章节 1：极速文化与物理极限——xAI 的核心哲学

📝 本节摘要：

本章作为访谈的开篇，包含了一段高能的“冷开场”剪辑和正式访谈的开始。受访者 Sulaiman Ghori (Sully) 分享了几个极具代表性的故事：Tyler 通过在24小时内完成 GPU 训练赢得 Cybertruck 的赌约、Sully 入职第一天“无人发号施令”的自由，以及每行代码提交（commit）带来的惊人价值估算。随后，Sully 阐述了 xAI 的核心运作模式：这里没有截止日期（因为“截止日期永远是昨天”），一切以物理极限而非人为限制为基准，并强调了公司目前正处于硬件受限但软件人才密度极高的独特阶段。

[原文] [Sulaiman Ghori]: tyler took this bet with Elon like get a Cybert truck tonight if you can get a training run on these GPUs in 24 hours and we were training that night did he get the Cyber Truck yeah he got Cyber Truck

[译文] [Sulaiman Ghori]: Tyler 和 Elon 打了个赌，大概是“如果你能在 24 小时内让这些 GPU 跑起训练任务，今晚你就拿辆 Cybertruck”。结果我们当晚就在训练了。他拿到 Cybertruck 了吗？是的，他拿到了。

[原文] [Sulaiman Ghori]: my first day they just gave me a laptop and a badge and I was like okay now what i don't even have a team i've not been told what to do ask Rock was spinning up at the time our integrations with X they're like can you help and I was like yes

[译文] [Sulaiman Ghori]: 我入职第一天，他们只给了我一台笔记本电脑和一个工牌，我就想“好吧，现在干嘛？”我甚至没有团队，也没人告诉我该做什么。当时 AsRock（注：可能指 Grok 相关的早期项目或代号）正在启动我们与 X（推特）的集成，他们问：“你能帮忙吗？”我说：“行。”

[原文] [Sulaiman Ghori]: what's the most fun thing about working there no one tells me no if I have a good idea I can usually go and implement it that same day and show it to Elon or whoever and we got an answer

[译文] [Sulaiman Ghori]: 在那里工作最有趣的事情是什么？没人会对我说“不”。如果我有一个好点子，通常当天就可以去实现它，然后展示给 Elon 或其他人看，我们会立刻得到一个答复。

[原文] [Sulaiman Ghori]: we did the math right now we're I think at about $2.5 million per commit to the main refill and I did five today so you added like 12 and a half million of value the levers are extremely strong

[译文] [Sulaiman Ghori]: 我们算了一笔账，现在我觉得我们每次向主仓库（注：原文 main refill 应指 main repo）提交代码大约价值 250 万美元，而我今天提交了五次。所以你大概增加了 1250 万美元的价值。这里的杠杆效应极强。

[原文] [Host]: today I have the pleasure of sitting down with Sully Kongori and he is one of the engineers at XAI i've been kind of fascinated by XAI since like 2023 when like Elon first started i think it's like one of the fastest growing companies of all time can you just talk about like what the fuck is happening at XAI

[译文] [Host]: 今天我很荣幸能与 Sully Kongori 坐下来聊聊，他是 xAI 的工程师之一。自从 2023 年 Elon 刚创立 xAI 以来，我就对这家公司很着迷。我认为它是史上增长最快的公司之一。你能聊聊 xAI 到底在搞什么名堂（what the fuck is happening）吗？

[原文] [Sulaiman Ghori]: yeah um we don't have really due dates it's always yesterday um there's no blockers for anything like at least nothing artificial uh the whole Elon thing about going down to the root uh the fundamental whatever the physical thing is we get there pretty quick if we can as quick as we can

[译文] [Sulaiman Ghori]: 好的，嗯，我们其实没有什么截止日期，因为截止日期永远是“昨天”。没有任何事情会有阻碍，至少没有什么人为的阻碍。你知道 Elon 那一套关于追根溯源的理念，无论基础是什么，无论物理实体是什么，只要我们能做到，我们会尽可能快地触达本质。

[原文] [Sulaiman Ghori]: which is funny in software it's not really like a thing that you think about is the physics too much but we do try quite a bit and we're not really fully a software company given all the infrastructure pull down um it's kind of hardware at this point it's like hardware constrained it's the probably our biggest edge is is is the hardware because nobody else is even close on on the deployment there

[译文] [Sulaiman Ghori]: 这很有趣，因为在软件领域，你通常不会太去考虑物理层面的东西，但我们确实做了很多尝试。而且鉴于所有的基础设施建设，我们并不完全是一家软件公司。在目前这个阶段，有点像是硬件公司，我们受限于硬件。这可能是我们最大的优势——硬件，因为在部署方面没有人能接近我们的水平。

[原文] [Sulaiman Ghori]: um although the talent density on software is like incredible i've never been anywhere like that it's it's really cool for Elon he is very good at figuring out like what the bottlenecks will be even like a couple months or even years in the future and then trying to work backwards from that and make sure that like he's in a really good position

[译文] [Sulaiman Ghori]: 嗯，尽管我们在软件方面的人才密度令人难以置信，我从未在那样的地方工作过，真的很酷。对于 Elon 来说，他非常擅长找出未来几个月甚至几年的瓶颈是什么，然后尝试从那个点倒推工作，确保他处于一个非常有利的位置。

[原文] [Host]: um how does that work dayto-day with just normal people like at XAI and like adopting that kind of mental framework

[译文] [Host]: 嗯，那在 xAI 的日常工作中，普通员工是如何适应这种思维框架的？

[原文] [Sulaiman Ghori]: usually when we spin something up new very quickly either one of us or he comes up with this uh metric that's usually very core to either the the financial or the physical return or both sometimes um and so everything is just focused on driving that that metric

[译文] [Sulaiman Ghori]: 通常当我们快速启动新项目时，要么是我们中的一员，要么是他（Elon），会提出一个指标（metric），这个指标通常对财务回报或物理回报（有时是两者）非常核心。所以一切工作都专注于推动那个指标。

[原文] [Sulaiman Ghori]: um there's never like a fundamental limitation to it or like whatever the fundamental limitation is it better be rooted deep down and not something artificial um and there is a lot of uh perceived limitations um especially in the software world coming from like especially in the last 10 years of like web dev and all these kinds of things people just assume or accept certain limitations especially when it comes to speed and latency and they're not true

[译文] [Sulaiman Ghori]: 嗯，从来没有什么根本性的限制，或者说，即使有根本性的限制，它最好是根植于深层的（物理极限），而不是某种人为的限制。有很多被“感知到”的限制，特别是在软件世界，源于过去 10 年的网络开发（web dev）这类的东西，人们只是假设或接受了某些限制，特别是在速度和延迟方面，但这些并不是真的。

章节 2：打破常规的迭代速度与“荒野求生”式入职

📝 本节摘要：

在本节中，Sully 揭示了 xAI 极速发展的秘诀：通过消除技术栈中的“愚蠢开销”，往往能实现 2 到 8 倍的效率提升。他提到公司的模型迭代速度达到了惊人的“每日更新”，甚至在硬件上架当天就能开始训练，这完全打破了行业惯例。随后，话题转向 Sully 的个人经历——他最初差点将联合创始人 Greg Yang 的招聘邮件当成垃圾邮件删除。入职第一天，他面对的是典型的“马斯克式”开局：没有团队、没有指令、甚至没有固定工位，全靠自己主动寻找任务，这种扁平化结构让他得以迅速融入核心开发工作。

[原文] [Sulaiman Ghori]: you can get rid of a lot of overhead like there's a lot of stupid stuff in in the stack and if you can knock out a lot of that you can usually two to 8x most anything at least anything invented relatively recently uh some stuff not so much but yeah

[译文] [Sulaiman Ghori]: 你可以去掉很多开销，比如技术栈里有很多愚蠢的东西，如果你能把这些去掉，通常能把大多数东西的效率提升 2 到 8 倍，至少对于那些相对近期发明的东西是这样。有些东西可能没这么多，但确实如此。

[原文] [Host]: when was the last time that you experienced this where there's some like conventional wisdom that says that there's this is the timeline and then you guys just were able to completely shred that

[译文] [Host]: 你最近一次经历这种情况是什么时候？就是那种传统观念认为“这是标准时间线”，然后你们完全粉碎了这个时间线。

[原文] [Sulaiman Ghori]: um most recently it's our model iterations on on macro hard um so we're working on some novel architectures actually multiple at the same time and uh we're coming out with new like iterations like daily sometimes multiple times a day which is from pre-train um in some cases

[译文] [Sulaiman Ghori]: 嗯，最近的一次是我们“Macro Hard”（注：xAI 内部项目代号，戏谑 Microsoft）上的模型迭代。我们正在研究一些新颖的架构，实际上是同时进行多个。我们几乎每天，有时甚至一天多次推出新的迭代版本，有些甚至是直接从预训练阶段出来的。

[原文] [Sulaiman Ghori]: uh which is not something you ordinarily really see but it comes from well a we have a pretty great supercomput team and they've knocked out a lot of the typical barriers it takes to train a lot of this stuff even with how variable our hardware like the hardware is like it's you know within a day of standing up a rack you can usually be training sometimes within the same day um even within uh a few hours in some cases and this is like not normal like normally the timelines are like days or or weeks it takes a lot well in most cases at least yeah in the last 10 years you abstract this away and let Amazon or Google take care of this um and so whatever their capacity is is what their capacity is but that's not like you can't have that be the case and when an AI now so the only solution is to die or uh or build it yourself

[译文] [Sulaiman Ghori]: 这通常是你很少见到的，但这源于：第一，我们有一个非常棒的超算团队，他们消除了很多训练这类东西所需的典型障碍。尽管我们的硬件情况多变，但在架起机柜的一天内，通常就可以开始训练了，有时甚至是当天，某些情况下甚至只需几个小时。这很不正常，通常的时间线是几天或几周。在过去 10 年里，大多数情况下你把这些抽象化了，让 Amazon 或 Google 去处理，所以他们的能力上限就是你的上限。但在现在的 AI 领域，你不能这样。唯一的解决方案要么是死掉，要么是自己建。

[原文] [Host]: can you tell me about like how what your experience was like joining why you joined and then kind of what the like onboarding process was for the first like couple weeks

[译文] [Host]: 能聊聊你加入时的经历吗？为什么加入？最初几周的入职流程是怎样的？

[原文] [Sulaiman Ghori]: yeah so um I was working on my own startup when I moved to the Bay um and actually during that time Greg Yang one of the co-founders of XA had reached out he's great at recruiting as it turns out um what did what did he say uh so I got an email and I thought it was spam because it was I was getting a lot of these like you know emails to founders at the time of like hey you want to chat or like I like what you're doing you want to chat whatever i was going to mark it as spam to like delete it and I saw the domain x.ai i was like oh wait a second i know these guys and they just uh I think it was probably eight months in at that point um and so I was like okay yeah let's chat

