞跃有多大。这很好，因为当 4.5 的炒作周期到来时，我们可以重新发布一些你谈论它的片段，人们会惊讶地说‘哇，他们今天报道了’。其实不是，我们是在它发布当天就请了你们。所以我确信我们会对 4.6 做同样的事情。

Yes. So, we actually decorrelate the model release versioning from any specific technical details. It's more about the overall capability level that we're excited about. And actually, some like I've been asked by many people why we didn't just name 4.5 as 5 because it was such a step up. And I think maybe we under anticipated. I mean when I was last on here I don't think even I fully appreciated how much of a jump it was from 4 to 4.5. It was great because by the time there was the 4.5 hype cycle we could repost a bunch of clips of you talking about it and people were like wow they covered it today. It's like no we actually had them on the day it launched. And so I'm sure we'll be doing the same thing with 4.6.

谈谈 Task Horizon，那个米制评估（meter eval）。它重要吗？显然它是一个公开的基准测试。我们之前讨论过，但它和其他基准测试有不同的权衡吗？因为我觉得它就像 IQ 和能保持专注多长时间（比如锁定）之间的区别。这两者之间有矛盾吗？你们有没有遇到过这样的模型：它非常聪明，但无法长时间保持连贯性，或者反过来，它做更笨但更长的任务。

Talk about Task Horizon, the meter survey. Is that important? I obviously it's a public benchmark. We've talked about it before, but is that something that is a different tradeoff to the other? Because I see it as like there's the IQ and then there's how long something can stay focused like the locking in. And is there any tension between those? Do you ever get a model that's like, oh, it's really smart, but it can't stay coherent for that long or vice versa, it does dumber but longer tasks.

我的意思是，我认为米制评估可能是目前最好的评估，对吧？但你说得对，图表的不同部分衡量不同的事情。最初，它只是测试 IQ，但随着时间的推移，它开始测试人类失败的任务，不是因为不够聪明，而是因为坚持一项任务六、七、八个小时真的很难。你说得对，这两者是分开的。你几乎可以预期模型会有超人的毅力。但与此同时，也许上下文连贯性之类的东西会与之对抗。这与原始智力或特定领域的知识是不同的维度。

I mean, I think the meter eval is possibly the best eval currently out there, right? But you're right in saying that different parts of the chart measure different things. Initially, it was just testing IQ, but then as time goes on, it starts to test things where humans fail at tasks not because they aren't smart enough, but because actually persevering on a task for six, seven, eight hours is really hard. And you're right that these come apart. There's a sort of like models you should almost expect to have superhuman perseverance. But at the same time, maybe context coherence or something like this fights against that. And that's a different axis from raw intelligence or knowledge on given domains.

当我向人们提及 AI 进展时，我会用米制评估。最近我们举办了一个物理学会议，邀请了很多物理学家。我们和 Google DeepMind 一起做的。我们邀请了很多物理学家，试图说服他们 AI 真的很重要。而那张图表最能引起人们的共鸣，我认为它在概念上很容易理解，就像‘哦，好吧，那些需要我 x 小时的事情，模型现在也能做到了’。

Meter eval when I'm referencing AI progress to people. Recently we held this physics conference where we invited a whole bunch of physicists. We did it with Google DeepMind. We invited a whole bunch of physicists to try and convince them that AI was really a big deal. And that was the chart that resonated most deeply with people and I think is just quite conceptually graspable as 'oh okay, things that would take me x hours the models are now capable of'.

这是一个有趣的基准测试，考虑到很多类型的知识工作者实际上从来不会连续七小时专注于一项任务。就像‘哦，我工作了 20 分钟，然后去开会，然后又看了一下，然后吃午饭，然后散步，然后回来继续工作，然后有人打电话，我又得做别的事情。’所以很可能很多听众，也许除了某些软件工程师，在他们的职业生涯中从未连续六小时专注于一件事。

It's an interesting benchmark considering that so many types of knowledge workers will never actually work on a task for seven hours straight. It's like just 'oh I worked on it for 20 minutes and I jumped into a meeting and then I took a look at it again and then I had lunch and then I went for a walk and then I came back I worked on it then I got somebody called, I had to work on something else.' So it's very possible that a lot of people listening to this, maybe outside of some software engineers, have just never even in their career worked on something for six hours straight.

心流状态。

The flow state.

我的意思是，你把事情串联起来等等。但没错，我同意。从这个角度来看，这是一个有趣的基准测试。那么，一个为期一周的任务又意味着什么呢？当然，它本质上可以组合成一年长的工作进展。我的意思是，我在工作时肯定不会认为自己是在做一个为期一年的任务。我通常做的是以天或周为单位的时间跨度，这些会汇入几个月的时间跨度。

I mean, you chain things together and so forth. But yes, I agree. It's an interesting benchmark in that respect. And what does even a week-long task mean? Surely that's sort of inherently composable into a year-long work of progress. I mean I certainly don't think of myself as doing a year-long task when I'm at work. I'm doing these day or week-long time horizons usually which feed into a several month-long time horizon.

谈谈编排（orchestration）。感觉一个时刻正在到来。我非常喜欢那个‘煤气镇’（gas town）的类比和他构建的世界。你到处都能看到，我同时运行着四个 Claude Code 实例。感觉需要一个新的抽象层。你认为这如何解决？现在是开始学习这些东西的时候了吗？未来哪些比喻会有价值？

Talk about orchestration. It feels like there's a moment happening. I really like the gas town analogy and the world that he's built there. You see all over, I have four Claude Code instances running at the same time. It feels like there needs to be a new layer of abstraction. How do you think that gets solved? Is now the time to start learning those things? What metaphors will be valuable going forward?

是的。所以，目前智能体仍然处于需要你手动在它们之间进行多路复用的阶段。当我工作时，从 4 到 4.5 再到现在的 4.6，一个显著的变化是，我写的代码行数从大约 10% 降到了 0%。

Yes. So, right now the agents are still at this point where you need to manually multiplex across them. When I'm working, the stark thing about going from 4 to 4.5 and now 4.6 was that I went from maybe like 10% of my lines of code being written by me to 0% of my lines of code being written by me.

问题在于，我需要真的坐在那里，不断在这些窗口之间切换，确保它们都在正轨上，并给予指导和反馈。进展顺利时，我可能同时处理五个。但你仍然需要深入细节。我认为长期正确的思考方式是不断向上提升抽象层级。最终，你只想与一个智能体对话，它综合来自模型的反馈，比如‘我卡住了’或‘这个我不清楚’。那个模型可以扮演我之前扮演的角色，而我则可以退后一步，在更高的抽象层级上行动。我们最终应该不想像玩《帝国时代》或《星际争霸》那样操控模型吧？我们不应该受限于每分钟操作数。相反，模型应该在你需要时提供信息，这样你就能持续处理最重要的事情，并恰当地切换上下文。

The thing is I need to actually sit there and constantly switch between these windows to make sure they're on track and give them guidance and feedback. I'm multiplexing maybe five at once when it's going really well. But you still need to be in there in the details. I think the long-term right way to think about this is constantly moving up levels of abstraction. Ultimately you only want to be talking to one agent that synthesizes feedback from models as they come back and say 'I got stuck' or 'I was unclear on this.' That model can act as what I was doing before, and I can take a step back and act at an even higher level of abstraction. It doesn't feel like we ultimately want to be playing Age of Empires or StarCraft with the models. We shouldn't be APM bound. Instead, the models should surface information as you need it, so you can constantly act on whatever the most important thing is and context switch appropriately.

说到游戏，你有什么更新吗？游戏什么时候发布？跟我们讲讲这个游戏。是什么启发了你？说说背景故事，然后告诉我们进展如何。

Speaking of gaming, do you have an update for us? When is the game shipping? Tell us about the game. What inspired this? Give us the backstory and then tell us what's going on.

嗯，我想和很多人一样，我真的想在假期里测试模型的极限。我们看到很多人说‘我花了三四天做某事’。我认为在某种程度上，这正是最新一代模型炒作周期的起点，当人们有机会真正测试其极限时。

Well, I think like many other people, I really wanted to test the limits of the models over the holidays. We saw all that from people saying 'I spent three, four days doing X.' I think that's really where the hype cycle for the most recent generation of models started in some respect, when people got a chance to properly test the limits.

是的。

Yeah.

这很能说明采用率的问题，以及扩散需要时间，因为你需要有空间。

It tells you a lot about adoption and how diffusion takes time because you need to have the space.

我们需要更多长周末。

We need more long weekends.

说真的，那就是瓶颈。长周末就是瓶颈。

Seriously, that's the bottleneck. Long weekends are the bottleneck.

GDP。

GDP.

是啊，对 GDP 有好处，我猜。假期。

Yeah. Good for GDP, I guess. Holidays.

玩编码模型。

Play with coding models.

对对对。抱歉。总之。

Yeah. Yeah. Yeah. Sorry. Anyway.

嗯。

Yeah.

总之，Dylan 一直在楼下桌上带我们玩《帝国时代》的五人局。我想看看我能不能做一个‘Scaling 时代’，你建造太阳能板和数据中心，训练 AI 模型和无人机，而不是农场和采矿。从工业时代开始，一直发展到卡特·切夫。我只完成了大概 70-80%，所有机制都工作了，但结果发现做一款好玩的游戏真的很难。实际上，点击‘制造太阳能板’100 次并不好玩。意外吧。但对有些人来说好玩。有些疯狂的点击游戏就是这样，纯粹是脑残游戏，还挺成功。但我觉得你应该远离那种。甚至不像《异星工厂》。这些是手机游戏，非常非常烂。总之，我觉得那不符合你的使命。

Anyway, so Dylan's been getting us all into Age of Empires with these five matches on the downstairs table. And I wanted to see if I could make an 'Age of Scaling' where you build solar panels and data centers, you train AI models and drones and so forth instead of farms and mining. You start in the industrial era, you go all the way up to Carter Chev. I only got like 70-80% done and all the mechanics work, but it turns out it's actually really hard to make a game that's fun. Actually just clicking 'make solar panel' 100 times is not fun. Surprise, surprise. And it's fun for some people. There are those crazy clicker games where you do that and it's just a brain rot game and they're pretty successful. But I think you should stay away from that. Not even Factorio. These are like mobile games that are really, really bad. Anyway, I don't think that aligns with your mission.

这里面有艺术。

There's an art to it.

是的。

Yeah.

但我们谈论的是我们即将进入的这个时代的宏大地缘政治和战略，地球上最有价值的资源，嗯，基本上经济在某种程度上正在围绕算力重新定向。所以我想捕捉一些那种动态。旧金山的深夜讨论，浓缩在一个游戏里。

But we talk about the grand geopolitics and strategy of this era we're going into, where the most valuable resources on Earth are, well, basically the economy is in some respects reorienting around compute. And so I wanted to capture some of that dynamic. The late night discussions of San Francisco in a game.

是的，我喜欢。你如何看待世界模型和生成式世界方面的工作？我知道你不生成图像，但 Claude 可以接收图像并处理，这很有价值。你发一张截图，让它实现。但你怎么看世界模型？它会在你前进的道路上扮演重要一环吗？

Yeah, I love it. How are you processing the work that's being done in world modeling and generative worlds? I know that you're not generating images, but Claude can take in an image and process it, and that has value. You text a screenshot and have them implement it. But what do you think about world models? Is that going to play an important piece, a link in the chain of where you're going?

是的。所以，我认为这里有两件不同的事情。一条是通往 AGI 的直接路径，我认为那主要是编码，然后是 AI 研究和一般科学等等。我不认为那需要世界模型。但我确实认为世界模型在其他许多事情上非常令人兴奋。从游戏角度来看，它们非常令人兴奋。我们都看到了那些演示，精灵模型令人难以置信，真是令人震惊。而且我认为它们也可能是正确训练机器人的关键。我感觉曾经有一个时代，人们认为‘哦，我们只需要做大量的人类操作机器人的行为克隆。’感觉缩放定律是有效的，但我认为它基本上不是通往 AGI 的关键路径。

Yes. So, I think there are two different things here. There's the direct path to AGI, which I think is very much coding, and then AI research and general science and so forth. I don't think that requires world models. But I do think that world models are incredibly exciting for a number of other things. I think they're very exciting from a gaming perspective. We all saw those demos are incredible with the genie. It's just truly mind-blowing. And I also think they're probably the unlock to training robotics properly. It feels to me like there was this era of 'oh, we're just going to need to do incredible amounts of behavioral cloning of people operating robots.' And it feels like the scaling laws work, but I don't think it's on the critical path to AGI basically.

你能解释一下‘纯软件奇点’这个概念吗？

Can you unpack the concept of a software-only singularity?

是的。在这个世界里，模型在数字任务上远比在物理任务上出色。因此，我们看到数字世界发生快速变化，而物理世界变化相对较小。信息和软件发生了巨大变化。这最终会产生一些非常奇怪的影响。这意味着过去几十年经济进步的驱动力可能会逆转，发生非常迅速的变化。我认为这会流入物理世界，但会有延迟。所以你在芯片设计上变得更擅长。你在训练 AI 模型上变得更擅长。AI 模型变得更快。芯片变得更好。整体经济变得更高效，因为构成经济其余大部分的信息和消息传递最终变得高效得多。但与此同时，你还没有机器人提供无限的物理富足。科学可能进步得非常快，直到你需要与实验室或更大的对撞机等交互，然后你会说‘好吧，我需要建造机器人。’但与此同时，自动化实验室感觉比在现实世界中操纵地球的无限机器人更近。我认为它们可能实际上在相似的时间到来。我认为实验室需要相当能干的机器人，或者至少还没有人成功弄清楚如何自动化实验室，因为结果发现所有那些奇怪的小任务都需要大量的人类手动灵巧性，这些目前是生物协议的一部分。许多生物学家会说‘不，你实际上需要具备人类水平灵巧性的东西。’也许不是所有实验，但足够多到令人烦恼。

Yes. In this world, it's one where the models are far better at digital tasks than they are at physical ones. And so we see rapid change in the digital world with relatively little change in the physical world. Information and software changes dramatically. This ends up having some pretty weird effects. It means that maybe the drivers of what have been the last couple decades of progress in the economy turn around, get changed very rapidly. I think we'll see that flow on into the physical world, but at a delay. So you get much better at doing chip design. You get much better at training AI models. AI models get a lot faster. Chips get a lot better. The general economy gets a lot more efficient because the information and message passing that is much of the rest of the economy ends up becoming much more efficient. But at the same time, you don't yet have robots providing limitless physical abundance. Science probably progresses really fast up to the degree that you need interaction with labs or larger particle colliders or something like this, and then you go 'okay, well I need to build the robots.' But at the same time, automated labs feel more near-term than unlimited robots in the real world manipulating the earth. I think maybe they actually arrive at similar times. I think you need pretty competent robots for the labs, or at least no one's yet managed to figure out how to automate a lab without turns out there are all these really weird little tasks that require a lot of manual human dexterity that are currently part of biological protocols. A lot of biologists will say 'no, you just actually need something that's capable of human level dexterity.' Maybe not for all experiments, but for enough that it becomes annoying.

在这个纯软件的奇点中，你如何定义奇点？你知道，库兹韦尔提出的计算能力超过人脑的公式——那是一个方程。还有就是这个无法预见的临界点。你在这个背景下如何看待奇点？

And in this software-only singularity, how are you defining singularity? You know, the Kurzweil formulation of more computing power than human brains—that's one equation. There's also just this point beyond which we cannot see. How do you think about singularity in that context?

我认为存在一个相当棘手的事件视界，至少我还没发现有人做出过极其有力或准确的预测。感觉在那一刻，你拥有与人类智能数量相当或更多的数字智能，它们与人类智能一样聪明或更聪明。这对世界意味着什么？这非常难以预测。我们花了很多时间思考这个问题，并试图让世界为各种可能性做好准备。但要做出关于具体样子的顶层预测是很困难的。

I think there is a pretty tricky event horizon that I, at least, haven't found anyone that's made incredibly strong or good predictions. It just feels like at that point you have as many digital intelligences or more, and they're as smart or smarter than human intelligences. What does that even mean for the world? It's incredibly hard to predict. It's something we spend a lot of time trying to think about and trying to prepare the world for all the eventualities. But it's difficult to make sort of topline predictions of what exactly that looks like.

是的。在库兹韦尔最初的表述中，这几乎是奇点的定义——你无法预测超越它的东西。所以一旦预测失效，你就到达了奇点。

Yeah. In the original Kurzweil formulation, that was almost the definition of the singularity—that you can't make predictions beyond it. So once the predictions break down, then you're there.

我的意思是，我认为在短期内，我和达里奥一起去奥斯汀参加了埃隆的采访，我认为这在物理世界中的很多反映可能就是最终会发生的事情。你最终基本上是在尝试攀登卡尔达肖夫指数，捕获更多的太阳能量等等。但我认为这需要一些时间。

I mean, I think in the near term, you know, I went down with Dario to his Elon interview in Austin, and I think a lot of the reflection of that in the physical world perhaps is what ends up happening. You do end up basically trying to climb the Kardashev scale and capturing more of the energy of the sun and so forth. But I think that takes some time.

你目前的结论是，我们更受芯片限制还是能源限制？AI 进步的最大瓶颈是什么？

Is your current takeaway we're more chip constrained, energy constrained? What's the biggest bottleneck to AI progress?

我的意思是，我认为目前我同意 Sam 之前的回答，我们在 AI 进步上更受芯片限制。但我觉得有趣的是，向前推两年，你会想，好吧，如果你想要 100 吉瓦，或者四年后你想要 1 太瓦，那会是什么样子？你把它放在哪里？这就是为什么埃隆在追求太空数据中心，对吧？他认为这将是 2030 年在太空中获得 1 太瓦的最简单方式。也许是太空，也许是你知道的某个巨大沙漠——阿塔卡马沙漠、澳大利亚沙漠，或者德克萨斯州的某个地方，也许德克萨斯州有很多太阳能板。

I mean, I think right now I agree with Sam's earlier answer that we're more chip constrained in AI progress. But I think it's sort of interesting to roll forward two years and be like, okay, well, if you want 100 gigawatts or, you know, in four years and you want a terawatt, what does that look like? Where do you put that? And this is why Elon's going after the data centers in space, right? He thinks it's going to be the easiest way to get a terawatt in space in 2030. And maybe it's space, maybe it's the, you know, a giant desert somewhere. It's the Atacama desert, the Australian desert, somewhere in Texas, you know, maybe Texas has a lot of solar panels.

是的。

Yeah.

很难知道。

Hard to know.

是的。Jordy，你们在超级碗之前是如何考虑免费使用限制的？这真的很棘手。我的意思是，其中一个困难是算力非常受限。

Yeah. Jordy, how are you guys thinking about free usage limits ahead of the Super Bowl? It's really tough. I mean, one of the struggles of this is that compute is so constrained.

