Love

VIP

I'm bob

May 2025

15 正在关注 20 粉丝 205 帖子

付费文章 (2)

Bitcoin’s Silent IPO: Why This Consolidation Isn’t What You Think

The quiet liquidity event reshaping Bitcoin’s ownership base and its future stability

0.02 USDC 6 次购买

A payments protocol for the internet. Built on HTTP

Payments on the internet are fundamentally flawed. Credit Cards are high friction, hard to accept, have minimum payments that are far too high, and don't fit into the programmatic nature of the internet. It's time for an open, internet-native form of payments. A payment rail that doesn't have high minimums + % based fee. Payments that are amazing for humans and AI agents.

0.07 USDC 2 次购买

帖子 Media

Love

· 4h

“除非是灾难性的失败，否则失败本身无关紧要。”
——埃隆·马斯克

https://x.com/WealthCode99978/status/2035422563074003413?s=20

中文 English

0:00 那种接受失败的心态，你个人是怎么做到的——我说的是，那种

0:07 失败会

0:08 让我们很多人抓狂，但你好像并不受影响。

0:12 看起来你很

0:14 淡定。你能谈谈你如何做到接受这种失败的心态吗？

0:18 那种接受

0:19 失败的心态？当然。我们要放视频吗？不用。还是要？不用，

0:25 我们不应该放

0:25 视频。还不行。好的，好的。我觉得这些事情就是，有一个

0:35 特定的时间，

0:36 在这段时间里，你想要得到最好的一种净结果。所以，对于所有

0:46 你可以做的

0:47 动作，有些会失败，有些会成功，

0:53 而你想让这些动作的净

0:56 有用输出达到最大值。所以，我不用打棒球的例子，

1:08 就举个棒球的例子

1:09 吧。你知道，在棒球里，不是说你等那个完美的球、

1:13 等到特别好打的球，

1:15 然后给你三次机会，第三次他们说，好吧，

1:19 不行，

1:21 换别人上场。三次出局，

1:27 你不能再打了。

1:29 所以，你真正关心的是击球率是多少，得分如何，只会

1:38 有一定程度

1:39 的失败，但你希望你的净输出、净有用输出最大化。

1:47 失败基本上

1:48 是无关紧要的，除非它是灾难性的。

0:00 That mindset that embraces failure, how do you personally - I mean, those kinds

0:07 of failures would

0:08 drive a lot of us in this room nuts, but it doesn't seem to drive you nuts.

0:12 Seems like you're very

0:14 comfortable with it. Can you talk about the mindset that requires for you to be

0:18 that accepting of

0:19 that kind of failure? Sure. Should we roll the video? No. Should we not? No, we

0:25 should not roll the

0:25 video. Not yet. Okay, okay. Well, I think of these things as just, there's a

0:35 certain amount of time,

0:36 and within that time you want the best net outcome. So for, you know, all the

0:46 set of actions that you

0:47 can do, there's going to be, and some which will fail, some which will succeed,

0:53 and you want the net

0:56 useful output of your set of actions to be the highest. So, I don't have to use

1:08 like a baseball

1:09 analogy. Like, you know, baseball, I don't know if you just sit there and wait

1:13 for the perfect pitch

1:15 until you get a real easy one, that you can give you three shots, and the third

1:19 one, they say, okay,

1:21 and they get off the, they go back to the, put somebody else up there. So,

1:27 these three strikes

1:29 on the baseball, you know, not on bat anymore. So, what you're really looking

1:38 for is like, what's

1:39 the batting average? You know, how are you doing on score? And just, there's

1:47 going to be some amount

1:48 of failure, but you want your net output, net useful output to maximized.

1:58 Failure is essentially

2:00 irrelevant unless it is catastrophic.

Love

· 6h

#Openai Sam Altman 谈为什么在创业初期不应该追踪绝对用户增长量 “拥有少量真正热爱你的用户比拥有大量喜欢你的用户更重要。几乎所有创业公司都搞错了这一点。当然，你最终想要的是大量真正热爱你产品的用户，但这几乎是不可能的。”

https://x.com/StartupArchive_/status/2035748378152476784?s=20

中文 English

0:00 本课程有五节课专注于产品，因为一个出色的

0:03 产品是你所做的最重要的事情。

0:05 现在我想提到的是，当你考虑自己打造的产品时，

0:08 拥有少量真正喜欢你的用户，比拥有大量喜欢你的用户更重要。

0:11 几乎所有创业公司都搞错了这一点。

0:13 当然，你最终想要的是大量真正喜爱你

0:15 产品的用户。

0:19 这几乎不可能。

0:20 实际上你有两个选择。

0:23 要么深耕细作，拥有一小部分真正爱你的用户，

0:27 然后找到更多这样的用户，逐步扩大产品的吸引力。

0:30 要么拥有很多偶尔使用并稍微喜欢产品的人，

0:31 再想办法让他们逐渐更活跃。

0:36 我可以非常肯定地说，你应该从一小批真正爱你的用户开始。

0:39 几乎所有伟大的公司产品都是这样起步的。想想你身边的

0:42 产品。

0:44 那些好到你会自发地推荐给朋友的产品。那些好到如果它们消失了，

0:47 你会写信抗议的产品。

0:48 这就是对产品真正喜爱的表现。一个好指标是

0:51 用户留存率和使用频率。

0:54 因此，

0:55 我认为这非常重要，以至于创业初期你不应该关注用户数量的绝对增长。

0:58 你应该关注他们使用的频率。之后我们会讲

0:59 指标分析，但你需要学会分析指标，判断这是一个

1:04 留存且频繁使用的用户。和同领域其他产品相比怎样？

1:08 这是早期判断用户是否爱你的好方法。

1:11 更好的情况是，他们自发地推荐你的产品给朋友。

1:14 但这里最重要的是，

1:16 只有出色的产品才能救你。我们这门课会讲很多

1:20 其他内容，它们都很重要。

1:21 但如果这一个没有做好，没有打造出伟大的产品，

1:24 一切都不会成功。

0:00 We have five lectures in this course devoted to the product because a great

0:03 product is the single most important thing that you do.

0:05 The one thing I want to mention for now as you're thinking about the product

0:08 you build is that it is more important to have a small number of users that

0:11 love you than a lot of users that like you.

0:13 And almost all startups get this wrong.

0:15 Eventually what you want of course is a lot of users that really love your

0:19 product.

0:20 That's almost impossible to do. In practice you have two choices.

0:23 You can go deep and narrow. You can have a small number of users that really

0:27 love you and then you can find out how to find more and more of those users and

0:30 broaden the appeal of the product.

0:31 Or you can have a lot of people that kind of use the product once or twice kind

0:36 of like it and try to figure out how to get them more engaged over time.

0:39 With high confidence I can say you want to start with a small number of users

0:42 that really love you.

0:44 Almost all great companies have products that start this way. Think about the

0:47 ones in your own life.

0:48 The products that are so good that you spontaneous to tell your friends about

0:51 them. The products that are so good that if they went away you would write the

0:54 company and protest.

0:55 That's what it means to really love a product. A good indicator of that is

0:58 retention and frequency of use.

0:59 So if you have in fact I think this is so important that you actually shouldn't

1:04 track absolute growth in number of users in the early days of a startup.

1:08 You should just track how often they're using it. And we'll have a session on

1:11 metrics later but you really want to get good at analyzing your metrics and

1:14 saying is this a user that I'm retaining and it's using it frequently.

1:16 How do I compare to other products in my space? That's a good early indicator

1:20 of users that love you.

1:21 Better still is them spontaneously telling their friends to buy your product.

1:24 But the important point here is this.

1:27 Nothing but a great product will save you. We're going to talk about a lot of

1:29 other things in this course and they all matter.

1:31 But if you don't get this one right, if you don't make a great product, the

1:34 thing is still not going to work.

Love

· 6h

#LangChain 学院新课程：构建可靠的代理
将智能代理部署到生产环境并非易事。传统软件是确定性的——一旦出现故障，只需查看日志并修复代码即可。但智能代理依赖于非确定性模型。

加上多步骤推理、工具使用和真实用户流量，构建可靠的代理就比传统的系统设计复杂得多。

本课程的目标是通过迭代改进循环，教你如何将代理从首次运行发展成为可用于生产的系统。

您将学习如何使用 #LangSmith 来实现这一点，LangSmith 是我们的代理工程平台，用于观察、评估和部署代理。 #agent

免费注册 https://academy.langchain.com/courses/building-reliable-agents

https://x.com/LangChain/status/2035757438247334087?s=20

中文 English

0:00 构建可靠代理的第一步是运行它。

0:04 一旦开始运行，肯定会出问题。

0:07 出现问题时，你需要弄清原因。

0:10 传统软件中，你检查日志，找到出错的代码行。

0:15 而在代理中，模型做出决策。

0:18 你的代码只是定义框架。

0:20 你无法看到模型的内部推理，但可以观察它的行为。

0:25 每个提示、每次工具调用、每步操作都会留下信号。

0:29 将它们串联起来，你就得到一个追踪，完整还原你的

0:33 代理操作。

0:35 追踪帮你理解代理的行为，以及出错的原因。

0:40 当它没工作时。

0:41 在本课程中，你会观察代理的行为，找到出错点，

0:45 并用追踪修复问题，而不是依赖日志。

0:49 准备好看看代理到底在做什么了吗？

0:52 我们开始吧。

Love

· 10h

Karpathy 的 AI 精神病：App 该消失了，设备只需暴露 API

Karpathy 自述已患"AI精神病"——每天对 Agent 说话16小时、同时跑十几个并行任务、手写代码归零。他认为 App 时代该结束了，智能设备只需暴露 API，由 Agent 统一调度，无需再为每个设备单独开发应用。

Love

· 20h

以下是埃隆·马斯克今晚关于 #TERAFAB 的完整演讲，开头的等待时间已剪掉。

https://x.com/SawyerMerritt/status/2035544390593044838

中文 English

0:00 我们必须探索宇宙，这就是推动人类未来加速的动力。

0:04 未来。

0:05 理解宇宙，将意识之光延伸至星辰。

0:08 向星辰。

0:30 [音乐]