[译文] [Sulaiman Ghori]: 好的。当时我搬到湾区做自己的创业公司，期间 xAI 的联合创始人之一 Greg Yang 联系了我——事实证明他非常擅长招聘。通过邮件联系的。我当时收到邮件以为是垃圾邮件，因为那时候作为创始人我会收到很多这类邮件，比如“嘿，想聊聊吗”或者“我很喜欢你在做的事，想聊聊吗”之类的。我正准备把它标记为垃圾邮件删掉，然后我看到了域名 x.ai，我就想“哦，等一下，我知道这帮人”。那时他们大概成立了八个月左右。于是我说：“好吧，聊聊看。”

[原文] [Sulaiman Ghori]: and so we chatted a bunch of times um then uh I wanted an aqua hire but uh I think we were too early at the time and that company kind of went mostly because it was fairly obvious that you can't build macro hard with like a million dollars um but the uh idea was sound so I spent the next like six seven months wasting all my money um building like aerospace projects and working on uh an aerospace astro mining concept um that also I realized like probably wouldn't work but it was worth a try and so I emailed Greg again like uh hey can uh like you want to chat again he's like yeah sure you want to interview tomorrow i was like okay and um uh I apparently did well and I moved on Monday and I started uh then

[译文] [Sulaiman Ghori]: 我们聊了好几次。当时我想做一次人才收购（acqui-hire），但我觉得那时候我们还太早期了。那家公司后来大概也因为很明显你不可能用 100 万美元建成“Macro Hard”而没做成，虽然想法是合理的。所以我接下来的六七个月把钱都浪费在做航空航天项目上了，研究一种太空采矿的概念，后来我也意识到那可能行不通，但值得一试。于是我又给 Greg 发邮件说：“嘿，想再聊聊吗？”他说：“好啊，明天面试怎么样？”我说：“行。”看来我表现不错，那个周一我就搬过去了，然后就开始了。

[原文] [Sulaiman Ghori]: and it was really great um nobody told me what to do so like my first day they just gave me a laptop and a badge and I was like okay um and I was like okay now what and so I went to go find Greg cuz I was like I don't I don't even have a team i've not been told what to do like Greg just brought me on cuz I think he liked what I was doing previously and it was related to what the long term was for macroard which wasn't really even a project at the time

[译文] [Sulaiman Ghori]: 感觉真的很棒。没人告诉我该做什么。我第一天去，他们就给了我一台笔记本电脑和一个工牌，我就想“好吧”。然后我想“现在干嘛？”于是我去把 Greg 找来，因为我甚至没有团队，也没人告诉我做什么。Greg 招我进来大概是因为他喜欢我之前做的事情，而且那与“Macro Hard”的长期目标有关，尽管当时那甚至还算不上一个项目。

[原文] [Sulaiman Ghori]: and I ended up working on actually uh Ascrock was spinning up at the time where our integrations with X and so they're like can you help and I was like yes I can help and so my first week was working uh with the one guy i found out very quickly like everything that we built like I could sit and I could stand up from my desk which I didn't even have a desk assigned to me i just sat at random people's desks that weren't there that day um and I could point to whoever built that thing at XAI um like from my desk it was very very very cool

[译文] [Sulaiman Ghori]: 最后我实际上参与了 AsRock（注：推测为早期项目名）的工作，当时我们正在启动与 X 的集成。他们问：“你能帮忙吗？”我说：“是的，我能帮忙。”所以我第一周是和一个人一起工作的。我很快发现，对于我们构建的所有东西，我只要坐在那里——或者从我的桌子站起来（我其实甚至没有分配到桌子，我只是随便坐在当天不在的人的桌子上）——我就能指得出来在 xAI 是谁做了那个东西。从我的位置看过去，这感觉非常非常酷。

[原文] [Sulaiman Ghori]: and there was like almost no people working there at this point just like a couple hundred right uh yeah about a hundred or so on the engineering staff and then I don't know what the uh uh infra buildout team looked like at that time and it's kind of hard to tell because some people move up the ladder from like the actual building and construction crew onto our payroll but um it was pretty small at the time like much much like an order of magnitude smaller than the other labs um and we had still just done Grock 3

[译文] [Sulaiman Ghori]: 那时候几乎没什么人在那里工作，大概就几百人吧？是的，工程人员大概只有一百人左右。我不清楚当时基础设施建设团队有多少人，这很难判断，因为有些人是从实际的建筑施工队晋升上来进入我们工资单的。但当时规模真的很小，比其他实验室（Labs）要小整整一个数量级。而我们在那时刚刚完成了 Grok 3。

章节 3：122天奇迹与分布式算力——当特斯拉变身超级计算机

📝 本节摘要：

本章深入探讨了 xAI 令人咋舌的基建速度与算力布局。Host 提到 xAI 仅用 122 天便建成了 Colossus 数据中心，这让 Jensen Huang（英伟达 CEO）都赞不绝口。Sully 解释了这种速度如何反过来让团队敢于制定更长远的规划（如提前布局 Grok 4 和 5）。

>

随后，对话揭示了一个极具破坏力的创新概念：“特斯拉计算机”（The Tesla Computer）。为了支撑“Macro Hard”项目所需的百万级“人类模拟器（Human Emulators）”，xAI 计划利用北美 400 万辆特斯拉汽车（特别是搭载 Hardware 4 的车辆）的闲置算力。Sully 将其描述为“数字版的 Optimus”——正如 Optimus 取代人类的物理劳动，这些模拟器将取代人类在电脑前的重复性数字劳动。

[原文] [Host]: one of the things that I kind of love um is how fast XAI went from being founded i remember Elon initially saying like we're not even sure if this can be a success with you know people having you know a multi-year advantage on on speed and like timing and then you guys got done with the first like Colossus data center in like 122 days um and that was just like unheard of and Jensen's out here saying singing the praises of XAI and Elon uh what kind of culture did that allow to be formed

[译文] [Host]: 我非常喜欢的一点是 xAI 从成立到现在的极速发展。我记得 Elon 最初说，考虑到别人在速度和时机上有多年的先发优势，我们甚至不确定这能否成功。然后你们大概只用了 122 天就建成了第一个 Colossus 数据中心，这简直闻所未闻，连 Jensen（黄仁勋）都在这儿对 xAI 和 Elon 赞不绝口。这种经历促成了什么样的文化？

[原文] [Sulaiman Ghori]: it definitely enabled like us on on model and product to kind of assume we would have the resources to do what we needed to do um and that's definitely the case like we're not super duper resource constrained like we've still found a way to push up against that wall um but that's just we have 20 different things going at the same time like more than that like many more things than that there's a absurd amount of of runs and training and all that stuff going on at the same time in parallel usually by like a few a handful of people um which is how we're able to iterate very quickly on on model and product side

[译文] [Sulaiman Ghori]: 这绝对让我们在模型和产品端能够假设：我们将拥有所需的资源去做我们想做的事。事实也确实如此，我们并没有受到非常非常严重的资源限制——虽然我们要做的东西多到还是能触碰到资源的上限——但这只是因为我们要同时进行 20 件不同的事情，甚至更多，多得多。有极其大量的运行和训练任务在同时并行，通常只需要极少数几个人来操作，这就是为什么我们能在模型和产品端快速迭代的原因。

[原文] [Sulaiman Ghori]: um and utilization has definitely been very high the the speed allows us definitely to I guess think more long term um so I think Grock 4 or five really what it was was already planned out and and designed in terms of size and what we expected um way early like before I joined i joined around Gro 3 so it's like thinking at least a year in advance you can yeah you can think much more in advance and assume that those estimates will be hit

[译文] [Sulaiman Ghori]: 嗯，利用率绝对是非常高的。这种速度确实让我们能够更长远地思考。所以我认为 Grok 4 或 Grok 5，实际上在规模和预期方面早就已经规划和设计好了，早在很早之前，甚至在我加入之前（我是 Grok 3 期间加入的）。所以这就像是至少提前了一年进行思考。是的，你可以思考得更长远，并假设那些预估目标将会被达成。

[原文] [Host]: um just because everyone's like pretty great and reliable which frees you up a lot in terms of like what your limitations are I guess so for us for example the assumed minimum latency was about three times higher than it actually needed to be and the buildout allowed for that basically um what do you mean by that

[译文] [Host]: 只是因为大家都非常出色且靠谱，这让你在限制条件方面解脱了很多。我想是的。所以对我们来说，比如，假设的最低延迟大概比实际需要的要高出三倍，而基础设施的建设基本上允许了这种情况。嗯，你这话是什么意思？

[原文] [Sulaiman Ghori]: so the one of the novel architectures we're working on um is not really possible unless you scale up your experiment rate because it's it's not building on any existing body of work you need a new pre pre-training body and you need also uh a new data set but that's not really constrained by the resources like the physical uh infrastructure resources mostly um although there's the uh the Tesla computer thing which I think maybe we'll get into maybe not

[译文] [Sulaiman Ghori]: 是这样的，我们正在研究的一种新颖架构，除非你能大幅提高实验频率，否则是不可能实现的。因为它不是建立在任何现有的工作成果之上的，你需要一个新的预训练主体（pre-training body），还需要一个新的数据集。但这其实并不太受限于资源，比如物理基础设施资源——虽然还有“特斯拉计算机”（Tesla computer）这回事，我想我们也可能会聊到，也可能不会。

[原文] [Sulaiman Ghori]: but um uh so actually this one's public so one thing that we're thinking about is okay like we're we're building this human emulator with macro hard um how do we deploy it because you actually need like if we want to deploy 1 million human emulators we need 1 million computers um how do we do that and the answer showed up two days later in the form of a Tesla computer because those things are actually very capital efficient as it turns out

[译文] [Sulaiman Ghori]: 不过，嗯，这其实已经是公开的了。所以我们在思考的一件事是：好吧，我们要用“Macro Hard”项目构建这种“人类模拟器”。我们该如何部署它？因为如果我们要部署 100 万个“人类模拟器”，实际上就需要 100 万台电脑。我们该怎么做？答案在两天后出现了，那就是“特斯拉计算机”，因为事实证明那些东西的资本效率非常高。

[原文] [Sulaiman Ghori]: and we can run um potentially like our our model and the like full computer that a human would otherwise work at on the Tesla computer for much cheaper than you would in on on a VM on AWS or Oracle or whatever or even just buying hardware from Nvidia that car computer is actually much much more capital efficient and so it enables us to assume that we can deploy much much faster at a much higher scale um and so we've adjusted our we adjusted our expectations for that basically