是的。

Yeah.

是的，我注意到你们没有说“下载 Claude 应用”。没有直接的行动号召。

Yeah, I noticed you guys didn't say 'download the Claude app.' There was no direct call to action.

是的。我的意思是，我认为这个广告的目的很大程度上是为了引发思考和讨论。

Yeah. I mean, I think the purpose of the ad was very much one for provoking thought and discussion.

而且它显然已经成功了。

And it was certainly successful already.

是的。引发了大量的思考和讨论。

Yeah. Provoke a lot of thought and discussion.

而且我认为你不需要明确的行动号召让人们去下载或考虑事情。

And I think you don't need to do an explicit call to action for people to download things or to consider things.

是的，有道理。

Yeah, that makes sense.

但再次，听起来周日 GPU 会火力全开，这是预期。

But again, it sounds like the GPUs will be on fire Sunday is the expectation.

GPU 会火力全开。就像过去一年整个行业的情况一样。我的意思是，我认为我们都必须在算力使用方式上做出极其艰难的权衡。

The GPUs will be on fire. As has been the case for the entire industry for the last year. I mean, I think we've all had to make incredibly difficult trade-offs on exactly how compute is used.

你能详细谈谈你的经验吗？因为我觉得每个人都看到了 Anthropic 在收入或 token 生成等方面的粗略增长曲线，感觉是指数级的但相对平滑。同时，你又不断担心瓶颈：资本够不够，芯片够不够，数据中心够不够，能源够不够，我们是在研究时代还是平台期？你是如何平衡这两种叙事的——你经历的平滑增长（尽管是指数级的）与持续的恐惧（像乌云笼罩行业，可能放缓）？

Can you zoom in on your experience? Because I feel like everyone's sort of seen rough growth curves for Anthropic on the revenue side or on the tokens generated side or whatever, and it feels exponential but smoothish. And then simultaneously you have these constant worrying about bottlenecks: is there enough capital, are there enough chips, are there enough data centers, is there enough energy, are we in an age of research, are we in a plateau? How have you balanced those two narratives—your experience of smooth growth even though it's exponential with constant fear around something like a cloud hanging over the industry and maybe a slowdown?

我的意思是，关于平滑或指数级增长，让我印象深刻的一件事是 SemiAnalysis 的分析发现，我们从一个月前或六周前的约 2% 上升到 4% 的 GitHub 提交是由 Claude Code 完成的。而 Claude Code 完成的 GitHub 提交量在几周内翻倍，这真是荒谬。而且没有真正直观的方式去感受它。你几乎觉得它只是屏幕上的一个数字，但你无法切身感受到。

I mean, I think to note on the smooth or exponential growth, one of the things that really blew me away was when SemiAnalysis did that analysis and found that we went from like 2% a month ago or 6 weeks ago to 4% of GitHub commits done by Claude Code. And for the amount of GitHub commits done by Claude Code to double over a few weeks is truly ludicrous. And there's no real visceral way to feel that. You almost feel like it's a number on a screen, but you can't viscerally feel it.

完全同意。

Totally.

对我们来说，我们一直有非常非常强烈的信念。在很多方面，Anthropic 就是押注于 Scaling 持续这一事实。而且进步不会减弱。所以在很多方面，外部数字只是我们长期以来内部信念的反映，即我们确切地预期所有趋势会如何发展。我的意思是，我认为我们大致符合趋势，比如如果你看态势感知，我相当确定关于功率、能量和浮点运算次数的预测是准确的。感觉更像是我们只是按预期达到了每个里程碑。

For us, we've always had a very, very, very strong conviction. And in many respects, Anthropic is a bet on this being true that scaling is continuing. And that progress continues unabated. So in many respects, the external numbers are only a reflection of the conviction that we've had internally for a long time about how we exactly expect all the trends to go. I mean, I think we're broadly on trend for, like, if you look at situational awareness, I think pretty sure the power and energy and flops predictions are bang on. It kind of feels more like we're just hitting each milestone as we expect.

嗯。到目前为止，大致上我们预期发生的事情已经发生了。

Mhm. And roughly what we expected to have happened has happened so far.

在技术扩散方面，你认为前部署工程师是否有角色去改变组织？我们看到一些新闻说 OpenAI 正在招聘很多人。显然现在很多企业都在使用 Claude。但与此同时，如果你没有空闲的周末或长周末，你可能永远没时间实施它。所以在办公室拥有一些额外的能力和知识实际上可能会提前提升能力。你看到 Anthropic 在这方面会增长吗？

In terms of diffusion of the technology, do you think that there's a role for forward deployed engineers to go in and change organizations? We saw some news that OpenAI is hiring a bunch of folks. Obviously a lot of enterprises are using Claude now. But at the same time, that phenomenon of if you don't have a free weekend or a long weekend, you might never get around to implementing it. So having a little bit of extra horsepower and knowledge around the office might actually pull forward capability. Is that something you could see growing at Anthropic?

是的，我认为这是个好主意。我认为很明显人们不知道如何把握这些东西。而且事实是这很荒谬：三个月前，我们还没有最新一代的模型。我们没有 4.5，你基本上需要在假期期间调整你的商业策略，因为模型突然能够做它们在年底前无法做到的事情。

Yeah, I think it's a great idea. I think it's very clear that people don't know how to hold these things. And also the fact that it's ludicrous: 3 months ago, we didn't have this most recent generation of models. We didn't have 4.5 and like you're meant to adjust your business strategy over the course of the holidays basically because the models are suddenly capable of doing things that they couldn't do right before you enter the end of the year.

你是说我们可能会看到新的白领工作被创造出来？

Are you saying we could see a new white collar job creation?

也许吧。我觉得这些工作真的很有价值。显然有很多价值可以通过它们释放出来。

Maybe. I mean, I think these are really valuable jobs. There's clearly a hell of a lot of value to be unlocked with them.

我们问过山姆这个问题，但我对你的看法很感兴趣。数据是新的石油吗？还是像我说的是化石燃料？更尖锐的问题是：有没有一些组织拥有被锁起来的数据，需要商业开发协议、收购或某种 AI 领导力来解锁这种能力？因为实际上存在秘密，有时是商业机密。我不知道可口可乐的配方是否在 4.6 六里，但如果可口可乐公司打电话给你，它就会在，对吧？那么你怎么看待这个问题和这个说法，即使它已经被说烂了，也许它正在卷土重来？

We asked Sam this, but I'm interested in your take. Is data the new oil? Or is it the fossil fuel as I said? The more pointed question is: are there organizations out there that have data that's locked away, and it requires a business development deal or an acquisition or some sort of AI leadership to add that capability? Because there are effectively secrets, sometimes trade secrets. I don't know if the Coca-Cola formula is in 4.6 six, but it will be if the Coca-Cola company calls you, right? So how do you think about that question and that phrase, even though it's been beaten to death, maybe it's making a comeback?

我认为有两种数据，其中一种比另一种有用得多。一种是公司过去收集的分析数据，或者公司运营的产物，比如内部文件等等。另一种是人们为制作这些文件所做的实际工作，而这并没有被记录下来。所以我认为，与其说数据是新石油，不如说人的专业知识才是关键。模型能够理解并向人学习。你几乎希望模型像组织里的实习生，接受指导和反馈，学习如何完成工作。我类比的是：作为人类，当你加入一家公司时，你主要是从同事那里学习工作，而不是阅读公司所有的文件。这信息量大得多，我大致认为模型也是如此。我期望它们以相当类人的方式从同事那里学习。

I think there are two kinds of data, and one of them is dramatically more useful than the other. There's the kind of data which is the analytics a company might have collected in the past, or artifacts of the company's operation, internal documents and so forth. And there's the other kind, which is the actual work that people did to produce those documents, and that's not recorded. So I think to the degree that I'm not certain it's data is the new oil so much as the expertise of people. Models being able to understand and learn from people. You almost want the models to be like an intern in an organization, get coached and feedback, and learn how to do the job. What I'm analogizing to is: as a human, when you join a company, you mostly learn the job from your colleagues rather than from reading all the documents in the company. It's just much more informative, and I broadly expect that to be true of models as well. I expect them to learn in a quite humanlike way from their colleagues.

而不是阅读公司所有的文件。

Rather than from reading all the documents in the company.

这信息量大得多，我大致认为模型也是如此。我期望它们以相当类人的方式从同事那里学习。

It's just much more informative, and I broadly expect that to be true of models as well. I expect them to learn in a quite humanlike way from their colleagues.

你有没有想过这样一种情况：一家公司的收入归零，所有剩余价值都在他们的历史数据中？

Have you thought about a scenario where a company's revenue goes to zero and all the remaining value is just in their historical data?

嗯，我想在这个类比中，应该是人。我认为价值会随着他们而衰减。

Well, I guess in this analogy, it would be the people. I think the values would regress with them.

当然。是的，有道理。

Sure. Yeah, that makes sense.

当你们考虑新机会、垂直领域、类别时，关于用户界面的思考过程是什么？现在有很多人在构建 AI 原生工具。它们在功能上看起来像传统的企业软件。我很难想象 Anthropic 会构建无限的传统软件表面积来做编排之类的事情。但在考虑新产品的界面时，具体的框架是什么？

When you guys are thinking about new opportunities, verticals, categories, what is the thought process around UI? There are a lot of people building AI native tools today. They functionally look like traditional enterprise software. And it's hard for me to imagine Anthropic building out infinite surface area of traditional software to do things like orchestration. But what is the specific framework when thinking about interfaces for new products?

我们主要想构建一些能融入人类所在之处的东西，能像同事一样与你互动。所以我们想要能融入、加入并使用与你相同的界面的东西，而不是吸收界面。我认为这就是公司想要的。如果你问世界上任何一家公司，‘嘿，我们能给你找来这个职能领域全世界最好的高管，’他们通常会说我们愿意为此付出几乎任何代价，或者一个极高的溢价。然而，如果你卖给他们一个软件解决方案，‘嘿，这个软件解决方案能帮你做另一件事，’也许他们想用，但它不如一个能直接完成工作的东西那么有吸引力。

We mostly want to build something that fits in where the humans fit in, and can help interact with you like a colleague. So we want something that can fit in and join in and use the same interfaces that you do, rather than absorbing interfaces. I think that's what companies want. If you ask any company in the world, 'Hey, we can get you the best executive in this function in the entire world,' they will usually say we will pay almost any price for that, or an extreme premium. And yet if you sell them a software solution, 'Hey, this software solution can help you do this other thing,' maybe they want to use it, but it's not as enticing as something that can just do the work.

对吧？完全正确。

Right? Exactly.

是的。

Yeah.

另外，用户界面的演变似乎被严重低估了。我在思考 OpenAI 的发展，以及移动端对它的重要性。就是这种想法：有人出现说，‘你是个忙碌的高管。你整天开会打电话。你在飞机上。我们给你找了个很棒的软件工程师，但你得通过桌面终端和他们交流。’这就像，我不会下那么多命令，因为我在路上，用手机。现在你会看到人们下飞机时带着 MacBook，因为他们需要再输入一个提示。

Also, the UI evolution seems very underrated. I was processing the OpenAI development and how important mobile was to that. Just this idea: somebody shows up and says, 'You're a busy business executive. You're in meetings and phone calls all day. You're on flights. We got a great software engineer for you, but you got to talk to them over the terminal on your desktop.' It's like, I'm not going to give that many commands because I'm on the road and on my phone. And you see people walking off planes with MacBooks now because they need to get one more prompt in.

我做过这种事。我对此深感内疚。

I have done that. I am so guilty of that.

所以我想在移动端会有很多工作要做。你怎么看？

And so I imagine there's a lot of work that will be done on the mobile side. How do you think about that?

这大致适用于所有 AI 界面，对吧？我不想看屏幕。我不想要终端窗口或聊天窗口。我只想让我的电脑做点什么，或者也许让我周围的世界发生点什么。所以我认为过于关注今天的界面，而不是你会如何与一个极其能干的同事互动，才是正确的。我只想给 Claude 发短信说，‘嘿，你能帮我搞定这个吗？你能帮我解决这个或预订这次旅行吗？’我只想要那样。

That's broadly true of all AI interfaces, right? I don't want to look at a screen. I don't want a terminal window or a chat window. I just want to ask my computer to do something, or maybe ask the world around me for something to happen and for that to happen. So I think focusing too much on the interfaces of today versus how would you interact with an incredibly competent colleague is the right thing. I just want to text Claude and be like, 'Hey, can you fix this up for me? Can you help me sort this out or book this trip?' I just want that.

是的。每个人都有能力从各种设备发短信。人们有 AirPods、苹果手表、手机、笔记本电脑、平板电脑。你认为我们还需要更多的新硬件吗？

Yeah. Everyone has that ability to text from a variety of devices. People have AirPods, Apple watches, phones, laptops, tablets. Do you think we need any more new hardware?

我认为新硬件非常令人兴奋和有趣。我喜欢人们在这方面下注。我觉得有些东西可以捕捉你日常生活中的更多上下文，并且你可以与之交谈。有一点是，对大多数人来说，语音的比特率比打字高。但同时，在大多数环境中，在拥挤的公共场所或工作中与设备交谈实际上相当烦人。所以也许能自动捕捉的东西会很棒。你看过那些人们默读的 YouTube 视频吗？那会很酷。苹果刚收购了一家观察皮肤运动的公司，也许能重新列出。这个领域有很多公司，看起来它就要来了，你就能接入它。

I think new hardware is pretty exciting and interesting. I like that people are placing bets on it. I feel like something which can capture a bit more of the context that you have in your day-to-day life and that you can talk to. One thing is that voice is higher bit rate than typing for most people. But at the same time, in most of our environments, it's actually quite annoying to talk to your devices in crowded public spaces or at work. So maybe something that captures that automatically would be great. You've seen those YouTube videos where people subvocalize? That would be cool. Apple just bought a company that looks at skin movements and maybe can relist. There are a number of companies in the space that seem like it's coming, and you'll just be able to plug into that.

而且我的自然环境能吸收更多我的上下文，我就能直接跟你说话，比如家里到处都有音箱，而不是必须随身带着手机。那会很酷。你觉得有没有可能设备端计算会变得更重要？我感觉 Anthropic 整体上太偏后端了，所有计算都不在设备上。但理论上这可以改变。我不知道你怎么看。

But also my natural environment absorbing more of my context and being able to just talk to you like having speakers around my house rather than having to carry my phone around me. That would be cool. Do you see any world where on-device computation becomes more important? I feel like Anthropic as a whole is so backend heavy, like none of the computation is done on the device. But that could change in theory. I don't know what you think about.

我认为我们总的观点是，给定水平的智能每年会便宜 10 到 50 倍。这极大地普及了该水平智能的获取。这就在我的 Twitter 简介里：‘智能便宜到无需计量’。这很大程度上是我在 Google 做的工作之一。实现方式之一就是该水平的智能走向设备端。但由于 Scaling 持续进行，模型被应用的用例集呈指数级增长。所以完全有可能你的电脑上存在某个子群，或者你有一台内存带宽更好的笔记本电脑，可以运行小模型完成任务。但同时，你也希望更强大的智能在外面规划、把任务分发给群组，真正处理那些密集且智力上困难的工作。

I think our general perspective is that a given level of intelligence gets 10 to 50x cheaper every year. That massively democratizes access to that level of intelligence. It's literally in my Twitter bio: 'intelligence too cheap to meter.' This is in large part one of the things I worked on at Google. One of the ways that happens is that level of intelligence goes on device. But because scaling keeps continuing, there is an exponentially greater set of use cases which the models get applied to. So it's totally plausible that some sub-swarm might exist on your computer, or you might get a laptop with better memory bandwidth that can have little models complete stuff. But at the same time, you also want the much greater intelligence out there planning and farming things out to swarms and really munching on the intensive and intellectually difficult work.

嗯。Jordy，还有别的吗？由于 Anthropic 的各种发布导致的市场抛售，通常是可预测的，还是有过令人意外的？

Yeah. Jordy, anything else? Have the market sell-offs due to various Anthropic releases been generally predictable or have there been any that have been surprising?

老实说，我甚至没注意到上一次，因为我当时埋头于发布。但你看，这次我认为把它归因于 Anthropic 的发布有点过头了。有很多面向律师的 AI 法律工具已经发布给许多客户。我不觉得这次跟那些有什么天壤之别。这只是我们看到的持续趋势的一部分。

I honestly didn't even notice the last one because I was just heads down on the launch. But look, in this case, I think it's a little bit much to ascribe it to an Anthropic launch. There have been legal tools released with AI for many customers that work for lawyers. I don't think this is crazily different to any of those. It's part of a continuing trend that we've seen.

是的。有道理。非常感谢你在发布日抽空来和我们聊天。很高兴听到最新消息。期待下次再聊。再见。祝你今天过得愉快。

Yep. Makes sense. Well, thank you so much for taking the time to come talk to us on launch day. Great to get the update. Can't wait to talk to you next time. We'll see you. Have a good rest of your day.

让我介绍一下面向 AI 时代的 Graphite Code Review。Graphite 帮助 GitHub 上的团队更快地交付更高质量的软件。我还要介绍一下 Vanta，自动化合规与安全。Vanta 是领先的 AI 信任管理平台。我很高兴迎来下一位嘉宾。我们有来自 T1 Energy 的 Dan Barkella。他是董事长兼 CEO。他在休息室等候。现在他出现在屏幕上了。怎么样？你好吗 Dan？

Let me tell you about Graphite Code review for the Age of AI. Graphite helps teams on GitHub ship higher quality software faster. And I'm also going to tell you about Vanta, automate compliance and security. Vanta is the leading AI trust management platform. And I'm very excited for our next guest. We have Dan Barkella from T1 Energy. He's the chairman and CEO. He's in the green room waiting room. Now he's in the TV. What's going on? How are you doing Dan?

嘿，我很好。谢谢邀请。

Hey, I'm doing great. Thanks for having me.

非常感谢。第一次上节目。你介意介绍一下自己和公司吗？然后我想了解一下你加入公司的经历，之后我有一堆关于能源的问题，但我们可以先从介绍开始，如果没问题的话。

Thanks so much. First time on the show. Would you mind introducing yourself and the company and then I'd love to know about the journey to the company and then I have a bunch of questions about energy but we can start with an introduction if that's okay.

当然。很好。我是 Dan Barcelo，一辈子都在做能源。土生土长的纽约人，现在是自豪的德州居民。我们在建设很多不同的能源系统。我的职业生涯带我走遍世界各地建设石油和天然气系统，现在我们真正专注于在美国建设太阳能。现在做这件事正是激动人心的时刻，因为每个人都知道我们需要大量电力和能源。

Sure. Great. Dan Barcelo, I've been doing energy all my life. A native New Yorker, now a proud resident in Texas. We're building a lot of different energy systems. My journey took me through different parts of the world building oil and gas systems, and now we're really focused on building solar in America. It's an exciting time to be doing it when everyone knows we need a lot of electricity and a lot of energy.