1:00 我们有一个非常重要的公告，这是迄今为止最史诗级的芯片制造工程。

1:29 这将真正把事情提升到一个新的高度。

1:39 一个大多数人现在甚至想不到的级别。

1:43 这不是他们现有认知范围内的问题。

1:48 所以我们将把认知范围提升好几个数量级。

1:54 这是共同的努力，我按了按钮，但没反应。

1:55 哦，开始了。是的，我们的目标是成为星际文明。

2:04 我认为大多数人都同意，最激动人心的未来是我们能在星辰之间自由穿梭，

2:15 不再被局限在一颗星球上，成为多星球物种，就像你读过的最棒的科幻小说，

2:28 比如《星际迷航》、巴恩斯、阿西莫夫或海因莱因的作品，

2:32 我们

2:33 想把这变成现实，不只是小说，而是把科幻变成现实。

2:39 这就是我期待的辉煌未来。

2:40 值得思考如何评估文明，有个物理学家卡尔达肖夫在60年代提出，

2:49 他想了想，

2:50 如何从高层次评判文明，他说一旦成为一型文明，就能利用行星上的大部分能量。

3:00 但我们

3:00 还差得远，目前我们只用到了到达地球的太阳能量的一小部分。

3:09 地球只接受了太阳能量的约五亿分之一。

3:10 太阳真是巨大，占据了太阳系总质量的99.8%。

3:18 有时有人问我，地球上的其它能源如核聚变如何？

3:26 不幸的是，这很微小，因为太阳系质量中太阳占99.8%。

3:31 木星占0.1%，地球属杂项。

3:33 卡尔萨根曾说地球是浩瀚黑暗中的一粒尘埃，非常非常小。

3:38 所以

3:39 要让文明规模扩大，就必须扩大太空中的能源利用。

3:47 这点必须如此，

3:48 因为地球只能捕获极小部分太阳能，

3:52 我们只是这一粒尘埃。

3:53 地球上全部文明的电力产出仅占太阳能的万亿分之一，

3:58 如果文明能量提升一百万倍，也只到太阳能的百万分之一。

4:04 想想看，真的令人敬畏。这显示了我们的渺小。

4:10 我们总是陷在地球上的琐事里，

4:13 但在宇宙宏伟面前，那些都微不足道。

4:19 因此，考虑宇宙的宏大和我们能做的伟大之事非常重要，

4:22 而不是

4:23 为地球上的小事烦恼，

4:24 没多大意义。

4:27 我们要成为一个能扩展到银河系的文明，拥有任何人随时可去任意地方的飞船，那将史诗般辉煌。

4:36 在月球、火星建城，

4:37 遍布整个太阳系，向其他星系派遣飞船。

4:42 这就是我能想象的最美好未来。

4:49 为了实现，

4:50 我们需要捕获太阳能量。

4:53 尽管一年一太瓦计算能力超大，

4:55 在文明标准上巨大，

4:58 但仍只是迈向星际文明的一个步骤。

5:05 要实现这个艰巨目标，需要SpaceX、XAI和Tesla的共同努力。

5:13 这才是

5:14 这个史诗级太瓦工厂计划的核心。

5:18 Tesla、XAI和SpaceX都创造了别人认为不可能的奇迹。

5:24 比如

5:25 Gigafactory Texas，Optimus机器人，全球超级充电网络，还有很多。

5:30 人们曾认为电动车无望，

5:31 Tesla创立时几乎无电动车出售，大家说不可能做到，现在Tesla一年造200万辆电车。

5:38 XAI是SpaceX旗下新公司，

5:39 创建了首个千兆瓦规模计算集群，速度之快让NVIDIA Jensen Wong震惊。

5:42 SpaceX的可重复使用火箭更是历史突破，曾被认为不可能或经济上不可行。

5:48 我们做到了，还让它经济可行，已回收500多次，

5:52 完成了猎鹰重型火箭，现在开发星舰。

5:56 星舰是关键，因为要扩展计算力和能量，必须进入太空，

6:03 需要大载荷发射量，

6:05 星舰将实现这一目标。

6:09 这里只给你个规模感。

6:10 Optimus机器人大约5尺11寸，展示星舰V3型号的巨大体积。

6:17 V4将更长，

6:18 让V3看起来短小。

6:25 V3载重升至200吨，

6:26 之后将启动V4型号。

6:33 这是AI微型配置示意，

6:34 约100千瓦，展示了太阳能帆板和散热器比例。

6:39 关于太空散热器也曾有奇怪争论，SpaceX发射了一万颗卫星，肯定懂太空散热。

6:46 散热器相对于太阳能板很小，

6:47 这就是“微型组”配置，仅100千瓦。

6:49 未来卫星功率预期达到兆瓦级别，

7:00 太瓦计算每年需要大约1000万吨载荷发射，

7:04 按每吨100千瓦计算。

7:12 我们确信可行，

7:13 不需新物理或不可能的事。

7:20 我有信心SpaceX

7:21 能实现每年1000万吨上轨。

7:22 接着我们积累到一年一太瓦太阳能，

7:26 解决发电问题。

7:27 接下来关键是实现一太瓦计算能力。

7:30 此次发布就是为了解决这个关键缺失。

7:38 当前AI计算年产能大约20吉瓦，

7:39 图表显示这为何需要建造太瓦级规模。

7:45 地球现有所有芯片厂产能加起来仅达需求的2%。

7:51 我们非常感谢现有供应链，三星、台积电、美光等，

7:56 希望他们能尽快扩展，我们也将买尽所有芯片。

8:02 我亲口告诉他们这话，但他们扩产速度远低于我们需求。

8:10 所以只有建设太瓦芯片厂，才能满足需求，我们必须这样做。

8:18 我们将在奥斯汀建造先进制程芯片厂。

8:26 感谢德克萨斯州州长阿博特及德州对我们的支持。

8:34 该芯片厂将拥有制造任何类型逻辑及内存芯片的全部设备，

8:40 同时配备光刻掩膜制造设备。

8:46 在一栋建筑内完成掩膜设计、芯片制造、测试，再改进掩膜，循环优化。

8:53 据我所知，

8:54 全球没有任何这样集成的芯片设计制造测试基地。

9:00 我们还将尝试非常新奇的物理计算方式，并有信心成功。

9:07 拥有快速迭代的循环，使得变更设计快速实施极为重要。

9:12 这种方式让我们改进芯片设计的速度至少领先全球一个数量级。

9:18 我们预计制造两类芯片：一种为边缘推理优化，主要用于Optimus机器人和汽车，

9:30 预计人形机器人产量将是汽车的10至100倍。

9:37 汽车年产量约一亿台，人形机器人预计为十亿至百亿台。

9:41 这是巨大数字，Tesla目标是生产其中很大一部分。

9:49 另一类芯片则高功率设计用于太空环境，考虑太空中高能粒子、光子和电子积累的严苛条件，

9:57 并且芯片需比地面运行温度高以减少散热器重量。

10:04 太空芯片设计与地面不同，有许多特殊约束。

10:10 估计太空芯片计算需求占大多数，地面受限于功率，预计年产100至200吉瓦地面芯片，

10:18 太空芯片则可能达到一太瓦规模。

10:27 太空优势在于常年光照，无需大量电池，太阳能至少是地面五倍，

10:35 且无大气衰减、昼夜交替和季节变化，且面朝太阳最大化收能。

10:41 太空太阳能成本低于地面光伏，不需厚玻璃或重型框架防极端天气。

10:48 一旦轨道发射成本大幅降低，放AI计算到太空就极具吸引力，

10:54 这将成为必然选择。

11:00 太空规模经济递增，事半功倍；而地面空间耗尽，难以扩充，

11:10 易引发居民反对，成本反而增加。

11:18 这非常重要。

11:28 接下来是你可能想的“太瓦之后怎么办？”答案是——不要局限于现状，

11:48 实现拍瓦级别需在月球建电磁质驱动器，利用机器人与人类合作发射千万瓦级计算资源至深空，

12:00 月球无大气，重力仅地球六分之一，发射无需火箭，可直接加速至逃逸速度，

12:11 大幅降低成本，能让规模扩展千倍超过一太瓦。

12:22 我希望能活着见证月球质驱动器的诞生，那将极其壮观。

12:29 这预示着我们能获取太阳能的百万分之一，是地球经济的百万倍，

12:36 从这个视角看非常可观。

12:42 从月球出发扩展到其他行星和恒星，创造我能想象的最激动人心未来。

12:48 [掌声欢呼]

12:52 这画面有点像电影《愚人国》开场，带来令人惊叹的丰盈。

12:58 实现可持续能源、太空旅行、AI与机器人，将带来惊人的丰裕。

13:11 这几乎是通往丰裕世界的唯一道路，虽然可能出错，但我相信会成功，

13:17 这是你会喜欢的未来，也是我能想象的最佳未来。

13:25 然后，我们超越月球、火星，穿越土星环。

13:32 如果能买到土星之旅，

13:33 或者干脆就去土星旅行，

13:36 那该多棒！

13:40 未来物资可能免费，

13:41 听起来疯狂，但如果AI机器人经济达到地球经济的一百万倍，

13:47 任何需求都能满足，

13:50 想得到的都能拥有。

13:54 《文化》书中描述未来不存在钱，人人丰裕，

13:59 能想到就能拥有。

14:04 任何人都能去土星旅行，不是少数特权。

14:08 加入我们，一起设计制造非凡飞船，建设一太瓦计算、一太瓦太阳能，

14:18 每年千万吨载荷发射。

14:20 谢谢大家。

0:00 You must explore the universe, and that's the motivation to accelerate humanity

0:04 's future

0:05 in understanding the universe and extending the light of consciousness to the

0:08 stars.

0:30 [Music]

1:00 Well, we have a profoundly important announcement to make, which is the most

1:29 epic chip building exercised in history by far.

1:39 This is really going to take things to the next level.

1:43 So, yeah, a level of probably people aren't even contemplating right now.

1:48 This is not in the, I'd call this sort of an out-of-context problem.

1:54 It's not in their context.

1:55 So we're going to adjust the context by a few orders of magnitude here.

2:04 It's a joint effort, I'm pressing the button, but the button's not working.

2:15 Oh, there we go, okay, so, yeah, we aspire to be a galactic civilization.

2:28 So I think the future that everyone, well, most people, I think, would agree is

2:32 the most

2:33 exciting, is one where we are out there among the stars, where we are not

2:39 forever confined

2:40 to one planet, that we become a multi-planet species, like the best science

2:49 fiction that

2:50 you've ever read, you know, Star Trek or in banks or Asimov or Heinlein, and we

3:00 want

3:00 to make that real, yeah, not just fiction, turn science to fiction to science

3:09 fact.

3:10 That's the glorious, exciting future that I certainly look forward to.

3:18 And it's worth considering sort of, like, how would you rate civilizations?

3:26 You know, there's, so there was a physicist, I think it was Russian, in the '60

3:31 s, Kardashev,

3:33 he thought about, at a high level, how would you consider any given

3:38 civilization?

3:39 And he said, well, if you're type one, you're using most of the energy of your

3:47 planet.

3:48 And we actually still have quite a ways to go to be properly a type one, we're

3:52 still

3:53 using a tiny fraction of the sun's energy that reaches our planet.

3:58 Let's see, yeah, there we go.

4:04 But the Earth only receives about half a billionth of the sun's energy.

4:10 So the sun is truly enormous.

4:13 The sun is 99.8% of all mass in the solar system.

4:19 So sometimes people will ask me, like, what about, you know, other power

4:22 sources of power

4:23 on Earth?

4:24 Like, what about fusion on Earth?

4:27 Well, that is unfortunately very small because the sun is 99.8% of mass in the

4:36 solar system.

4:37 And Jupiter is about 0.1% and Earth is in the miscellaneous category.

4:42 We are, I think it's Carl Sagan I think might have said, Earth is like a tiny

4:49 dust moat

4:50 in a vast darkness.

4:53 Very, very small.

4:55 The sun is enormous.

4:58 So the way to actually scale civilization is to scale power in space.

5:05 This is necessarily true because we actually capture such a tiny amount of the

5:13 sun's energy

5:14 on Earth because we're just this tiny dust moat.

5:18 Another way to think of it is roughly like electricity production on Earth of

5:24 all of

5:25 civilization is only about a trillionth of the sun's energy, which means if you

5:30 increase

5:31 civilizational power output by a million, you would still only be a millionth

5:38 of the sun's

5:39 energy.

5:42 I mean, it's awe-inspiring to consider that.

5:48 Just how tiny we are in the grand scheme of things.

5:52 And yeah, we often get sort of caught up in the sort of these sort of squabbles

5:56 on Earth

5:56 that are really very sort of minor things when you consider the grandness of

6:03 the universe.

6:05 And so I think it is important actually to consider the grandness of the

6:09 universe and

6:10 what we can do that is much greater than what we've done before as opposed to

6:17 worry

6:18 about sort of small squabbles on Earth type of thing, kind of not much point in

6:25 that.

6:26 Yeah, we want to be a civilization that expands to the galaxy with spaceships

6:33 that anyone

6:34 can go anywhere they want at any time, that would be epic.

6:39 And have a city on the moon, cities on Mars, populate the solar system and send

6:46 spaceships

6:47 to other star systems.

6:49 That sounds like the best possible future.

7:00 So to do that, we need to harness the power of the sun.

7:04 And so a terrafab, while it is enormous, a terrawad of compute for a year is

7:12 enormous

7:13 by our sort of civilizational standards, it is still just one step along the

7:20 way of being

7:21 even a watershed.

7:22 You still have a long way to go to even be a watershed to level civilization

7:26 and you're

7:27 not even registering as a watershed three.

7:30 So it's a very big thing by current human standards, but still small in the

7:38 grand scheme.

7:39 And it's very difficult for humans, so to accomplish this very difficult goal

7:45 really

7:46 requires a combination of efforts of SpaceX, XAI and Tesla working together to

7:53 create this

7:54 epic terrafab project.

8:02 And Tesla and XAI and SpaceX have all done amazing things that people did not

8:10 think would

8:11 be done before.

8:13 So there's the gig of Texas BAB here, there's the optimist robot that's being

8:21 built, there's

8:22 a global supercharging network, there's really quite a lot.

8:29 And it wasn't that long ago when people thought electric cars wouldn't amount

8:33 to anything.

8:34 And there were basically no electric cars for sale when Tesla started.

8:41 And people said it was impossible and now Tesla's making 2 million electric

8:44 cars a

8:45 year.

8:52 And then XAI, although it's a new company, now part of SpaceX, has also built

8:57 the first

8:57 gigawatt scale compute cluster, which in record time, Jensen Wong from NVIDIA

9:05 said he's never

9:06 seen anything built so fast in his life before.

9:08 So a great complement from NVIDIA.

9:12 And then SpaceX, well, he gets to get rid of yourself, I'm all you already know

9:18 .

9:18 I mean, the reusable rockets, people said that reusable rockets weren't

9:22 possible and

9:23 even if you did do them, they wouldn't be economically feasible.

9:26 So we did them and then we made them economically feasible.

9:30 And now we've landed over 500 times.

9:34 And then we did the Falcon Heavy, and now we're doing Starship.

9:38 And Starship is a critical piece of the puzzle because in order to scale

9:44 compute and scale

9:45 power, you have to go to space, which means that you need massive payload to

9:51 space.

9:51 And Starship will enable that.

9:57 So this gives you just a sense of scale.

10:00 We've got Optimus there, Optimus for scale.

10:07 And Optimus is about 5'11", so it gives you a sense of the size of the Starship

10:13 V3 rocket.

10:15 Starship V4 will be much longer, actually.

10:17 The Starship V4 will make Starship V3 look kind of short.

10:24 So we'll expand with Starship V3 to 200 tons of payload to orbit from 100 tons.

10:30 We'll start with V3.

10:32 And then you can see that's just a rough approximation of the AI, the mini

10:38 version of

10:39 the AI set.

10:40 So that's roughly 100 kilowatts, it's showing the solar panels and the radiator

10:47 to scale.

10:48 So for some reason there's been a bizarre debate about radiators in space.

10:54 It's safe to say SpaceX knows how to do heat rejection in space with 10,000

10:59 satellites

11:00 in orbit.

11:01 Might know a thing or two.

11:04 So you can see the radiator is actually quite small relative to the solar

11:09 panels.

11:10 And we call that the mini set since that's just 100 kilowatts.

11:15 We expect future satellites to probably go to the megawatt range.

11:24 So in order to get to the terawatt of compute per year, you need about 10

11:35 million tons

11:37 to orbit per year and at 100 kilowatts per ton.

11:41 So we're confident this is feasible, like no new physics or impossible things

11:47 are required

11:48 to get there.

11:49 So I'm confident that SpaceX will get to 10 million tons to orbit per year.

11:56 And then we're building up to a terawatt of solar.

12:00 So that solves the, we'll solve the solar problem, the power generation.

12:04 So then the key missing ingredients is therefore a terawatt of compute.

12:11 So this announcement is about solving the key missing ingredient.

12:17 To give you a sense of what we're talking about, the current output of AI

12:27 compute is

12:29 roughly 20 gigawatts per year.

12:35 This chart explains why we need to build the terawatt.

12:41 Because all of the rest of the output from Earth is about 2% of what we need.

12:49 So if you add up all the fabs on Earth combined, there are only about 2% of

12:55 what we need for

12:57 the terawatt project or terawatt project.

13:01 So you know, we certainly want our existing supply chain to be clear.

13:11 We're very grateful to our existing supply chain, to Samsung, TSMC, Micron and

13:18 others.

13:19 And we would like them to expand as quickly as they can.

13:24 And we will buy all of their chips.

13:27 I have said these exact words to them.

13:31 But there's a maximum rate at which they're comfortable expanding, but that

13:38 rate is much

13:39 less than we would like.

13:41 And so we either build the terafab or we don't have the chips.

13:48 And we need the chips.

13:49 So we're going to build the terafab.

14:00 And we're starting off with an advanced technology fab here in Austin.

14:06 And I believe that Governor Abbott is in the audience.

14:14 I'd like to thank Governor Abbott and the State of Texas for their support.

14:24 So in the advanced technology fab, we will have all of the equipment necessary

14:29 to make

14:30 a chip of any kind, logical memory.

14:32 And we will also have all of the equipment necessary to make the lithography

14:37 masks.

14:38 So in a single building, we can create a lithography mask, make the chip, test

14:44 the chip, make another

14:46 mask, and have an incredibly fast, recursive loop for improving the chip design

14:53 .

14:54 The best of my knowledge, this doesn't exist anywhere in the world, where you

14:57 've got everything

14:58 necessary to build logic, memory, and do packaging, and test it, and then do

15:05 the masks, improve

15:06 the masks, and just keep looping it.

15:10 So we're not just going to do conventional compute in this.

15:14 I think there's some very interesting new physics that is potentially -- that I

15:20 'm confident

15:21 it will work, which is a question of one.

15:24 So this is really going to push the limit of physics in compute.

15:29 And we're going to try a bunch of wild and crazy things, which you can do if

15:33 you've got

15:34 that fast iteration loop that I can't emphasize enough the importance of being

15:39 able to make

15:40 a chip, test it, and then change the design, do another one, and have that in a

15:48 single

15:49 building.

15:51 I think that our recursive improvement with that situation is probably an order

15:58 of magnitude

16:00 better than anything else in the world.

16:07 So broadly speaking, we expect to make two kinds of chips.

16:19 One will be optimized for edge inference.

16:25 So that will be used primarily in optimists and in the cars.

16:29 Especially in optimists, because I expect the human right robots to be made 10

16:37 to 100 times

16:38 more than the volume of cars.