[译文] [Sulaiman Ghori]: 我们有可能在特斯拉计算机上运行我们的模型，以及人类工作所需的完整计算机环境，其成本远低于你在 AWS 或 Oracle 等平台的虚拟机（VM）上运行，甚至比直接从 Nvidia 购买硬件还要便宜。那个车载计算机实际上资本效率要高得多。所以这使我们能够假设我们可以以极高的规模、极快的速度进行部署。基本上，我们也因此调整了我们的预期。

[原文] [Host]: are you basically able to just bootstrap off of the like car network so that's one of the one of the potential uh solutions basically yeah so like okay well we want 1 million VMs um there's like 4 million uh Tesla cars in North America alone um and like let's say 2/3 or half of them have hardware 4 um and like somewhere between 70 80% 80% of the time they're sitting there idle probably charging

[译文] [Host]: 你们基本上可以直接利用现有的汽车网络启动吗？

[Sulaiman Ghori]: 那是潜在的解决方案之一。基本上是的。比如，好吧，我们需要 100 万个虚拟机。仅在北美就有大概 400 万辆特斯拉汽车，假设其中三分之二或一半搭载了 Hardware 4（硬件 4.0），而在 70% 到 80% 的时间里，它们是闲置的，很可能正在充电。

[原文] [Sulaiman Ghori]: we can just potentially pay and they have you know networking they have cooling they have power um we can just pay pay owners to lease time off their car and let us run um like a human emulator uh digital Optimus on right on it and uh they get you know their lease paid for and we get uh a full human emulator we can put to work um and that's something without any buildout requirement it's a purely software implementation that's required

[译文] [Sulaiman Ghori]: 我们可能只需要付钱——你知道，它们有网络，有冷却系统，有电源——我们可以付钱给车主，租赁他们车辆的时间，在上面运行“人类模拟器”，也就是“数字版的 Optimus（擎天柱）”。车主可以（通过这种方式）抵扣他们的租车费用，而我们得到了一个可以投入工作的完整人类模拟器。这不需要任何额外的基建，只需要纯软件的实现。

[原文] [Host]: yeah the the asset is sitting there and you can just go and use it yeah amazing what for the human emulators uh in macro hard what is the like purpose of that of scaling up you know millions of many humans

[译文] [Host]: 是的，资产就在那里，你可以直接拿来用。太神奇了。关于“Macro Hard”里的“人类模拟器”，扩容到数百万个“小人类”的目的是什么？

[原文] [Sulaiman Ghori]: um I mean the basic con concept is very simple right with with Optimus you're uh taking any physical task a human can do and allowing a robot to do it automatically at a fraction of the cost at 20 you know with 24/7 uptime um we're doing the same with anything that a human does digitally so any anything where they need to digitally input uh a keyboard and mouse inputs which is usually what humans do um and look at a screen back and make decisions uh we just emulate what the human is doing uh directly so no adoption from any software is required at all um we can deploy in any situation in which a human is in potentially currently

[译文] [Sulaiman Ghori]: 嗯，基本概念其实很简单，对吧？对于 Optimus（人形机器人），你是把人类能做的任何物理任务，让机器人以极低的成本、24小时全天候地自动完成。我们正在对人类在数字领域做的事情做同样的事。任何需要数字输入——比如键盘和鼠标输入（这通常是人类做的事），以及看着屏幕做出决策的事情——我们直接模拟人类在做什么。所以完全不需要任何软件层面的适配，我们可以在任何目前人类所处的场景中进行部署。

[原文] [Host]: um interesting what is what is that actually going to look like uh for rolling it out

[译文] [Host]: 嗯，有意思。那实际推广起来会是什么样子？

[原文] [Sulaiman Ghori]: um I don't think we've detailed our plans publicly yet specifically on uh on how we'll roll out it'll be slowly at first and then very quickly basically like uh like the difference for us given that infrastructure buildout already has happened or we can go on the Tesla network or we can build out our own data center Tesla computers actually um the difference for us from from going from 1,000 human emulators to a million is actually not very big it's not it's not the biggest part of the challenge

[译文] [Sulaiman Ghori]: 嗯，我想我们还没有公开具体的推广计划。基本上会是先慢后快。对我们来说，既然基础设施建设已经完成，或者我们可以利用特斯拉网络，或者我们可以建立自己的“特斯拉计算机”数据中心，那么从 1,000 个“人类模拟器”扩展到 100 万个，这中间的区别其实并不大，这并不是挑战中最大的部分。

章节 4：一通电话的解题效率与“Imagine”的 24 小时生死时速

📝 本节摘要：

在本章中，对话深入探讨了 Elon Musk 解决问题的风格：他不仅仅是发号施令，更是通过一通电话就能消除硬件与软件团队之间的隔阂，瞬间解决那些通常需要数周扯皮的“阻碍（Blockers）”。Sully 描述了 xAI 内部“仗剑而生，亦死于剑下（live by the sword, die by the sword）”的责任文化——只要你能搞定，你就能拥有这部分技术栈。此外，他详细讲述了负责“Imagine”功能发布时的经历：当时 iOS 团队仅有三人，他们进入了疯狂的“24 小时迭代循环”，白天改代码，晚上发布，次日清晨通过用户反馈修复 Bug，周而复始，没有周末。

[原文] [Host]: elon I know one of the things that he does best is he basically just goes from fire to fire on whatever the company is and just kind of like puts it out and unfucks whatever problem is exists uh what has that been like what when have you like seen some problem exist and just had it unfucked very rapidly do this kind of process

[译文] [Host]: 我知道 Elon 最擅长的事情之一，就是他在公司里哪里有火就去哪里救火，基本上就是把问题扑灭，把存在的烂摊子迅速搞定（unfucks）。那是种什么体验？你有没有见过什么问题存在，然后通过这种过程被迅速搞定？

[原文] [Sulaiman Ghori]: um definitely on the infra build out this is the biggest um on model side we've been like we've had hiccups but it's more or less been smooth but on model side especially cuz there's a lot of uh I mean infra side there's a lot of very specific uh operations that each of these basically AS6 these GPUs are are built for and when we roll out new products like when we pick up new products from Nvidia or whoever um not everything works

[译文] [Sulaiman Ghori]: 嗯，绝对是在基础设施建设方面，这是最大的部分。在模型方面，我们遇到过一些小插曲，但大体上还算顺利。但在基础设施方面，特别是考虑到每一块 ASIC 或 GPU 都是为了非常特定的操作而构建的，当我们推出新产品，或者从 Nvidia 或其他地方拿到新产品时，并不是所有东西都能直接正常工作。

[原文] [Sulaiman Ghori]: so in some of the meetings that we had with him uh early last year uh he would hear these and he would make a phone call and the software team would deliver a patch the next day and we would work like side by side until that was resolved um and then we could run a model uh or a train training run uh on the hardware uh very very quickly where otherwise it would have taken weeks of back and forth so those kind of blockers are usually very quickly resolved with one phone call

[译文] [Sulaiman Ghori]: 所以在去年年初我们和他的一些会议中，他听到这些问题后，会直接打一个电话，然后（对方的）软件团队第二天就会发来一个补丁，我们会并肩工作直到问题解决。然后我们就能非常非常快地在硬件上运行模型或训练任务，而如果没有这样做，这通常需要来回扯皮好几周。所以这类阻碍通常通过一通电话就能迅速解决。

[原文] [Sulaiman Ghori]: um or just us bringing it up to him or him just offering like frequently when uh a meeting is ending or there's a lull in in the conversation he'll be like okay how can I help how can I make this faster whatever and someone will come up with with an answer

[译文] [Sulaiman Ghori]: 或者只是我们向他提出来，或者他主动提出。经常是当会议快结束或对话出现停顿时，他会问：“好吧，我能帮什么忙？我怎么能让这个更快？”之类的问题，然后就会有人提出解决方案。

[原文] [Host]: I know you guys are doing many different products in parallel and I get that it's kind of like you have to do that but also it's sometimes in most organizations it's like very difficult to stay focused on a single thing and like a single objective how does that kind of work uh for just executing on multiple different fronts at the same time

[译文] [Host]: 我知道你们在并行做很多不同的产品，我理解你们必须这样做，但在大多数组织中，有时候很难保持对单一事物或单一目标的专注。这种同时在多个不同战线上执行任务的模式是如何运作的？

[原文] [Sulaiman Ghori]: very frequently we actually uh and this is increasing with scale we don't have a full picture until like the all hands or we just chat with people what everyone is doing and how far everyone is on these different projects like for example on on on when we did our our our voice model and our voice deployment um we actually had a lot of the work built for extremely low latency uh extreme low latency end like uh packets to be sent to the client it was already built out and um it was a matter of flipping the right switches and the right configs basically to cut our latency pretty significantly um like 2 3x

[译文] [Sulaiman Ghori]: 实际上很多时候——随着规模扩大这种情况越来越多——我们要等到全员大会（all hands）或者和人聊天时，才能全面了解大家都在做什么，以及各个项目的进展如何。比如，当我们做语音模型和语音部署时，其实我们已经构建了很多用于极低延迟的工作——极其低延迟的数据包发送到客户端——这些都已经建好了。基本上只需要拨动正确的开关和配置，就能显著降低我们的延迟，大概降低 2 到 3 倍。

[原文] [Sulaiman Ghori]: uh and end um this is actually the case a lot of the time is there is a stupid thing that uh exists somewhere in the software or the hardware and someone has come up with a solution um and you find it when you go to look for it in in our codebase somewhere or you ask around and someone's like oh yeah this XYZ person has done this you should talk to them and they will hook you up

[译文] [Sulaiman Ghori]: 最后，很多时候情况是这样的：软件或硬件的某个地方存在一个愚蠢的问题，而某人已经想出了解决方案。当你去代码库里找，或者四处打听时，你会发现它，或者有人会说：“噢对了，某某人已经做过这个了，你应该找他聊聊。”然后他们会帮你搞定。

[原文] [Sulaiman Ghori]: um there's not a lot of time spent syncing up with anyone or asking for permission or um waiting for anyone at all like the answer is like when you propose someone someone says a good idea like usually you propose something and the the answer is either no that's dumb or why isn't it done already like um and then you go and do it and then it's done

[译文] [Sulaiman Ghori]: 我们不会花很多时间去和任何人同步，或者请求许可，或者等任何人。通常情况是，当你提出一个好主意时，得到的回答要么是“不，这很蠢”，要么是“为什么还没做完？”然后你就去做，接着就搞定了。

[原文] [Host]: with Elon companies you can kind of just ask for responsibility and then you basically just live by the sword die by the sword and if you get things done then you can just ask for more responsibility and you can keep on doing that or you're just like out what's been your experience like with that