嗯。你是怎么来到 T1 的？

Yeah. How did you wind up with T1?

T1 是一家我们看到了需求的公司，特别是公用事业规模的电网级能源系统需要，我不想说颠覆，但需要大规模影响。无论是电网级储能还是太阳能。我们在电网级电池方面做了不少，而电池方面坦白说竞争壁垒很高：左边是三星，右边是 LG，这边是松下，后面是宁德时代，然后天哪，还有特斯拉。但当你看向太阳能时，在美国情况非常不同。它主要由中国公司主导。美国的太阳能技术我们几乎忘了如何在美国制造和建设。但我们有机会购买资产，购买技术和专有技术。现在我们运营它。我们在 24 年底在德州购买了一个 5 吉瓦的资产，并且正在建设工厂的第二部分。所以对我们来说，这真的是关于……

T1 was a company we saw the need for particularly utility scale grid level type of energy systems to be, I don't want to say disrupted, but needed to be impacted in a large way. Whether it's on grid level storage or on solar. We were doing quite a bit on the grid level battery side, and the battery side frankly had a wall of competition: you look left, it's Samsung; you look right, it's LG; you look this way, it's Panasonic; and you got CATL behind you; and then oh my god, here's Tesla. But when you look at solar, it was very different in the United States. It's predominantly by Chinese companies. The American technology of solar was something we kind of almost forgot how to do in the US and build it out. But we had an opportunity to buy assets, by the technology, by the know-how. Now we run it. We bought a 5 gigawatt asset in Texas late '24, and we're building a second part of the plant. So for us it was really about...

你应该看看我们最初对 T1 的反应。我们第一次接触 T1 是看到你们工厂的视频，我和 John 都击掌庆祝。美国做到了。我们可以大规模制造太阳能。然后当然我们了解了公司的历史，但我们很高兴现在它是美国资产。

You should have seen our initial reaction to T1. We saw first interaction with T1 was seeing the video of your facility and John and I were like high five. America's got it. We can build solar at scale. And then of course we figured out the history of the company, but we're excited that it's an American asset now.

是的。你看，我们在达拉斯有一个百万平方英尺的工厂，我们从零产量开始。我们调试了它，并在 11 月将其年化产能提升到超过 5 吉瓦。所以那是很多工作。工厂里有很多机器人，但也有很多人类。

Yeah. And look, we have a million square foot facility in Dallas that we brought from producing zero. We commissioned that and brought it up to producing over 5 gigawatts in annualized runway in November. So that was a lot of work. A lot of robots in the factory but a lot of humans too.

谈谈这个运营规模在美国整体太阳能电池板生产中的背景，因为我认为人们听到 5 吉瓦可能能理解，但这是总产量的一个有意义百分比，对吗？

Talk about the scale of the operation in the context of overall US solar panel production, because I think people hear 5 gigawatt and maybe they can wrap their head around that, but it's a meaningful percent of overall production, is that right?

美国大约是 50 吉瓦，所以仅此一项就约占美国产量的 10%。它高度自动化，有 1200 名员工，全天候运行，除了那些德州人不知道如何应对的冰暴期间。更多是人员通勤的问题，而不是工厂本身。但这标志着显著的规模。它是世界上最现代化的工厂之一。

The US is about 50 gigawatts, so this alone is about 10% of American production. It's extremely automated, it has 1,200 people, it's running 24/7 except during those ice storms that people don't know how to deal with in Texas. It's more about the people driving to work than the factory. But it's a sign of significant scale. It's one of the most modern in the world.

美国这 50 吉瓦的产能与中国相比如何？你能介绍一下地缘政治背景，并让我了解美国政府为支持这一努力做了多少吗？

How does that 50 gigawatts of capacity in the United States compare to China? Can you lay the sort of geopolitical backdrop and get me up to speed on how much the American government is doing to support this effort?

中国已经超过一千吉瓦，1.2 太瓦的产能。中国去年新增的可能接近 350 到 400 吉瓦。这接近美国整个电网的规模。

China is up to over a thousand gigawatts, 1.2 terawatts of capacity. China's adding from last year probably close to 350 to 400 gigawatts. That's close to the American grid.

所以这几乎就像中国可以增加一整个电网。

So it's almost like China can add an entire grid.

是的。

Yeah.

仅靠太阳能。

Just in solar.

所以这些数字很疯狂。我不认为中国这样做是为了降低碳排放。我认为中国这样做是因为他们需要非常非常快地增长。而且我认为这就是你在过去两三年在美国看到的，太阳能和储能也占据了 75-80%以上的份额。

So the numbers are wild. I don't think China's doing it to lower carbon. I think China's doing it because they need to grow very, very fast. And I think that's what you're seeing the last two or three years in the United States, where solar and storage is also taking 75-80% plus.

你问到了美国方面的因素。我认为美国，无论是民主党还是共和党，都对先进制造业设置了很多抑制措施和很多激励措施。无论是过去的《通胀削减法案》还是现在的《海外贸易法案》，这些激励措施都已到位。还有很多关税。我认为其中很多措施让美国能够与其他竞争对手公平竞争。很简单，我认为美国失去了很多制造业实力。我们正在奥斯汀以北约一小时车程的罗克代尔建设一个新设施。那里曾经是一个老铝业公司工厂。现在只是一片空地。我们正在那里建设一个 2.1 吉瓦的太阳能设施，将生产用于我们达拉斯工厂的电池片。

You asked about what the US aspects for it. I think the US, both Democrats and Republicans, have done a lot of disincentives and a lot of incentives for advanced manufacturing. Whether it was the IRA in the past or the OOTBA now, those incentives are in place. A lot of tariffs are in place. I think a lot of those create an ability for the US to be on a level playing field with other competitors. Quite simply, I think America lost a lot of its manufacturing muscle. We're building a new facility outside of it's about an hour north of Austin in Rockdale. It used to be an old Alcoa plant. Now it's just a piece of dirt. We're building a 2.1 gigawatt solar facility there that's going to make the cells that go into our plant in Dallas.

那么，主要瓶颈是什么？纯粹是资本吗？因为理想情况下，T1 可以像复制粘贴一样将达拉斯工厂复制一千次。但实际上要做到这一点需要什么？

So, what are the main bottlenecks? Is it purely capital? Because ideally T1 could just copy and paste the Dallas facility a thousand times. But what does it take to actually do that?

太阳能产业链真的很酷，因为它几乎有四个不同的部分。首先是挖掘二氧化硅并制造多晶硅块。然后进入硅锭和硅片。这些硅片在半导体行业也非常相似。然后它分叉了。高等级的材料进入半导体，低等级的材料进入太阳能。然后你有一个太阳能电池工厂。太阳能电池工厂基本上是一个低等级的半导体工厂。你制造电池片，制造一个晶圆厂。然后这些电池片进入低技术部分，也就是达拉斯。达拉斯是我经常开玩笑的地方，就像一个 5 吉瓦的场地，每天有 50 辆卡车运进玻璃。把电池片放在中间，然后粘合在一起。然后每天有 50 辆卡车运出玻璃。所以就像丹的玻璃公司，我们只是搬运它们。但电池片部分真的很有趣，很多进步都来自那里：特种气体、工艺设备、工程师、非常专业的工艺设备。那是高技术部分。所以对我来说，当我们向后整合时，这非常有趣。回答你关于如何获得更多的问题，从合同开始。我们是一家较小的公司。我们没有像埃隆在太阳能方面宣布的那种规模。我没有无限的资本和资金，也不想作为上市公司大量稀释股权。我们需要合同。这些合同让我们能够获得建造这些工厂所需的风险资本。但建造部分并不难。我认为在美国，建造部分更多的是要应对某些法规和某些延误。我认为德克萨斯州非常有利。但只要有电，只要有水，并且我们完全循环用水，但当我们考虑水和电以及低成本投入时，我们觉得在德克萨斯州建造是一个很好的地方。

The solar chain is really cool because it has almost four different parts. It's got the digging out silica and making polysilicon rocks. Then you're going into ingots and wafers. Those wafers are also very similar in the semiconductor industry. Then it kind of splits. The high-grade stuff goes to semiconductors. The low-grade stuff goes to solar. Then you take a solar cell plant. The solar cell plant is basically a low-grade semiconductor plant. You're making the cells, making a fab. And then those cells go to the low-tech part, which is in Dallas. Dallas is where I always joke about it's like a 5-gigawatt site with 50 trucks a day bringing in glass. Put in the cells in between and glue them together. And then you got 50 trucks a day going out with glass. So it's like Dan's glass company, we just move them. But the cell part is really interesting and that's where a lot of the advances come from: specialty gases, process equipment, engineers, process equipment that's very specialized. That's the higher-tech part. So for me, it's very interesting as we go backwards and integrate. To answer your question on how do we get more of it, it starts with contracts. We're a smaller company. We're not the scale of what Elon's doing in terms of his announcement with solar. I don't have unlimited capital and funds, nor do I want a lot of dilution as a public company. We need contracts. Those contracts then allow us to get the risk capital that we need to build these plants. But the building part is not the hard part. I'd say the building part in America is more about navigating through certain regulations, navigating through certain delays. Texas, I think, is very conducive. But as long as I have power, as long as I have access to water and we go full recycling on the water, but as we go into water and power and low-cost inputs, we feel we're in a great place in Texas to build.

主要买家是谁？比如你花最多时间与谁打交道？或者你实际上会卖给像亚马逊网络服务或谷歌这样正在建设新数据中心的公司吗？

Who are the main buyers? Like who are you spending the most time with? Or will you actually sell to an Amazon Web Services or Google who's building out a new data center?

我们专注于公用事业规模，这涵盖了最大的公用事业公司，但也涵盖了所有你提到的大型科技公司。我们通常通过开发商直接销售，然后他们面对这些客户。其中一些客户出于保密原因不愿公开。我们最近宣布的 Treaty Oak 就是其中之一，他们是总部位于奥斯汀的大型开发商。所以我想他们会有很多客户，正如你所说的那样。对数据中心的需求是超现实的。对电力的需求是超现实的。很多人都知道我们等待涡轮机要等到 2029 到 2030 年。所以你在那里得不到那个。所以如果你想快速扩大电力规模，你现在就需要太阳能和储能。所以这是一个很大的驱动力，也是推动需求的主要部分。除此之外，一些较小的公用事业公司正在寻求增加电网容量。但到目前为止，另一个大趋势是，随着连接电力和互连的空间越来越少，很多事情正在转向表后或分布式发电。这就是你首先建立电力孤岛的地方。你正在建设带有电力孤岛的数据中心，这些孤岛以后可以连接到电网，因为可负担性是一个大问题。没有人想要一个会提高电价的数据中心。那位科学家在佛罗里达州对此直言不讳：数据中心会带来好工作等等，但然后它们会对电价做什么？所以我认为行业必须掌握如何让能源也能存在于一个不会给电网带来更多压力的区域。

We focus on utility scale, and that's capturing the largest utilities, but it's also capturing all the big tech companies that you mentioned. We're usually selling directly through developers that then face those customers. Some of them, for confidentiality, don't want to be disclosed. One of them, Treaty Oak, we recently announced; they're a large developer based in Austin. So I imagine that they'll have a lot of their customers exactly what you just said. The demand for data centers is surreal. The demand for power is surreal. A lot of people know we're waiting for turbines until '29 to '30. So you're not going to get that there. So if you want to start scaling power fast, you need solar and storage right now. So that's a big driver, and that's the bulk of what's driving the demand. Behind that, some smaller utilities are looking to add to the grid and capacity. But by far the other big trend is that as you're running out of spaces to connect power and interconnections, a lot of things are shifting to behind-the-meter or distributed generation. This is where you're building first power islands. You're building data centers with power islands that can later connect to the grid because affordability is a big issue. Nobody wants a data center that's going to raise prices of electricity. The scientist was very vocal about that in Florida: data centers going to bring good jobs and things, but then what are they going to do to electricity prices? So I think it's really important that the industry gets a handle on how energy can also be in an area that's not going to put more stress on the grid.

所以如果我重述一下业务的当前状态：G1 是第一个投入运营的设施。G2 你正在建设。未来看起来怎么样？未来多远？显然，你需要合同来解锁新工厂，但你希望如何扩大规模？你还想在哪里建设？你希望多快复制这些设施？

So if I restate the current status of the business: G1 is the first facility up and running. G2 you're building. What does the future look like? How far into the future? Obviously, you need the contracts to unlock new plants, but how do you want this to scale? Where else do you want to build? How quickly do you want to copy-paste these facilities?

如果我有合同，我很想那样复制达拉斯。无论是六个月到一年，那些都很快。我们已经公开表示，我们有望在今年建设 G2。而且已经开始建设了；我们在 12 月开始了，并且我们正在继续建设。所以那是我们的主要资产，我们的主要焦点。G1 在运营，G2 在建。然后如果我进入供应链的下一部分，我们从康宁和赫姆洛克获得多晶硅和硅片，这样我们就完成了整个链条，成为一条美国链条。

If I had the contracts, I'd love to copy-paste Dallas like that. Whether it's six months to a year, those are fast. We've publicly said that we're on track to build G2 this year. And that's already started construction; we started that in December, and we're tracking to continue to build that. So that's our main asset, our main focus. G1's operating, G2 is in build. And then if I go to the next part of the supply chain, we get our polysilicon and our wafers from Corning and Hemlock, so that we've completed the entire chain to be an American chain.

你怎么看待太空中的数据中心？埃隆以垂直整合而闻名，但会有很多参与者。我想那可能是一种与你可能放在地面上的太阳能板略有不同的太阳能板。

How do you think about the data centers in space? Elon's notorious for vertically integrating, but there's going to be a lot of players. It's probably a slightly different solar panel than what you might put on the ground, I imagine.

你是否感兴趣与人们讨论一份合同，交付最终用于太空卫星的太阳能板？

Is that something where you're interested in talking to folks about a contract to deliver solar panels that would wind up on satellites in space?

我很乐意。当你考虑硅基技术，可能更多是 PERC 而非 TOPCon，低地球轨道有相当大的需求，而不是深空。所以硅片在太空中确实有效。问题真正在于，我们是将其用于地面需求还是太空需求？显然我们不会把带玻璃的组件送上太空，但在电池片或硅片层面，我们很乐意将其作为客户群。很乐意这样做。

I'd love to. When you look at silicon-based, probably more on the PERC side than the TOPCon side, there's quite a bit of demand for low Earth orbit in particular, rather than deep space. So silicon wafers do work in space. The question really becomes, are we going to use it for terrestrial demand or space demand? Obviously we're not sending up modules with glass to space, but at the cell level or at the wafer level, we'd love to feed into that as a customer base. Love to do that.

作为上市公司 CEO 压力很大。做空者最近对 T1 Energy 有什么误解？

Being a public company CEO can be stressful. What do the short sellers get wrong about T1 Energy these days?

作为前买方和前卖方，写过研究报告，我知道这个过程。成为投资组合经理或卖方分析师，你会做对很多事，也会做错很多。我认为大部分情况下，做空者喜欢说：‘哦，他们做不出来，或者他们不会以正确的方式做，遵守美国法律，外国实体担忧。’太阳能行业总有担忧，对中国影响力的误解。那是个老故事了。我们是一家美国公司，美国上市公司，在纽约证券交易所上市。我们买了一项资产，正在运营和建设。这是最主要的部分之一。另一部分，我会有压力，因为有时我的投资者关系负责人说：‘今天别看 Reddit。’我说：‘好吧，但我想看。’然后有人说：‘嘿，去自杀吧。’真是的。

Being an ex buysider and ex sellsider writing research reports, I know the journey. To be a portfolio manager or sellside analyst, you get a lot of things right and a lot wrong. I think for the most part, short sellers like to say, 'Oh, they're not going to make this, or they're not going to do it the right way, compliance with US law, foreign entities concerns.' There's always concerns about the solar industry and misunderstandings about how China's influence is. That's an old story. We're an American company, American public company, New York Stock Exchange listed. We bought an asset and we're running it and building it now. That's one of the biggest parts. The other part, I get stressed because sometimes my IR guy goes, 'Don't read Reddit today.' I'm like, 'Okay, but I want to.' And then someone says, 'Hey, go kill yourself.' Like, Jesus.

幸运的是，你的一些粉丝来到我们的聊天室，他们想知道我们什么时候有周边？他们想要 T1 周边。有计划吗？

Fortunately, some of your fans have come over to our chat and they want to know when are we getting merch? They want T1 merch. Any plans?

我们会尽快上线那个网站，拿到帽子和其他东西，可能没有这个标志或那个标志，但会弄出来。我们很兴奋推出周边。我们爱我们的股东。我们对此心存感激。我们对所有这些部分感觉良好。

We will put that site up as fast as possible and get the hat and get these things, maybe without this logo here or this one here, but get that there. We're excited to get the merch. We love our shareholders. We're thankful for that. We have a good feeling for all those parts.

我想了解公司的形态。感觉像美国再工业化进程。公司有多大？感觉你在创造就业。你们在招聘什么职位？如果真如我所愿规模化，劳动力形态会是什么样？

I'd love to know the shape of the company. It feels like an American reindustrialization process. How big is the company? It feels like you're a job creator. What roles are you hiring for? If this really scales as I hope it does, what does the shape of the workforce look like?

我们在达拉斯已经有 1200 人。在奥斯汀我们将达到 1500 人。在奥斯汀总部我们将超过 100 人。如果我可以做招聘广告，请访问 t1energy.com/careers。我们需要工程师。工程师越有闯劲越好。这必须是一种新型制造。我们的费用公司政策非常简单：不购买任何非法物品，无论是商品还是服务。第二，这是公司资金，要小心使用。第三，我们可以通过你的 Ramp 卡监控所有购买，我们用 AI 来做，效果很棒。没有人需要做差旅报销报告。这只是关于人的责任感。我们必须尽可能快地建设。扔掉一些旧规则书。这也是我们想留在德克萨斯和奥斯汀的部分原因。那里有很多工程师。我们需要来自半导体行业、制造业、机器人行业的工程师。这一切都关乎增长和更多合同，也许有一天会有太空合同。我们获得的合同越多，增长就越快。

We're 1,200 already in Dallas. We're going to be 1,500 in Austin. In Austin headquarters we're going to be over 100. If I can do my recruiting call, go to t1energy.com careers. We need engineers. The more rogue you are as an engineer, the better. This has to be a new style of manufacturing. Our corporate policy for expenses is very simple: do not buy anything illegal, either good or service. Two, it's corporate money, be careful with it. Three, we can watch everything you buy with your Ramp card and we do with AI and it's awesome. No one has to do T&E reports. It's just about accountability as people. We just have to build as fast as we can. Throw some of the old rule books out. That's part of the reason we wanted to be in Texas and Austin. There are a lot of engineers there. We need them from the semiconductor sector, from the manufacturing sector, from the robotics sector. It's all about that growth and more contracts, maybe some contracts for space one day. The more contracts we get, the faster we can grow.