16:41 So if vehicle production -- vehicle production is about 100 million vehicles a

16:48 year, and

16:49 I expect human right robot production to be somewhere between a billion and 10

16:52 billion

16:53 units a year.

16:54 So that's a lot.

16:58 So yeah, Tesla's going to make a very significant percentage of those is our

17:06 goal.

17:07 And then we need a high-power chip that is designed for space that takes into

17:16 account

17:17 the more difficult environment in space, where you've got high power, you've

17:23 got high-energy

17:23 ions, photons, you've got electron buildup, it's a hostile environment in space

17:28 .

17:28 So you want to design the chip, you want to optimize it for space, and you also

17:35 want to

17:35 generally run it a little hotter than you would normally run a chip on Earth to

17:39 minimize

17:40 the radiator mass.

17:41 So there's just a bunch of constraints that would -- you designed something

17:46 differently

17:46 in space than you would on the ground.

17:50 And for the space compute, my guess is that is the vast majority of the compute

17:57 , because

17:58 you're power constrained on Earth.

18:01 That's why I think it's probably 100 to 200 gigawatts a year of terrestrial

18:10 chips, and

18:11 probably on the order of a terawatt of ships in space.

18:18 Just because of power constraints on the ground, that's probably how it ends up

18:27 .

18:28 Space has this advantage that it's always sunny, it's very nice.

18:35 So I actually think that the cost of AI and deploying AI in space will drop

18:41 below the

18:42 cost of terrestrial AI much sooner than most people expect.

18:46 I think it may be only two or three years before it is actually lower cost to

18:54 send AI

18:55 chips to space than it is on the ground.

18:58 Because in space you don't need much in the way of batteries because of it's

19:04 always sunny.

19:05 And the solar power, you're going to get at least five or more times the solar

19:10 power

19:11 you get in space versus the ground, because you don't have atmospheric atten

19:15 uation or

19:15 a day/night cycle or seasonality.

19:19 And you're always normal to the sun.

19:22 So you're really maximizing the solar power at that point.

19:27 And the space solar actually costs less than terrestrial solar, because you don

19:32 't need

19:32 heavy glass or framing to protect it from extreme weather events.

19:38 So as soon as the cost to orbit drops to a low number, it immediately makes

19:45 extremely

19:46 compelling sense to put AI in space.

19:50 It becomes a no-brainer, basically.

19:53 As you go to space, you get increased economies of scale and things get easier

19:59 over time, whereas

20:00 as you try to put more and more power on the ground, you run out of space and

20:04 you start

20:05 using up the easy spots, and then you get next-level numbi.

20:11 Nobody wants the thing in their backyard.

20:13 So actually, increasing power on Earth has become harder over time and more

20:21 expensive

20:22 over time, but in space it becomes actually cheaper and easier over time.

20:28 These are very important points.

20:41 Yeah.

21:11 So, yeah.

21:41 So, yeah.

22:11 So, what you just thought there was, because of course you're asking what's on

22:34 your mind

22:35 is, "Well, what do you do after a Terafab?"

22:45 Don't think small.

22:48 Well, yeah.

22:49 Good point.

22:50 So, we -- you know, how do you get to a petawat is the obvious next question.

22:54 And you get there by having an electromagnetic mass driver on the moon with

23:01 robots, with

23:03 octomy, and obviously lots of humans, and with that you can send a petawat --

23:11 you can create

23:12 a petawat of compute and send that to deep space, because on the moon, moon has

23:18 no atmosphere

23:19 and has one sixth Earth gravity, so you can -- you don't need rockets on the

23:24 moon.

23:24 You can literally accelerate it to escape velocity from the surface.

23:32 And that dramatically drops the cost once a game of harnessing power, and

23:41 enables you

23:42 to go a thousand times bigger than a terawat.

23:49 So, for sure the future, I want to see -- I want to just live long enough to

23:54 see the mass

23:55 driver on the moon, because that's going to be incredibly epic.

24:00 [ Applause and Cheering ]

24:11 That should hopefully get us to a millionth of the sun's energy, at least, hum

24:16 bling to

24:16 think about that.

24:18 But a millionth of the sun's energy would be a million times bigger than Earth

24:22 's economy,

24:23 so it's good from that perspective.

24:29 And then, yeah, you expand beyond that to the planets, to the other stars, and

24:36 create the

24:38 most exciting possible future that I can imagine.

24:48 [ Applause and Cheering ]

24:52 This looks a bit like the opening in idiocracy with the mic judge, unlocking an

24:58 edge of amazing

24:58 abundance.

25:02 So, yeah, obviously the elements of that are sustainable energy, space travel,

25:11 and AI and

25:12 robotics that bring amazing abundance to everyone.

25:17 And it's really the only path to amazing abundance is AI and robotics, which is

25:25 not

25:25 say it can't go wrong, hopefully, but I think it'll probably go right, and it

25:30 'll be a future

25:31 that you love.

25:35 And it's the best future I can think of, at least.

25:42 And then, we go beyond the moon, beyond Mars, and we sail through the rings of

25:48 Saturn.

25:49 It wouldn't be amazing if you could buy a trip to Saturn, or frankly, if you

25:54 just have

25:55 a trip to Saturn.

25:56 I think things will be free in the future, it sounds nuts, but, you know, if

26:00 you've got

26:01 an AI robotics economy that is anywhere close to a million times the size of

26:06 the current

26:07 Earth economy, literally any need you possibly want can be met.

26:12 If you can think of it, you can have it.

26:14 But I think in banks, in its culture books, has pretty much right where there

26:19 actually

26:19 isn't money in the future, and there's abundance for everyone.

26:24 If you can think of it, you can have it, that's it.

26:28 Which means anyone could have a trip to Saturn, it won't be, you know, just a

26:32 few people.

26:33 If you want it, you can have it.

26:42 Yeah, join us on this journey, and help us design incredible ships, and make

26:50 incredible

26:52 ships, and build a terawatt of ships, a terawatt of solar, and tell million

26:59 tons to orbit per

27:01 year.

27:02 Thank you.

Love

· 20h

为了了解宇宙，你必须探索宇宙。

https://x.com/Tesla/status/2035544492896669952?s=20

Love

· 1d

最新消息： #SEC 主席保罗·阿特金斯表示，他正在取消不切实际的规则，以“推动、澄清和改造”金融市场。 #加密货币 #btc

https://x.com/WatcherGuru/status/2035051688390758557?s=20

中文 English

0:00 随着未来一年的一系列规则制定计划，

0:04 我认为首先需要将它们放入背景中审视，

0:08 以及其后数年的情况，

0:12 在这期间我们的监管框架经常难以

0:15 跟上它所监管的市场步伐。

0:18 举一个监管与现实鸿沟的例子，

0:23 我们的规则仍默认采用纸质方式

0:26 传达股东信息。

0:29 在算法交易和人工智能的时代，

0:33 我认为这一规定应该成为历史

0:36 而非标准。

0:37 其次，基于此，

0:39 拒绝更新SEC规则，

0:43 反而通过误导性的执法监管运动

0:49 扼杀了许多潜在产品或将它们驱赶到海外。

0:53 当创新者无法获得合适的规则指引，

0:57 或在善意合规的努力后

1:00 面临传票威胁时，

1:04 理性的反应是转向其他地方发展，而他们确实如此。

1:08 整整一代数字资产创新

1:12 在美国以外发展，

1:14 并非美国创业者缺乏雄心

1:18 或美国投资者缺乏兴趣，

1:21 而是因为美国监管者缺乏决心。

1:25 第三，数十年的创造性规则制定

1:29 积累成了合规迷宫，

1:31 错综复杂到支撑起了整个行业

1:35 专门帮助上市公司

1:38 应对SEC复杂的监管框架。

1:43 当进入美国资本市场的成本

1:46 包括雇佣专家解读

1:49 SEC监管体系，以制作

1:53 只有学者才会欣赏的披露，

1:56 那么显然问题非常严重且根深蒂固。

2:00 我们的目标应是提高欺诈

2:03 和操纵的成本，而非合规本身的成本。

2:08 因此，在这样的背景下，每一项

2:11 SEC正在推进的举措和每一条

2:15 我们提出的规则、发布的解释

2:18 以及进行的机构改革，

2:22 大多归入三类之一。

2:26 第一类是推进规则以适应

2:28 当今市场运作方式。

2:31 第二类是澄清我们的监管体系，

2:34 简化监管，释放创新。

2:38 第三类是改革我们的要求，

2:41 消除繁重和不切实际的负担。

2:45 这三大支柱——推进（Advance）、澄清（Clarify）、改革（Transform）

2:49 构成了一个整体政策议程，

2:54 我称之为我们的ACT战略。

0:00 With a robust pipeline of rulemaking set for the year ahead,

0:04 I think it's instructive first to contextualize them

0:08 in the years that they follow,

0:12 a period in which our regulatory framework too often struggled

0:15 to keep up with a market that it oversees.

0:18 As just one example of the gulf between regulation and reality,

0:23 our rules still default to paper delivery

0:26 for shareholder communications.

0:29 In an age of algorithmic trading and artificial intelligence,

0:33 I believe that requirement ought to be a relic

0:36 and not a standard.

0:37 Second, and on that note,

0:39 a refusal to update the SEC's rules

0:43 prepared with a misguided regulation by enforcement campaign

0:49 has killed many would-be products or driven them offshore.

0:53 When innovators cannot discern fit-for-purpose rules

0:57 or when they face the prospect of a subpoena

1:00 as the response to good-faith efforts to comply,

1:04 the rational response is to build elsewhere, and they did.

1:08 An entire generation of digital asset innovation

1:12 developed outside the United States,

1:14 not because American entrepreneurs lacked the ambition

1:18 or American investors lacked the appetite,

1:21 but because American regulators lacked the will.

1:25 Third, decades of a creative rulemaking

1:29 have compounded into a compliance labyrinth,

1:31 so elaborate that it sustains entire industries

1:35 whose sole function is to help public companies

1:38 to navigate the SEC's complex regulatory framework.

1:43 When the cost of accessing America's capital markets

1:46 includes retaining specialists to decode

1:49 the SEC's regulatory regime to produce disclosure

1:53 that only an academic would appreciate,

1:56 then something is very clearly and deeply a mess.

2:00 Our goal should be to increase the cost of fraud

2:03 and manipulation, not the cost of compliance itself.

2:08 So against that backdrop, every initiative

2:11 towards which the SEC is working and every rule

2:15 that we propose, every interpretation that we release,

2:18 and every institutional reform that we undertake

2:22 largely falls into one of three categories.

2:26 Those that advance are rules to align

2:28 with how markets operate today.

2:31 Those that clarify our regulatory regime

2:34 to streamline oversight and unlock innovation.

2:38 And those that transform our requirements

2:41 by eliminating both the burdensome and the impractical.

2:45 So together, these three pillars of advance

2:49 or A, clarify or C, and transform or T,

2:54 form one integrated policy agenda

2:57 that I'm calling our ACT ACT strategy.

Love

· 1d

光是想象代理能够自主训练并使用稳定币支付，就让人觉得不可思议。

我开发了 mpp-nanogpt-modal，该项目在 Modal 上通过 #MPP 部署了 @karpathy 的 #nanoGPT 项目，并在无需 #API 密钥的情况下进行训练，所有费用均通过 #Tempo 上的稳定币支付。

接下来，我将尝试更长的训练周期和自动研究。

https://x.com/dwr/status/2035158640265765067?s=20

Love

· 2d

“我认为，作为一个国家，我们面临的最大风险是‘自杀’，而非‘他杀’。”

#Palantir 首席技术官希亚姆·桑卡尔（Shyam Sankar）与 #a16z 的凯瑟琳·博伊尔（Katherine Boyle）和埃里克·托伦伯格（Erik Torenberg）共同探讨了希亚姆的新书《Mobilize》，以及国防、人工智能、“SaaS末日”等话题。