[译文] [Host]: 在 Elon 的公司里，你似乎可以直接索要责任，然后基本上就是“仗剑而生，亦死于剑下”（live by the sword, die by the sword，意为风险与机遇并存，靠实力说话）。如果你把事情做成了，你就可以索要更多的责任，继续这样做；或者你就出局了。你在这方面的体验是怎样的？

[原文] [Sulaiman Ghori]: very much so yeah like um I've jumped around a lot of different projects and mostly just because someone asked for my help and I kept helping and then I ended up owning some of the stack or a lot of the stack um and this is the case for everyone like this is just how it is um if you have any particular experience or um can iterate on something very quickly within days you own that component

[译文] [Sulaiman Ghori]: 确实非常如此。是的，比如我在很多不同的项目之间跳来跳去，主要是因为有人找我帮忙，我就一直帮，最后我就变成了拥有那部分技术栈（stack）或者很大一部分技术栈的人。每个人都是这样，这就是这里的情况。如果你有特定的经验，或者能在几天内快速迭代某样东西，你就拥有那个组件。

[原文] [Sulaiman Ghori]: um yeah there's no formal anything I think officially on our HR software I I'm on voice and iOS or something and our security software thinks I still work on RX integration and um which never updated yeah no no one ever updates this stuff like um it's kind of ridiculous

[译文] [Sulaiman Ghori]: 嗯，是的，没有什么正式的东西。我想在我们的 HR 软件上，官方显示我是做语音和 iOS 或者什么的；而我们的安全软件认为我还在做 RX 集成。这些从来没更新过，是的，从来没人更新这些东西，有点荒谬。

[原文] [Host]: what's been the journey from like the starting to to now like what what projects have you worked on

[译文] [Host]: 从开始到现在，你的经历是怎样的？你参与了哪些项目？

[原文] [Sulaiman Ghori]: yeah so specifically I started um I first worked on like ASRock uh and our integration there and I worked with our backend team a bit on like reliability and scaling up because we were scaling up a lot at that time uh and then after that I took on solo building up our our desktop suite um and took that went to internal completion

[译文] [Sulaiman Ghori]: 好的，具体来说，我刚开始是做 ASRock（注：原文如此，推测为项目代号或 Grok 相关集成）以及我们那边的集成工作。我和后端团队在可靠性和扩容方面合作了一段时间，因为当时我们正在大规模扩容。在那之后，我独自承担了构建我们桌面套件（desktop suite）的工作，并把它做到了内部完成的阶段。

[原文] [Sulaiman Ghori]: uh and then I got asked for help on our imagine roll out and iOS which yeah our iOS team is small for like how many people use it like it's ridiculous you won't guess the number um like five people for three it was three and I was the third person at the time when we were rolling that out it was like it was ridiculous and everyone's like really really good

[译文] [Sulaiman Ghori]: 然后有人找我帮忙做“Imagine”（图像生成功能）的发布和 iOS 端。是的，考虑到用户量，我们的 iOS 团队小得离谱，你绝对猜不到那个数字——大概五个人？不，是三个人。当时我们发布的时候，我是第三个人。简直离谱，而且每个人都非常非常强。

[原文] [Host]: what was the first uh experience that you had where you thought to yourself like you're actually being kind of used to your full you know potential

[译文] [Host]: 你第一次觉得自己的潜力真正被充分发挥出来的经历是什么时候？

[原文] [Sulaiman Ghori]: and I think that imagine rollout was definitely like it was a really good push cuz like we had this 24-hour iteration cycle um you all would get feedback every night on whatever we were doing um and yeah we we would push out that night uh in the morning we would have all the feedback we would immediately knock out all the bugs um implement the new stuff that that people were asking for whatever model had come up with we implemented that too like it was a very very fast cycle and it was uh I think it was the longest like continuous stretch of me being in the office like every day

[译文] [Sulaiman Ghori]: 我觉得那次“Imagine”的发布绝对是一个很好的推力。因为我们当时处于一种 24 小时的迭代周期中。无论我们做什么，每天晚上都会收到反馈。是的，我们当晚就会发布更新。到了早上，我们拿到所有反馈，立刻消灭所有 Bug，实现人们要求的新功能，或者模型组搞出来的任何新东西，我们也一并实现。那是一个非常非常快的循环。我想那是我连续在办公室待的时间最长的一段日子，就像每天都在那里。

[原文] [Host]: what was that like at the time it was like two or three months two or three times yeah yeah okay um yeah like there weren't weekends for a while which was uh it was good to know that I could do that and I was pretty happy doing that

[译文] [Host]: 当时是什么感觉？那是持续了大概两三个月吗？

[Sulaiman Ghori]: 是的，大概两三个月。嗯，有一段时间是没有周末的。不过知道自己能做到这一点感觉很好，而且我做得很开心。

章节 5：Colossus 战记——移动电站、电池墙与“嘉年华”式基建

📝 本节摘要：

本章揭秘了 Colossus 数据中心背后的“战争故事”。从 Tyler 再次被提及的 Cybertruck 赌约（这一次补充了更多现场细节），到为了应对极端的电力需求，xAI 不得不调动 80 多台移动发电机卡车，并在电网负载过高时瞬间切换电源。Sully 深入解释了为何在该场景下电池（化学能）比发电机（物理旋转惯量）在响应毫秒级功率波动时更具优势。最精彩的部分在于，Sully 透露了 xAI 能够以创纪录速度（122天）建成的秘密武器：他们利用了通常用于“嘉年华”活动的临时土地租赁条款来绕过常规的建筑许可审批——从技术上讲，xAI 目前是一家“嘉年华公司”。

[原文] [Sulaiman Ghori]: um and after that I got pulled onto Macro hard product which was just one other person at the time so it was the two of us uh for a while and I've been on that since uh since that project off basically

[译文] [Sulaiman Ghori]: 在那之后，我被调到了“Macro Hard”产品组，当时那里只有另外一个人，所以有一段时间就我们两个。自从那个项目启动以来，我就一直待在那里。

[原文] [Host]: i don't know how much you like know about this but uh the like Colossus build and all the ridiculous stuff that the like early XAI team had to do to turn on Colossus and like get power and all the necessary inputs to making that work and even today I think like it's just bottlenecks across the entire thing you just want more you you want more like uh chips and GPUs and all the stuff working and faster um what was that like there's a lot of war stories um and a lot of bets um want to go into a few

[译文] [Host]: 我不知道你对此了解多少，关于 Colossus（数据中心）的建设，以及早期 xAI 团队为了启动 Colossus、获取电力和所有必要输入以使其运转所做的那些离谱的事情。甚至直到今天，我觉得整个过程到处都是瓶颈，你只想要更多——更多芯片、更多 GPU，让所有东西运转得更快。那是一种什么体验？听说有很多“战争故事”（war stories）和很多赌约，想聊聊吗？

[原文] [Sulaiman Ghori]: yeah so I think Tyler was took this bet uh with Elon like uh one we were setting up new racks I think of I forget what which GPUs we were rolling out at that time um we took a bet uh Elan's like "Okay you get a cybert truck tonight if you can get a training run on these GPUs uh in 24 hours." Uh and we were training that night

[译文] [Sulaiman Ghori]: 是的，我想 Tyler 和 Elon 打了个赌。当时我们正在架设新的机柜，我忘了当时我们在部署哪种 GPU 了。我们打了个赌，Elon 说：“好，如果你能在 24 小时内让这些 GPU 跑起训练任务，今晚你就拿辆 Cybertruck。”结果我们当晚就在训练了。

[原文] [Host]: um did he get the cyber yeah he got I think it's Yeah I see it from our lunch window mhm cafeter yeah he's cool

[译文] [Host]: 嗯，他拿到 Cybertruck 了吗？

[Sulaiman Ghori]: 是的，他拿到了。我想是的，我能从我们午餐窗口——食堂那里看到它。是的，他很酷。

[原文] [Sulaiman Ghori]: um uh you know what the I so for power we actually have we have to collaborate very tightly with the like municipal uh and state power companies uh because when load goes high on their end we have to shut off and go fully on the like 80 or maybe it's more than that I think more more than that 80 uh mobile generators we brought in on trucks um and go fully on on those um just so that we don't like impact power uh anywhere are like within

[译文] [Sulaiman Ghori]: 嗯，你知道吗，关于电力，实际上我们必须与市政和州电力公司非常紧密地合作。因为当他们那边的负载变高时，我们必须切断（市电），完全依靠那 80 台——或者可能更多，我想不止 80 台——我们要完全依靠那些用卡车运来的移动发电机供电，仅仅是为了确保我们不会影响到区域内的电力供应。

[原文] [Sulaiman Ghori]: and we have to do that like seamlessly without interrupting anyone's uh extremely volatile training runs uh on extremely volatile uh you know GPUs and and hardware which scales up and down by like megawws in milliseconds it's it's a lot

[译文] [Sulaiman Ghori]: 而且我们必须做到无缝切换，不能打断任何人的训练任务。这些训练任务极其不稳定，运行在极其不稳定的 GPU 和硬件上，其功率会在毫秒级内上下波动数兆瓦（megawatts）。这真的很夸张。

[原文] [Host]: um is that also part of the logic of like basically putting massive battery packs right next to the uh desenters cuz then you can kind of like go up and down much faster without batteries can scale up a lot uh scale up and down and uh balance that load a lot faster

[译文] [Host]: 基本上在数据中心旁边放置巨大的电池组，也是出于这个逻辑吗？因为这样你可以更快地调节？

[Sulaiman Ghori]: 没有电池的话……电池可以大幅度地、更快地进行上下调节和平衡负载。

[原文] [Sulaiman Ghori]: um cuz with a generator you're literally asking a physical thing to speed up or or slow down like a spinning spinning physical thing that's obviously just going to take a certain amount of time the batteries can uh react to the light much much faster and then yeah it's like actually from the phys from physical standpoint I think there's the uh local capacitors the station like data hall side capacitors the batteries and then generators and then the public municipalities although we might have changed that infrastructure at this point things very quickly especially on the cooling side

[译文] [Sulaiman Ghori]: 因为如果是发电机，你实际上是在要求一个物理实体加速或减速——那是一个正在旋转的物理东西，显然需要一定的时间。而电池的反应速度要快得多。从物理角度来看，我想顺序是：本地电容、站点（数据大厅侧）电容、电池，然后是发电机，最后是市政公用电网。虽然我们现在可能已经改变了那套基础设施，毕竟事情变化得很快，特别是在冷却方面。

[原文] [Host]: do you have any other really good like war stories that are just like uh I don't know things that shouldn't have been possible that became possible