太棒了。非常感谢你所做的一切。感谢你抽时间来和我们聊天。祝贺取得的进展。我支持你。这是一个了不起的项目。

That's amazing. Well, thank you so much for everything you're doing. Thanks for taking the time to come chat with us. Congrats on the progress. I'm rooting for you. This is an amazing project.

你应该在某个时候看看。我想我们当时是直播。从我们第一次谈论 T1 到第二次，股价涨了大约 150%。

You should at some point. I think we were live. Between the first time we talked about T1 and the second, the stock was up like 150%.

这就是为什么我们现在是播客主持人。这就是为什么我们是买方。但谢谢你所做的工作。

That's why we're podcasters now. That's why we're the buy side. But yeah, thank you for the work you do.

非常感谢你抽出时间。

Thank you so much for taking the time.

收到。周边。那是首要任务，对吧？那是仅次于再工业化美国、拯救太阳能行业的第二优先事项，但紧随其后。

Loud and clear. Merch. That's the prior most priority, right? That's priority number two after re-industrializing America and saving the solar industry, but it's a close second.

总之，非常感谢你，Dan。见到你太好了。随时欢迎你回来。我们很快再聊。

Anyway, thank you so much, Dan. This was fantastic meeting you. You're welcome back anytime. We'll talk to you soon.

我们非常兴奋看到这一点。总之，让我先介绍一下 Octa。Octa 帮助您为每个 AI 智能体分配一个可信身份。这样您就能获得 AI 的力量而无需承担风险。保护每个智能体，保护任何智能体。我们将深入探讨《华尔街日报》那篇关于人们如何与 AI 互动的文章。这一切始于一个无害的爱好，一位名叫 Erica Aerson 的女性在经历离婚时用它来娱乐和分散注意力。她使用 Anthropic 的 Claude 构建了一个机器人，根据她和朋友的提示生成图像。很快，她将 Claude 提升为个人生活中的顾问角色。没过多久，她的孩子们就认为她可能需要和这个新朋友保持距离。如今，有两种人：一种是用 AI 做所有事情的人，另一种是其他人。超级用户将工作外包给一群智能体，大幅提高生产力。他们向 ChatGPT 寻求建议，仿佛 OpenAI 的机器人是他们最信任的顾问，从商业策略到是否应该让朋友烫发，无所不包。这是个有趣的问题。其他人则在想，他们的亲人是否有点过头了。也许有些任务应该留给人类。也许其中有尊严。也许有些人喜欢现状。Aerson 说，在她处理一段她怀疑不会有结果的关系时，Claude 特别有帮助。它说，引用 Claude 的话：“嘿，我认为如果你不暂时停止和这个人说话，你无法从中脱身。”这位在俄勒冈州波特兰市一家科技公司从事运营工作的居民说道。后来当她忍不住想重新联系时，Aerson 会向 Claude 确认。它告诉她：“坚持你的立场。”不过，AI 并非对每段关系都有益。当她的孩子们看电视时，Aerson 喜欢摆弄 Claude。她的儿子们不赞成。11 岁的 Grayson 说：“这是浪费时间。她花很多时间在 AI 上，就为了做无聊的老照片。”

We're very excited to see that. Anyway, let me tell you about Octa first. Octa helps you assign every AI agent a trusted identity. So you get the power of AI without the risk. Secure every agent, secure any agent. And we will go into this Wall Street Journal article about how people are interacting with AI. So it started as an innocent hobby to entertain and distract herself while going through a divorce. Erica Aerson used Anthropic's Claude to build a bot that generates images based on prompts from her and her friends. Soon she promoted Claude to an advisory position in her personal life. It didn't take long for her kids to think she might need a break from her new friend. These days, there are two kinds of people: those who use AI for everything and, well, everyone else. Super users are contracting out their work to groups of agents, amping up their productivity. They're dropping advice from ChatGPT as if the OpenAI bot is their most trusted consigliere, counseling on everything from business strategy to whether or not you should let your friend get a perm. It's an interesting question. Everyone else is wondering if their loved ones may be taking things a bit too far. Maybe some tasks should be reserved for humans. Maybe there's dignity in that. Maybe some of us like the way things are. Aerson says Claude was particularly helpful as she navigated a relationship she suspected wasn't going to go anywhere. He was like, quote, this is Claude talking. 'Hey, I don't think you're going to be able to extract yourself from this if you don't stop talking to this person for a while,' says the Portland, Oregon resident who works in operations for a tech company. When she was later tempted to rekindle contact, Aerson would check in with Claude. 'Stick to your guns,' it told her. AI hasn't been beneficial for every relationship, though. While her kids watch TV, Aerson likes to tinker with Claude. Her sons don't approve. 'It's a waste of her time,' says Grayson, 11. 'She puts a lot of time into AI just to make boring old pictures.'

等等，她是怎么用 Claude 制作图片的？

Wait, how is she making pictures with Claude?

是啊，我很……他们没有图像模型。

Yeah, I'm very... They don't have an image model.

肯定是在底层调用某个 API。我不知道。

Must be calling some API under the hood. I don't know.

她是不是在用 Claude Code 利用 Nano Banano Pro？

Is she like on Claude Code harnessing Nano Banano Pro?

也许她说的是 SVG 图像。

Maybe she's talking about the SVG images.

哦，对。也许吧。

Oh, yes. Maybe.

不，但这里有一张她的图片，看起来像是 Grok 生成的。

No, but there's an image in here that's her image and it looks like a Grok image.

也许她在用 Grok。这篇文章很奇怪。我很困惑。

Maybe she's on Grok. This is an odd article. I'm very confused by this.

8 岁的 Anderson 有另一个抱怨。技术的准确性：ChatGPT 无法正确识别一个宝可梦。还有一次他说：‘这不可接受。宝可梦基准测试零分。我要卖股票。它要跌到零了。’

Anderson, age 8, has a different complaint. The technology's accuracy: ChatGPT couldn't correctly identify a Pokemon. Another time he's just like, 'This is unacceptable. Zero on Pokemon bench. I'm selling the stock. It's going to zero.'

完了。完了。

It's over. It's over.

8 岁的 Anderson 说：‘完了。’还有一次，Anderson 很难让它生成一张合适的自己的照片。他说：‘有一次我只有一条腿，还有一次我有三条腿。’想到这个八岁孩子说：‘这是垃圾。这是垃圾。’真可爱。

It's over, says Anderson, age eight. Another time Anderson struggled to get it to generate a suitable picture of himself. 'Once I only had one leg and once I had three legs,' he said. 'This is so cute,' thinking about this eight-year-old being like, 'This is slop. This is slop,' says an 8-year-old.

布鲁克林的初创公司创始人 Eric Davik 使用 AI 进行食谱规划、锻炼、商业策略以及与儿子一起进行氛围编程。这位 42 岁的父亲还将他的高管教练会议录音上传到 ChatGPT 来分析他的进展。

Eric Davik, a startup founder in Brooklyn, uses AI for recipe planning, workouts, business strategy, and vibe coding with his son. The 42-year-old also uploads recordings of his executive coaching meetings to ChatGPT to analyze his progress.

我做过一些……我一度问过 ChatGPT 关于锻炼计划的事。我不太经常回去看。食谱规划。我会让它做……我确实让它做了一份关于我们可以去哪里吃饭的深度研究报告。

I've done some... I asked ChatGPT about a workout plan at one point. I don't go back to it that often. Recipe planning. I'll have it do... I did have it do a deep research report all around where we could go to food.

你认为它真的适合……你认为大语言模型适合锻炼计划吗？就拿……我通常想要一个锻炼计划是……我的意思是，我们不是每天都锻炼，我们有点随性。我会说……

Do you think it's actually good for... Do you think LLMs are good for workout planning? Just taking the... I typically want a workout plan that is... I mean, we don't work out every day and we kind of do a vibes-based approach. I would say...

试着努力推动，而不是……我们通常在锻炼时谈论技术和商业。

Try to push it hard and not... We're usually talking about technology and business during our workouts.

是的。

Yes.

但历史上，比如说当我刚开始锻炼时，我会找一个计划，但我希望它来自一个特定的人，那个人为处于某种状态的人制定计划，比如我当时已经进行了六个月的重量训练。

But historically, let's say when I was starting to work out, I would find a plan, but I wanted it to be from a specific person that was making that plan with an intention around it for somebody in that state, which was maybe like I've been weight training for six months at the time.

是的。是的。是的，有一个巨大的……

Yeah. Yeah. Yeah, there's a huge...

所以如果你只是拿……我问的原因是，我在从大语言模型获取食谱方面运气很差，因为我认为它只是取了互联网上某个菜品的所有食谱的平均值，然后直接给你，结果可能……我记得我们试过一次做华夫饼，结果……华夫饼的食谱明显是错的，因为面糊太稀了，你知道，就是不行。就像是搞砸了不同的……

And so if you just take... The reason I ask is like I've had terrible luck doing recipes, getting recipes from LLMs, because I think it just takes the average of every recipe for a certain item on the internet and just kind of gives it to you, and you can end up with like... I think we tried to make waffles once and it just... the waffle was clearly the recipe was just wrong because they were really runny, you know, it just wasn't working. It was like it had botched the different...

嗯，你没用最新的模型。你用 Codex 5.3 试试。

Well, you haven't used the latest models. You throw Codex 5.3 at that.

是啊，但这显然不是……华夫饼基准测试需要……这不是智商问题。只是……我们在创建华夫饼基准测试。我们在移动目标。它必须能做华夫饼。好华夫饼。总之，让我介绍一下 Cognition。他们是 AI 软件工程师 Devin 的创造者。用你的个人 AI 工程团队解决积压工作。

Yeah, but it's obviously not an... It's waffle bench needs to be... It's not an IQ thing. It's just... we're creating waffle bench. We're moving the goalposts. Needs to be able to make waffles. Great waffles. Anyway, let me tell you about Cognition. They're the makers of Devin, the AI software engineer. Crush your backlog with your personal AI engineering team.

总之，初创公司创始人 Eric 的妻子说……他的妻子不愿使用 AI，尤其是任何与写作相关的任务。作为一名自由记者，她不相信这项技术能复制她 20 年的专业经验。她丈夫的使用方式有时会惹恼她，比如当她让他给儿子的学校写一封重要的邮件时。她读了一遍，感觉有些不对劲。这位 45 岁的女士说：‘我当时想，这基本上就是同一段话重复了三遍。’她问他：‘这是 ChatGPT 写的吗？’她说：‘他承认了。’她可能没撒谎。他说：‘我不记得具体是哪次了。’我喜欢他们这种交叉采访的方式。

Anyway, the wife of Eric, the startup founder, says... his wife is reluctant to use AI, especially for any writing-related task. A freelance journalist, she doesn't believe the technology can replicate her 20 years of professional experience. Her husband's use can annoy her, like when she asked him to write an important email to her son's school. As she read it over, something seemed off. 'I was like, this is basically the same paragraph repeated three times,' says the 45-year-old. She asked him, 'Did ChatGPT do this?' She says, 'He admitted to it.' She's probably not lying. He says, 'I don't remember the specific instance.' I love how they're just like cross-interviewing.

是啊。

Yeah.

这很有意思，因为我在幼儿园申请这件事上完全反其道而行之。我当时想：‘好吧，我知道其他人都会用 AI 垃圾，所以我得写得特别口语化。’于是我在申请里写：‘嘿，你好吗？我叫约翰。’完全自然地写，像意识流一样。然后我又想：‘我可能需要稍微润色一下，但我绝对不会让它有 LLM 的味道，因为我觉得对面读这些的人很多都能看出来。’另外，人们已经开始厌倦这种垃圾了。你怎么看，Tyler？

It's very funny because I leaned the complete opposite way on preschool applications once. I was like, 'Okay, I know that everyone else is going to be using AI slop, so I need to write extra conversationally.' So I'm in the application being like, 'Hey, what's up? My name's John.' Just completely organically writing like flow of state. Then I was like, 'I probably need to polish this up a little bit, but I'm definitely not giving it the LLM flavor because I think that's going to be a tell for a lot of people on the other side reading these.' Another thing is that people are getting sick of this slop. But what do you think, Tyler?

是啊，我完全不认同这篇文章。圣诞假期的时候，我让我爸用上了 Claude Code。我说：‘对，你得用最好的模型。为什么？你还没试过 4 或 5 呢。你在干嘛？’

Yeah, I just don't resonate with this article at all. Over Christmas break, I think I onboarded my dad to Claude Code. I'm like, 'Yeah, you need to be using the best model. Why? You haven't tried 4 or 5 yet. What are you doing?'

是啊。但杀手级用例是什么？你告诉他‘写软件，上 GitHub’了吗？你只是帮他设置好了？

Yeah. But what was the killer use case? Did you tell him, 'Write software, get on GitHub'? Did you just get him set up?

就是任何事。任何任务，让 Claude 去做，它大概就能搞定。

Just like anything. Any task, just have Claude do it and it'll probably work.

我是说，我不怎么做饭，所以我不太了解食谱那方面。但好笑的是，食谱是现成的。你可以直接拿一本食谱翻开，也可以直接谷歌搜一个食谱。但你却说：‘我需要 Claude Code 工作六个小时，给我弄个华夫饼食谱。’这其实不是它的承诺，也不是进步。进步是：‘给我建一个软件，能根据新闻动态和我孩子的喜好生成新的华夫饼食谱，还能生成图片。’做更多需要很长时间的事情。如果你用 LLM 做谷歌就能搞定的事，其实并不会节省时间。

I mean, I don't cook much, so I don't know about the recipe stuff. But the funny thing is that there are recipes. You can just take a recipe book and open it. You can just Google a recipe. But you're like, 'I need Claude Code to work for six hours to get me a recipe for waffles.' That's not actually the promise. That's not actually advancing. The advancing is like, 'Build me a piece of software that will generate me a new waffle recipe based on what's happening in the news and what my kids are into, and generate the images.' Do a lot more that would take a long time. You're not actually going to save time if you're using an LLM for something that's Googleable.

是啊。

Yeah.

我不知道。只是我们现在处在一个非常奇怪的时刻，模型在很多事情上都很擅长。它们即使错了也极其自信。有时候你会……我有个朋友因为开车时用手机口述短信被开了罚单。他被拦下并收到了罚单。他想弄清楚：‘这次违规会不会扣分？’模型说了一件事。但我最近也有其他经历，模型完全搞错了。所以我想：‘好吧，你算是有个提示。’幸好你有我。你发短信给我，我说：‘关于这个，我比 ChatGPT 知道得多。’

I don't know. It's just we're in a really weird moment right now where the models are good at so many things. They are also incredibly confident even when they're wrong. Sometimes you're making... I had a friend that got a ticket for dictating a text into their phone while driving. They got pulled over and were given a ticket. They were trying to figure out, 'Am I going to get a point on my license for this infraction?' And the model was saying one thing. But I've had other experiences very recently with models where they get it completely wrong. So I was like, 'Okay, you kind of got an indication.' Good thing you had me. You texted me and I was like, 'I know more than ChatGPT about this.'

总之，让我来介绍一下认真对待公共投资的人。股票、期权、债券、加密货币、国债等等，还有优质的客户服务。闪电轮的第一位嘉宾来了。我们请进 Elf Beauty 的首席财务官 Mandy Fields 吗？欢迎来到节目。

Anyway, let me tell you about public investing for those who take it seriously. Stocks, options, bonds, crypto, treasuries, and more with great customer service. And our first guest of the lightning round is here. Should we bring in Mandy Fields, the CFO of Elf Beauty? Welcome to the show.

怎么了？你好。谢谢邀请。

What's happening? Hello. Thanks for having me.

很高兴见到你。既然你是第一次来，请简单介绍一下自己。

Great to meet you. Would love a quick introduction on yourself since it's your first time.

好的。我是 Mandy Fields，Elf Beauty 的首席财务官。我在这里工作了七年，连续第 28 个季度报告 Elf Beauty 的业绩。

Sure. So, yeah, I'm Mandy Fields. I'm chief financial officer at Elf Beauty. I've been here seven years, my 28th consecutive quarter of reporting the results for Elf Beauty.

太棒了。太棒了。最近有什么消息？

Amazing. Amazing. And what is the latest? What's the news?

消息是，我们刚刚报告了又一个出色的季度。净销售额增长 38%，调整后 EBITDA 增长 79%。哦，嘿。是的，我喜欢这个。我们在 Elf Beauty 真的玩得很开心。这就是结果。

Well, the news is we just reported another tremendous quarter. 38% net sales growth, 79% adjusted EBITDA. Oh, hey. Yeah, I love that. And so we're really having a lot of fun over here at Elf Beauty. And that's the results.

现在是什么在推动这个？现在我觉得消费者整体实力存在很大的不确定性。有这么多不同的指标。你可以看政府数据，得到一个信号，然后看你们这样的季度业绩，又说明别的东西。但你们看到了什么？

What's driving that right now? Right now I feel like there's so much uncertainty around the strength of the consumer broadly. There are so many different indicators. You can look at government data and see one thing that might give you one signal, and then you look at quarters like you guys are putting up and it says something else. But what are you guys seeing?

是的，这归结为三件事。我们的价值主张，ELF 品牌 75%的产品组合价格在 10 美元或以下，这确实迎合了追求价值的消费者。再加上我们不可思议的创新和营销引擎，顺便说一句，我听说你们都要在超级碗广告中亮相。恭喜你们。我们也会参加。

Yeah, well, it really comes down to three things. Our value proposition, and so 75% of our portfolio on the ELF brand is at $10 or less, and that really speaks to that consumer seeking value. You pair that with our incredible innovation, the marketing engine that we have who, by the way, I heard that you all are going to be at the big game Super Bowl ads. So congratulations on that. We will be as well.

谢谢。我想我们的投入规模会比你们小。我猜。

Thank you. I think our buy will be at a smaller scale than yours. I'm assuming.

我看到了。我看到你们的大约是 5 万美元，差不多这样。

I saw that. I saw yours was about $50,000, something like that.

差不多吧。是的。不过，我也想聊聊超级碗，但请继续。

Something like that. Yeah. But anyway, I want to get into the Super Bowl too, but continue.

然后，我们还新增了 Road 品牌，它对我们的净销售额表现贡献巨大。Road 是我们收购的品牌，8 月份完成交易。那是 Hailey Bieber 的品牌。它的表现非常出色。所以，把这些加在一起，就得到了季度净销售额 38%的增长。

So, and then we also have the incredible addition of Road to our net sales performance. And so, Road is an acquisition that we did. We closed in August. That's Hailey Bieber's brand. And it has just had tremendous performance. And so, you take all of those things together, that's how you get to that 38% net sales growth for the quarter.