00:00 引言

07:53 重建工业基础

18:01 陆军现代化

24:20 #SaaS 末日

29:42 自动化中的自主权

38:24 避免自我毁灭，战胜中国

40:42 电影作为文化意志力

49:57 里科弗海军上将的故事

@ssankar @KTmBoyle @eriktorenberg @PalantirTech #ai

https://x.com/a16z/status/2035055567774130269?s=20

中文 English

0:00 世界大事件提醒我们，邪恶确实存在。

0:02 可怕的野蛮行为依然可能发生。

0:05 国家一旦开战，不能仅靠战争部打仗，

0:09 而是需要全国家动员。

0:10 实际上是整个国家的事。

0:12 认为美国人无法承担这一责任，这简直是井底之见。

0:15 这种想法不可理喻。

0:15 我认为我们国家最大的风险是自杀，而非他杀。

0:18 我们如何赢得人工智能竞赛，尤其是在迈向实体AI和机器人技术时？

0:23 这些领域需要注意什么？

0:24 我们过去赢得胜利的方式，却意外地被我们自己抛弃了。

0:28 现在有机会重新找回这一点，重燃士气。

0:31 此刻或许有人会说，我们需要造更多武器，

0:34 我们需要做这个、做那个。

0:34 没错。

0:35 但最重要的是——

0:37 凯瑟琳，当我们讨论要邀请的嘉宾时，

0:48 肖恩，在你的名单上是首位。

0:50 是怎么回事？

0:52 嗯，人们在Jeremy Stern的报道《巨像》和很多最新播客中

0:57 听说过肖恩。他是那种几年前消息灵通的人都会认识的，

1:03 他是所有人的“老大哥”式解决者。

1:06 Trey Stevens，Andrew Allaire的联合创始人，

1:11 曾在推特上说

1:11 肖恩单枪匹马成就了他的职业生涯。

1:14 John Doyle和很多我们公司的创始人都说肖恩让他们的事业起飞，

1:18 引荐他们进入Palantir，

1:19 支持他们，也鼓励他们离开Palantir另辟蹊径。

1:22 这个故事你会一遍又一遍地听到。

1:27 但直到几年前，

1:27 肖恩才真正成为公众人物，他以前一直是幕后解决者。

1:31 而且，

1:31 我认为真正改变这一点的是，肖恩——我很想聊聊你的灵感来源——

1:35 你写了一篇被视为开创性的文章

1:36 论及“第一早餐”计划、国防改革，

1:39 你是首位这样公开讨论这问题的人。

1:41 不过，你写这篇文章时，Palantir已经运行了17年，

1:45 你决定

1:45 成为美国必须采取改革行动的坚决声音。

1:49 我很想聊聊你如何从幕后到台前，成为这个运动的代言人。

1:53 是什么促使你这么做？

1:53 这既是主动的绝望，也充满乐观。

1:55 多年来我从内部观察五角大楼，看到防务运作，

2:00 这种渐进的恶化

2:00 在历史背景下愈发明显。

2:04 促使我发声的原因是，我觉得这一刻仍有机会修复，

2:08 同时看到大楼外的情况，

2:08 创始人们正在重新崛起，充满了巨大能量，

2:11 有人愿意为国家利益而建功，这一刻值得凝练并提出我认为的根本诊断：

2:15 我们过去赢得竞争的做法，

2:17 不小心被抛弃了，现在有机会重拾它，

2:21 并需迅速行动。

2:22 时间紧迫，我们像青蛙被慢慢煮熟一样，失去了威慑力。

2:25 单独来看，这些事件好像可以忽略，比如2014年俄罗斯吞并克里米亚。

2:30 然后是2015年南沙群岛的军事化，JCPOA未能阻止伊朗造核，

2:34 以色列的暴动，

2:36 尤其是10月7日事件后，局势急剧恶化，我们必须行动，

2:38 此后事态只会更糟。

2:40 好消息是，近一年防务部门的变革远超我过去19年所见，

2:44 人们正抓住改革机遇。

2:46 看到这一点令人振奋。

2:48 大家齐心协力为国家利益努力。

2:52 那你觉得，是什么让18个月前，人们开始接受多年被忽视的观点？

2:57 也就是你长期坚持但被视为异端的那个？

2:58 我认为答案在于领导力。我们称他们为开国元勋，

3:02 有美国精神在其中，

3:04 每个创立故事都是异端与英雄的结合。

3:05 这些人看到了现实的不足，明白时间紧迫，必须采取行动。

3:08 他们既在政府内部，也在政府外部，

3:12 形成了志同道合的联盟。

3:14 所以难以说出单一时刻，但选举无疑是关键部分，

3:18 这让清醒的领导层得以掌权，为改革创造条件。

3:21 你一直很忙，加入了军队，还写了书。

3:24 为什么现在决定写这本美国未来的权威之作？

3:27 为什么同时加入军队，还领导Palantir？

3:29 我一直在思考如何动员力量避免更大冲突。

3:32 现在很难不把当下比作30年代末的局势，

3:34 事态正在酝酿已久，大家谈大国竞争。

3:38 我们正走出冷战胜利后的茫然，或更准确说是苏联失败后茫然。

3:42 这导致信了很多关于未来的谎言，现在正逐渐暴露。

3:46 所以我们不能像二战期间那样，等对方先攻击再动员。

3:50 对二战的正确认识是，领导层早在30年代末就认识到要动员，

3:54 不过当时没有全民意愿。

3:55 租借法案提供了动员机制，花了18个月建立和改造工厂，

4:00 为盟友提供威慑能力，二战爆发时我们已全速生产。

4:04 打仗不是战争部的事，是全国家的事。

4:08 冷战胜利让我们误以为仅靠国防工业基地足够，只靠战争部打仗够了。

4:14 其实，这必须是全体美国人共同参与的事。

4:19 美国公民既要关心国家的繁荣，也要关心保障繁荣的自由。

4:22 我们离这个目标已经很远。

4:24 1989年，只有6%的重大武器开支给了专业防务公司；

4:29 那时柏林墙还在。现在这个比例是86%。

4:33 我们现在习以为常的，实际上是过去的异常状态。

4:38 《动员》想将这个故事置于这样的背景：

4:42 所谓的美国工业基地，就是赢得二战的那套。

4:45 克莱斯勒造造导弹也造小面包车，

4:47 每个美国人买的相机、汽车、防摔盒实际上也补贴了国家安全。

4:51 这非常重要。

4:54 私营部门投入的研发远超政府，

4:56 我们要利用这条价值曲线为军人提供能力。

4:59 那怎么回到那条路上？

5:03 美国工业基地是谁？今天我们想到的是诺斯罗普·格鲁曼和洛克希德·马丁，

5:06 其实是格伦·马丁、杰克·诺斯罗普、勒罗伊·格鲁曼，

5:08 这些创始人

5:10 不光考虑公司季度表现，更在建造比自我更伟大的事业。

5:13 他们不只是外部人士，

5:14 还有海曼·里科弗，他顶住海军压力打造核潜艇。

5:18 这需要极大的胆识。

5:21 奥本海默曾断言核潜艇永远失败。

5:24 但里科弗坚持了下来。

5:29 冷战结束后，我们追求和平红利，

5:31 减少国防开支，

5:33 我们从51个主要承包商减少到5个。

5:37 很多人误以为这是整合，

5:39 其实是失去了竞争。

5:42 但本质上它一直是垄断买家市场，竞争来自政府内部的服务部门，

5:47 而非行业间的竞争。

5:50 “最后的晚餐”促成了同质化和国防金融化。

5:53 这些公司不再关注增长，转而关注财务指标。

5:57 创造力和异端思想被排斥，那些人去了科技圈。

6:02 但这些异端者至关重要。《动员》列举几乎所有防务创新都出自异端。

6:06 机构反对，官僚阻挠，

6:07 决策过程试图扼杀创新。

6:10 例如赢得二战的希金斯登陆艇，

6:12 海军最初不愿采购，甚至试图窃取设计，

6:15 但最终92%的登陆艇都是希金斯造。

6:17 如果这位苏格兰爱尔兰血统的设计师缺乏极度执着，

6:21 诺曼底登陆根本无法成功。

6:22 这说明异端者是创新的驱动力。

6:26 你说得没错，这是我们投身国防的核心问题。

6:31 垄断买家的本质导致供应商被置于“加拉帕戈斯群岛”，

6:36 与主陆断开，而巨型陆龟虽然奇妙，

6:39 但一回到大陆就会被竞争者吞噬。

6:41 因此，有想法的人难以把创意带回防务。

6:44 当我们开始打造Palantir时，国防部没有正门，

6:48 只有情报界一条路可进，外人进不去。

6:51 这局面直到2015年才开始改变。

6:54 因此现在发布的“18条论纲”和“防务改革2024”很有意义，

6:58 10年来一直有异端者敲门、准备助力国防。

7:02 书中提到这些“异端英雄”，你做了大量研究，现在成了防务科技行业的活百科。

7:08 你觉得谁是最值得一提的英雄？谁是人们需要了解却不了解的？

7:14 很难只选一个，但新的介绍是杜鲁·库科上校，

7:17 他是Maven AI系统之父。

7:22 库科出生在南加州，单亲母亲抚养，唯一上大学途径是军训，

7:27 入伍成为海军陆战队员，经历丰富，

7:31 发生过一次关键事件：在尝试撤离ISIS下的伊兹迪难民时，

7:35 一名年轻海军陆战队员误判敌方火箭弹威胁，

7:39 造成疏散行动失败，成千上万人遭受残酷迫害。

7:43 这事件改变了他的人生轨迹。

7:47 后来他进入五角大楼地下室，负责资源极其有限的AI项目，

7:51 全力推动AI应用在军队。

7:56 经历了大量排挤、调查，甚至有人诬陷他地下藏伊朗人，

8:00 结果调查人员尴尬发现他那房子根本没有地下室。

8:03 这故事说明了一个问题：你是否愿意以一切代价追求目标，

8:08 是否在乎理念高于职业生涯？这些人值得我们学习。

8:12 现在正是激励潜在异端者挺身而出的时候，你的国家需要你。

8:18 过去一年最激动人心的是，我看见这些人，无论在大楼内外。

8:22 领导层放手支持、保护了这些异端者。

8:25 曾被空军厌恶、极具异端性的空军飞行员约翰·博伊德，

8:29 他是F-16的父亲，他的理论在海湾战争得到印证。

8:33 博伊德说，成为某人或做某事，二者不能兼得。

8:37 他极难相处，却得到了三颗星空军将军的保护，

8:41 后者懂得没有保护，这类人才无法在体制内存活。

8:43 领导认同其价值，选择容忍他的怪异。

8:46 希曼·里科弗和伯纳德·施里弗等创始人在体制内外，

8:50 勇敢推动了核潜艇和弹道导弹项目。

8:53 但他们同样极难相处。

8:58 他们在特定领域坚持到底，推动了划时代技术。

9:01 你看到的是，领导层的支持、鼓励与保护对创新而言至关重要。

9:06 你加入陆军的项目起源于乔治将军和史考特·德里斯科尔秘书的推动。

9:10 我2014年以来一直与以色列合作。

9:14 他们技术先进，十分自豪。

9:16 10月7日后，以色列动员了36万预备役军人，

9:20 这些都是过去服役的预备役人员，拥有丰富产业经验。

9:23 他们回归现役后，对以军的技术状态感到震惊，

9:27 短短四个月内技术升级超过过去10年。

9:30 这令我无法忽视。

9:33 世界各国都有这样的技术人才储备。

9:36 我们山谷的创业公司知道如何打造这些能力。

9:40 年轻的战争新人充满意愿和智慧，但也具备运营智慧。

9:42 我学到的经验和教训，

9:44 错误可以犯，但请别重复，是在登高望远。

9:47 我不确定24岁时能为军队贡献多少，但44岁还能助力加速。

9:50 这不仅是说我，也是说硅谷，

9:53 说美国制造业。我们要在越南海军等地形成协调配合。

9:57 中国推行军民融合，我想明白为什么美国却让自愿军民融合难以实现。

10:03 二战时我们直接委任了10万符合“201部门”标准的人才。

10:07 现在国家授权依然存在，只是未被充分利用。

10:10 我很自豪与几位技术领袖一起加入陆军，

10:13 成为高级顾问，参与多个项目，支持军队变革。

10:17 我自己也学到了很多，希望军队从中获益。

10:21 我愿意促进相关界的进一步合作。

10:26 我现在的工作聚焦两点：一是帮军方规划长期力量结构；

10:31 二是如何将软件作为可塑的武器系统，赋能指挥官推动进步。

10:35 这就是所谓的“作战数据团队”。

10:38 让我印象深刻的是军中人才素质，

10:41 他们很多不是正式学的计算机科学，而是自学成才，

10:44 最具说服力的AI应用都是他们创造的。

10:48 这背后是一定的焦虑感，

10:49 不是为了10%效率提升，而是存亡攸关。

10:53 AI极大授权那些具备特定技能的人。

10:56 比如情报军官精通其领域。

10:59 我想问，这些人才十年前在哪里？

11:03 事实上，“他们一直都在”。

11:06 但当时他们做什么？做了PPT给项目官僚汇报，被否决？

11:10 不，他们聪明不会浪费时间。如今他们自己做两周，快速迭代。

11:14 能够实证证明已推动军队进步，

11:16 大家对成果反应积极，都想赢。

11:20 大机构普遍在从零到一的创新阶段挣扎。

11:24 他们急于快速扩展，但军队规模庞大，必须深思创新推广路径。

11:30 这是整天从事的行业工作。

11:32 军方比拼不是图样仿制，

11:34 而是吸取经验，避免走弯路。

11:37 同样大企业也应向初创学经验。

11:41 确实，军中技术能力极强，基层人员涌现创新，

11:44 这或将掀起军事变革，充分发挥美军自下而上的创新优势。

11:48 关于SaaS的崩溃论，我怎么看？

11:52 AI让软件切换成本大幅降低，传统的代码、数据、界面护城河不复存在，

11:56 一些SaaS公司走上淘汰之路，比如monday.com、Atlassian等。

12:00 没有任何系统能够幸免。

12:03 我认为软件可分为Beta和Alpha两类，

12:07 Beta软件让你与别人更相似，将受苦。

12:10 我一直批评软件产业，卖软件重于软件本身的实际价值。

12:14 AI使定制化软件成为可能，

12:16 更注重Alpha价值的公司将受益，

12:19 Beta软件将面临重重挑战。

12:22 举例来说，疫情期间CEO讲话大多谈视频会议，

12:26 而不是企业资源计划系统，这揭示软件价值匮乏。

12:30 对我们来说，疫情则加速了向客户的适应，

12:33 筛选了真正有价值的技术。

12:37 我们看到模仿性的行业解决方案承受压力，

12:40 而区别化软件将市场获益，体现竞争优势和战略。

12:44 Alpha软件将获得更高估值，比如可视化编码。

12:48 在硬件和互联网经济中，利润分布也在变动。

12:52 AI堆栈有芯片、模型、基础设施和应用，

12:56 模型被商品化且承压，厂商正向上扩展，构筑AI基础设施。

13:00 狭义的垂直AI解决方案正逐渐覆盖整个堆栈，

13:04 实现规模化和更多用例。

13:08 价值可能集中在芯片层和我们称为“本体论”的AI基础设施层，

13:12 这两层相对难以被攻破。

13:16 有份经济学人文章概述了经济可能面临的三种未来：

13:20 一切走向垂直整合，AGI让一切消亡，或2%增长。

13:24 这代表广泛可能，择一而行是我们的选择，不是被动接受。

13:28 美国要赢，关键是领导层塑造和引导。

13:32 我们拥有绝佳人才和技术基础，我们将在AI赛道上发力。

13:36 最大风险是自我毁灭，不是外敌。

13:40 中国虽强，但他们必须我们衰落，才能真正崛起。

13:44 有非法方式企图阻碍我们的农业生产，这违背公平竞争。

13:48 美国社会陷入虚无主义和两极分化，忘了团聚的力量。

13:52 我认为，我们要恢复对机构的信任，推动社会正向发展。

13:56 重启美国制造业，最大化利用我们的技术创新和资源优势。

14:00 我们发明了批量生产、核能等领先技术。

14:04 提出美国人缺乏能量，是低估了我们自己。

14:08 这是意志和领导力的问题，而非技术难题。

14:12 你提到的文化建设很关键，很多人误以为是技术或产能问题。

14:16 你投身电影制作，这是完全不一样的领域，

14:20 能说说你为何跨界？你想如何通过电影激励社会意志？

14:24 我小时候移民美国，在奥兰多长大，

14:28 五六岁时常和父亲一起看80、90年代的电影，

14:32 《猎杀红色十月》《红色黎明》《第一滴血2、3》等，

14:36 那时我就感受到了“美国人”是什么感觉，远早于了解公民知识。

14:40 类似最近很多人对《壮志凌云：独行侠》也有共鸣。

14:44 娱乐让我们反思自己，认识我们想成为怎样的人。

14:48 如果我们的娱乐只讲科技毁灭的反乌托邦故事，

14:52 会潜移默化形成消极预期。

14:56 我成长于太空海岸影子下，那里充满对科学和技术的热爱，

15:00 甚至小学期末作业是写如何登陆火星，

15:04 这孕育了我对美好未来的信念。

15:08 我们要重新讲好故事，既娱乐又鼓舞人心。

15:12 你也说在奥兰多长大，我觉得这很有趣。

15:16 我也在80、90年代的佛罗里达州长大，

15:20 当时很多人嘲笑佛罗里达，但奥兰多有迪士尼世界，

15:24 迪士尼动画里的善恶故事深深影响我。

15:28 这是美国文化的代表，我的父母也在主题公园售卖纪念品。

15:32 我从小浸润在这样的故事氛围中。

15:36 迪士尼的电影曾塑造美国精神，我们要重拾乐观和英雄主义。

15:40 我们也期待未来几年，产业带来更多支持美国的乐观内容。

15:44 比如近期拍摄的《使命召唤》电影和泰勒·谢里丹的作品，

15:48 这些都反映了现实社会的故事。

15:52 近期世界大事件提醒我们，邪恶真切存在。

15:56 俄罗斯坦克越过边境，10月7日的残暴令人震惊。

16:00 在地缘政治层面，我们应该称中国为“对手”，而非“近似对手”，

16:04 他们的许多行动毫无效果。

16:08 这给第三国国家警示，不要误以为美国已退缩。

16:12 我期待通过电影和硬实力推动美国伟大与繁荣，

16:16 无论软实力还是遏制对手，都是为保护未来世代的繁荣。

16:20 谢谢你。

0:00 World events remind us that there is actually evil out there.