[译文] [Host]: 你还有其他真正精彩的“战争故事”吗？就是那种你觉得本来不可能实现，却最终实现了的事情。

[原文] [Sulaiman Ghori]: Uh so the the lease for the land itself was actually technically temporary it was the fastest way to get the permitting through and actually start building things um I assume that it'll be permanent at some point but yeah it's I think a very short-term lease at the moment technically for all the data centers it's fastest way to get things done

[译文] [Sulaiman Ghori]: 嗯，这块土地的租赁合同本身在技术上其实是临时的。这是通过许可审批并开始建设的最快方式。我假设某个时候它会变成永久的，但在目前，从技术上讲，所有数据中心签的都是非常短期的租约。这是把事情做成的最快方法。

[原文] [Host]: and how do they how do they do that

[译文] [Host]: 他们是怎么做到的？

[原文] [Sulaiman Ghori]: um I think there's basically a special exception uh within like the local and state government says okay if you want to just uh modify this ground temporarily I think it's like for like uh carnivals and stuff you can Xi is actually just a carnival company currently and so that was the way to get done quickly I mean it was done yeah 122 days

[译文] [Sulaiman Ghori]: 我想基本上地方和州政府有一项特殊的豁免条款，规定如果你只是想“临时”改造这块土地——我想那是针对“嘉年华”之类的活动的——你就可以做。所以，xAI 目前实际上只是一家“嘉年华公司”。这就是快速搞定它的方法。我的意思是，确实搞定了，122 天。

章节 6：逆向规划与超级杠杆——从“删除算法”到 AI 代理军团

📝 本节摘要：

xAI 的规划逻辑与众不同：不从现有资源出发，而是从“最高经济杠杆”的目标（如 1000 亿收入）倒推物理和软件需求。Sully 确认了公司内部严格执行类似 SpaceX 的“算法”——先删除要求，若有必要再加回来。他举了一个关于视频编码器（Video Encoders）的生动例子，说明如何在删除特殊情况后发现问题并快速修正。

>

此外，本章揭示了 xAI 惊人的人效比：Sully 计算出他当天的 5 次代码提交创造了约 1250 万美元的价值。更令人震撼的是，公司正在利用 AI 彻底改变开发模式——核心生产 API 的重构工作，正由一名工程师指挥着 20 个 AI 智能体（Agents）独立完成。

[原文] [Host]: for like internal planning I know things are just going to keep on scaling up like crazy and Elon's talked about energy being the biggest bottleneck and then you know just being able to get chips um how do you guys plan when it's very difficult to like predict 12 to 24 months in the future exactly what projects you're going to be working on or what their like resource requirements are going to be

[译文] [Host]: 关于内部规划，我知道事情会持续疯狂地扩张，Elon 说过能源是最大的瓶颈，然后还有芯片供应的问题。当很难预测未来 12 到 24 个月具体的项目是什么，或者资源需求会是多少时，你们是如何进行规划的？

[原文] [Sulaiman Ghori]: we try we try very hard to work backwards from like what's the highest leveraged thing we can be doing and then we determine the physical requirements later so like if we want to get to 10 or hundred billion in revenue by this date uh what are the highest leverage things we can do like from an econ economic perspective how can we actually build systems to do that and then what does it take on the physical and software side to roll that out and and get it done um just roll down roll backwards the whole way so we don't usually start with the with the physical requirement that's usually actually at the end

[译文] [Sulaiman Ghori]: 我们非常努力地尝试从我们能做的“最具杠杆效应”的事情开始倒推，然后再确定物理需求。比如，如果我们想在某个日期前达到 100 亿或 1000 亿的收入，从经济学角度看，我们可以做的最高杠杆的事情是什么？我们实际上如何构建系统来实现这一点？然后在物理和软件方面需要什么来部署并完成它？我们就这样一路倒推。所以我们通常不会从物理需求开始，那通常实际上是在最后。

[原文] [Host]: is there like a SpaceXesque um like algorithm for making things happen as in like the usual delete yeah i mean that's the case all the time um and we do do the thing where Yeah we delete something and then add it back later

[译文] [Host]: 有没有那种类似 SpaceX 风格的“算法”（The Algorithm）来促成事情发生？比如常规的“删除”步骤？

[Sulaiman Ghori]: 是的，一直都是这样。我们确实会这样做：是的，我们删除某些东西，然后（如果有必要）稍后再加回来。

[原文] [Host]: um what was the like last time that you did that today today

[译文] [Host]: 你们最近一次这样做是什么时候？

[Sulaiman Ghori]: 今天。就在今天。

[原文] [Sulaiman Ghori]: um today yeah so with macro hard we deploy on um a lot of like physical hardware that changes and um the testing harness for that is hard um so we try to minimize how much how many special cases are downstream of where it needs to be and um for example like with display scaling um we need to be able to support displays that are you know 30 years old as well as the latest like 5K Apple whatever displays and that has to happen on the same stack

[译文] [Sulaiman Ghori]: 嗯，今天。对于“Macro Hard”项目，我们部署在大量不断变化的物理硬件上，而为此建立测试工具（harness）很难。所以我们试图最大限度地减少下游的特殊情况。比如关于显示缩放（display scaling），我们需要支持那些 30 年前的老显示器，同时也得支持最新的 5K Apple 显示器之类的，而且必须在同一个技术栈上实现。

[原文] [Sulaiman Ghori]: um turns out not all the systems are happy with that at all times like you have to you have to fiddle with the encoders at a certain level like uh video encoders um was the specific thing basically we I didn't know but uh as it turns out there are limits to the maximum amount of pixels that certain encoders can take so we have to now have I removed this special case for multiple encoders and turns out we found a problem at at plus 5K resolution and so we added that back

[译文] [Sulaiman Ghori]: 结果并非所有系统都能时刻兼容。你必须在某种层面上调整编码器，具体来说是视频编码器。基本上——我之前不知道，但事实证明某些编码器能处理的最大像素数是有限制的。所以我之前移除了针对多编码器的特殊情况处理（即执行了“删除”步骤），结果我们在 5K 以上分辨率时发现了问题，于是我们又把它加了回来。

[原文] [Sulaiman Ghori]: um like things that I thought would be stupid are okayed and we just do them and we try them and it's like we we'll do a hackathon and if we get five people in as a result it's worth it um because just their like expected return on on the company's like revenue or valuation is higher than the cost of running this hackathon for 500 people um like the verhead value is actually very high which is like funny

[译文] [Sulaiman Ghori]: 有些我觉得会很蠢的事情也被批准了，我们就直接做，去尝试。比如我们会举办一场黑客马拉松，如果结果能招到五个人，那就值了。因为仅凭他们对公司收入或估值的预期回报，就高于为 500 人举办这场黑客马拉松的成本。每个人头带来的价值实际上非常高，这很有趣。

[原文] [Sulaiman Ghori]: we did the math um earlier this week uh right now we're like I think at about $2.5 million per commit is to to the main to the main repo um and I did five today so you added like 12 and a half million of value um light day light days exactly it was a good day

[译文] [Sulaiman Ghori]: 我们这周早些时候算了一笔账。现在我想我们每次向主仓库（main repo）提交代码大约价值 250 万美元。而我今天提交了五次。所以你大概增加了 1250 万美元的价值。轻松的一天？对，轻松的一天，确实是个好日子。

[原文] [Sulaiman Ghori]: um it's funny things like that um like the levers are are extremely strong like you you can get a lot a lot done with a lot less effort and time than you used to be able to for sure just because of who you work with the internal tooling that we built up

[译文] [Sulaiman Ghori]: 这种事情很有趣。这里的杠杆效应极强。你绝对可以用比以前少得多的精力和时间完成多得多的事情，这仅仅是因为和你共事的人，以及我们要建立的内部工具。

[原文] [Sulaiman Ghori]: some people want to own huge parts of the technical stack so like right now we're doing a big rebuild um of like our core uh production APIs it's being done by one person with like 20 agents um and they're they're very good and they're capable of doing it and um like it's working well uh so you can own huge chunks of the code base no problem

[译文] [Sulaiman Ghori]: 有些人想拥有技术栈的很大一部分。比如，目前我们正在对核心生产 API 进行大规模重构，这是由一个人带着 20 个智能体（agents）完成的。他们非常厉害，而且完全有能力做到这一点，效果也很好。所以你可以毫无问题地拥有巨大的代码库板块。

章节 7：极客猎手——寻找“十行代码”解题者与挑战不可能

📝 本节摘要：

在本章中，Sully 描述了 xAI 的核心人才画像：充满热情、任务导向，且极具野心（无论是想走管理路线还是想独自掌控巨大的技术栈）。为了筛选出这类人才，Sully 分享了他的面试“杀手锏”：一道看似复杂的计算机视觉题，实则只需极简的方案。他强调，在 AI 能轻松生成 200 行代码的时代，能找到 10 行代码解决方案的人才比以往任何时候都珍贵。此外，他借鉴了同事 Chester 的一种“压力测试”法：故意在面试题中植入错误的或物理上不可能的需求，如果候选人不敢指出错误，就不会被录用。

[原文] [Host]: what's like an example of the type of person that like wants to work here cuz I know when when you're talking about it you kind of show up and the first day you're just like I want to work on the weekends i want to work on you know during the night all this stuff uh go all in on this um what kind of special characters are are working there

[译文] [Host]: 想在这里工作的人大概是什么类型的？因为听你的描述，你第一天出现，感觉就像“我想在周末工作，我想在晚上工作，我想全力以赴（all in）”。那里都在这群什么样的特殊角色？

[原文] [Sulaiman Ghori]: people are definitely very enthusiastic when they come in like um very very enthusiastic uh just like like mission oriented um there's I guess different types of ambition for sure some people want to move up like the leadership ladder and own more in terms of a managerial like how many people report to me sense some people want to own huge parts of the technical stack

[译文] [Sulaiman Ghori]: 人们进来时绝对是非常热情的，非常非常热情，就像是以使命为导向的（mission oriented）。当然有不同类型的野心：有些人想沿着领导阶梯往上爬，在管理层面拥有更多（比如有多少人向我汇报）；有些人则想拥有巨大的技术栈板块。

[原文] [Host]: absolutely for hiring um what's what unusual practices outside of just hackathons uh does XI do

[译文] [Host]: 确实。在招聘方面，除了黑客马拉松之外，xAI 还有什么不寻常的做法吗？

[原文] [Sulaiman Ghori]: uh so we're pushing very hard on Macro hard like for two or three weeks I was doing upwards of 20 interviews a week so that's like some of them are like quick 15 minutes some of them are full 1 hour technicals so a lot of my time uh is dedicated towards bringing in new people and a lot of people are very good so it's it's actually very hard to judge them