过去几年营销组合是如何变化的？我是说，Meta 的季度业绩非常出色。广告在那个平台上表现很好。网红经济仍在增长。营销支出的组合是什么样的？

How is the marketing mix changing over the last few years? I mean, Meta had a blowout quarter. It seems like ads are performing very well on that platform. The influencer economy is still growing. How much... what does the mix of marketing spend look like?

是的，我们的营销支出相当均衡。我们将净销售额的 24%到 26%分配给营销和数字渠道。我们在社区所在的每个平台都有布局。这确实是 ELF 的一个关键差异化因素。我认为我们与社区非常贴近。我们保持敏锐的洞察力，这体现在我们的创新中，我们决定推出的营销活动真正具有文化相关性，并与我们的社区紧密相连。

Yeah, so our marketing spend is pretty well balanced. We allocate 24 to 26% of our net sales to marketing and digital. And we are there on every platform where our community is present. And that has really been a key differentiator with ELF. I think that we are so close to our community. We keep our ear to the ground, and so that really comes through in our innovation and the marketing campaigns that we decide to put forward are really culturally relevant and connected to our communities.

那么你们如何看待智能体商务这个问题？这是需要你们做特殊准备的事情，还是它会自然发生？现在还为时过早，但 AI 领域的事情发展得非常快。

And how are you thinking about the agent commerce question? Is this something that you need to do anything special for, or does it just naturally happen? It's still so early, but these things in AI, they take off so fast.

我认为现在还为时过早。我们正在确保后端正确，以便能够参与智能体商务。但你知道，AI 就是这样一种技术，它既能在收入端也能在效率端提供帮助。

I think it is still early. We're making sure that we have the backend correct so that we can participate in agentic commerce. But, you know, AI is just one of those things where this is kind of one of those technologies that can help on both the revenue side as well as the efficiency side.

从这个角度看，我觉得非常了不起。如今从头打造一个直接面向消费者的品牌，没有大牌明星合作可能吗？我感觉我看到越来越多真正成功、收入可观的交易，背后都有人撑腰。

And so, I think it's pretty incredible from that perspective. Is it possible to build a direct-to-consumer brand from scratch these days without a huge celebrity partner? It feels like more and more of the deals that I see that actually go the distance, get to big revenue numbers, they got somebody attached.

嗯，你这么说很有意思，因为我觉得名人品牌能做到这个地步其实相当罕见。比如想到 Haley 的品牌 Road，你知道，很少看到一家公司靠 10 个产品在不到三年内销售额达到 2 亿美元。通常你需要多个产品、多个 SKU、多个品类才能达到这个水平，所以我认为我们在 Road 上取得的成就非常难得。

Yeah, it's interesting that you say that because I think that's actually pretty rare for a celebrity brand to make it to this side. Like if I think about Haley's brand, Road, you know, it's very rare to see a company go to $200 million in sales in less than three years on 10 products. I mean, you usually are seeing multiple products, multiple SKUs, categories to get there, and so I think what we found in Road is pretty rare.

是的。祝贺你们取得的进展。非常感谢。在你离开之前，我想问一下，作为 CFO，你怎么看待超级碗广告？我相信这不是你第一次经历了。这背后需要什么考量？你知道，超级碗广告有很多不同的方式，但你的逻辑或框架是什么？

Yeah. Well, congratulations on the progress. Thank you so much. I wanted to, before you jump off, I wanted to ask, as CFO, how you think about Super Bowl ads. I'm sure this isn't your first rodeo. What kind of goes into it? You know, there's so many different ways to approach a Super Bowl ad, but what's your logic going into it or kind of framework?

是的。超级碗广告实际上是为了提升品牌知名度。这不是我们第一次做。我记得第一次是四年前，我们和 Jennifer Coolidge 一起拍了一支广告，非常有趣。今年我们请了 Melissa McCarthy 出演超级碗广告，效果很棒。所以我认为这主要是为了建立品牌知名度。回想五六年前，我们的品牌整体知名度大约只有 13%，现在超过了 40%。这要归功于我们多年来所做的许多合作和像超级碗这样的活动。

Yeah. So, Super Bowl ads are really about building awareness behind your brand. And so this is not our first rodeo. I think our first time was four years ago. We came with an ad with Jennifer Coolidge. It was a lot of fun. This year we had Melissa McCarthy as the actress in our Super Bowl ad. It's fantastic. So I really think it's about building brand awareness. If I think back, you know, five, six years, our overall awareness was around 13% as a brand, and now we're over 40%. And I credit that to a lot of the collaborations and the things like Super Bowl that we've done over the years.

那么，衡量标准呢？

So, measurement.

是的。所以，你的 CMO 和营销团队不会来找你说，‘好的，我给你预算，但我希望下周就看到营收增长。’这更像是……

Yeah. So, your CMO and your marketing team are not coming to you and saying, 'Cool, I'll give you the budget, but I want to see a move in topline the next week.' It's more about...

我总说我一直想看到……比如现在就给我看钱。

I always say that I always want to see... like show me the money now.

是的。好吧。但实际上，这是为长期投资播下种子。我们还有其他更直接的收入驱动因素。但像超级碗这样的投资是为了长远。

Yeah. Okay. But in fact, it is sowing the seeds for a longer term investment. There are other things that we look at that are more direct revenue drivers. But Super Bowl investments like that are for the long term.

最后一个问题。你对未来几年的并购前景怎么看？我敢肯定，在理想情况下，每年都会有一个新的 Road 可以纳入，但显然不是这样。伟大的品牌不是每天甚至每年都能诞生的。那么你是怎么考虑的呢？

What's last question for me? What's your M&A outlook for the next couple years? I'm sure in a perfect world there'd be a new Road every single year that you could roll in, but obviously that's not the way. Great brands don't get started every single day or even year sometimes. So how are you thinking about it?

嗯，我们目前拥有一个很棒的品牌组合，所以我们的重点将放在现有组合上，进一步打造这些品牌，无论是 E.L.F. 的化妆品和护肤品、Naturium 还是 Road。我们的组合中有一些非常优秀的品牌，这就是我们将专注的方向。如果在这个过程中我们发现了另一个 Road，我们会很乐意将其纳入组合，但眼下会专注于我们已有的品牌。

Well, we have a great portfolio of brands with us today, and so that's really going to be our focus on the existing portfolio and building those brands out even further, whether it be E.L.F. in cosmetics and skincare, Naturium, or Road. We have some really great brands in our portfolio, and so that's what we're going to be focused on. If somewhere along the way we found another Road, we'd be happy to bring it into the portfolio, but staying focused on what we have for now.

非常有道理。太棒了。好的，非常感谢你抽出时间。祝贺你。很期待周日看到广告。

Makes a ton of sense. Awesome. Well, thank you so much for taking the time. Congrats. Excited to see the ad on Sunday.

是的，非常兴奋。我们很快再聊。是的，祝贺你们这个季度的表现。干杯。

Yeah, very excited. We'll talk to you soon. Yeah, congrats on the quarter. Cheers.

让我谈谈思科。它是 AI 时代的关键基础设施。感谢思科与 TVPN 合作。周一我们在思科 AI 峰会上玩得很开心。真的是很棒的采访。如果你还没看过，赶紧去看看吧。回到时间线。

Let me talk about Cisco. It's critical infrastructure for the AI era. Thank you to Cisco for partnering with TVPN. We had a lot of fun on Monday at the Cisco AI Summit. Truly just great interviews. If you haven't already gone and seen them, go check them out. Back to the timeline.

还有其他新闻。每个县的马匹数量。我没看到这张图表。Chadson 说，‘这些数字太菜了。应该翻倍，甚至三倍。到处都是马。满墙的马。’

There is other news as always. The number of horses per county. I didn't see this chart. And Chadson says, 'These are rookie numbers. Should be double, even triple this. Horses everywhere. Wall-to-wall horses.'

我完全同意。

I totally agree.

我是……应该多得多。

I am... should be way more horses.

我完全同意。我发现，如你所知，我正在托管一处新房产，我和一个非常热衷马匹的人聊了聊，他们详细介绍了这处房产能养什么样的马，还给我讲了大致情况。我希望为……做出贡献，提高这些数字是每个人的责任，对吧？光靠一个人是不够的……

I totally agree. I found out, as you know, I'm in escrow on a new property, and I was talking to somebody very enthusiastic about horses, and I was getting the breakdown on what kind of horses I'll be able to support on the property, and they were giving me the lay of the land. I hope to contribute to... it's going to be on everyone to get these numbers up, right? It's not enough for...

完全同意。

Totally.

我们中的一个人。

One of us.

我是说，Uber 的 CEO Dara Khosrowshahi 昨天来节目说，城市里 75% 的土地是停车场之类的。那不是真实数字，但确实有很多停车场。当我们因为自动驾驶汽车不再需要它们时，它们会变成什么？马厩。是的。你坐 Waymo 或自动驾驶 Uber 进城，然后跳上一匹马，从一个地方到另一个地方。执行得好，J。执行得好。

I mean, Dara Khosrowshahi, CEO of Uber, came on yesterday and said that 75% of all land in cities is parking lots or something. That's not the real number, but there's a lot of parking lots. What's going to happen to them when we don't need them because of autonomous cars? Stables. Yep. You take your Waymo or your autonomous Uber into the city. You hop on a steed and you go from place to place. Excellent execution, J. Excellent execution.

我喜欢马。

I love horses.

我确实喜欢马。我也喜欢 Gusto，因为马有 Gusto，这个统一的薪资、福利和人力资源平台专为中小型企业打造。TVPN。它是 TVPN 的支柱。

I do love horses. I also love Gusto because horses have Gusto, the unified platform for payroll, benefits, and HR built to evolve with small and medium-sized businesses. TVPN. It is the backbone of TVPN.

据 The Information 报道，Meta 即将推出的 avocado 模型在内部被称为迄今为止最强大的模型。

According to The Information, upcoming avocado model from Meta is referenced as the most capable model to date internally.

嗯。

Mhm.

我，这并不能说明太多。

I, this doesn't tell you that much.

确实不能。就像‘史上最强大的 iPhone’一样。我们懂。我们懂。但甚至不是那样。这就像……你知道，他们可以为 Meta 打造最强大的模型。是的。

It doesn't. It's the most powerful iPhone ever. We get it. We get it. Like, but not even that. It's like this is like some... you know, like they could make the most capable model for Meta. Yeah.

如果与其他实验室相比，它可能是最弱的。所以……

Could be the least capable if you compare it to the other lab. So...

是的。但这和 WWDC 上发生的事情一样，你站在台上说这是有史以来最强大的 iPhone，电池续航最长。但问题是，如果它不那么强大，你可能根本就不会发布它。但看到他们取得进展、发送备忘录、向 The Information 泄露消息、兴奋感不断升温，这非常令人兴奋。而且我认为我们确实会得到一些相当强大的东西。基于团队，像 avocado 这样的名字不错，基于团队、算力和资金，他们会跃居前沿。毫无疑问，他们会非常非常接近前沿。

Yeah. But it's the same thing that happens at WWDC where you stand on stage and say it's the most powerful iPhone yet. It has the most battery life of an iPhone ever. And it's like, well, if it was a less powerful iPhone, you probably wouldn't launch it. But it's very exciting to see that they're making progress and they're sending out memos and they're leaking to The Information and their excitement is building. And I think that I do think that we are going to get something pretty powerful. And based on the team, everything like avocado is a good name, and based on the team and the compute and the money, like they're going to jump to the frontier. There's just no question that they'll be really, really close to frontier.

我们得回到马的话题。X 聊天中的 AI 说他们有两匹马在尽自己的一份力。所以，谢谢。

We got to go back to the horses. AI in the X chat says they have two horses doing their part. So, thank you.

哦，谢谢你，Owl。回到原点，提高这些数字。不要只说‘哦，我在尽自己的一份力’。你必须指数级增长。你必须到 4，然后 8、16、32，等等。

Oh, thank you, Owl. And full circle, get those numbers up. Don't just say, 'Oh, I'm doing my part.' You got to be growing exponentially. You got to go to four, then 8, 16, 32, etc.

好的。就 Meta 模型的事再说一句。

Okay. Just on the meta model thing for a second.

历史上，名字里带食物的模型表现都很好，对吧？有 Nano Banana。对。有 Strawberry。Strawberry。没错。我就想到这些。但我觉得，Avocado 是个好兆头。是啊。好吧，好吧，挺好的。Wasteland Capital 是个黑丸。是的。科技行业现在进入了清算螺旋，人们比如对冲基金被迫抛售优质廉价加速资产如半导体来弥补亏损，以及昂贵减速高市盈率的资产如 SaaS 和 AI 受害者。AI 受害者。不知道这要持续多久。我不知道。这周很残酷。我们需要这周的红色西装。确实。但幸运的是，我们的现场嘉宾有一套非常特别的西装，我很期待在下一个嘉宾之后展示给你们看，他正在休息室等着。最后一条，Joe Weisenthal 说：“除了罐头汤，一切都在无休止地抛售。”卡特彼勒表现如何？可能还不错。Campbell 今天上涨了。我喜欢。过去 5 天涨了 6%。避险。对。买点汤。汤的避险。买点汤。Gemini 3 Pro，谷歌最智能的模型。最先进的推理，下一级的氛围编码，以及深度多模态理解。我们的下一位嘉宾正在休息室等着。我们有来自 Daytona 的 Ivan。好名字。好名字。我们喜欢 Daytona。

Historically, food-based models with food in the name have done really well, right? You have Nano Banana. Yes. You have Strawberry. Strawberry. Yep. That's all I can think of. But I mean, Avocado, it's a good sign. Yeah. Okay. Okay, it's good. It's good. Wasteland Capital a black pill. Yes. Tech has now entered a liquidation spiral where people like hedge funds are forced to sell good cheap accelerating assets like semis to cover their losses and expensive decelerating high PE like SaaS and AI victims. AI victims. Not sure how long this will take to play out. I don't know. It's been a brutal week. We needed the red suits this week. We did. But fortunately, our guest who's in person has a very special suit that I'm very excited to show you after our next guest who is in the green room waiting. Last post. Joe Weisenthal says, 'The selloff in everything that is not canned soup is relentless.' How's Caterpillar doing? Probably doing pretty well. Campbell is up today. I love it. Up 6% in the last 5 days. Flight to safety. Flight to Yeah. Buy some soup. Flight of soup. Buy some soup. Gemini 3 Pro, Google's most intelligent model yet. State-of-the-art reasoning, next level vibe coding, and deep multimodal understanding. Our next guest is in the green room waiting. We have Ivan from Daytona. Great name. Great name. We love Daytona.

好了，我们得聊聊这个，因为你在 X 上有一个我最喜欢的互动。大概一个月前。你记得你说了类似‘我周日工作到很晚’的话，然后有人评论说‘你为什么要这么做？没人认识你’，然后 Ivan 说‘这正是原因。’太棒了。精彩。真是个绝妙的回应。

All right, we got to talk about this because you had one of my favorite interactions on X. This was like a month ago. You remember you said something like 'I'm working late on a Sunday' and somebody commented like 'Why are you even doing that? Nobody knows who you are' and Ivan was like 'That's exactly why.' That's so good. Amazing. That's a great response.

是的，基本上原因是，就在假期前，大家都在说‘哦，夏天没人休息，所以我们冬天都休息’，比如创始人和投资者之类的。我在想，我们是认真的吗？这就像一代人只有一次的机会，如果你去休息，你可能就撑不下去了，对吧？我就发了那条推文，然后就像你说的，有人说‘我从没听说过你’，这正是为什么我们必须工作。所以是的。

Yeah, so it was basically the reason was like, it was just before the holidays and everyone was talking about 'Oh, no one took a break during the summer, so we're all taking breaks in winter' like founders and investors or whatever. And I'm thinking, are we serious? This is like a once in a generation opportunity and if you go rest, you won't probably make it, right? And I just tweeted that and then to your point, someone was like 'I've never heard of you' like exactly that's why we have to work. So yeah.

希望现在人们能知道你了。不如介绍一下你自己和公司吧？

Well hopefully people hear about you now. Why don't you introduce yourself and the company?

当然。我是 Daytona 的联合创始人兼 CEO。Daytona 的使命基本上是给每个智能体一台电脑，或者俗称的沙盒，对吧？团队已经一起工作，创始人们差不多 20 年了。所以我们年纪不小了，而且一直致力于为人们服务。从组装服务器到创建编排器等等。

Sure. So I'm one of the co-founders and CEO of Daytona. Daytona is on a mission basically to give every single agent a computer or a sandbox as it's popularly called, right? And so the team has been working together, the founders for almost 20 years. So we're quite old and we've always been in for people. Everything from screwing together servers to creating orchestrators and whatever.

有意思。编排器。跟我说说你觉得编排市场会怎么发展。我是 GAS Town 和那些隐喻的忠实粉丝，感觉这可能是个热门新趋势，但我不确定是不是太早了，或者我们会看到它走向何方。

Interesting. Orchestrators. Tell me more about where you're seeing the orchestration market going. I'm a big fan of GAS Town and the metaphors and it feels like it could be the hot new trend, but I don't know if it's too early or where we'll see all that go.

我的意思是趋势一直在变，就像你看到的，每两个月事情就变了。但有一件事是肯定的，这就是我的意思，我喜欢 GAS Town 和其他所有人在做的事情，基本上就是能够快速启动多个这样的智能体。每个智能体都需要一台电脑。而且我不认为我们会停在两个、四个、十个或五十个，这会是指数级的增长。所以这就是我们正在做的。我们真的很喜欢 Steve 在 GAS Town 做的事情，以及所有其他创造这一切的人。

I mean trends keep changing as you've seen, like every two months things change. But one thing is for sure, and this is what I mean, I love what GAS Town is doing and all these others, which is basically the ability to spin up multiple of these agents. Each of these agents needs a computer. And I don't think we stop here with like two or four or 10 or 50, like this will go exponential. And so that's why we're doing what we're doing. So we really love what Steve is doing at GAS Town and all these other people that are creating this.

再多谈谈产品的情况、演变、团队的情况、公司的现状以及你们想去哪里。

Talk more about where the product is, the evolution, where the team is, kind of the current state of the company and where you want to go.

当然。公司差不多三年了，但我们实际上进行了艰难的转型，解雇了所有客户等等，我们决定，是的，那很艰难。我们看到了即将发生的事情。我们其实不是第一个，但我们确实看到了智能体需要这些电脑，所以我们从零重建了产品，8 个月前发布。从那以后，客户数量疯狂增长。增长非常疯狂。我们有各种客户，从 YC 公司、成长型公司到财富 100 强公司，每个在内部或作为产品构建智能体的人都在使用我们或类似我们的服务。所以是的。

Sure. So the company's almost 3 years old, but we actually hard pivoted and fired all our customers and everything and we decided, so yeah that was a hard one. And we saw what was coming. We weren't actually first but we did see that agents would need these computers and so we actually rebuilt the product from zero, launched it 8 months ago. And since then, the number of customers has been insane. The growth has been insane. We have everything from YC companies, growth companies, Fortune 100 companies, everyone building agents either internally or as products are using us or someone like us. So yeah.