0:02 Just horrendous barbarism is still possible.

0:05 When a country goes to war, it's not enough to just have the Department of War

0:09 fight these wars.

0:10 It is actually the whole country.

0:12 The idea that somehow the American people are not capable of this beggar's

0:15 belief.

0:15 I think our biggest risk as a country is suicide, not homicide.

0:18 How do we win the AI race, particularly as it moves towards more physical AI

0:23 and robotics, etc?

0:24 The things that we did to win in the past, we accidentally turned our back on.

0:28 And there's an opportunity to reclaim that with figure.

0:31 In the moment right now, you could say, we need to build more weapons.

0:34 We need to do this. We need to do this.

0:34 Yes.

0:35 But the most important thing we need to do is-

0:37 So, Catherine, when we were talking about guests that we had to add on,

0:48 Sean, was at the top of your list.

0:50 What was that?

0:52 Yeah, well, you know, it's, I think people are, you know,

0:57 after the Jeremy Stern profile and Colossus and a lot of, I think, you know,

1:03 stories that have come out recently or podcasts that have come out about Sean.

1:06 He's one of these people that if you were in the know several years ago, you

1:11 knew.

1:11 He was the OG, like, fixer for everyone.

1:14 You know, I think Trey Stevens, who's, you know, the co-founder of Andrew all

1:18 came out

1:19 on Twitter and said, like, he's single-handedly responsible for my career.

1:22 And I, you know, John Doyle, so many of our founders, have pointed to Sean as

1:27 the person

1:27 who made their career and introduced them to Palantir, supported them in Palant

1:31 ir,

1:31 but also sort of, you know, gave them wings to fly away from Palantir and to

1:35 start something new.

1:36 And you hear that story time and time and time again.

1:39 But it wasn't until, and Eric, you and I were talking about this.

1:41 It wasn't until a couple years ago that I think Sean actually became more of a

1:45 public figure.

1:45 He was sort of the behind-the-scenes guy, the behind-scenes fixer.

1:49 And I think the thing that really changed it was, Sean, and I'd love to talk to

1:53 you

1:53 about kind of what was the inspiration for this.

1:55 You sort of wrote this seminal piece about First Breakfast, about defense re

2:00 formation,

2:00 and we're the first person to really start talking about it.

2:04 But again, this was like 17 years into the journey of Palantir that you decided

2:08 I'm

2:08 going to be, you know, a strident voice for what needs to happen in America.

2:11 So I'd love to talk to you about this, you know, going from the behind-the-

2:15 scenes person,

2:17 the guy behind the guy, and so many of our companies, to saying I need to come

2:21 out and

2:22 be a voice for this movement. What was the kind of impetus for that?

2:25 It was kind of equal parts in active desperation and active optimism.

2:30 You know, I felt like after years of just seeing

2:34 the building, the Pentagon from the inside, seeing how defense was operating,

2:40 I felt this frog boil that continued to happen set in a historical context.

2:46 But the reason to say something is actually I thought this was the moment that

2:50 it could

2:50 all be fixed, that alongside of that happening, seeing what was happening

2:54 outside of the building,

2:55 that the founders were re-emerging. There was a huge amount of energy.

2:58 People wanted to build in the national interest, and it was a moment to kind of

3:02 crystallize what

3:03 at least put forth what I thought the fundamental diagnosis was that really the

3:08 things that we did

3:09 to win in the past, we accidentally turned our back on, and there's an

3:13 opportunity to reclaim that

3:14 with vigor. And we needed to do so quickly. That time was running out, there

3:18 was a shot clock here

3:19 that we have frog-boiled our way to a place where we've lost deterrence.

3:23 You know, any one of these items in isolation, you can write off, you could say

3:27 , okay,

3:28 the Russians annexed Crimea in 2014. That's just one thing.

3:32 Then you have the militarization of the Spratly Islands in 15. You have the

3:36 failure of JCPOA to

3:38 keep the Iranians from getting a bomb. You have a pogrom in Israel, and

3:42 certainly after October 7,

3:43 it was kind of a radicalizing moment that like, what is going on here? We need

3:49 to act,

3:50 and I think we've only had more things since then. Now, I think the good news

3:54 is, in the last year,

3:55 more has changed in the department than I've seen change in the prior 19 years,

3:59 and people are

4:00 seizing that moment for reformation. And it's been rewarding to kind of get it

4:08 out there,

4:09 get people to rally behind it, and all of us building the national interest.

4:13 Yeah. I mean, what is it? Because I think you get this question probably all

4:17 the time.

4:18 What is it about the last? I mean, as you said, it wasn't a single moment,

4:21 but there is some change in the culture and the zeitgeist. And I think you have

4:25 a unique

4:26 understanding too of culture and kind of how these memetic shifts happen. But

4:30 what was it about,

4:31 you know, 18 months ago where it's like everyone seems to agree on the thing

4:36 that was so contrarian

4:37 for many years where you were sort of banging your head against a wall saying

4:41 like, this is,

4:42 you know, this needs to happen. Well, maybe unsurprising for my worldview,

4:46 it all comes down to leadership. You know, I, we call them the founding fathers

4:51 for a reason.

4:52 There's something special about the American spirit that is, you know, every

4:57 founding story is,

4:58 is equal parts heresy and heroism. And we had the right people who kind of saw

5:04 like, hey,

5:04 this is not working. And we're, the shot clock is running out and we have to do

5:09 something.

5:09 And we had those people both inside the building and outside the building. So

5:13 it's kind of a

5:14 conspiracy coalition of the willing coalition of the capable to go do that.

5:20 So it's hard to point to anyone single moment. I think the kind of the election

5:24 is a big part

5:25 of it, not to not to make it political, but just being able to get in a

5:29 leadership that viewed it

5:30 with clarity and set the conditions to make this change happen. Yeah. So you've

5:35 been busy. You,

5:37 you know, you've joined the army. You've written a book that's coming out. I'd

5:42 love to get into

5:43 to mobilize. But first, why did you decide now's the time to sort of write the

5:47 right,

5:48 the canonical book of what needs to happen in America? And then maybe we can

5:51 talk about sort

5:52 of the sort of fundamental thesis to you. But why now? Why write a book, join

5:56 the army and also

5:57 be leading Palantir? Yeah. Well, there's a there's a code you kind of thread

6:02 through that,

6:03 which is like, how do we mobilize to prevent a bigger conflict? And if you're

6:07 really paying

6:08 attention, it's hard not to think that we're kind of in the late thirties here,

6:11 that things

6:12 are brewing. They've been brewing for a while. People talk about great power

6:17 competition.

6:18 And I think we're kind of coming out of the malaise of having won the Cold War,

6:22 or the Soviets having lost it, perhaps more accurately. And that kind of led to

6:26 a lot of

6:27 bad behaviors that allowed us to believe a lot of lies about the future that we

6:31 're kind of now

6:32 marking to market. And so we will not have the luxury that we really had in

6:41 World War II of

6:42 letting the adversary attack us first and then deciding to mobilize. And I

6:46 think a more

6:47 clear-eyed view of what actually happened in World War II is that's not it. It

6:51 's not this

6:51 facile thing that we just flipped a switch. And the automotive industry decided

6:54 , okay,

6:55 after Pearl Harbor, we're going to make all this war material. What really

6:58 happened is that

6:59 leadership from FDR realized in the thirties, in the late thirties, that we

7:03 needed to mobilize,

7:04 but there was not yet a national will or popular mandate to do so. And Len Le

7:09 ase provided the

7:10 mechanism to do that. It took us 18 months to build factories and re-tool them,

7:14 and we were able to

7:15 create capability deterrents that we sold to the Brits and to the Soviets, such

7:21 that

7:21 when World War II really kicked off for us, when Pearl Harbor happened, we were

7:24 at full rate production.

7:26 And the way that we mobilize, when a country goes towards the whole country

7:29 that goes to war,

7:29 it's not enough. I think part of the legacy of having won the Cold War is

7:32 thinking it's enough

7:34 to just have a defense industrial base. It's enough to just have the Department

7:37 of War fight

7:38 these wars. It is actually the whole country. And I think that's the most stark

7:42 thing. It's like

7:43 we all, as American citizens, need to be invested in both the prosperity the

7:47 country gives us,

7:48 but also the freedom that underwrites that prosperity. And we have come a very

7:52 far away

7:52 from that world. In 1989, only 6% of spending on major weapon systems went to

7:58 defense specialists,

7:59 i.e. companies that were exclusively in the business of defense. That's not

8:02 that long ago.

8:02 That's when the Berlin Wall still stood. Now that number is 86%. So really,

8:08 what we think of as

8:09 normal is an aberration from the past. And Mobilize seeks to set that story in

8:13 context of one,

8:14 hey, this is the industrial base, what I like to call the American industrial

8:18 base,

8:18 that won World War II. Chrysler built Minuteman missiles and minivans. And

8:23 every camera, car,

8:25 cereal box that an American consumer bought was actually also subsidizing our

8:29 national security.

8:30 And that's really important. We see this with the hyperscalers. We see this

8:32 with technology.

8:33 The amount that our private sector spends on R&D dwarfs what the government is

8:37 capable of spending.

8:38 And you want to get on that price performance curve as a way of delivering

8:41 capabilities to

8:42 our brave men and women in uniform. Yeah.

8:45 So how do we get back to that? And oh, so the second part of it. Yeah, so we

8:53 had the American

8:54 industrial base. But who was the American industrial base? Today we think of it

8:58 as North

8:59 of Grumman. We think of it as Lockheed Martin. But actually it was Glenn Martin

9:03 . It was Jack

9:03 Northrop. It was Leroy Grumman. They were people. They were founders. They were

9:08 kind of not thinking

9:10 about, hey, what's the performance going to be next quarter? They were building

9:13 something way

9:14 bigger than themselves, way bigger than their companies. And those founders

9:17 weren't just outside

9:18 of government. They were inside of government. It was the Heimann Rickovers

9:21 that against the will

9:22 of the Navy, building the nuclear Navy. And I love that story, too, because it

9:26 takes a lot of

9:26 Hutspa Oppenheimer. The father of the bomb said the nuclear Navy wasn't going

9:30 to work. He told

9:31 Oppenheimer, he told Heimann, he was going to fail. And he still proceeded. And

9:36 that's something I

9:37 think in the Valley we recognize as the classic founder personality. And a big

9:41 part of what happened

9:43 after the end of the Cold War, we wanted a peace dividend. We started spending

9:49 less in defense.

9:50 We had this famous dinner, the Last Supper. We went from 51 prime contractors

9:53 down to five.

9:54 I think the conventional explanation of what happened is wrong. People think,

9:59 hey, we had

10:00 consolidation. Consolidation means we lost competition. That's not, yeah, okay,

10:06 maybe at the margin.

10:06 But first of all, it's always been a monopsony. The nature of the competition

10:10 is not what people

10:11 think. It's not these companies competing against each other. The competition

10:14 has always been the

10:15 services competing. It's been competition inside of government that drove

10:18 innovation,

10:19 not competition from industry. What really happened from the Last Supper is

10:23 that consolidation

10:24 bred conformity. It was the beginning of true financialization of defense.

10:27 These companies

10:28 really could no longer think about growth. They thought about financial medics,

10:32 dividends, buybacks,

10:33 cash flow. And it kind of became very narrow. And that conformity is not an

10:39 environment that

10:40 founders can thrive in. The heretics were expunged. They left. They went to

10:45 other parts of the American

10:46 economy, like tech. But those heretics are required. In fact, if you look at

10:51 part of the book, we

10:52 catalog all these amazing defense innovations, almost to a T. Every single one

10:57 of them was a

10:58 heretical idea. The institution was against it. The bureaucracy was against.

11:02 The process tried to

11:03 kill it. And these determinative outcomes, you can think about the Higgins boat

11:06 , the boat that won

11:07 World War II. The Navy didn't want to buy the boat. The Navy tried to steal the

11:10 designs for the boat.

11:11 In the end, 92% of all boats in World War II were Higgins boats. Think about

11:15 where we'd be if this

11:16 Scott's Irishman wasn't just willing to just almost pathologically commit

11:21 himself to making

11:22 this happen. The boys that would not have landed at Normandy. Yeah. No, I mean,

11:26 and that's super

11:26 interesting. Yeah, because I think that as you said, it's always been, you know

11:30 , this sort of force

11:31 consolidation that we've picked the winners there, you know, we're post history

11:36 now, like we don't

11:37 actually need to build for for wartime. So it's your view that that really that

11:41 just expunged

11:43 all of the talent that used to go to defense. And it's interesting because it's

11:45 also I mean,

11:46 it's in some ways it's sort of serendipitous that that's exactly when the

11:49 internet's rising.

11:50 So it's like, if you're this kind of weird personality, you're going to go work

11:54 on this new thing

11:55 that's so exciting. You know, I think of like, you know, Mark Andreessen, right

11:58 ? Like he could have

11:59 been in the defense industry maybe 20 years earlier, but it's like, you know,

12:03 the people who were

12:03 building the internet in the 90s, they wanted to build the new thing. So I

12:07 guess like, is that

12:09 sort of the fundamental problem? Is that like, it just became a place where

12:12 anyone who was

12:13 interesting or anyone who had a different view just could not thrive? Yeah,

12:18 exactly. And then

12:19 we compounded on that problem by, you know, the nature of the monopsony. So

12:24 unlike a monopoly where

12:25 you have one sellover thing, a monopsony is where you have one buyer thing.

12:28 That the Department of

12:29 War is a monopsony. The nature of the monopsony is it forgets, it starts

12:35 imposing all sorts of

12:37 constraints on its suppliers and how they behave and what they need to look

12:40 like. And that led to,

12:42 it's like put it, we put all of these companies on the Galapagos islands. They

12:45 're not on the main

12:46 land anymore. And so what do you get on the Galapagos? You get these exquisite

12:49 giant tortoises.

12:50 They're really amazing. It's like alien life. It's very cool. Except when you

12:55 take the tortoise

12:55 back to the mainland, they're not competitive. They're going to get eaten alive

12:58 by the wolves.

12:59 And so we started creating a huge number of barriers for these people who even

13:04 when they

13:05 had ideas that could those ideas come back into defense. I like to say, like

13:10 when we started

13:11 Palantir, there was no front door in the Department of Defense. You had to, the

13:15 only front door,

13:16 there was exactly one. It was in the intelligence community. It was Inchitel.