[译文] [Sulaiman Ghori]: 嗯，我们在“Macro Hard”项目上推进得非常猛。有两三个星期，我每周要做 20 多场面试。其中有些是 15 分钟的快速面试，有些是整整 1 小时的技术面试。所以我很多时间都花在招新人上，而且很多人都非常优秀，所以其实很难评判他们。

[原文] [Host]: how do you uh I have a very specific problem that I have solved i'm not going to reveal it because then people will use it but I have solved a very specific computer vision problem a few years ago for one of my startups and I uh I give people half an hour to try to implement the solution it's actually very very simple this deceptively simple solution people always overthink it

[译文] [Host]: 你是怎么做的？

[Sulaiman Ghori]: 我有一个非常具体的问题——这是我在几年前为我的一个创业公司解决过的计算机视觉问题。我不会透露具体是什么，否则人们会去刷题。我给候选人半个小时去尝试实现解决方案。实际上它非常非常简单，这是一个具有欺骗性的简单方案，但人们总是会想得太复杂（overthink）。

[原文] [Sulaiman Ghori]: um and this is something I like to index for on my team especially is like can you not overthink it and come up with a simple solution um it helps a lot because we're deploying on such a wide variety of like uh on a wide variety of hardware as a result of the wide variety of of customers like literally 30 years 40 years of uh different hardware different operating systems everything like that you have to come up with simple solutions or you're going to have a 10 million line code base uh next week

[译文] [Sulaiman Ghori]: 这就是我在团队中特别看重的特质：你能不能不要过度思考，而是拿出一个简单的解决方案？这非常有帮助，因为我们要部署在种类繁多的硬件上，面对各式各样的客户——真的是跨度达 30 年、40 年的不同硬件、不同操作系统等等。你必须想出简单的解决方案，否则下周你的代码库就会膨胀到 1000 万行。

[原文] [Sulaiman Ghori]: so you you this is like very important um and especially now relying more and more on on agents and and an AI and and such for writing code um an AI will happily train out 200 lines when a 10line solution will do um and probably do better so you have to look for that like I want people and I look and actively hire for people who can find the 10line solution first

[译文] [Sulaiman Ghori]: 所以这非常重要。尤其是现在我们越来越依赖智能体（Agents）和 AI 之类的东西来写代码。当一个 10 行的方案就能解决问题且可能效果更好时，AI 会很乐意给你整出 200 行代码来。所以你必须寻找那种特质。我想要的人，以及我积极寻找并雇佣的人，是那些能首先找到那“10 行代码解决方案”的人。

[原文] [Sulaiman Ghori]: um we're totally fine with people using AI to code things like you should you should use that as a force multiplier but uh for now we're smarter we'll see next year

[译文] [Sulaiman Ghori]: 嗯，我们完全不介意人们使用 AI 来写代码，你应该把它当作一种力量倍增器（force multiplier）来使用。但就目前而言，我们（人类）还是更聪明一点。至于明年怎么样，咱们走着瞧。

[原文] [Host]: what other like force multipliers do you kind of like look for

[译文] [Host]: 你还会寻找其他什么样的“力量倍增器”特质？

[原文] [Sulaiman Ghori]: i like people who will challenge uh challenge requirements and challenge me so often uh I got this from uh Chester Zai German for he told me this and I thought it was great he throws in usually um an incorrect requirement or question or an impossible like uh line in uh his challenges for people when he's hiring like coding challenges and he expects people to come back and say like hey this is wrong this is not possible you made a mistake and if he doesn't then uh he doesn't hire them same thing for me I picked that up it's a great idea

[译文] [Sulaiman Ghori]: 我喜欢那些会挑战需求、挑战我的人。我从 Chester Zai（注：音译，可能指 xAI 某位工程师）那里学到了这一招，我觉得这招很棒。他在招聘做代码测试时，通常会故意扔进去一个错误的需求、一个错误的问题，或者一条不可能实现的指令。他期望人们回过头来说：“嘿，这是错的，这不可能，你搞错了。”如果对方没有这么做，他就不录用他们。我也一样，我学会了这招，这是个很棒的主意。

章节 8：健身房里的作战室与“一夜抵数月”的疯狂突击

📝 本节摘要：

在本章中，Sully 描述了适应 xAI 极速节奏的方法：由于团队发展太快根本来不及写文档，新员工必须具备直接通过“阅读代码”来理解系统的能力。他提到 xAI 的一大优势是工程师拥有对 Grok（自家 AI）的“无限访问权”，这让他们可以零成本地尝试各种“愚蠢的实验”。

>

随后，对话转向了更为硬核的“作战室”文化。Sully 透露，“Macro Hard”项目组已经在一个由健身房（Gym）临时改造的作战室里高强度工作了四个月。在这种环境下，经常会出现“突击（Surge）”时刻——比如 Elon 半夜发现模型输出错误，醒着的工程师会立即响应修复。Sully 引用了联合创始人 Igor 的话说：“有些月份只有几天的进展，而有些夜晚却发生了几个月的进展。”

[原文] [Host]: the pace is insanely fast and like you said you kind of have worked on a number of different things How do you kind of come up to speed on something as quickly as possible when you're on a new task or project

[译文] [Host]: 这种节奏快得疯狂。正如你所说，你参与过很多不同的事情。当你接手一个新任务或新项目时，你是如何尽可能快地跟上速度的？

[原文] [Sulaiman Ghori]: it depends on what thing it is if there's a lot of code to read read the code by hand um like GD go to definition over and over again and you'll find things out very quickly actually it's not that hard um for most things the implementation is like less lines of code than you would otherwise see which is nice

[译文] [Sulaiman Ghori]: 这取决于具体是什么事。如果有很多代码要读，那就手动读代码。比如不断地使用“GD”（Go to Definition，转到定义），你很快就会搞清楚状况。其实没那么难。对于大多数事情来说，实现代码的行数比你预期的要少，这挺好的。

[原文] [Sulaiman Ghori]: which is good because we don't like write a lot of docs we write things we do things too fast to write docs really um actually yeah we're trying to figure out some systems on on my team to like automatically generate docs as we like build stuff

[译文] [Sulaiman Ghori]: 这很好，因为我们不怎么写文档。我们写代码、做事情的速度太快了，真的来不及写文档。实际上，是的，我们团队正在尝试弄清楚一些系统，以便在我们构建东西时自动生成文档。

[原文] [Sulaiman Ghori]: um and with Grock which is cool that we have unlimited access to uh very smart AI because then we can try a bunch of stupid things see if it works which otherwise you know at a startup would cost you maybe like$100 or a million dollar 100k or a million dollars in in credits or whatever we do it for free

[译文] [Sulaiman Ghori]: 而且有了 Grok（注：xAI 的大模型），很酷的一点是我们拥有对这种非常聪明的 AI 的“无限访问权”。这让我们能尝试一堆蠢事，看看行不行得通。如果在其他初创公司，这可能会花费你 100 美元、10 万美元甚至 100 万美元的额度（credits），或者随便什么代价，但我们可以免费做。

[原文] [Host]: i remember in the like early days of uh SpaceX there was this internal I think Elon would say internally like every day that we delay is like 10 million in loss revenue and I have no idea what it would be like for XAI like things are moving so fast it's like is there kind of an internal thing in your head of every day that we don't like push push hard or make something happen um we're losing out on x amount of value that could be created

[译文] [Host]: 我记得在 SpaceX 的早期，我想 Elon 在内部说过，每延迟一天就像是损失了 1000 万美元的收入。我不知道 xAI 是什么情况，事情发展得太快了。你脑子里会不会有那种内部概念，觉得如果我们某天没有全力推进，或者没把事情做成，我们就损失了 X 数额的潜在价值？

[原文] [Sulaiman Ghori]: yeah for for macroart specifically we do have a few pretty specific revenue targets i can't delete the number specifically but um like in my head whenever something gets delayed or accelerated I can pretty quickly calculate how much money we just made or lost um just wild swings you just Yeah I mean the numbers are huge uh just because the expected return is so huge and um the timeline is so fast so a few days is actually proportionately fairly large compared to how much you would you would otherwise expect the revenue to be

[译文] [Sulaiman Ghori]: 是的，特别是对于“Macro Hard”项目，我们确实有几个非常具体的收入目标。我不能透露具体的数字，但在我脑子里，每当某件事被推迟或加速时，我能很快算出我们刚刚赚了多少或赔了多少钱。那是巨大的波动。是的，我是说数字非常巨大，因为预期回报非常巨大，而且时间线非常快。所以相比于你预期的收入，几天的偏差按比例来看其实相当大。

[原文] [Host]: so there's been like moments in time where I've seen um Elon on X and someone has said like this is obviously not right and it's like some Grock output and he's like we're going to fix this and then you know 12 hours later 24 hours later he's like all right it's fixed when that happens like what happens internally

[译文] [Host]: 我有过这种时刻，看到 Elon 在 X（推特）上，有人说“这显然不对”，比如是 Grok 的某种输出，然后 Elon 说“我们会修复这个”。接着你知道，12 小时或 24 小时后，他说“好了，修好了”。当这种情况发生时，内部发生了什么？

[原文] [Sulaiman Ghori]: uh he shows us what went wrong and then quickly whoever um is awake at the time it will uh start up a thread to go and solve it uh usually individually pull in a few few others if need be um and then give a postmortem on what happened and everyone will understand then what uh what went wrong and how to avoid it in the future

[译文] [Sulaiman Ghori]: 嗯，他会向我们展示哪里出了问题，然后当时醒着的任何能够处理的人，会立刻开启一个线程去解决它。通常是个人行动，如果需要的话会拉几个人进来。事后会给出一个复盘报告（postmortem），说明发生了什么，大家就会明白哪里出了错，以及未来如何避免。

[原文] [Host]: do you guys have like a sign on the door that says war room amazing actually well yeah we we outgrew the original war room um and so we moved everything out and uh I'm told like walks in to the war room and it's totally empty and he's like "Where is everyone what?" and he walks over to where we are now which is just the gym which we cleared out and put everyone in now and then conducts his impromptu questions of what's going on

[译文] [Host]: 你们门口有挂着“作战室（War Room）”的牌子吗？

[Sulaiman Ghori]: 实际上很有趣，是的。我们原来那个作战室不够用了，所以我们把所有东西都搬了出来。有人告诉我，（Elon）走进原来的作战室，发现里面空空如也，他就问：“人都去哪了？什么情况？”然后他走到我们现在所在的地方——其实就是健身房，我们把它清空了，把所有人都塞了进去。然后他在那里进行即兴问答，询问进展如何。

[原文] [Sulaiman Ghori]: that was a long night what is it like on on one of those nights where a lot of things kind of get shaken up and and moved forward or like there's there's one of these searches what does that feel like