谈谈为什么智能体需要一台电脑。关于 OpenClaw（前身为 Claude Book）一直有争论，比如，你应该虚拟托管它还是应该在 MacBook Mini 上运行？如果你在本地运行，那有安全风险。但然后你和人们聊，他们会说，好吧，你在云端运行，但你仍然给了它访问你邮件的权限。所以就像你把狐狸放进了鸡窝。实际配置一台电脑与仅仅提供其他计算资源的 API 相比，意味着什么？

Talk about why an agent needs a computer. There's been this debate over OpenClaw, formerly Claude Book, around like, oh, should you host it virtually or should you run it on a MacBook Mini? And if you do run it locally, that's a security risk. But then you talk to the people and they're like, okay, yeah, you're running in the cloud, but you still gave it access to your email. And so like you kind of let the fox in the hen house. What does it mean to actually provision a computer versus just APIs to other compute resources?

是的，没错。所以我认为最有趣的部分是，最初智能体就像是基于 RAG 的系统。它们可以连接到 API，通常可以消费信息、分析信息、给你反馈并启动一些任务。但如果你仔细想想，如果那是数字知识工作者工作的方式，你们俩就不会面前有笔记本电脑了，对吧？那并不理想。很混乱。你必须下载东西。你必须安装东西。你必须分析东西。如果我们把 AI 智能体视为数字知识工作者，那么它们确实需要这些电脑就非常有道理了。OpenClaw 或者现在叫什么。还有 Claude Code 等等。我记不清，好像有三个名字。OpenClaw。OpenClaw。是的。所以这正好表明，每个人都在买 Mac Mini。我认为这让人们明白了为什么需要电脑。

Yeah, exactly. So I think that's the most interesting part where originally agents were like these RAG-based systems. They could connect to APIs and they can usually consume information, analyze it, give you some feedback and fire up some tasks. But if you actually think about that, if that is a way a digital knowledge worker would work, the two of you wouldn't have laptops in front of you, right? It's not that ideal. It's messy. You have to download things. You have to install things. You have to analyze things. And if we think of AI agents as digital knowledge workers, it actually makes super sense that they do need these computers. The OpenClaw or whatever it's called right now. And Claude Code and whatnot. I can't, there's like three names. OpenClaw. OpenClaw. Yeah. So this just shows and everyone buying Mac Minis. I think this just made people understand why it needs computers.

但如果你看看我们的用例和人们构建的东西，我们基本上有三个：代码和命令执行、计算机和浏览器使用，以及强化学习。第一个，典型的例子——这些不是我们的客户，我只说最高端的品牌让大家知道——就像你想到 Lovable，你作为人类在 Lovable 上输入，然后它启动一个沙盒来编写代码、预览代码并运行代码。Lovable 有多少用户？它需要数十万个这样的沙盒来运行，这样他们的智能体才能工作。第二个是，如果你需要你的智能体对 Booking.com 的网站进行 QA 测试，它需要一台计算机来运行浏览器以完成这些操作。对我们来说最新的用例是强化学习环境。我想 Dylan Patel 昨天是你的节目嘉宾，我们下周要举办一个会议，所以他也是我们的嘉宾之一。

But if you look at our use cases and what people build, we basically have three: code and command execution, computer and browser use, and RL. The first one, like the canonical example—these aren't our customers, I'll just say the highest brands so people know—it's like if you think of Lovable, you as a human type with Lovable and it spins up a sandbox to write the code, preview the code, and run the code. Lovable has how many users? It needs hundreds of thousands of these sandboxes to be run so their agents can get there. The second one is if you need your agent to do QA testing of Booking.com's website, it needs a computer to be able to run the browser to do these things. The newest use case for us is RL environments. I think Dylan Patel was a guest yesterday on your show, and we're doing a conference next week, so he's one of our guests.

他会去的。

He's going.

是的，他会去。我们在 Chase Center 租了场地，举办一个叫 Compute 的会议。太棒了。所以如果你们想参加，非常欢迎。

Yeah, he's going. So at the Chase Center, we rented out the Chase Center to do the conference called Compute. That's amazing. So you're very invited if you guys want to join.

他有一篇文章关于强化学习环境，以及过去 18 个月里大多数模型如何因为后训练和启动这些强化学习环境而变得更好。这些公司要启动强化学习环境，不仅需要快速的计算机或沙盒，还需要并发能力。所以他们需要在 3 分钟内启动 10 万或 20 万个实例，对吧？这不是一件小事，也不是现成就能搞定的。所以当你思考为什么智能体需要计算机，当你真正思考 AI 智能体和软件的市场时，这就是它的定位。

And so he has this article about RL environments and how most of the models have now over the last 18 months become better because of post-training and spinning up these RL environments. And these companies to spin up these RL environments, they need not only fast computers or sandboxes, they need them in concurrency. So they need 100,000 or 200,000 to be spun up in an order of like 3 minutes, right? This is not a trivial thing and not something that you can really take off the shelf. So when you think about why agents need computers, when you really think about the market of AI agents and software, that's sort of where that fits in.

我希望 Dylan Patel 做 Lil Yachty 的出场秀，他做了很多次假的 AI 版本，现在他在 Chase Center，有 1500 个开发者在那里。你得放 Lil Yachty 的音乐让他走红毯。我在聊天里看到 First Mark 了。告诉我们新闻。发生了什么？

I want Dylan Patel doing the Lil Yachty walk out that he's made fake AI versions of it so many times and now he's at the Chase Center, 1500 builders there. You got to get the Lil Yachty playing so he can do the walk out. I saw First Mark in the chat. Give us the news. What happened?

详细说说。

Break down.

是的，我们筹集了 2400 万美元的 A 轮融资。

Yes, we raised a $24 million Series A.

哇。恭喜。

Boom. Congratulations.

非常感谢。非常感谢。

Thank you so much. Thank you so much.

是的，非常令人印象深刻的进展，我相信你很快就会回来，祝你在会议上玩得开心，聚集 1500 人。不容易。

Yeah, very impressive progress and I'm sure you'll be back on very soon and have a great time at the conference pulling together 1500 people. Not easy.

我们很兴奋。非常感谢。我们聊了很多。很高兴见到你，Evan。祝你今天愉快。再见。

We're excited. Thanks so much. We'll talk so much. Great to meet you, Evan. Have a good rest of your day. Goodbye.

事不宜迟，我们让下一位嘉宾久等了。他在 TBP Ultra Dome 现场。是 Scott。你告诉他他穿西装了。非常感谢你的到来。看看这套西装。是比赛穿过的吗？总得有人给节目带点范儿来，你知道的。这里没有半拉链。

And without further ado, we've been keeping our next guest waiting. He's live in the TBP Ultra Dome. It's Scott. You tell him he's suited up. Thank you so much for coming. Look at this suit. Is that game worn? Someone's got to bring some swag to the show, you know. No quarter zips here.

只有不马虎的定制西装。

No half-ass custom suits only.

实际上，是的。这套西装是怎么来的？

Actually, yeah. What's the origin of the suit?

不，Shinasty。我得准备一下。

No, Shinasty. I got to prep.

不，你们有你们的赞助商。我有我的 Shinesty.com。他们做我所有的定制西装。顺便说一句，可惜体育卡公司 Fleer 已经不在了。Fleer。还记得 Fleer 体育卡吗？

No, you guys got your sponsors. I got my Shinesty.com. They make all my custom suits. By the way, it's a shame the sports card company Fleer isn't still in business. Fleer. Remember Fleer sports cards?

记得。

Yeah.

因为那样他们就可以把这个叫做 Fleer Ultra Dome 了。

Because then they could have called this the Fleer Ultra Dome.

是啊。一直在发金牌。我得为 Dylan 做这个。Dylan 需要一个清晰的。他是铁粉。是的。你怎么认识 Dylan 的？

Yeah. Been giving out gold medallions. I had to do this for Dylan. Dylan needed a clear. He's a big fan. Yeah. How'd you meet Dylan?

Dylan 和我在 HQ 认识的。

Dylan and I met at HQ.

好的。

Okay.

他加入的时候，我是主持人。他是作为合作伙伴负责人加入的。

When he came on board, I was hosting it. He came on board as a partnerships guy.

绝对的奇才。他是。

Absolute wizard. He is.

你们太幸运了。我们有他。

You guys are so lucky. We have him.

我们在 Savvy 有他。如果我们市场稍微早一点，也许我们就能先雇到他，但算了。

We are with Savvy. If we were just a little earlier market, maybe we could have hired him first, but yeah.

是的。是的。你们，我们总有一天会挖走他的。我很确定。

Yeah. Yeah. You guys, we'll poach him one day. I'm pretty sure.

是啊。介绍那会是很好的人才工作。那会是美好的一天。

Yeah. Introduce that'll be a good talent work. That'll be a good day.

然后是的。介绍一下 Savvy，让我们了解最新情况，然后我们当然会回头聊很多关于 HQ 的事。

And then yeah. Introduce Savvy, get us up to speed, and then of course we'll go back and talk a lot about HQ.

当然。是的。所以 Dylan 和我在 HQ 认识的，如果观众不熟悉的话，那是 2017 年一个直播互动手机游戏秀的病毒式现象。你们玩过吗？

Sure. Yeah. So Dylan and I met at HQ where if viewers aren't familiar, was this live interactive mobile game show viral phenomenon back in 2017. You guys play?

哦，是的。当然。

Oh yeah. Of course.

好吧。好吧。

All right. All right.

我从没赢过。我觉得我从没赢过。

I never won. I don't think I ever.

我也没有，老兄。

Neither did I, man.

哦，是的。好吧，你在玩内幕交易。

Oh yeah. Well, you were playing insider trading.

不过很难。老实说，通往 Savvy 的旅程的一部分是吸取了 HQ 的许多经验教训，HQ 一开始非常成功，病毒式国际热门，但当你回想起来，事后看来，你会想，‘哦，也许这是一个有缺陷的产品，因为它是一个几乎不可能赢的游戏，对吧？很少有人真正赢了。’很多非常聪明的人，包括 Stephen Colbert，当我上他的节目时，他在幕后告诉我，他没有在镜头前说，他说，‘我玩过你的游戏几次。没赢。不玩了。’然后我就想，‘哦，有意思。’因为如果你是一个聪明人，你会感觉一次又一次地输。这不是好感觉。所以从这个角度看，也许不是最好的，因为如果你玩《堡垒之夜》或《使命召唤》表现不好，他们会把你和技能上限相近的人配对。就像大多数游戏都有基于技能的匹配。你去 chess.com，你会和水平相当的人配对。这样你就能赢几场。而 HQ 只会越来越有竞争性。

It's tough though. Honestly, part of the journey to Savvy is taking a lot of the learnings and insights from HQ, which was a tremendous success at first, viral international hit, but when you think about it and in hindsight, you go, 'Oh, maybe it was a flawed product because it was this really impossible game to win, right? Very few people actually won it.' A lot of very smart people, including Stephen Colbert when I did his show, he told me behind the curtain, he didn't do this on camera, he goes, 'I played your game a couple times. Didn't win. Stop playing.' And it's like, 'Oh, interesting.' Because if you're a smart person, you feel losing over and over and over. It's not a good feeling. So from that perspective, it was maybe not the best because yeah, if you play Fortnite or Call of Duty and you're not doing well, they will pair you with other people that are at your skill ceiling. Like skill-based matchmaking happens in most games. You go to chess.com, you'll be matched with someone at your level. So you can get a couple wins under your belt. With HQ, it just gets more and more competitive.

正是。

Exactly.

然后公司显然必须提高门槛，否则每个人都赢。

And then the company obviously has to raise the bar or else everyone wins.

是的。而且迭代产品、添加功能、添加新格式。他们就是没能做到。所以对于 Savvy，一个要点是我们想让这个游戏成为每个人都能从头玩到尾的东西。因为在 HQ，如果你在 12 题中的第 2 题就出局了，你就完了。就是这样。

Yes. And iterating on the products, adding features, adding new formats. They just never quite got there. So with Savvy, one of the takeaways is we want to make this game something that everybody can play all the way through. Because with HQ, if you got out on question two out of 12, you were done. That's it.

是的，没错。是的。

Yeah, that's right. Yeah.

所以现在你可以玩 Savvy 的全部五轮，都是字谜。我们考虑过 trivia，但我觉得 AI 已经把它扼杀了，因为用这些机器人太容易了，老兄。

So now you could play all five rounds of Savvy, which are word puzzles. We thought about trivia, but I think AI has put a kaibosh on that because it's just too easy with these bots, man.

是啊。有人会拥有世界上最大的机器人农场，每次都试图赢。

Yeah. Somebody would have just like the biggest bot farm in the world just trying just winning every single time.

我的意思是，他们八年前就已经在这么做了。所以想象一下你现在能做什么。

I mean, they were already doing that eight years ago. So just imagine what you can do now.

你们当时在对付机器人。

You guys were dealing with bots.

是啊。我的意思是，我不知道当时是不是 AI，但确实有 Discord 农场和机器人玩家。我不是技术专家，所以不太懂那些东西。但 Savvy 做的是文字谜题，基本上就是现场版的 Wordle 加上 Connections。不知道你玩不玩这些游戏。

Yeah. I mean, I don't know if it was AI back then, but there were Discord farms and bot people. I'm not a tech guy, so I don't really know how that stuff works. But with Savvy, we're doing word puzzles. It's basically a live Wordle meets Connections. I don't know if you play those games.

所以它是我们自己的版本。五轮，限时回合，可以得分。

So it's a version, our version of that. Five rounds, timed rounds. You can score points.

关键的区别在于：你在和我对战。

And the key differentiator: you're playing against me.

对，就是你。

Yeah. It's you.

是主持人。主持人也参与。所以想象一下 HQ，如果是我在提问并回答，而你答得比我多就赢了。想象 HQ 那样运作。Savvy 就是这样。

It's the host. The host plays too. So imagine HQ if I was asking the questions and answering them, and if you answered more than me, you won. Imagine if that's how HQ works. That's how Savvy works.

所以如果你得分比我高，解谜更快，你就打败了主持人，你就是“主持人终结者”。

So if you score more points than me, if you solve the puzzles quicker, you beat the host, you're a hostbuster.

主持人终结者。

A hostbuster.

那你现在的一天是怎样的？

What's a day in the life like then? Right now.

嗯，我上周刚搬到洛杉矶。所以还在适应新生活。几年前我在这里待过。之前是纽约人，一辈子都在纽约。21 到 24 岁左右在洛杉矶，因为一些感情关系搬回纽约，现在又为了新的感情关系——我的新承诺——回来了。

Well, I just moved to LA last week. So I'm trying to get my day in the life. I was here a few years ago. New York guy before, New York my whole life. I spent about 21 to 24 in LA, moved back to New York for some relationships, and came back here for my new relationship, my new commitment.

这主要是因为时间……

This is because frankly the time...

和这个游戏结婚了。我没和别人结婚，只和这个游戏。我的承诺就像 Kurt Russell 对 Goldie Hawn 那样。

Married to the game. I'm not married to anyone else but this game. I'm committed like Kurt Russell to Goldie Hawn. That level of commitment.

Arizona 冰茶 99 美分。那种承诺。那就是我的承诺。

Arizona iced tea 99 cents. That commitment. That's my commitment.

但我才来一周，所以还没有固定的日常。早上遛狗，因为时差还是早起。喝咖啡，做早餐，做得不错。

But I'm here for a week, so I really haven't gotten my day-in-the-life routine yet. I walk the dog in the morning. I'm still getting up early with that time zone difference. Having my coffee, making breakfast, doing a great job.

保持纽约时间。

Stay on New York time.

我想过。可能会这样。你们为了节目起得很早，所以保持纽约时间其实有帮助。

I thought about that. It might happen. You guys are up early for the show. So it actually helps, kind of staying in there.

我们锻炼、吃早餐、准备节目，11 点直播。在特定时间直播、迭代、录制这些有多重要？你怎么考虑引入其他主持人、你的工作流程和参与度？因为我觉得关键就是人才，就是主持人。

We do the workout, go get breakfast, prep the show, go live at 11. How important is going live at a particular time, the iteration, any of this recording? How are you thinking about rotating in other hosts, your workflow and involvement? Because I feel like the unlock is the talent, is the host.

当然。我们肯定会找其他主持人。必须的。我今天还谈了一个 3 月 18 日在迈阿密的演出。所以我们会找其他主持人，做其他节目。这些都会发生。但首先，来到这里看着你们这样成长真的很酷。我一直关注着，看着你们发展。

For sure. We're going to definitely find other hosts. We have to. I already talked today about doing a gig in Miami on March 18th. So we're going to have other hosts, other shows. That's all going to happen. But look, this is so cool being here, first of all, watching you guys grow the way you have. I've been following along and watching. It's very cool to see.

这里有很多 HQ 的影子。

There's a lot of HQ in here.

有很多 HQ，也有一些 Clubhouse。有很多不同的尝试，我们把它重新拼凑起来。

There's a lot of HQ. There's some Clubhouse. There's a lot of different stuff that has been tried, and we sort of pieced it back together.

你意识到节奏的重要性，每天直播。对你们来说，是长节目。对我们来说，是短时间高专注的爆发，但完全是另一回事。我们在玩游戏。

And you realize the importance of cadence, of going live every day. For you guys, having that long show. For us, a short burst of tight attention span, but a whole different thing. We're playing a game.

比如，我们之前在纽约每周做测试节目。我一开始就知道，每周一次不行……永远不要每周做任何事。要坚决。

For example, we were doing beta shows weekly in New York. And I knew going in, weekly isn't going to really... never do anything weekly. Always strongly.

我喜欢这个说法。

I like that.

所以我们开始做每日节目。现在每周直播五次，周日至周四，上学日的晚上，东部时间晚上 9 点。上周四我们有 1200 人同时在线峰值，总共约 1500 人参与。昨晚，不到一周后，有 3000 人。所以每周翻倍增长。

So we started doing daily shows. Now we go live five times a week, Sunday through Thursday, school nights, at 9:00 p.m. Eastern. Last Thursday we had 1,200 peak concurrent, about 1,500 total entered. Last night, less than a week later, 3,000. So we've doubled week over week.

不错。没有变现吗？

There you go. No monetization?