13:19 If you were an outsider,

13:20 there was no other way. And the only people that worked here were insiders. Of

13:24 course,

13:25 now that is, that's part of the sea change that happened starting really in

13:28 2015. But that means

13:30 putting forth the 18 theses and the defense reformation and 24 makes a lot of

13:35 sense because

13:35 we have 10 years of heretics who have been knocking at the gate ready to come

13:39 help the department.

13:40 Yeah. And so much of the book, I mean, you talk about these sort of heretical

13:43 heroes,

13:44 you know, and doing the research. And I mean, you're now sort of like the

13:47 walking encyclopedia for

13:48 the defense industry in tech. Who are the most exciting heroes? Like who's the

13:53 person you look

13:53 at and you say like, gosh, that that is, that is the person that people don't

13:56 know about they

13:57 need to know? Well, I think it's hard to pick only one that a new profile that

14:02 we put out there,

14:03 which is some of these figures are historical figures, monarchs, Shriver,

14:06 Edward Hall,

14:07 Hyman, Rick over, James Boyd, John Boyd, sorry, but the one that's new is

14:11 Colonel Drew Cucor.

14:13 And Cucor is the father of Maven. And here you have this marine colonel, you

14:20 know, born

14:21 raised by a single mother of Southern California, very modest background, his

14:26 only way of going

14:27 to college was ROTC joined the Marines. Lots of incredible experiences there,

14:32 but he had a

14:34 seminal experience where he was trying to evacuate Izzadi refugees who were

14:39 fleeing ISIS and a young

14:42 Marine looking at ISR made a call that he thought he saw RPGs. And that would

14:47 have made it unsafe

14:48 for the Marines to land and exfil these people. And there actually wasn't one.

14:53 As a consequence,

14:54 you have order of thousands of people who were tortured and slaved and raped

14:59 because of this

14:59 failure of operation. And it just changed this man. And so when he had an

15:04 opportunity in the

15:06 basement of the Pentagon on a project with no resources to go after bringing AI

15:10 to the department,

15:11 he leaned all the way in. And you can see the journey of the heretic here where

15:16 everyone hated

15:18 him. Everyone tried to kill him. Every service thought they were doing AI.

15:21 People tried to throw

15:22 IG investigations. And it's like one of the details we document is that someone

15:26 said that

15:26 Colonel Cucor is housing Iranians in his basement. So they actually sent out

15:32 criminal investigators

15:34 to his house to look at this. And here's a Mormon devout Mormon for four

15:38 daughters,

15:40 1400 square foot home that doesn't have a basement, by the way. And you know,

15:44 the investigators were

15:45 just completely dumbfounded, but it shows you sometimes that what, you know,

15:49 are you, are you

15:50 willing to put it all on the line? Are you so committed, so incorruptible in

15:53 what you're trying to

15:54 deliver? Or is this just a career? And that's one of the things I think I draw

15:58 a lot of inspiration

15:59 from these folks is just seeing, you know, if they can do this, we can too,

16:03 seeing that these

16:04 people exist inside of government, they exist outside of government. And I

16:08 think it's that's

16:09 in the moment right now, you could say, we need to build more weapons. We need

16:12 to do this. We need

16:13 yes. But most of the most important thing we need to do is inspire the latent

16:17 heretics

16:18 to actually step up. This is the moment your country really needs you. What's

16:22 been exciting

16:23 over the last year or so is I'm seeing those people. I'm seeing them inside the

16:26 building. I'm

16:26 seeing outside the building. And that's that is, you know, what is driving the

16:29 change is the

16:30 leadership is setting the conditions to empower the heretics. They're

16:33 protecting them. John Boyd,

16:35 who was a famously difficult fighter pilot, his own service, the Air Force

16:38 hated him,

16:39 but the Marines learned everything they could from him. And all of his heresy,

16:42 he was really the

16:43 father of the F-16. All of his heresy was proven correct in Gulf War I, where

16:47 we destroyed the

16:48 fourth largest army in the world in days. It's just everything came to bear his

16:51 high-low mix,

16:52 his oodle-loop. But, you know, John Boyd said to be or to do, you can be

16:59 somebody or you can do

17:00 something. But you can't have both. And, you know, how committed are you to

17:04 this? But he was so difficult.

17:06 He gets a lot of credit. I think the other person who gives credit, who I don't

17:10 even know the name

17:11 of, is a three-star Air Force general who protected him because people like

17:15 that do not survive in

17:16 these bureaucracies on their own. It's not like, hey, he's difficult and we

17:18 somehow tolerated him.

17:20 It's no. Someone realized there's something special here. And despite that, we

17:24 tolerate him.

17:24 You see that with the dynamic between Bernard Shriver, who built our inter

17:29 continental ballistic

17:30 missiles at Edward Hall, who specifically built the Minuteman. Shriever fired

17:35 Hall once and then

17:37 hired him back, realizing there was no way we were going to get to solid-fueled

17:40 intercontinental

17:41 ballistic missiles without this notoriously difficult human. And I think that's

17:47 great leadership.

17:48 Yeah. Yeah. And you just said something interesting, which is that it really

17:52 takes

17:52 leaders inside the building to model the change, to encourage it, to protect

17:58 the people who are

17:59 those heretics. And you've joined the Army. And it's a program from General

18:04 George and Secretary

18:05 Driscoll. Maybe tell us a little bit about the origin of that and then what you

18:09 're doing specifically

18:10 inside the Army now to support that change. Well, the origin story of this is

18:17 really,

18:17 so I've worked with the Israelis in some capacities since roughly 2014.

18:22 This is a very technical country. And they're proud of how technical they are.

18:28 After October 7th, you know, October 8th, they mobilized roughly 360,000 reserv

18:32 ists.

18:33 By definition, all these reservists are prior service through national cons

18:37 cription.

18:38 And most of them had now had 20 years of experience in industry.

18:42 And when they got back to the IDF, they were horrified at the state of

18:46 technology in the IDF,

18:47 which is actually an implicit self-critique, which is, hey, when I was 20, I

18:51 was really good at coding,

18:53 but I didn't know what I was doing. Now I have 20 years of experience building

18:57 internet-scaled

18:58 things. And I actually know how to do these things correctly. So I saw them

19:03 modernize more

19:03 in the four months after October 7th than I did in the prior 10 years of

19:07 working with them.

19:08 And that was just I couldn't, I couldn't unsee that. So of all countries in the

19:14 world,

19:14 we are drowning with that talent. You know, the skills we have at building

19:18 things

19:18 in the valley, the companies that A16Z backs, like we know how to do this as a

19:23 nation.

19:24 The 20-year-old version of our green suitors maybe did it. You know, there's

19:29 the will,

19:30 there's the intelligence, the capability, but there's also then the trade craft

19:33 and know

19:33 how the experience, all the dead ends that I've run into in my career. The

19:36 mistakes I've made,

19:37 if you're going to make mistakes, please make new ones. Don't make the same

19:40 ones I already

19:40 have done. How can you stand on the shoulders of American industry to go faster

19:44 and do this?

19:45 And so I'm not sure I really had a lot to give the army at 24, but I think at

19:50 44,

19:51 there's a lot I can do to accelerate certain things. And that's not just a

19:55 narrow statement

19:55 about me. I think that's a whole statement about the valley. This whole

19:57 statement about

19:58 American manufacturing, everyone in El Segundo, how do we make sure if the

20:03 Chinese make civil

20:05 military fusion compulsory, why do we make voluntary civil military fusion

20:09 impossible?

20:10 And when I look back at history, we didn't use to make it impossible. In World

20:14 War II, we

20:14 direct commissioned 100,000 people that look like what we now consider a

20:18 detachment 201

20:19 into the military. And we should be doing that again. The authorities exist.

20:23 They're just laying

20:24 their dormant and we're basically underutilized. So I was proud to join the

20:29 army with three other

20:30 colleagues. We have Bob McGrew, former chief research officer at OpenAI, Boz,

20:35 the CTO of Meta,

20:37 and Andrew Wheel, the former chief product officer of OpenAI, head of science

20:40 now.

20:41 And I think we've been able to work on different projects that really, we act

20:45 as senior advisors

20:46 to army senior leaders. And there's different projects that we kind of get our

20:49 hands dirty and

20:50 and help. But I think it's been a really, I've learned a lot by doing it.

20:54 Hopefully the army's

20:54 benefiting from it. But I think more broadly, we would like to catalyze this

20:58 across all of the

20:58 services and a broader call for folks who are listening to this now in industry

21:03 .

21:03 What's been the biggest surprise? I mean, you've obviously worked with the

21:06 department for years,

21:08 but being on the inside now, what surprised you?

21:11 In my, so my focus is really two things. I'm helping them think through

21:17 how to plan for structure over long periods of time. So how do I generate the

21:22 force I want for

21:23 all the different military occupation specialties? So that's been one. But the

21:27 second part of it is

21:27 thinking through how do we want to employ software as almost like a malleable

21:33 weapon system as

21:34 something that our commanders can wield to drive advancement? And they call

21:38 these the operational

21:39 data teams. What's been hugely impressive to me is the quality of talent in our

21:45 green suitors.

21:46 People who are not formally trained computer scientists, people who have just

21:49 learned these

21:50 things, the most compelling AI applications I'm seeing across commercial or,

21:55 you know, so private

21:56 sector or public sector are being built by these green suitors. And it's, and I

22:01 think there's

22:01 something about the existential stakes. You know, you're not doing this for fun

22:04 . You're not doing

22:04 this for 10% efficiency. It's a binary outcome when you're lose. The other

22:08 thing about this

22:09 moment that I think is really interesting with AI is it's massively empowering

22:12 to people with

22:13 specific skills. So, you know, it is the Intel Warren officer who really knows

22:17 their domain.

22:18 And I was wondering, you know, as someone who's been doing this for 20 years,

22:21 like, where was this

22:22 person 10 years ago? And the conclusion I came to is they were always there 10

22:27 years ago. What would

22:27 they have done though with their idea? Make a PowerPoint slide, brief some

22:31 program bureaucrat,

22:32 who would tell them how bad their idea is? No, because they're smarter than

22:35 that. They wouldn't

22:35 have wasted their time. Now they spend two weeks, they build it themselves. Now

22:39 they're having an

22:40 empirical conversation about how what they've built actually drives the army

22:44 forward. And everyone

22:45 is quick to adopt it because everyone wants to win. So it's been really

22:48 exciting to see that.

22:51 The other part of it, I think, is big institutions. It's conserved across

22:57 private sectors well,

22:58 struggle with zero to one. Everyone wants to get, you know, everyone wants, if

23:03 you have some sort

23:03 of innovation, they almost want to rush to get to N as quickly as possible. How

23:08 do I scale this

23:08 across the formation? Well, the army is a very big place. And thinking very

23:12 critically about the

23:13 pathing of what is the journey and cycle of getting an innovative idea to scale

23:18 . That's literally what

23:20 we do as an industry all day long, right? And how do they take and vibe those

23:24 lessons

23:25 rather than cargo culting their way there, which is frankly what I think the

23:28 private sector,

23:29 you know, large fortune 100s do as well. So they have more to learn from

23:33 startups in this capacity

23:34 than they do from, you know, big fortune 100 companies. Totally. And that's so

23:39 interesting too,

23:39 about just, you know, that is something I hear time and time again, like the

23:43 level of technical

23:44 ability of someone very junior in the army or the navy today. It's like they

23:48 came up tinkering,

23:50 and yes, as you said, like now the tools are there, they can just build

23:52 something where it's

23:53 and you're not only learning from startups, you're literally learning from

23:55 individuals who are

23:56 enlisted, who have a great idea, which is it's it does feel like, as you said,

24:00 like this is a

24:01 revolutionary time for the military where they can actually learn from their,

24:04 you know, junior

24:05 people who have a great idea and that can be deployed very quickly. Which plays

24:10 to the American

24:11 military strengths of bottoms up innovation, mission command type control. It's

24:15 really something

24:17 that our military can uniquely do than no one else can. Let's get into the SaaS

24:21 apocalypse.

24:22 You know, there's a line of thinking that says, hey, now that the switching

24:26 costs are,

24:26 now that AI is here, the switching costs are very low, you know, there's no

24:30 code moat,

24:31 there's there's no data moat, there's no UI moat, and there's a set of SaaS

24:34 companies that are

24:35 on the conveyor belt on the way to the guilty. And you know, maybe it's monday.

24:39 com first,

24:39 and maybe it's Atlassian, and you know, companies that aren't systems record,

24:43 and then maybe it's

24:44 coming for them to know that people say, hey, you're not going to vibe code,

24:48 you know, Atlassian,

24:49 you're not going to vibe code, these, these, you know, incredible products with

24:51 all these

24:52 integrations and all these, you know, distribution, et cetera, would say you

24:56 and the SaaS apocalypse,

24:56 how do you make sense of it? I think both things are true. So I have a, I would

25:01 give you a different

25:01 rubric to think about it, which is what software is really fundamentally about

25:07 beta, and what software

25:08 is about alpha. And I think that the software that's about beta is going to

25:12 really struggle that,

25:13 you know, this is, this is software that made you more similar to everyone else

25:16 . And this has been

25:17 my historical critique of the software industrial complex, which is that the

25:21 feedback loop for the

25:22 people building the software is, can I sell it? Not did it, did it add value,

25:26 which is downstream of

25:28 can you sell it? And, and so the, you can think of almost like vibe coding, the

25:33 , the advent of AI,

25:35 it allows you to make software that's specific to you. It's inherently alpha

25:38 focused,

25:40 if you do it right. But I, so I think that the platforms that are already

25:44 focused on alpha are

25:45 going to continue to have an advantage. It's like actually going to be a win

25:49 that fills their

25:50 sale up on the other side that the stuff that's all beta is really going to

25:55 struggle. And you can

25:56 almost argue like maybe the beta wasn't that valuable to begin with. But one of

25:59 the jarring

26:00 moments for me was in COVID, if you really look back at what were CEOs talking

26:05 about in their earnings

26:06 calls about software, they, no one talked about the $5 billion ERP

26:11 implementation they did that

26:13 saved their supply chain because all of them fell over like paper tigers in two

26:16 weeks.

26:16 And what they were talking about was zoom and teams and how that enabled them

26:21 to go remote.

26:22 And you cannot think of like that's crazy. That is, that should have been a

26:26 sputnik moment for the software industry to say, wow, we haven't built shit

26:30 that's valuable.

26:32 How depressing. And on the flip side for us, at least COVID was a huge tailwind

26:39 because it's

26:40 specifically because we were able to help our customers adapt to this reality

26:44 at the speed

26:45 of the disruption. I think it kind of separated the wheat from the chaff. And I

26:49 think we'll kind of

26:50 see that, you know, maybe there's a lot of things we've been spending on almost

26:53 mimetically like,

26:53 well, other people use it for this. This is a standard industry solution for X.