[译文] [Sulaiman Ghori]: 那是个漫长的夜晚。在那种很多事情被彻底重组、推进，或者发生这种“突击（surges）”的夜晚，是什么感觉？

[原文] [Sulaiman Ghori]: um I think actually I saw this from one of the co co-founders uh of XA posted this recently um Igor uh who was great to work with i liked him a lot... uh I saw him uh post this thing a few days ago where he's like "Okay there there are some uh you know months where uh only a few days go by and then there's some nights where months happen." And that was like one of them for sure

[译文] [Sulaiman Ghori]: 嗯，我想我最近看到 xAI 的一位联合创始人 Igor 发过这个——和他一起工作很棒，我很喜欢他……我几天前看到他发了一条动态，大意是：“有些月份只有几天的进展，而有些夜晚却发生了几个月的进展。”那天晚上绝对就是其中之一。

[原文] [Sulaiman Ghori]: um months might be an exaggeration i think we would have gotten to the technical result we would have in a few weeks anyway but doing it in one night was a huge push and it was a long night

[译文] [Sulaiman Ghori]: 嗯，“几个月”可能有点夸张。我想按正常节奏，我们在几周内也能达到那个技术成果，但把它压缩在一夜之间完成是一个巨大的推进。那是个漫长的夜晚。

[原文] [Host]: has there been any moments where the company just didn't leave the office for like 5 days or like a week yeah the surges for the models usually results in a lot of people staying in overnight um and you mentioned there's like five or six pods that people can sleep in and they like toggle out yeah yeah there's some there's some sleeping pods and we have some bunk beds now too um which are less less nice but they exist um and then when the tent picture came out everyone kept sending that to me and I was like honestly yeah we have tents but I've never seen that many out at once

[译文] [Host]: 有没有那种公司全员连续 5 天或一周都没离开办公室的时刻？

[Sulaiman Ghori]: 是的，模型训练的“突击”期间，通常会导致很多人在公司过夜。

[Host]: 你提到有大概五六个睡眠舱，大家轮流去睡？

[Sulaiman Ghori]: 是的，是的。有一些睡眠舱，我们现在也有一些双层床了——虽然没那么舒服，但确实有。之前那张（办公室搭满）帐篷的照片流出时，大家都发给我看。我说，老实讲，是的，我们有帐篷，但我从来没见过一次性搭出来那么多。

章节 9：少年极客的野蛮生长——从切断手指的 3D 打印机到被查封的陀螺工厂

📝 本节摘要：

在本章中，话题转向了 Sully 的少年创业史。他 11 岁时通过编写游戏外挂赚到了人生第一桶金，体会到了技术变现的快感。随后，他迷上了 RepRap 3D 打印机，在从阿里巴巴买零件组装的过程中，发生了一起骇人的意外：铜线深深刺入拇指，为了不耽误进度，他没有去医院，而是自己用剪刀处理了伤口继续工作。

>

故事的高潮在于他的“指尖陀螺帝国”：他在卧室搭建了 24 小时运转的“黑工厂”，利用学校同学作为分销商建立商业网络。生意火爆到最终被县政府以“食品特许经营权”的荒谬理由查封。这段经历让他养成了一种对体制和权威的“良性蔑视（healthy disrespect for authority）”。

[原文] [Host]: so yeah I know you worked on a bunch of different projects as a kid and I think I don't know if this was their first one but it was like fidget spinners and and and making fidget spinners um I don't think it was in your garage but maybe it was like in your room yeah what kind of stuff like that tinkering mindset how much of that have you kind of taken to this

[译文] [Host]: 是的，我知道你小时候做过很多不同的项目。我不确定这是否是第一个，但好像是指尖陀螺（fidget spinners），制作指尖陀螺。我想那不是在你家车库，可能是在你房间里？是的，那种捣鼓东西的心态，你有多少带到了现在的工作中？

[原文] [Sulaiman Ghori]: uh quite a bit quite a bit yeah um so I learned programming when I was quite young um my dad got me a book when I was like 11 and I liked it a lot um well I liked it a bit but I really started to like it when I realized you can make money from it and so um I I met some people online who were basically writing scripts for games as hacks and would sell them online um for small amounts of money but you know making a couple hundred bucks online was huge for me

[译文] [Sulaiman Ghori]: 相当多，相当多。我很小的时候就学了编程。大概 11 岁的时候，我爸给了我一本书，我很喜欢。嗯，虽然只是一点点喜欢，但当我意识到这能赚钱时，我才真正开始喜欢它。我在网上遇到了一些人，他们基本上是为游戏编写脚本作为外挂（hacks），然后在网上出售。虽然钱不多，但你知道，在那个年纪能在网上赚几百块钱对我来说是巨大的。

[原文] [Sulaiman Ghori]: um I think the first time that you like have someone give you money it's the strangest feeling crazy yeah i remember having to ask my dad for like a PayPal like custody account or whatever and uh getting the money in and it was like the coolest thing of all time ever for me

[译文] [Sulaiman Ghori]: 嗯，我觉得第一次有人给你钱的时候，那种感觉是最奇怪的。很疯狂。我记得当时还得求我爸帮我开一个 PayPal 托管账户之类的，然后钱打进来的时候，那对我来说简直是史上最酷的事情。

[原文] [Sulaiman Ghori]: um yeah it was really big and so uh I did that for um like a couple months and saved up enough money to at the time I was really interested in uh added manufacturing like 3D printers reprap was the big thing then so that was kind of where what kicked off the modern 3D printing revolution uh RepRap was like this built your own right yeah you had to that was the only option

[译文] [Sulaiman Ghori]: 是的，那真的很重要。所以我干了几个月，存够了钱。当时我对增材制造（additive manufacturing），比如 3D 打印机非常感兴趣。那时 RepRap 非常火，它算是开启了现代 3D 打印革命的东西。[Host]: RepRap 是那种通过自己组装的，对吧？[Sulaiman Ghori]: 是的，你必须自己组装，那是唯一的选择。

[原文] [Sulaiman Ghori]: um RepRap is literally just a bunch of university students basically um who said like let's see if we can build a machine that can build almost all the components for itself um which was that why it was called RepRap and uh they basically built in a variety of universities these rooms where you start with one printer um and then it prints the parts for the next printer and you go all the way up and you scale up

[译文] [Sulaiman Ghori]: RepRap 实际上基本上就是一群大学生，他们说：“让我们看看能不能造出一台能为自己制造几乎所有零件的机器。”这就是它为什么叫 RepRap（Replicating Rapid Prototyper）。他们基本上在各个大学里建立了这样的房间，你从一台打印机开始，然后它打印下一台打印机的零件，就这样一直做下去，规模不断扩大。

[原文] [Sulaiman Ghori]: but I was very obsessed with it and so I took one of their parts list and bought everything from Alibaba and a month later things came in and I assembled it all one night which went poorly actually when I was uh unbundling the copper cable for the power supply um which was a very sketchy power supply and did catch fire in the end um the all the copper windings came like loose and frayed everywhere and one went like 2 in into my thumb

[译文] [Sulaiman Ghori]: 当时我很迷这个，所以我拿了他们的零件清单，从阿里巴巴买了所有东西。一个月后东西到了，我花了一个晚上把它组装起来。但过程其实很糟糕。当我拆开电源的铜缆时——那个电源非常劣质，最后确实着火了——所有的铜线圈松开并散得到处都是，其中一根大概刺进了我拇指里 2 英寸深。

[原文] [Host]: did you just can't your thumb just doesn't work or did you go to the hospital or something

[译文] [Host]: 你的拇指是不是废了？还是你去了医院之类的？

[原文] [Sulaiman Ghori]: no so it was a school night and it was like 3:00 a.m cuz I wasn't very good at building things at 13 um and I spent like an hour in the bathroom trying to pull it out with tweezers and it just wasn't it was like it was bad so I just cut it off and I was like eh and so bit by bit over the next few weeks it came out and I would snip it off in the mornings um it was fun

[译文] [Sulaiman Ghori]: 没有。因为那天是上学的晚上（第二天要上学），而且已经是凌晨 3 点了，毕竟我 13 岁的时候还不太擅长做东西。我在浴室里花了一个小时试图用镊子把它拔出来，但根本拔不动，情况很糟。所以我直接把它（露在外面的部分）剪断了，我想着“诶，就这样吧”。在接下来的几周里，它一点一点地长出来，我就每天早上把它剪掉一点。嗯，很有趣。

[原文] [Sulaiman Ghori]: um yeah uh but I got the printer assembled um and uh around that time yeah the fidget spinner craze was going off so I bought 1 th00and skateboard bearings from China and basically set up a little factory uh in my bedroom where every two hours at night I would wake up and I would clear the print bed start a new print of fidget spinners and I would sell them online

[译文] [Sulaiman Ghori]: 是的，不过我把打印机组装好了。就在那个时候，指尖陀螺的热潮开始了。所以我从中国买了 1000 个滑板轴承，基本上在我的卧室里建了一个小工厂。每天晚上每隔两小时我就醒来一次，清理打印床，开始打印新一批指尖陀螺，然后在网上卖。

[原文] [Sulaiman Ghori]: and then um before school I had a little assembly line in my garage where I would uh put in the bearings spray paint dry and then run around to all the other bus stops of the other schools um sell them to my distributors which were just uh other kids of other schools sell all day at school come back collect from my distributors and then um sell online ship uh built a little healthy business

[译文] [Sulaiman Ghori]: 然后在上学前，我在车库有一条小装配线，我会把轴承装进去，喷漆，晾干。然后跑遍其他学校的所有校车站点，把货卖给我的“分销商”——其实就是其他学校的小孩。他们在学校卖一天，我回来后从分销商那里收钱。然后我在网上销售、发货。建立了一个很健康的小生意。

[原文] [Sulaiman Ghori]: and uh after 2 months they ended up getting shut down by the county um their official quoted reason was that the uh companies that sell the school food have technically an exclusive license to sell anything in school property but I think they just didn't like that I was distracting everyone and making money doing it

[译文] [Sulaiman Ghori]: 两个月后，这生意最后被县政府（county）查封了。他们官方给出的理由是，那些在学校卖食物的公司在技术上拥有在学校资产内销售任何东西的独家许可。但我觉得他们只是不喜欢我让大家分心，并且还以此赚钱。

[原文] [Sulaiman Ghori]: um but it taught me a good like healthy disrespect for authority i think that that has kind of been like a constant theme

[译文] [Sulaiman Ghori]: 嗯，但这教会了我一种很好的、对权威的良性蔑视（healthy disrespect for authority）。我觉得这已经成了我不变的主题。