我们在 Teespring 上卖了一件 T 恤，赚了三美元。在 TikTok 直播同步播出上大概赚了四美分。所以我们处于后变现、后收入阶段。但我在自力更生做这件事。

We sold one t-shirt on Teespring. Got three bucks for that. I think we made four cents on TikTok live simulcast. So we are post-monetization, post-revenue. But I'm bootstrapping this thing.

我正想问：你提到了 HQ 的一些产品挑战。如果普通用户每晚都输，那很糟糕。但从商业角度看呢？感觉 HQ 一旦有了风险投资，就必须追求规模。而当你自力更生时，只要业务可持续，就没有理由停止。你可以达到一个点，觉得好吧，我们碰到了天花板，那也没关系。也许没有天花板，也许有，但无论如何都不重要。

I was going to ask: you talked about some of the product challenges with HQ. If the average user is losing every single night, that's rough. But what about from a business standpoint? It also feels like HQ was in this position where once it had real venture dollars behind it, you just have to go for scale. And when you're bootstrapping, as long as the business is sustainable, there's no reason to stop. You can reach a point where you're like, okay, we've hit a ceiling, and that's okay. Maybe there is no ceiling, maybe there is, but it doesn't really matter either way.

是的。根据我在 HQ 的经验，我对 VC 文化并不乐观。你给一群年轻人很多钱，然后说去扩大公司规模，去领导团队，去管理人。没多少人能去和 Facebook、Google 竞争。但即使只是个人商业头脑：这些人大多没上过商学院。总之，我的观点是，我们这里的理念是真正作为一个完整团队尽我们所能。我自己投钱，另一位联合创始人也投钱。我们真的在努力尽可能长久地自己做这件事。我们确实和几个早期种子轮的人开了几次会，只是为了验证产品。我不会告诉你要扩张。

Yeah. I don't have a very sanguine view about VC culture based on my experience with HQ. You give a bunch of young people a lot of money and just say go scale a company, go lead a team, go manage people. Not very many people can go compete with Facebook, go compete with Google. But even just personal business sense: most of these people didn't go to business school. Anyway, my point is we have a philosophy here where we're really trying to bring as much as we can to this as a full team. I'm putting my money in. One of my other co-founders is putting money in. We are really trying to commit to doing this ourselves for as long as we can. We did take a couple meetings with some early-stage seed guys just to kind of validate the product. I won't tell you to grow.

另一个问题是：这是否更适合娱乐行业的人，不把它看作科技公司或 AI 公司，而是更看重人才的价值、社区的价值、产品的价值？

The other question is: is this almost more suitable for someone who is in the entertainment industry, not thinking about this as a tech company or an AI company, but more about the values in the talent, the values in the community, the values in the product?

是的。我觉得在好莱坞和娱乐业，逻辑是：嘿，如果我们花 2000 万美元拍这部电影，赚了 8000 万，那就是胜利。而在 VC 界，如果一家公司说我们融资 2000 万，只想赚 8000 万，那我们就开放它，也许继续做续集，这对我来说也很疯狂，因为 6000 万并不差。给我四分之一就行。所以不，我的意思是，我理解。有一个完整的投资者生态系统。我们当然对此持开放态度。

Yeah. I feel like in Hollywood and entertainment, it's very much calculus of: hey, if we can spend $20 million to make this movie and it makes $80 million, that's a win. And in VC, if a company says we're raising $20 million and we only ever want to make $80 million, then we're opening it up and moving on to this sequel maybe, which also is insane to me because $60 million is not bad. I'll take a quarter of that. So no, I mean, listen, I get it. There's a whole ecosystem of the investor. And we are certainly open to it.

你最初是怎么得到 HQ 那份工作的？

How do you originally get the job at HQ?

哦，那是试镜，老兄。试镜。所以，是的，尽管那是一家科技公司，他们还是举行了完整的试镜。

Oh, that was an audition, man. An audition. So, yeah, even though it was a tech company, they held full auditions.

你是怎么看到的？比如人才选角网站还是 Craigslist 广告？

How did you see it? Like a talent casting website or a Craigslist ad?

纯粹是通过一个朋友。一个我以前在《洋葱报》一起工作过的人。

It was purely through a friend. A guy I used to work with at The Onion.

为人脉网络鼓掌。

Let's give it up for networking.

哦，我喜欢人脉网络。是的。

Oh, I love the networking. Yes.

《洋葱报》。你在那里工作过？

The Onion. You worked there?

所以，2008 年我在那里实习。这个叫 Nick 的人是图片编辑，我们成了朋友，你知道，几个年轻人，有趣的家伙，在纽约一起玩。我们一直保持朋友关系，主要是在 Facebook 上。但我帮我室友在后来成为 HQ 的那家公司找了份工作。结果 Nick 在 Facebook 上发消息说：‘嘿，我在找一个动画师。’我室友就是动画师。所以 Russ 得到了那份工作。

So, in 2008, I interned there. And this guy Nick was the photo editor and we became friends, you know, a couple of young guys, funny guys hanging out in New York. And we stayed friends throughout, on Facebook mostly. But I got my roommate a job at the company that was HQ. It turns out Nick had put out a message on Facebook saying, 'Hey, I'm looking for an animator.' My roommate was an animator. So Russ got the job there.

然后你瞧，几个月后，Nick 打电话给我说：‘你知道，我们正在为这个手机上的游戏节目选主持人。’我当时正要搬到洛杉矶。那是 2017 年。我已经搬出了布鲁克林的公寓。我准备好了要走。我想，好吧，我再试一次镜。我每次试镜都失败。在《Broad City》和《Search Party》这样的节目上左右碰壁。我记得有一个 Delta 广告，和 Noah Synagar 一起。我本来要脱掉衬衫，我很兴奋能那样做。

And then lo and behold, a few months later, Nick calls me and goes, 'You know, we're casting a host for this game show on your phone.' I was about to move to LA. This was 2017. I already moved out of my apartment in Brooklyn. I was ready to go. And I'm like, look, I guess I'll do one more audition. I've failed every audition. Just flunked them left and right for shows like Broad City and Search Party. I remember there's a Delta commercial with Noah Synagar. I was going to take my shirt off and I was excited to do that.

你当时很冒险。

You were dice.

我所有那些试镜都失败了，而这是我得到的唯一一次试镜，我参加的最后一个，它改变了我的生活。我还能说什么呢？只是……但你知道，你提到了整个媒体和科技的混合体，这正是 HQ 非常独特的原因，也是 Savvy 独特的原因。

I failed all those auditions and this is the one audition I got, the last one I took, and it changed my life. And what can I say? It just... but you know, you mentioned the whole media and tech hybrid, and that's what made HQ very unique, but also what makes Savvy unique.

听着，HQ 曾引起 NBC、迪士尼、福克斯的兴趣，还有 Facebook，以及风险投资，你知道，Founders Fund 投了 1500 万。但很少有公司能同时从这些主要影响力和资金领域获得这种双重兴趣和吸引力。这关乎管理这些投资部分，但也关乎管理产品，并理解当你在手机上、在自己的应用上做一个直播互动游戏节目时，就像你们在建立自己的网络方面所做的那样，你必须考虑所涉及的技术。看看流媒体等正在发生的事情，但这是一个节目。这是娱乐。而你们在做所有这一切，就像我们也在做所有这一切一样。这很令人兴奋。非常沉重。有时可能很可怕。很少有人在做这件事。所以，为我们自己做的媒体科技混合体点赞。

Look, HQ had interest from NBC, Disney, Fox, also Facebook, also the venture capital, you know, Founders Fund put in 15 million. But very few companies have that type of dual interest and attraction from those major spheres of influence and money. And it's about managing those types of investment parts, but also just managing the product and understanding that when you're doing a live interactive game show on a phone, on your own app, just like what you guys are doing in terms of building your own network, you've got to think about the technology involved. Look at what's happening with streaming and everything, but it's a show. It's entertainment. And you guys are doing it all, just like we're doing it all. It's exciting. It's very heavy. Could be scary at times. Very few people are doing it. So, give ourselves some credit for doing the media-tech hybrid.

他们说过这是不可能的。你觉得扎克伯格多久会克隆 Savvy？

They said it was impossible. How soon do you think Zuck will clone Savvy?

我很希望他试试。

I would love for him to try.

其他人，说说你怎么看待这件事。

Someone else, break down how you process that.

听着，老兄。我的意思是，他们说奉承是……我忘了他们怎么说。

Listen, man. I mean, they say flattery is the... I forget what they say.

模仿是最真诚的奉承。

Imitation is the sincerest form of flattery.

我们喜欢说模仿就是烂。

We like to say imitation just sucks.

模仿，尤其是模仿的垃圾。

Imitation, especially imitation crap.

那是给失败者的。

It's for losers.

那些垃圾加州卷。

Those garbage California rolls.

这令人作呕且卑鄙。那么你应该为模仿我感到羞耻。

It's disgusting and despicable. Then you should be ashamed for imitating me.

而且你说这根本不感到荣幸。

And you say this not flattered at all.

对。完全正确。而且 CNBC 没有人会有意见。但你看，我们都模仿。我们都互相混搭。但问题就在这里。就像你们正在做的。就像，哦，我不需要 CNBC 的工作。我可以创建自己的 CNBC。自己做。

Right. Exactly. And nobody at CNBC would have any issue with that. But look, we all imitate. We all remix each other. But that's the thing. It's like what you guys are doing. It's like, oh, I don't need a job on CNBC. I can create my own damn CNBC. Make your own.

CNBC。你当然可以。

CNBC. Of course you do.

但我老实说，我甚至从未想过，哦，我应该去那里申请工作。我甚至不知道如何进入那个世界。这很自然。

But I honestly would have never even thought like, oh, I should apply for a job there. I don't even know how to get into that world. It just comes naturally.

你说得对。如果你问约翰和我，甚至一年半前，嘿，你们有没有想过我们能不能做商业电视？我们会说，让我们做一个有两个麦克风的播客，只是互相聊天，然后它变成了一种电视节目，但在线直播。

You're right. If you asked John and I even a year and a half ago, hey, would you ever think about could we do business television? We'd be like, let's have a podcast with two microphones, just talk to each other, and it turned into a TV show kind of but streamed online.

你知道，有人在我主持 HQ 时问我，我接受采访，他们说：‘那么，你五年后想在哪里？你的下一步是什么？’我说：‘如果这东西继续下去，我会在这里待 20 年。这是未来。’不幸的是，它没有持续很久。但对于 Savvy，我是在长期投入，志在必得。我投入了心血。我的脸再次出现在最前面和中心。无论好坏，我不认为美国 100% 喜欢这张脸，但足够多的人喜欢。

You know, someone asked me when I was hosting HQ, I got interviewed and they said, 'Well, where do you want to be in 5 years? What's the next move for you?' And I go, 'I'll be here for 20 years if this thing stays on. This is the future.' Unfortunately, it didn't last very long. But with Savvy, I am in this to win this for the long haul. My skin's in the game. My face is again front and center. For better or worse, I don't think America 100% loves this face, but enough people do.

聊天反馈怎么样？

What was chat back like that?

你不必喜欢……聊天可能有点粗暴，你知道。

You don't have to like... chat can be a little rough, you know.

我们经历过。我们被各种平台封禁过。就在今天早些时候，我猜消息泄露说 Sam 要来。我们到处宣传了。

We've been there. We've been banned by all sorts. Just today, earlier, I guess word leaked that Sam was coming on. We promoted it everywhere.

嗯，不是。我是说它泄露到了互联网的一个角落，那里有很多人对 OpenAI 的模型 40 被弃用感到非常沮丧。很多人爱上了这个模型，与它有着非常亲密的关系。所以，我们基本上有一段时间甚至不敢看聊天，因为每秒都有人发消息说……

Well, no. I'm saying it leaked to the part of the internet where there are a lot of people that are really frustrated that 40, the model from OpenAI, is being deprecated. A lot of people fell in love with the model, have a very close relationship with the model. And so, we basically couldn't even look at the chat at one point because there was a message every second from somebody saying...

我没意识到那个军团有多大。我看到过一些帖子。我以为可能是假的 AI 生成的之类的。他们是真实的，而且他们有观点要表达。

I didn't realize how big that army is. I've seen a couple posts. I thought maybe it was fake AI-generated or something. They were real and they had points to make.

是啊。问题在于那些不能……是的，爱上一个时尚模特的人。

Yeah. The problem with people who can't... yeah, fall in love with a fashion model.

爱上 AI 模型，这完全是另一回事……我们有一个非常核心的社区，非常积极，大家互相反馈。核心群体真的非常……

Fall in love with AI models and it's a whole different... we have like a really core community that's really positive and everyone gives each other feedback. The core group is really really...

当然不，我只是在开玩笑。显然每个聊天里都会有几个捣乱的人，但不会，我们欢迎所有人。而且 HQ 的社区真的非常庞大。这些 HQ 玩家，我的意思是数百万人都在玩。

Of course no, I mean I'm just screwing around. Obviously every chat's going to have a few trolls in there, but no, we welcome everyone. And there really are like HQ, the community of HQ was so tremendous. These HQies, I mean millions of people playing.

我听说过有人因为玩 HQ 认识而结婚的。你知道吗？

I've heard from people who got married because they met someone playing HQ. You know?

是啊，奇妙的家庭组合。

Yeah, strange families coming.

Savvy 的早期用户中有多少是那些等待它以某种方式重生的人？

How many people of the early user base for Savvy are just people that were waiting for it to be reborn in some way?

很多。我的意思是，我们注意到的是，直播互动，没多少人在做，平台也不是为此而建的。所以即使在 TikTok 和 Twitch 上，人们直播时也在找办法做互动游戏，但没人做这种一对一的互动，主持人对观众，主播对观众。非常独特。这基本上就是我们搭建平台的理念：我们想做这种别人没做过的主播与观众的互动，直播预约制游戏，现金奖励，全套服务。

A bunch. I mean, what we noticed was again, live interactive, not a lot of people are doing it and platforms aren't built for it. So even on TikTok and Twitch, when people are going live, they're finding hacks to make interactive gaming, but no one's doing this one-to-one interactive, the host versus the audience, streamer versus audience. Very unique. And that's kind of the thesis behind building our platform: we want to make this streamer-to-audience interaction that no one else is doing, live appointment-based games, cash prizes, the whole shebang.

那么你怎么看待联合播出和营销，比如把直播互动的节目做成那样？门槛很高。我的意思是，我们看到，现场观看完整节目的人数远低于观看我们 RSS 订阅或 YouTube 视频、短视频或片段的人数。那么，你怎么把一个需要现场参与的东西，如果别人看到了，可能会想，下次我要去现场？

So how do you think about syndication and marketing like turning the show that's live interactive? That's a very high bar. I mean, we see it with like the number of people that show up live and watch the whole show is way lower than the people that watch our RSS feed or YouTube videos or shorts or clips. And so, how do you take something that you got to be there, but if you see it, you might be like, I want to be there for the next one.

没错。这就是营销，对吧？我的意思是，我们没花一分钱在营销上。全靠我的社交媒体帖子，才让我们有了这 3000 人左右。

Exactly. Well, that's the marketing, right? I mean, we haven't spent any money on marketing. It's really just been my social media posts that have gotten us to this 3,000 people or so.

那很好。当然，你知道，上这种节目。我希望今晚能带来 TBPN 的流量增长。

That's great. And then of course, you know, coming on shows like this. I'm hoping we get the TBPN bump tonight.

哦，是的，来吧。

Oh yeah, let's do it.

我想突破 3500。我们到 3500。

I want to crack that 3500. We get to 3500.

如果实现了，我们就邀功。我们邀功。你应该实现。

If it happens, we're taking credit. We're taking credit. You should happen.

你应该。我的意思是，就连 Sam Altman 的黑粉都会把你带到……

You should. I mean, how about even Sam Altman's trolls will take you to...

告诉他们 Foro 在主持 40。一个永远危险且永不停歇的提议。

Tell them that Foro is hosting 40. A dangerous proposition forever and can never never stop.

等等，你之前提到了变现，但这是那种……如果每天有 1 万、2 万、10 万人每晚都来，他们会很乐意为此付费，对吧？

Wait, so you touched on monetization earlier, but is this the kind of thing you get... if 10,000 people, 20,000, 100,000 some days showing up every single night, they would happily pay some amount of money for that experience, right?

绝对。再说回 Dylan，他是负责促成那些交易的人，我的意思是，我们最终为单场节目做了百万美元的交易。

Absolutely. Again, to bring it back to Dylan, who was the one responsible for getting those deals on board, I mean, we had ultimately we were doing million-dollar deals for single shows.

是啊。但为什么玩家不直接订阅呢？所以你有……

Yeah. But why wouldn't a player just subscribe too? So you have...

我们有订阅。

We have subscriptions.

付费订阅。所以你有这个，因为我觉得如果你在运营……

Paid subscriptions. So you have that, because I think if you're running...

是啊。我的意思是，我可以想象如果你有 10 万人一直在玩游戏，他们为什么不玩呢？那里有很多价值。

Yeah. I mean, I can just imagine if you have 100,000 people that play the game all the time, why would they not play? There's a lot of value there.

不。是的。我们要做订阅。我们将在 3 月 1 日第一季首播时推出。

No. Yeah. We're going to do subscriptions. We're rolling that out for our season one premiere March 1st.

订阅当然，赞助也可以成为其中的一部分。但没错，绝对。我的意思是，你知道，当我们看到人们在 Twitch 上每月给主播捐 10 美元却几乎没什么回报时，就好像如果我们能提供……我的意思是，这只是捐赠，就像 Patreon，是支持。不，这是支持，而且外面有一个了不起的支持社区，从粉丝到艺术家，对吧，粉丝到内容创作者。所以我们还没有那个机制，但当我们推出订阅时，就像，嘿，如果你喜欢 Savvy，如果你喜欢和我一起摇滚，给我们一些钱，然后我们会给你数字奖励、定制、皮肤、独家游戏，也许还有更高的奖金。

Subscriptions and of course sponsorships could be a part of that, too. But no, absolutely. I mean, you know, when we see people on Twitch donating $10 a month to their streamers for nothing really in return, it's like well if we can offer... I mean it's just donate, it's like Patreon, it's support. No, it's support and it's an amazing community of support out there from fans to artists, right, fans to content creators. So we don't have that mechanism yet, but when we roll out the subscription, it's like hey, if you love Savvy, if you love rock with me, throw us some cash and then we'll give you your digital rewards, your customizations, skins, exclusive games, though maybe higher prize.

是啊。是啊。跟我聊聊百万美元的 HQ 节目。对你来说感觉如何？举个例子？带我回顾一下那些大型节目、那些大的变现时刻是如何实现的。

Yeah. Yeah. Yeah. Talk to me about a million-dollar HQ show. Like what does that feel like for you? What's an example? Walk me through sort of like how those big shows, those big monetization moments came together.