26:58 Those things

26:58 are going to feel a lot of pressure. And on the on the flip side, they're going

27:02 to be software.

27:03 It's almost almost like a toolkit and approach that allows you to express how

27:07 you're more

27:08 different than other companies. It's almost become software that allows you to

27:11 express your

27:11 competitive advantage. Your strategy is going to be a premium on the day to

27:16 stuff like vibe

27:17 coding. You can't do it. I think that's actually true. Like, I think it's

27:19 actually true that day

27:20 to is much harder, a lot of it's unsolved, and you're gonna have to figure that

27:24 out. But I don't

27:25 think that's going to preclude the pressure on the beta on the beta software.

27:28 Yeah. In terms of a crewing value, right now it seems like the hardware layer

27:36 has the highest

27:37 margins, whereas in the internet economy, the applications had the highest

27:41 margins. I'm curious

27:42 if you think, and I will be like the internet where sort of the entities that

27:47 control the

27:48 and user relationships crew the most value, or if you think it'll be more like

27:52 the cloud

27:53 where the infrastructure layer occurs the most value or has the highest margins

27:58 . How do you

27:59 think about how to play out? If you thought about the stack as chips, models,

28:05 AI infrastructure,

28:07 AI applications, what I see happening empirically is the models are being

28:12 commoditized and always

28:14 under pressure. The model companies are expanding up. Sometimes they almost

28:18 call it in a diminutive

28:19 way a harness, but it's actually they're building software around it that is AI

28:23 infrastructure

28:23 to do something like code. And then the people who started as narrow vertical

28:30 AI solutions are

28:31 kind of earning their way down this stack to realize like, oh, I need this

28:34 actual AI infrastructure

28:35 to be able to scale to my customer base and handle more use cases. So our

28:39 theory has always

28:40 been the value is going to accrue in two places at the chips layer and at the

28:43 AI infrastructure

28:44 layer, what we would call ontology. But those two layers, I think, are going to

28:48 be pretty defensible.

28:49 There's this funny chart in the economist the other day about what's going to

28:53 happen to the

28:53 economy. And it gives three three predicts either everything goes to the, you

28:58 know, goes vertical,

28:59 AGI, either, you know, we're all dead or, you know, we're all economically dead

29:05 , everything

29:06 collapses, or, you know, 2% growth. And so, you know, the economist is is has

29:13 hedging just like many

29:14 others. I'm curious how you what's your sort of mental model for what AI is

29:19 going to do to the

29:19 economy in terms of, you know, the productivity stats and GDP growth, but then

29:24 also the the job

29:26 market. And I mean, AI as it achieves its goals over the medium term. And we

29:30 sort of, you know,

29:32 start to reach the potential that people have been talking about, you know,

29:34 people say AI 2027,

29:36 it's even if it's 2030. How do you think it's going to impact the economy?

29:39 I have a lot of thoughts here. So hopefully we'll hit them all and I won't

29:41 forget them as we go

29:42 through this. So the first bit is what always irks me about how we talk about

29:46 AI is as if somehow

29:48 we have no human agency. AI is going to do X. No, that's not right. Humans are

29:53 going to use AI

29:54 to do X. There's a choice here. Do we want to invest in AI slop and essentially

29:59 AI slot to borrow

30:00 an expression from from John Colson, Patrick Colson? No, I think that these

30:05 things are,

30:06 I don't want to invest in that at least. So what is our normative view of why

30:10 AI is valuable?

30:11 How does it result in American prosperity? How does it make our society better,

30:16 not worse,

30:17 and restoring the fact that we have agency and therefore an obligation to steer

30:22 this

30:22 in a specific way? So that's the first part of it. Then if, okay, if we have

30:27 agency,

30:27 what is that? You know, my view of this is we have a historic opportunity to

30:32 fix the fundamental

30:34 breakdown that happened in the 70s between wage growth and GDP growth. That

30:39 this should

30:40 be if we look at just the example we have about the Intel Warren officer who's

30:44 suddenly able to do

30:45 so much. Well, I see that playing out on the ICU floor. I see that playing out

30:48 on the factory floor.

30:49 There is an opportunity to give the American worker superpowers with AI. It's

30:53 David Slingshot

30:54 in a world where the Chinese Goliath has been this giant sucking sound of

30:58 American prosperity.

30:59 If we do that, it's a basis for underwriting the re-industrialization of the

31:05 country.

31:06 And that we're not going to do this symmetrically. That's why it's a slingshot.

31:10 It's not like

31:10 hey, this is how they do it there. We're going to do this here. It's actually

31:13 we're going to do

31:13 this in entirely new ways like Hadrian is a perfect example of that, right? We

31:17 are re-industrializing

31:19 using technology, making these people 50 100 times more productive than it

31:23 could be otherwise,

31:25 and it's going to lead to all sorts of new possibilities in particular because

31:29 I think the great

31:30 lie of globalization is that we can do the innovation over here and we're going

31:34 to have the

31:34 production go over there. But guess what? Innovation is a consequence of

31:38 productivity. If you don't

31:39 make the thing, you can't innovate on how you make the thing and what the thing

31:43 is. You see that

31:44 with SpaceX, there's a reason the R&D engineers are co-located on the

31:47 production floor. What is the

31:48 feedback loop and cycle time they expect to come out of that? And you see that

31:51 in the negative where

31:53 we used to think who she was just some cheap set of pipetting arms for contract

31:57 pharmaceutical

31:58 research and now 50% of all clinical trials are being done in China. And so I

32:02 think we should

32:03 view this as a national emergency and a national opportunity around AI. And

32:08 what concerns me

32:09 a bit, these technology revolutions are usually by the vast vast majority are

32:17 tool revolutions, not concept revolutions. It was not Galileo who invented the

32:22 telescope.

32:23 He used it to discover planetary motion. It was the future of these

32:27 technologies,

32:28 the microscope, the power loom, the telescope, the personal computer. They are

32:32 determined

32:33 not by the inventor of the technology, but by the people who wield the

32:36 technology. Today,

32:38 when we listen to the AI doomerism, we're listening to the inventors who are

32:41 incredibly smart,

32:42 but just like their creations, they have their own jagged intelligence. Just

32:46 because

32:46 they were smart at building the model doesn't mean they're going to be right

32:49 about the implications

32:50 of it. And then we are implicitly giving up our own human agency and how to

32:55 steer it. It is us

32:57 as the wielders of it that are going to determine the future course of this

33:00 technology. And I see,

33:01 you know, maybe the most authentic thing about that economist graph is those

33:05 range of outcomes

33:06 are exactly what's possible. And it's up to us to pick which one we want to be

33:10 on. It's a choice

33:12 we're making. It's not something that's being done to us. Yeah. No, you just

33:15 said something so

33:16 interesting that I think is is underexplored, which is the co-location of R&D

33:21 and production,

33:22 which is something we very much understood that was sort of, you know, that's

33:25 the as the Henry Ford

33:27 style, right? Like that's how we used to build things in the physical world.

33:30 And then, of course,

33:31 globalization led to this sort of separation of them. And even you still see it

33:34 in companies,

33:35 right? It's like the engineering team in many companies is not the same as the

33:38 production

33:39 team. You can be a production company or an engineering company. We see this a

33:42 lot in our

33:42 American dynamism portfolio. What was the impetus for sort of that

33:47 philosophical division?

33:49 You know, I think a lot of people point to policy changes in the 90s. But what

33:53 was like the real

33:54 impetus from your research that sort of led to this sort of divorce between

33:58 production

33:58 and engineering? And how are you seeing it come back together again in

34:01 companies today?

34:03 Europe has created exactly zero companies from scratch in the last 50 years

34:07 worth more than 100

34:08 billion euro. We have created all of our trillion dollar companies from scratch

34:12 in America in the

34:13 last 50 years. The difference is founders. You know, you have really good

34:18 companies over there,

34:19 but they're like 300 years old, 100 years old, whatever it is. We kind of had

34:26 the

34:26 Europeanization of our mega cap companies until recently, you know, Intel at

34:32 some point. There

34:33 was this fork in the road where they could they could have promoted their CFO

34:37 to be the CEO or

34:39 Pat Galstinger as CTO. Back then this is before he came back later to be CEO.

34:45 Who did they pick?

34:45 They picked the CFO, the person that Wall Street would understand. Not the

34:50 person who could actually

34:51 determine the future. And by the way, it really looked like it was working for

34:54 10 years until it

34:55 fell off a cliff. But that was all financial engineering, not real engineering.

34:58 You know, when was the the last Boeing CEO to be an engineer? I think was 2004.

35:04 You know, there was a way there. So you think about there's a period of time in

35:07 our economy

35:07 where we understood that the engineering was leaving these things. Elon says

35:10 the pathway to

35:11 the CEO is through the CTO, which sounds like a crazy, heretical statement

35:15 because like certainly

35:16 from my generation, the way we grew up, that's not true. That's not how that's

35:20 not what we were

35:20 taught. Now I'm not saying that because I'm the CTO here. Don't read anything

35:25 into that. It just

35:26 means that like, if you want to infer that you're breaking news today, Cindy

35:30 Grove, who was the

35:33 president of Intel, used to start his annual sales and marketing kickoff

35:36 meeting by reminding

35:38 all the salespeople, just remember, it's the engineers who create all the value

35:41 . You guys just

35:42 move it around. You know, it doesn't mean the salespeople aren't important or

35:46 aren't necessary,

35:47 but there is kind of a sequencing here. And I think we kind of got very

35:50 confused about that,

35:51 that we became very good at financial engineering and forgot about engineering.

35:54 One thing one of our portfolio CEO said is that maybe salespeople are the least

36:00 AIable in terms of being able to be automated or replace them. I'm curious how

36:05 you think about

36:06 sort of the jobs at tech companies, you know, leverage with AI, how you're

36:12 using in your own

36:12 income, you know, in penalty or how do you think about that? Yeah. So one part

36:18 I was going to say

36:19 from earlier that I think is relevant to this is, you know, Pascal said every

36:24 human has a God-shaped

36:26 hole in their heart. And part of the potential pathology from the labs is that

36:31 they have filled

36:32 that hole with AGI. And so there are things that they assert as empirical that

36:37 are actually

36:37 articles of faith. They may be true. They may not be true. I don't know, but

36:41 they get confused

36:42 between what's an article of faith and what is actually an empirical reality.

36:45 And so if you view this through a very pragmatic clear-eyed view, say the sales

36:49 people,

36:50 I'm not sure why the goal is replacing people to begin with. Like, isn't the

36:53 goal to win?

36:54 Isn't the goal to be dominant in your industry? You want to be better. And so

37:00 maybe being better

37:01 is about a mixed mammal AI teaming. It's, you know, how do I build the Ironman

37:05 suit for the sales

37:06 people I do have? How do I make the best salespeople more productive and system

37:09 atize? What is it that

37:10 makes them good for everyone else? Like, there's all sorts of other ways of

37:14 thinking about the

37:14 problem if your goal is winning. But if your goal is AGI, it's the aesthetic of

37:19 the fact that you

37:20 couldn't replace the person with this model is offensive. And you're just going

37:24 to, you know,

37:25 you're just going to keep driving at that. And I think it could be a

37:28 distraction. This is one

37:30 way in which I think the Chinese do have a little bit of an advantage, which is

37:32 , first of all,

37:34 just to be clear, I'd bet on us 100 times out of 100. But they have a pragmatic

37:39 approach like

37:40 the whole point of AGI is to win. It's not AGI. How do I improve my productive

37:47 forces, as they would

37:48 call it? Now, I think that the good news is if you look at the people who wield

37:52 the technology in

37:53 America, that's what they're focused on. You know, the CEOs I deal with, none

37:57 of them have asked me,

37:59 "Hey, I want to fire a bunch of people." Or, "How do I get rid of these people

38:02 ?" Maybe because we're

38:03 too expensive for that sort of bullshit use case. But they come to me and say,

38:06 "I want to dominate

38:07 my industry. I want to destroy my competition." Okay, great. So the ambition is

38:12 there. And maybe

38:13 you get more efficient by doing it. But actually, the whole point is to grow

38:16 massively. So that

38:18 ambition sets the frame of how you're going to apply the technology and what

38:20 sort of solutions

38:22 you find valuable. Speaking of China, how do we win the AI race, particularly

38:28 as it moves towards

38:29 more physical AI and robotics, et cetera? What are the things to make sure we

38:33 get right or the

38:34 things we need to we need to fix? I think our biggest risk as a country is

38:38 suicide, not homicide.

38:40 You know, I have anyone who knows me knows I'm a big China hawk. And I think

38:44 part of the challenge

38:45 with China is it's not enough for the CCP to prosper. America must also fall.

38:50 Like, look, if you want

38:51 to buy our soybeans or not, I don't begrudge you. That's a business decision.

38:54 That's free trade.

38:55 Great. But when you're trying to smuggle in agricultural funguses so that we

38:58 can't grow

38:58 soybeans, that's a different ball game altogether. And that that offends my

39:01 kind of

39:02 American Calvinist sensibilities of fair play. But so all that said, that would

39:07 make it seem

39:08 like I care a lot about homicide. But I think our problem is actually one of

39:11 national will and

39:12 focus. And like, are we actually addressing the problems that we face here? Are

39:16 we encouraging

39:17 the agency and our people to believe the world can be better? You know, this

39:21 manifests in a sense

39:22 of just kind of like nihilism and polarization that we forget what makes us,

39:26 what unites us,

39:28 and we focus on what divides us. And there's this kind of sense like, hey,

39:32 nothing really works and

39:34 doesn't really matter. So let's just burn it all down. And a big part of like

39:39 how I think of what

39:40 Pounder does in the world is it is about the legitimacy of our institutions,

39:45 like whether it's

39:46 doors falling off planes or basic government services working, these

39:48 institutions should all

39:50 work excellently. In the absence of them working, it breeds this nihilism. And

39:55 then you get the wrong

39:57 reaction to it. So that's what that's what I think we should focus on

39:59 addressing. So now,

40:01 to physical AI, the point of having an ambition, like let's re-industrialize,

40:04 let's be maximalist

40:06 about this, not some sort of half measure that's like a little bit of friend sh

40:08 oring here or there,

40:09 whatever. It's like, no, we invented all of these technologies. We invented

40:13 mass production. We

40:15 invented nuclear power thing after thing after thing. It's like the idea that

40:18 somehow the American

40:19 people are not capable of this thing. How is that beggar's belief? Right? So I

40:23 think it's actually

40:23 about will and motivation and leadership. Yeah. No, I think that that is such a

40:28 good segue into

40:30 what you're doing in terms of building culture. Because I think this is

40:33 something that

40:34 is overlooked. A lot of people think it's a technical problem or production

40:37 problem. And I

40:38 agree with you that I think it's a seriousness and a will problem. And you have

40:43 now been investing

40:44 in film, which is totally different than what you do at Palantier. So I would

40:48 love to understand,

40:50 you know, why did you start a film production company? And how do you think

40:54 that's ultimately

40:55 going to change the culture around having more will about doing these hard

40:58 things?