这是访谈整理的第十章。本章节从 Sully 的个人哲学延伸到又一个硬核的工程故事。他阐述了为何要追求“非常规结果”就必须对抗体制化的惯例，随后讲述了他如何在感恩节前夜，为了赶在回家前完成任务，冒着爆炸风险在 6 英尺距离内点燃了一台手搓的液体燃料火箭发动机。

章节 10：蔑视权威的哲学与手搓液体火箭——当外套着火时

📝 本节摘要：

在本章中，Sully 深入探讨了他对“权威”和“体制”的看法。他认为，想要获得非凡（非常规）的成就，就必须打破体制所强制执行的惯例。他引用了 John Carlson 的观点，认为世界本质上是由少数人的“激情项目”（如 YKK 拉链）支撑奇迹。

>

随后，访谈进入了另一个高潮故事：Sully 曾在感恩节前夕决定制造并点火一台液体燃料火箭发动机。虽然前期花了四周通过教科书学习设计（因为火箭没有 GitHub 代码可抄），但最终的组装和点火是在一个疯狂的夜晚完成的。由于缺少长电源线，他被迫站在离引擎仅 6 英尺的地方用笔记本电脑供电点火，最终导致外套因泄漏的乙醇燃料而着火——但这件烧焦的外套如今成了他的奖杯。

[原文] [Host]: what what does that actually how is that materialized in your life with like the healthy disrespect for authority like what and you even mentioned um institutions like you don't like necessarily trust institutions um how did you kind of come to that and what does what does that look like

[译文] [Host]: 那实际上意味着什么？这种对权威的“良性蔑视”在你的生活中是如何具体体现的？你甚至提到了“体制（institutions）”，比如你未必信任体制。你是怎么形成这种想法的？那看起来像是什么样子的？

[原文] [Sulaiman Ghori]: um I I've always known from very young like I I want an unconventional outcome and so going through a conventional path would pretty much necessarily not lead you to an unconventional outcome so I grew opposed to any form of convention and institutions necessarily enforce convention

[译文] [Sulaiman Ghori]: 嗯，我从很小的时候就知道，我想要一个非传统的（unconventional）结果。所以，走传统的道路几乎必然不会带你走向非传统的结果。因此，我变得反对任何形式的惯例，而体制必然是强制执行惯例的。

[原文] [Sulaiman Ghori]: um I think creativity and interesting outcomes come mostly from free-spirited individuals um in almost every case if not all of them um I guess it's a bit of a like high highminded way of saying it but yeah like individuals are the most creative you can get and so staying true to that is the way to go

[译文] [Sulaiman Ghori]: 我认为创造力和有趣的结果，在几乎所有情况下（如果不是全部的话），主要来自于拥有自由精神的个体。我想这说法可能有点唱高调（high-minded），但没错，个体是你所能获得的最具创造力的源泉，所以坚持这一点才是正道。

[原文] [Sulaiman Ghori]: i do love uh John Carlson's idea of like everything is so hard to build and so hard to make especially you know put into the real world that if you look around it's basically like the world is just filled with some you know people's passion projects yeah it's a total miracle um there's a story behind every little thing um way more than you would think

[译文] [Sulaiman Ghori]: 我非常喜欢 John Carlson 的观点：建造任何东西都很难，尤其是把它真正带入现实世界。如果你环顾四周，这世界基本上充满了人们的“激情项目”。是的，这完全是个奇迹。每件小事背后都有一个故事，远比你想象的要多。

[原文] [Sulaiman Ghori]: i remember reading about the um I think it was YKK zippers apparently every good zipper like there's two or three companies in the world that make zippers which are actually pretty little little miracles they're very cheap but also mechanically comp like relatively complicated for how much you pay for them and there's only a few companies that are capable of building or have have set up to build them

[译文] [Sulaiman Ghori]: 我记得读到过关于——我想是 YKK 拉链的故事。显然，对于每条好拉链，世界上大概只有两三家公司在生产。拉链其实是微小的奇迹，它们非常便宜，但相对于你支付的价格，其机械结构又相对复杂。世界上只有少数几家公司有能力制造，或者建立了制造它们的体系。

[原文] [Sulaiman Ghori]: um and it's it's Yeah yeah basically this one Japanese guy's passion project over 40 years uh to figure out how to do this properly um and this is the case for pretty much everything um anything very specific and at scale is probably only done by a few companies or a few people in the world

[译文] [Sulaiman Ghori]: 嗯，是的，基本上这就是一个日本人在 40 年里为了搞清楚如何正确做好这件事而投入的激情项目。几乎所有事情都是如此。任何非常具体且具规模的事情，可能世界上只有少数几家公司或几个人在做。

[原文] [Host]: um so yeah I mean you hear about it every so often right like some company in Germany arbitrary company in Germany shuts down and Volkswagen has to halt all their lines or something like that um happens all the same it was a big thing in co right before we met you had made a liquid fuel I think rocket engine um it was like a very small thing i saw it upstairs um but you said we were talking before this that you did it in like 24 hours just on a whim um how did that happen

[译文] [Host]: 嗯，是的，我是说你经常听到这种事，对吧？比如德国某家不知名的公司倒闭了，大众汽车（Volkswagen）就不得不停掉所有生产线之类的。这事总是发生。在我们见面之前，你做过一件很厉害的事，我想是一个液体燃料火箭发动机。我在楼上看到了，是个很小的东西。但我们在聊之前你说过，你是在一时兴起下，大概用 24 小时做出来的。那是怎么回事？

[原文] [Sulaiman Ghori]: yeah um so it was a project over like roughly four weeks um and I started by literally just buying a bunch of textbooks um and trying to figure out like what are the design principles behind a rocket engine like how do I design it there's not like um you it's totally different from learning software where you can just go on GitHub and download people's code and modify it there's no file for a rocket engine

[译文] [Sulaiman Ghori]: 是的，那实际上是一个大概持续了四周的项目。我开始的时候真的是买了一堆教科书，试图搞清楚火箭发动机背后的设计原理是什么，比如我该如何设计它？这和学软件完全不同，你不能直接上 GitHub 下载别人的代码然后修改。火箭发动机没有现成的文件给你。

[原文] [Sulaiman Ghori]: you have to learn how to like what are the material properties what's the chemical properties how do you actually machine it um how do you design the parameters and know what to expect from in terms of thrust output and how do you not over pressure the engine and all these kinds of things

[译文] [Sulaiman Ghori]: 你必须去学习材料属性是什么？化学属性是什么？你实际上如何加工它？你如何设计参数并知道预期的推力输出是多少？你如何不让引擎超压？诸如此类的事情。

[原文] [Host]: um how did you design the injector which is uh the injector was very hard that was probably 50% of the time was that the hardest thing yeah the injector was very hard and it was like the biggest flaw in the end

[译文] [Host]: 你是如何设计喷注器（injector）的？

[Sulaiman Ghori]: 喷注器非常难。那大概占了 50% 的时间。

[Host]: 那是最难的部分吗？

[Sulaiman Ghori]: 是的，喷注器非常难，而且最后它也是最大的缺陷所在。

[原文] [Sulaiman Ghori]: um so yeah I spent like 3 4 weeks doing this and uh expedited a bunch of parts from China like CNC and all that stuff um and uh it was right before Thanksgiving i was going to go fly back to the east coast and visit my family and I was like "Okay either I fire it like build it and fire it tonight it was all just a bunch of parts at that time uh or I uh do this in two weeks and I'm like I'm not going to do this in two weeks i'm going to do this right now

[译文] [Sulaiman Ghori]: 嗯，所以我花了大概三四周做这个，还从中国加急订购了一堆零件，比如 CNC 加工件之类的。当时正好是感恩节前夕，我要飞回东海岸看望家人。我想：“好吧，要么我今晚就把它组装好并点火（当时它还只是一堆零件），要么我就得两周后再做。”然后我想：“我才不要两周后再做，我现在就要做。”

[原文] [Sulaiman Ghori]: so uh I drank a lot of coffee in the morning and then spend the whole day like hacking away at at uh aluminum extrusions and built out the test frame and then the the engine itself and uh let it off that night

[译文] [Sulaiman Ghori]: 所以那天早上我喝了大量的咖啡，然后花了一整天时间在那儿狂砍铝型材（aluminum extrusions），搭建了测试架，然后组装了引擎本身，并在当晚点火了。

[原文] [Sulaiman Ghori]: um yeah which had a lot of um we'll say uh concessions made to make it happen that night um I did find it absolutely hilarious that you like you said were you like a couple feet away

[译文] [Sulaiman Ghori]: 是的，为了在当晚搞定，我们做出了很多——怎么说呢——妥协（concessions）。

[Host]: 我觉得超级搞笑的是，你说你当时离它只有几英尺远？

[原文] [Sulaiman Ghori]: yeah so I designed it like I wasn't stupid i designed it so that I could remotely fire it but um I didn't the power supply hadn't come in yet to to remotely power the computer that was on board so I had to use a USB cable from my laptop to power the onboard computer and I didn't have a long enough USB cable uh the longest one I had was like 6 foot so I had to stand right next to it and light it up and I was like there's like a 30% chance that this thing explodes or or launches fire everywhere

[译文] [Sulaiman Ghori]: 是的。我是这样设计的——我不傻——我设计成可以远程点火。但是，给板载计算机远程供电的电源还没到货，所以我必须用我笔记本电脑的 USB 线给板载计算机供电。但我没有足够长的 USB 线，我最长的一根只有 6 英尺（约 1.8 米）。所以我不得不站在它旁边把它点着。我当时想：“大概有 30% 的几率这玩意儿会爆炸，或者把火喷得到处都是。”

[原文] [Sulaiman Ghori]: and actually um I don't know if it shows in the video i think it does show in the video but my jacket did catch fire because I I wasn't that great at designing the injector and it did create a lot of over pressure events which meant there was a lot of basically uh unburnt fuel spewing out which was ethanol and so that's liquid and just landed some landed on my on my jacket and caught fire um so yeah that's a trophy still the burnt jacket

[译文] [Sulaiman Ghori]: 实际上，我不知道视频里有没有拍到——我想应该拍到了——我的外套确实着火了。因为我的喷注器设计得不是很好，确实导致了很多超压事件，这意味着有大量未燃烧的燃料喷涌而出。那是乙醇，所以是液体，有些溅到了我的外套上，然后着火了。所以，是的，那件烧焦的外套现在还是我的奖杯。

📝 全文档总结 (Epilogue)

【核心主题】

本次访谈深度揭示了 xAI 如何在极短时间内成长为 AI 领域的巨头。贯穿始终的核心哲学是：对物理极限的极致追求与对人为规则的彻底蔑视。

【关键看点回顾】

[End of Document]