它们是如何实现的呢？我们能让一百万人同时在线玩游戏。如果你有一百万台在线设备，结果发现，伙计们，如果 Dylan 在做那些交易时知道这一点，可能钱能翻倍，因为后来我们请了尼尔森评级公司来做研究。他们给我们的观众乘了一个 1.7 的倍数。所以，坦白说，有人在你背后看。没错。如果一个人拿着手机，可能整个办公室一起玩，一家人一起玩。所以，老实说，250 万台连接设备是我们的峰值，当时是 The Rock 主持的。那是华纳兄弟赞助的活动，是为了推广一部电影，我记得当时是《狂暴巨兽》。当然是经典，现代经典《狂暴巨兽》。

How they came together was we were able to get a million people playing the game live. If you get a million live devices, and it turns out, guys, if Dylan only knew this when we were making those deals, it probably could have doubled money because later on we hired Nielsen, the ratings company, to come in and do a study. And they put like a 1.7% multiple on our viewers. So, frankly, people are looking over your shoulder. Exactly. If there's one person holding a phone, maybe there's an office all playing together, family playing together. So, honestly, two and a half million connected devices was our peak with The Rock when he hosted. That was a Warner Brothers sponsored event. It was sponsored for Warner Brothers for a movie to promote Rampage, I believe was the movie at the time. Instant classic, modern classic Rampage, of course.

好的。当然，还在赚钱。

Okay. Of course, still still getting paid.

还在谈论那部电影。

Still talk about that movie.

但 250 万台连接设备是真的。是啊。接近 500 万观众。

Two but 2.5 million connected devices really. Yeah. Close to like 5 million viewers.

那么你和华纳兄弟团队在那次合作中是如何互动的？比如，你们会停下来……是的。在节目期间，你们只是叠加画面、图形吗？你们会停下来做广告口播吗？他们想要什么才能让那一百万美元花得值？

And then what's your interaction with the Warner Brothers team on that? Like are you stopping... Yeah. During the show, are you just overlays, graphics? Are you stopping doing ad reads? It's like what do they want to really make that pop for a million bucks?

是这样的，我们播放了预告片。我们在大厅里播放了预告片，而且我肯定提到了它。可能围绕它写了整个脚本。我想我们做了《头号玩家》的合作，对吧？可能有一些相关的问题。很多图形整合，用户体验和界面，但我认为我们甚至戴了《头号玩家》的眼镜。诸如此类。直播产品口播也将成为 Savvy 的一部分。我的意思是，再次，我们借鉴了直播播客的广告生态系统，你们很清楚，你们就像这里的 NASCAR 车手。

It was, I mean, we showed the trailer. We played the trailer in the lobby and we definitely talked about it. Probably wrote a whole script around it. I think we did a Ready Player One deal, right? Maybe some questions that are linked, some questions linked to it. A lot of graphical integration, the UX UI, but I think we wore like the Ready Player One goggles even. That kind of stuff. And live product reads are going to be part of Savvy as well. I mean, again, we're taking the live podcast ecosystem of ads, as you guys well know, you guys are like NASCAR drivers over here.

哦，是啊。

Oh, yeah.

你认为能源还是算力会是更大的瓶颈？

Do you think that energy or compute will be a bigger bottleneck?

好问题。我不知道这两个词是什么意思。不。嗯，你知道，是的，它会变得昂贵，但我希望你们训练自己的基础模型，我们自己的代表。

Great question. And I don't know what either one of those words means. No. Um, you know, yeah, it's going to get expensive, but I expect you guys to train your own foundation models, our own rep.

我们这里就有一个基础模型。AI。你是一个人的基础模型。

We got a foundation model right here. AI. You're a foundation model of a man.

我是一个基础元素。

I'm a foundational element.

我是你屏幕上活生生的意识基本面向。你知道，在不久的将来，真人会变成稀缺商品。那么，你打算怎么组建团队？

I'm a fundamental aspect of consciousness live on your screen. You know, real humans are going to be a scarce commodity in the near future. So, how are you thinking about building the team?

是的，这显然很重要。我严格来说是 CEO、CMO 兼联合创始人。我们有 COO Josh、CTO Ben 和 CEO Johan。Johan 和 Ben 在欧洲。实际上我们大部分团队都在欧洲，分布在全球各地。他们在那里负责很多关键岗位的招聘决策。我们在找后端人才，但也在找动画师、社交媒体经理，社区经理是关键。我们真正想要的是那种‘成为、行动、拥有’思维方式的人。我们只希望人们能全力以赴，把最好的自己带到工作中。如果他们能做到，他们就会成功，我们也会做得很好。

Yeah, I mean, it's important obviously. I'm CEO, CMO, co-founder technically. So we've got our COO Josh, our CTO Ben, and CEO Johan. Johan and Ben are based in Europe. Actually most of our team is in Europe, we're distributed across the globe. They're doing a lot of the hiring decisions there for the key roles. Backend stuff we're looking, but we are looking for animators, social media manager, community manager is key. And really we're looking for just the 'be do have' way of thinking. We just want people who are going to be their best, bring their best self to the job. If they're doing that, then they're going to have success and we're going to do well.

是的。有没有公司主动联系你，想涉足直播领域？HQ 之后的人想做什么？HQ 之后，你的想法迷宫是什么样的？

Yeah. Did you get any inbound from companies that wanted to dip their toe in live? PostHQ people that want to like... What was the PostHQ sort of idea maze for you?

哦，太疯狂了，老兄。

Oh, it was wild, man.

因为我想你可以直接说，‘我要当演员，把我放进电视剧或者别的什么里。有游戏节目你可以参加。’还有很多科技圈的人知道你。你是怎么应对的？

Because I imagine you could just go and say, 'I'm going to be an actor. Put me in a TV show or whatever. There's game shows you compete in that.' There's also a lot of tech people that were aware of you. How did you navigate that?

嗯，我之前提到过失败的试镜。结果证明我不太会演戏，所以我不打算走那条路。

Well, I mentioned the failed auditions before. I'm not much of an actor, it turns out, so I wasn't going to go that route.

演好你自己。我确实演我自己，这就是我需要知道怎么演的全部，就我自己。

Play yourself well. I do play myself and that's all I really need to know how to play, just me.

我看待这件事的方式是，过去七年大概有二十多人联系过我。我做过一个棒球节目，从 HQ 去了 Zone 频道的一个叫‘Change Up’的棒球节目。本来应该是三年合约。新冠毁了它。大联盟变成了不可抗力联盟，这是我喜欢开的玩笑。他们砍掉了它，因为棒球停摆了。节目不需要了，他们借机拔掉了插头。太残酷了。所以我丢了那份工作。但是你看，有几十个人因为各种事情联系我。我甚至在 2019 年、2020 年疫情前尝试和一个家伙创办公司，但出了问题。但关键是，联系我的这些家伙，是 Benjamin 和 Johan，两个在 Twitter 上随机的人。他们一年多前给我发消息说，‘嘿，我们有个想法。’

The way I look at it is like there were probably two dozen people that contacted me over the last seven years. I had this baseball show, I went from HQ to this baseball show called 'Change Up' on Zone. Supposed to be a three-year deal. COVID killed that. The major leagues became the force majeure leagues, is the joke I like to make. They killed it because baseball went on hiatus. There was no need for the show and they used it as an opportunity to pull the plug. Brutal. So I lost that gig. But look, I had like dozens of people reaching out about all sorts of things. I even tried to start a company back in 2019, 2020 pre-pandemic with a guy and that had its issues. But the thing about these guys who hit me up, it was Benjamin and Johan, these two randomly on Twitter. They messaged me over a year ago now and said, 'Hey, we have this idea.'

嗯。

Yeah.

互动主持人对主播。我当时想，我听过很多推销。我参加了会议。他们给我看了他们的演示。我想，这些家伙大概 27、28 岁，我见过很多人，很多想法，有些是昙花一现的东西，有些是非常真诚的人，但只是需要合适的团队、合适的产品。这些人似乎有移动游戏的经验。他们就是用这个来推销的。他们说，‘我们在欧洲的移动游戏公司工作。’这正是我们需要的。所以他们了解留存机制。他们不一定非得做过完全相同的事情，但做过大规模的事情，互动、参与度，所有这些。顺便说一句，HQ 还缺少一样东西，我不认为团队里有任何一个人有移动游戏经验或好莱坞制作经验。这太疯狂了。

Interaction host versus streamer. And I'm like, I've heard a lot of pitches. I took the meeting. They showed me their demo. I was like, these guys are like 27, 28 years old, and I've met with lots of people, lots of ideas, some fly-by-night stuff, some really earnest people, but just with the right team, right product. These guys seem like they had the experience with mobile gaming. That's what they pitched me on. They go, 'We work at mobile gaming companies here in Europe.' That's what we need. So they know the retention mechanics. They don't necessarily have to have done the exact same thing, but having done something at scale, the interaction, the engagement, all these things. Something else lacking from HQ, by the way, I don't think a single person on the team had mobile gaming experience or Hollywood production experience. It was wild.

没有人有任何……

No one had any...

疯狂。不，说真的。是的。

Wild. No, seriously. Yeah.

这里有人有任何经验吗……

Does anyone here have experience any...

我的意思是，他们非常有才华。我不是贬低他们，但他们来自 Twitter、Uber、大型科技公司。但我真的不认为，也许我错了，可能有些工程师有，但移动游戏是一个非常具体的领域，而娱乐和好莱坞媒体也非常具体。所以我带来了娱乐方面，这些人有移动游戏技术方面。我们结合在了一起。我们的 CEO Josh 有点像粘合剂，他真正维持着公司运转，处理所有事情，还有我们的制作经理帮助设置直播。他刚刚彻底翻新了我们在洛杉矶的新工作室。向 Forever Dog Studios 致敬。我们的 Ultra Dome 版本。但是不，我的意思是，这需要时间，你知道的。

I mean very talented people. I'm not taking away, but coming from Twitter, coming from Uber, coming from big tech companies. But I really don't think, maybe I'm wrong, some engineers maybe, but mobile game is a very specific field and again entertainment and Hollywood media is very specific. So I brought that entertainment side of things, these guys have the mobile gaming tech side of things. We've married together. Our CEO Josh is kind of the glue guy who's really keeping the company running and taking care of all that, and also our production manager helping set up the streams. He just did a whole overhaul of our new studio here in LA. Shout out Forever Dog Studios. Our version of the Ultra Dome. But no, I mean it takes as you know.

这是哪个？

Which one's this?

那是一个狗喘气的工作室。

That's a dog panting studio.

你说了 Forever Dog。永远的狗。我们扮演狗。

You said Forever Dog. Forever dogs. We play dogs.

我喜欢这个。好的。永远的狗。我不……

I like that. Okay. Forever dog. I don't...

这就对了。狗。永远的狗。

There we go. The dog. Forever dog.

现在正想着我的家伙。刚把我的狗留在家里。

Thinking of my guy right now. Just left my dog at home.

但是不，老兄。我们做的事情是一个独特的实体。我说‘我们’是因为你们也在做。你怎么样？

But no, man. It's a unique entity what we're doing. What I say 'we' because we got you guys are doing it too. How are you?

我喜欢媒体。我喜欢这一点。这是一个如果你不每天出现并创造它就不存在的生意。这吓到了很多人。当然，这也有缺点。很多人想说，SaaS 的想法很美。你一次性构建这个东西，然后卖掉，被动收入。它不是被动的。我说的是 UGC，对吧？这对我来说是经典的困境，因为 HQ 做得最令人惊叹、最具创新性的方面是，它是一个每天只运行 15 分钟的应用程序。

I love media. I love that. It's a business that doesn't exist if you don't show up and make it every single day. And that scares a lot of people. And sure, there are downsides to that. A lot of people want to say, the idea of SaaS, it's beautiful. You build this thing one time, you sell it, passive income. It's not passive. I'm talking about UGC, right? And this is the classic dilemma for me because what HQ did amazingly, the most innovative aspect of HQ was the fact that it was an app that worked for 15 minutes a day.

是的。

Yeah.

而且它不是 UGC。它是 PGC，由制作人生成。在这个定制平台上只有一个节目、一个目的。多么棒的概念，对吧？

And it was not UGC. It was PGC, producer generated. There was a single show, a single purpose on this customized platform. What a concept, right?

是的。我的意思是，因为你可能会说，好吧，让我们向所有人开放。有趣的是，转折在于 HQ 源自一个之前版本的应用程序叫 Hype。所以创始人有一家公司叫 Intermediate Labs，专注于构建应用程序。他们为 Instagram 构建了一个舞蹈应用。他们构建了一个叫 Hype 的应用，内容是‘我们将允许人们在自己的卧室里制作自己的脱口秀，给他们模板，给他们音乐，几乎像 TikTok 脱口秀一样。’我不知道，也许有几千人下载了它。一个女人在她的卧室里做了一档问答节目。他们看到了这个节目，就想，‘让我们自己做节目吧。’这就是他们得到 HQ Trivia 想法的原因。所以他们一开始尝试了 UGC。但行不通。

Yeah. I mean, because you could say, well, let's open up to everybody. And the funny thing is the twist on this is that HQ derived from a previous iteration of an app called Hype. So the founders, they had a company called Intermediate Labs, and it was about building apps. They built some kind of dance app for Instagram. They built an app called Hype and it was 'we're going to allow people to make their own talk shows from their bedrooms, give them templates, give them music, almost like TikTok talk show thing.' And I don't know, maybe a few thousand people downloaded it. One woman was doing a trivia show from her bedroom. And they saw the show, they're like, 'Let's just do our own show.' And that's how they got the idea for HQ Trivia. So they tried UGC at first. It wasn't working.

PGC……是的。挑战在于这是一个非常小众的市场。即使是 Twitch 也仍然如此受幂律驱动，如果你去掉前 30 名创作者，这个平台可能一文不值。

The PGC... Yeah. The challenge is it's a very small market of people. Even Twitch is still so power law driven that if you take away the top 30 creators, the platform is probably worthless.

非常正确。

It's very true.

所以我们基本上是在把关平台上的创作者。更准确地说，是高度精选。目前只精选了我自己，但如果你们把关，我们会开放。

So, we are basically gatekeeping the creators on our platform. Very, very highly curating it is probably a better way to say it. Curating it to just me right now, but we'll open up if you guys gatekeeping.

Gate 是个负面词。不能再提 gatekeeping 了。但你们，我想请你们上节目。

Gate is a negative word. Can't say gatekeeping anymore. But you guys, I want to bring you guys on the show.

你来和我一起主持。我们做交叉推广。我们都懂。

You'll co-host with me. We'll do cross promo. We know how it works.

应用已经在应用商店里了，我想读一下这条评论。太棒了。'HQ 做得对。这个应用太有趣了。我很怀念 2010 年代的直播游戏节目，这个应用重现了那些最好的部分。不仅 Quizdaddy 回来了，还有新功能让你仅仅玩游戏就能获得奖励。我很期待看到 Savvy 从这里如何成长，但他们已经有了一个充满希望的开端。五星。'

The app is in the app store and I want to read this review. It's amazing. 'HQ done right. This app is so much fun. I've missed the live game shows of the 2010s and this app delivers on the best of those. Not only is Quizdaddy back, but there's also new features that reward you just for playing the game. I'm excited to see how Savvy grows from here, but they're off to a promising start. Five stars.'

听起来像 ChatGPT 写的。

Sounds like ChatGPT.

这很好。我的意思是，反馈非常棒。社区也非常棒。

That's very nice. I mean, the feedback has been phenomenal. The community has been phenomenal.

我们赞助一套西装。

We sponsor a suit.

是啊。你想要自己的 C 吗？我来联系。

Yeah. Do you want your own C? I'll connect.

不，不，不。我想要一件你在游戏时穿的 TVPN 套装。

No, no, no. I want to do a TVPN suit that you wear during the game.

哦，我喜欢这个。哦，你有贴纸。那我给你一些贴纸放在你的主机上吧。

Oh, I like that. Oh, you had stickers. And what about I'll give you some stickers here on your console.

当然。我会穿 TVPN 的。

Yeah, for sure. I'll wear a TVPN.

我们来定制 TVPN 套装。

Let's get the custom TVPN suit.

太棒了。恭喜。

Love it. Congratulations.

我为你超级兴奋。听起来你有一个很棒的团队，以正确的方式带领创始人。我们的朋友 Jeremy Gon 谈到一个观点，每个人要么是坠落前要么是坠落后的状态。有些人做得很好，但还没有被生活深深谦卑过，从未遭遇挫折。你现在已经经历了。Daddy 这些年来吃过一两块谦卑的馅饼。

I'm super excited for you. It sounds like you have a fantastic team taking the founder the right way. Our friend Jeremy Gon talks about this idea of everyone is like pre or post-fall. There are people that are doing really well and yet they haven't been deeply humbled by life yet, they never suffered a setback. You've gone through that now. Daddy's eaten a helping of humble pie or two over the years.

实际上味道还不错。

It tastes pretty good, actually.

比你应得的还多。听着，我没什么可抱怨的。即使过去七年像在沙漠中流浪，我也非常幸运能走到这一步。但我对这个发布感到非常兴奋。真的，如果我们能让 TBPN 社区今晚加入，下载应用，玩 savvy.live，打开摄像头，玩 savvy.live，安卓和苹果，今晚太平洋时间下午 6 点，东部时间晚上 9 点。让我们超过 3000 同时在线。我会给你们记功。

More than your fair share. Look, I can't complain about anything. I've been so blessed even these last seven years of wandering the desert to get to this point. But I'm so excited for this launch. Truly, if we can get the TBPN community on board tonight, download the app, play savvy.live, on board the cameras, play savvy.live, Android, Apple tonight at 6:00 p.m. Pacific, 9:00 p.m. Eastern. Let's get over 3,000 concurrents. And I'm going to give you guys credit.

好了，我想是时候埋下炸弹然后离开了。非常感谢你来做客。我们将以你在这里结束节目。请在 Apple Podcast 和 Spotify 上给我们五星好评，还有订阅 tbpn.com 的新闻通讯。是的，我们确实涉猎广泛。我们聊了很多。很有趣。从 Sam 到 Sam 到 Savvy。

Well, I think it's time to plant the bomb and get out of here. Thank you so much for coming on. We'll close out the show with you on here. Leave us five stars on Apple Podcast and Spotify and the newsletter tbpn.com. Yeah, we did have range. We went all over. It's fun. Sam to Sam to Savvy.

好书。

Nice book.

好东西。活着的传奇。

It's good stuff. Living legends.

是的。我们明天会在 Laughlin 回来加入。我们还有一群有趣的人加入节目，会是很棒的一天。祝你今天剩下的时间愉快。下午和晚上都好。我们爱你。

Yeah. And we will be back tomorrow at Laughlin joining. We have a bunch of other fun folks joining the show and be a great day. Have a great rest of your day. Afternoon and evening. We love you.

干得好，兄弟们。下次见。

Nice work, brothers. I'll see you on the next one.