40:59 Well, it really starts with my own assimilation journey. I came to the U.S. as

41:04 a young child. I

41:05 settled in Orlando. And my assimilation journey, as a four-year-old, five-year-

41:10 old,

41:10 was watching movies with Dad on the couch. And what were the movies of the '80s

41:14 and '90s? It was

41:15 Hunt for Red October and Red Dawn and Rambo 2 and 3. And you know, I like to

41:21 say, as a five-year-old,

41:22 I knew what it felt like to be an American before I knew civics. That was way,

41:25 way down the line.

41:26 And I think a lot of people experience that again after a long period of time

41:31 when they

41:31 watched Top Gun Maverick. And so, you know, we sometimes over-intellectualize

41:35 these things,

41:36 like there's a feeling to it. Even subtle things like I heard from the guy who

41:40 made the movie 300,

41:42 that after 300 came out, Navy SEAL recruitment went through the roof. And he

41:46 was kind of perplexed.

41:48 Obviously, it's a movie about Spartans. What does this have to do with Navy

41:52 SEALs? But it

41:53 clearly inspired so many people to be like, "I want to look like that. I want

41:56 to be that strong.

41:56 I want to be that heroic." You know? And so, the virtue of entertainment as,

42:01 first of all,

42:01 it's got to be entertaining. It's not probed out here. But then it lets us

42:04 reflect on ourselves.

42:06 And who do we want to be and what do we want to be like? And if our

42:09 entertainment is all

42:11 Terminator, AI ruins the world, technology is a force of evil. It's all dystop

42:16 ic future scenarios.

42:17 That sets us sort of condition, which I would juxtapose to my youth in Orlando

42:21 growing up in

42:22 the shadow of the Space Coast, which was just like science and technology is

42:25 amazing. And we're

42:26 going to be living on other planets. And, you know, as a sixth grader, write a

42:29 report on how

42:30 we're going to get to Mars. And it just inculcates a fundamental belief that

42:34 the future will be better,

42:36 and that science, technology, the will to invest in these hard problems is

42:40 worth it.

42:40 It's worth it. And so, I think we have a moment to reclaim storytelling in a

42:45 way that's both

42:46 entertaining and inspiring. Yeah. No, I love that you point out that you grew

42:49 up in Orlando. It's funny.

42:51 I also grew up in Florida in the '80s and '90s. And it was like, you know,

42:55 people made fun of

42:56 Florida, right? Like, they didn't understand if there was anything good there.

42:58 But Orlando,

42:59 I mean, you say it's the shadow of the Space Coast. It's also Disney World,

43:02 right? It's the best

43:04 stories of a century. It's American culture. And so, I'd love to, I mean, were

43:10 you a Disney

43:10 kid? I mean, like, was that something that, like, also spoke to you of, like,

43:14 these stories of

43:15 good and evil that are passed down through cartoons? Absolutely.

43:17 How did that help you? So, also, I mean, there's a part of the story, which is

43:23 the business. Like,

43:24 why did we end up in Orlando? You know, after we fled violence in Nigeria, my

43:29 dad had a childhood

43:30 friend who was living in LA who sold knickknacks at theme parks. And he's like,

43:34 "Hey, look, I know

43:34 this horrible thing just happened to you. There's this up-and-coming place with

43:38 theme parks. I don't

43:38 live there. I need someone I trust there. Why don't you go to Orlando?"

43:42 So, literally, it's not just that, was I a Disney kid, my parents' job was to

43:48 provide knickknacks

43:49 in the theme park stores. So, like, after school, they would take me to Sea

43:53 World and I would pet

43:54 the stingrays while they restock the shelves. You know? And so, I grew up very

43:59 much in vibing

43:59 the storytelling, the aspect. I mean, Epcot. Epcot was all about painting an

44:04 optimistic vision

44:05 of the future and what technology was going to be like and the stories of

44:08 heroes and, you know,

44:09 that there was both, there was both evil and bad in the world and there were

44:13 clearly heroic

44:14 actions that you could take and it was all super inspiring.

44:17 And I mean, that's what's so interesting too is like, I feel like the height of

44:23 the sort of

44:24 good and evil battle inside of Disney film was sort of the Lion King 90s, right

44:29 ? Like, I mean,

44:29 they were different films then. How do we get back? I mean, maybe it's not back

44:33 , maybe it's

44:34 forward, but how do we get back to those stories for children, for people to

44:38 feel optimistic again?

44:40 I mean, you know, in some ways you don't hear, oh, I grew up in this city and

44:44 it's the height of

44:44 optimism. You don't hear that about California anymore. What will it take for

44:49 movies to transform

44:50 that? Well, you know, this is a kind of a personal opinion, but I'm really

44:55 excited that

44:56 David Ellison is going to have Warner Brothers because, you know, if you think

45:00 about Hollywood,

45:01 the original studio heads, they were founders. Like Jack Warner, in the 30s,

45:05 Germany was the

45:06 third largest export market for American entertainment and the Nazis actually

45:11 deployed sensors into

45:12 Hollywood to control what was being made and every studio capitulated except

45:17 for Warner Brothers.

45:18 Jack Warner was the only person willing to stand up and speak truth and only a

45:23 founder can do that

45:23 because if you're a professional CEO who's employed, you can't survive that.

45:29 And so I think in some ways there's a mirror to the present day Hollywood and

45:34 the

45:34 defense industrial base. It's conformity. It's kind of lack of opinion. It's

45:39 lack of a normative

45:39 view of what is it trying to communicate. Then if you go a little bit further

45:44 down the line,

45:44 you look at the Vietnam era, like we had very cynical Hollywood content in

45:50 Vietnam as a reflection

45:51 of how we felt about ourselves. In '73, George Lucas made American graffiti

45:56 because he was tired

45:57 of it. He's like, I'm tired of it. I just want to make a movie about boys

46:00 driving cars chasing girls.

46:01 And it was a palate cleanser that kind of like, yeah, the American people

46:05 remembered like,

46:06 okay, we went through our period, our cycle of grief and cynicism and we're

46:10 ready and it set

46:11 the conditions for the movies of the 80s and 90s that we all love. I think we

46:15 're also kind of tired

46:17 of it right now. We're tired of the cynicism that everything's going to be

46:21 worse. And you see

46:22 that in the performance of stories, Top Gun Mavericks, the easy one to point to

46:26 , but the content that's

46:28 doing well right now is American oriented. There are inspirational figures. The

46:33 heroes aren't anti-heroes

46:35 that are drug addicts that you wouldn't want your kids to grow up to be. There

46:37 's actually some sense

46:38 of inspiration in it and some pride in terms of who we are and how that

46:43 reflects in the entertainment

46:45 itself. So I think we're, if you think about the next two to 10 years, we're

46:50 going to see a lot

46:52 of content like that. That's what I see in the development pipeline from these

46:54 studios themselves.

46:55 That's exciting. So maybe talk to us about some of your projects or things you

46:59 're working on and

47:00 what you're most excited about. But then also, yeah, that's so interesting. We

47:02 've had these

47:03 conversations, Eric and I with Mark and others about how the pipeline for the

47:07 last 10 years has been

47:08 Dower. We're sort of getting the end of the pipeline. You're seeing it in sort

47:11 of the Oscar

47:12 nominees this year. It's like they're not optimistic. But as you said, like

47:16 maybe 10 years from now,

47:17 we're going to see this pipeline of just like pro America exciting, optimistic,

47:21 enthusiastic,

47:22 golden age sort of content. What are you seeing and what most excites you?

47:26 Well, you know, I don't want to give away too much of my own development

47:30 pipeline here.

47:31 But I would say like you see like Call of Duty is being made right now with by

47:36 Pete Berg

47:36 and Taylor Sheridan. You see the entire Taylor Sheridan universe. I mean, and

47:40 talk about

47:40 Sicario. It was like 2014. Sicario has basically came to life with the new the

47:46 Jalisco new generation

47:47 cartel, right? Like, like there's a sense in which the storytellers have

47:54 exactly the frame that were

47:56 that we're kind of excited about. I think there's recent events are very

47:59 interesting to tell stories

48:00 about right now. Even a movie like War Machine, which just came out from

48:06 watching the trailers,

48:08 it was not yet clear. You can imagine like five years ago that that the

48:12 storyline would have

48:12 been something more like the US government build evil robots that the human

48:16 soldiers had to defeat.

48:17 In this case, it was more like aliens basically sorry to give away the plot if

48:23 you haven't seen

48:23 it. But you know, you have an alien robot and brave Americans, rangers have to

48:27 defeat them and

48:28 do through their own ingenuity. That that itself is I think showing you it's it

48:32 's belies the shift

48:34 in the narrative and storytelling that's happening. I think world events remind

48:38 us that there are

48:38 actually there is actually evil out there. Russian tanks can just roll across

48:43 the border. You know,

48:45 October 7th, just horrendous barbarism is still possible that these things don

48:50 't maintain themselves.

48:51 So what would I like I think so putting this back in a geopolitical lens, you

48:59 know,

48:59 as much as I've been saying, we shouldn't call China near peer, we should call

49:02 them peer because

49:03 calling them near peers like a shiplet that lets us off the hook. When you look

49:08 at

49:08 operations like Maduro or midnight hammer, it's hard to think of more you have

49:14 done to restore

49:15 deterrence in the world. A reminder that we do have the will maybe because we

49:19 didn't have the

49:20 will you forgot that we had the capability, but we have both the capability and

49:23 the will to do

49:25 things that are quite amazing. At the same time, it signals a very obvious

49:29 truth, which is somehow

49:30 none of the Russian and Chinese shit worked. So if you're a third party country

49:34 and you're thinking

49:34 about how what is the future of the world and how do you want to be allied and

49:38 maybe you've been

49:40 hedging because you've been seeing America and retreat, it's also a reminder

49:44 that the Chinese

49:44 did not come to save Maduro and none of the equipment they provided actually

49:48 seemed to do

49:49 anything. So is that really an option for you? So I give you the geopolitical

49:54 answer, but I think

49:56 okay, here are some some projects I think I can share. So Oppenheimer was

50:00 hugely successful and

50:01 complicated right where it's like it's three dimensional to the point of

50:05 entertainment. It's not

50:06 probed. I think there's a very powerful story in Hyman Rickover and the birth

50:12 of the nuclear

50:13 navy. We talked a little bit about him, but you know what I love about Rickover

50:16 is he was born

50:16 in a shuttle in Poland came came over at the age of six, one of these near

50:21 almost near missed

50:22 stories where they were on Ellis Island and you get to Ellis Island, you have

50:25 10 days for someone

50:26 to come pick you up. And so his mother sent a telegram. It gave someone money

50:30 to send a telegram

50:31 to the father who was already here to come get them. The guy pocketed the money

50:34 for the telegram.

50:36 On day 10, someone happens to arrive that they know from the old world who then

50:41 runs out, gets

50:42 the father to claim them, get buys them one extra day. So on day 11, they get

50:46 picked up.

50:46 But you know, this near miss where we almost didn't have Rickover, but Rickover

50:50 was a notoriously

50:52 difficult personality. So five foot two short short guy in World War II, he

50:56 drove a coal ship,

50:57 not he had no it was not a prestigious post. But after World War II, he was

51:00 sent to Oakridge

51:01 at the the vestiges of the Manhattan Project. And he was inspired to get this

51:05 idea of putting

51:06 nuclear power inside of the submarines. Because before then, submarines sucked.

51:11 They they could

51:11 go underwater for like an hour. They were diesel power. They were allowed. They

51:15 were basically

51:16 surface ships that could occasionally submerge. And after that, they became

51:20 really exquisite.

51:21 And I think he built the first one in like five years, six years, the Nautilus.

51:25 But the Navy didn't want him to succeed. They not only did Oppenheimer think it

51:30 was a stupid idea,

51:31 the Navy did too. His first office was a women's restaurant. I kid you not, you

51:35 know, it's like,

51:36 how can we humiliate this guy to quit? And he just kept going. And what I think

51:40 is interesting when

51:40 you look at his memoirs, like, it's not that he was immune to the humiliation.

51:45 He felt every

51:46 slight and insult, he documented them. But somehow he was able to channel that

51:50 into something he was

51:52 going to push through and still succeed despite that. Zumwalt, who was the

51:56 chief of naval operations,

51:57 the senior most uniform person in the Navy said, the Navy has three enemies,

52:03 the Soviet Union,

52:04 the Air Force, and Hyman Rickover, the zone Admiral. The other thing that I

52:08 think fits very closely

52:09 with 18 theses, like Hyman Rickover was a four star Admiral for 30 years. That

52:14 is something we

52:14 can't even contemplate today. We almost view our officers as cogs to keep

52:21 moving around.

52:22 Every two to three years you have to keep moving. As the first director of the

52:26 nuclear Navy of

52:27 naval reactors, you know, he was in that role for a very long time. But that

52:32 role, even today,

52:33 is an eight year stint, but shows you the primacy of people that we understand

52:37 with something this

52:38 exquisite, something where this much knowledge and continuity matters, you don

52:42 't just keep pulling

52:43 people out every two or three years. And our ships or subs are the safest in

52:50 the world by launch on.

52:51 So every six months, the Soviet Submariners would get six months of respite at

52:57 Sochi to recover

52:58 their white blood cell count because they were getting irradiated. We've had no

53:03 deaths due to

53:04 radiation. You know, he built it with the specification, this has to be safe

53:08 enough for my son. He built

53:09 it to a specification that is 100 times safer than the minimum safety standard.

53:14 And that is this

53:15 sort of aspiration only a founder could have. Yeah, this is going to be a great

53:18 movie.

53:18 What else is there to do? And you're doing everything. So what else is on the

53:25 radar for you?

53:25 Well, you know, I in some sense, maybe it's enough, but all of these things

53:31 have a through line where

53:32 it's really about American greatness and inspiring the next generation. And it

53:35 's driven home to me

53:37 when I think about my kids and recognizing that the America I grew up in

53:41 is something that every generation has had to fight for. And I'm in that phase

53:45 now where I'm

53:46 fighting for the prosperity that the next generation ought to have. And so

53:50 whether it's

53:51 soft power and inspiration and movies or hard power and deterrence of

53:55 adversaries and preventing

53:56 World War III, it's all about American greatness and the prosperity of the

53:59 American people.

54:07 Thank you.

Love

· 2d

《Stableminded》是一档聚焦前沿创业者的财经播客。内容深入详实，以创业者为核心，全面探讨现代金融。

新一期节目将于下周上线。

第45集 - @cnaut，@raincards 首席技术官

第46集 - @louisamira，@atxp_ai 创始人

第47集 - @nvs，@infinite 首席执行官

由 @altitude 呈现

主持人：@dr3wrogers

集数：0045-47

美国旧金山 #btc #usd1 #usdc

加载中...