Reflections on Palantir 對於 Palantir 的反思
Published: October 15, 2024 (Substack link)
發佈日期：2024 年 10 月 15 日 (Substack 連結)
Palantir is hot now. The company recently joined the S&P 500. The stock is on a tear, and the company is nearing a $100bn market cap. VCs chase ex-Palantir founders asking to invest.
Palantir 現在非常熱門。該公司最近 加入了 S&P 500。股票表現強勁，市值即將達到 1000 億美元。風險投資家追逐前 Palantir 創始人，要求投資。
For long-time employees and alumni of the company, this feels deeply weird. During the 2016-2020 era especially, telling people you worked at Palantir was unpopular. The company was seen as spy tech, NSA surveillance, or worse. There were regular protests outside the office. Even among people who didn’t have a problem with it morally, the company was dismissed as a consulting company masquerading as software, or, at best, a sophisticated form of talent arbitrage.
對於公司的長期員工和校友來說，這感覺非常奇怪。特別是在 2016 年至 2020 年期間，告訴人們你在 Palantir 工作並不受歡迎。這家公司被視為間諜技術、NSA 監控，甚至更糟。辦公室外經常有抗議活動。即使在那些道德上對此沒有問題的人中，這家公司也被視為一個偽裝成軟體的諮詢公司，或者在最好的情況下，只是一種複雜的人才套利形式。
I left last year, but never wrote publicly about what I learned there. There’s also just a lot about the company people don’t understand. So this is my effort to explain some of that, as someone who worked there for eight years.
我去年離開了，但從未公開寫過我在那裡學到的東西。還有很多關於這家公司的事情人們並不理解。因此，作為一名在那裡工作了八年的人，這是我努力解釋其中一些內容。
(Note: I’m writing this in my personal capacity, and don’t have a formal relationship with the company anymore. I’m long $PLTR.)
（注意：我以個人身份撰寫此文，並且不再與該公司有正式關係。我持有 $PLTR 的多頭頭寸。）
1. Why I joined 1. 為什麼我加入了
I joined in summer 2015, initially in the newly-opened London office, before moving to Silicon Valley, and finally DC – as a forward deployed engineer. For context, the company was around 1500 people at the time; it had offices in Palo Alto (HQ), NYC, London, and a few other places. (It’s now 4000 or so people, and headquartered in Denver.)
我於 2015 年夏季加入，最初在新開的倫敦辦公室，然後轉到矽谷，最後到華盛頓特區，擔任前線部署工程師。當時公司大約有 1500 名員工；在帕洛阿爾托（總部）、紐約市、倫敦和其他幾個地方都有辦公室。（現在大約有 4000 名員工，總部位於丹佛。）
Why did I join? 我為什麼加入？
First, I wanted to work in ‘difficult’ industries on real, meaningful problems. My area of interest – for personal reasons - was healthcare and bio, which the company had a nascent presence in. The company was talking about working in industries like healthcare, aerospace, manufacturing, cybersecurity, and other industries that I felt were very important but that most people were not, at the time, working on. At the time the hot things were social networks (Facebook, LinkedIn, Quora, etc.) and other miscellaneous consumer apps (Dropbox, Uber, Airbnb) but very few companies were tackling what felt like the real, thorny parts of the economy. If you wanted to work on these ‘harder’ areas of the economy but also wanted a Silicon Valley work culture, Palantir was basically your only option for awhile.
首先，我想在“艱難”的行業中解決真正有意義的問題。出於個人原因，我的興趣領域是醫療保健和生物技術，而該公司在這方面有著初步的存在。該公司談論著在醫療保健、航空航天、製造業、網絡安全等行業工作，而這些行業我認為非常重要，但當時大多數人並未投入其中。當時的熱門話題是社交網絡（如 Facebook、LinkedIn、Quora 等）和其他各類消費者應用（如 Dropbox、Uber、Airbnb），但很少有公司在解決經濟中那些真正棘手的部分。如果你想在這些“更艱難”的經濟領域工作，同時又希望擁有矽谷的工作文化，Palantir 基本上是你唯一的選擇。
My goal was to start a company, but I wanted (1) to go deep in one of these industries for a while first and learn real things about it; (2) to work for a US company and get a green card that way. Palantir offered both. That made it an easy choice.
我的目標是創辦一家公司，但我想先在這些行業中深入一段時間，學習真正的知識；還有，我希望能為美國公司工作，藉此獲得綠卡。Palantir 提供了這兩者，因此這是一個簡單的選擇。
Second, talent density. I talked to some of the early people who started the healthcare vertical (Nick Perry, Lekan Wang, and Andrew Girvin) and was extremely impressed. I then interviewed with a bunch of the early business operations and strategy folks and came away even more impressed. These were seriously intense, competitive people who wanted to win, true believers; weird, fascinating people who read philosophy in their spare time, went on weird diets, and did 100-mile bike rides for fun. This, it turned out, was an inheritance from the Paypal mafia. Yishan Wong, who was early at Paypal, wrote about the importance of intensity:
第二，人才密度。我與一些早期創建醫療垂直領域的人（Nick Perry、Lekan Wang 和 Andrew Girvin）交談，深感印象深刻。隨後，我與一些早期的業務運營和策略人員進行了面試，結果讓我更加印象深刻。他們都是非常專注、競爭激烈的人，渴望成功，是真正的信徒；有些奇特而迷人的人，閒暇時閱讀哲學，嘗試奇怪的飲食，並為了娛樂進行 100 英里的自行車騎行。這顯然是來自 Paypal 黑手黨的遺產。早期在 Paypal 工作的 Yishan Wong 曾寫過關於強度的重要性：
"In general, as I begin to survey more startups, I find that the talent level at PayPal is not uncommon for a Silicon Valley startup, but the differentiating factor may have been the level of intensity from the top: both Peter Thiel and Max Levchin were extremely intense people - hyper-competitive, hard-working, and unwilling to accept defeat. I think this sort of leadership is what pushes the "standard" talented team to be able to do great things and, subsequently, contributes to producing a wellspring of later achievements."Palantir was an unusually weird place, too. I remember my first time I talked to Stephen Cohen he had the A/C in his office set at 60, several weird-looking devices for minimizing CO2 content in the room, and had a giant pile of ice in a cup. Throughout the conversation, he kept chewing pieces of ice. (Apparently there are cognitive benefits to this.)
「一般來說，當我開始調查更多的初創公司時，我發現 PayPal 的人才水平在矽谷的初創公司中並不罕見，但區別的因素可能是來自高層的強度：彼得·蒂爾和馬克斯·列夫欽都是極具強度的人——競爭激烈、勤奮工作，並且不願接受失敗。我認為這種領導風格能夠推動“標準”人才團隊做出偉大的成就，並進而促成後來的豐碩成果。」
Palantir 也是一個非常奇怪的地方。我記得第一次和 Stephen Cohen 談話時，他的辦公室空調設在 60 度，裡面有幾個看起來奇怪的裝置用來降低房間內的二氧化碳含量，還有一杯裡面堆滿了冰塊。在整個對話過程中，他一直在咀嚼冰塊。（顯然這對認知有好處。）
I also interviewed with the CEO, Alex Karp and talked to other members of the leadership team. I probably don’t need to convince you that Karp is weird - just watch an interview with him. I can’t say what Karp and I talked about, but he gives a good flavor for his style in a 2012 interview:
我也與首席執行官亞歷克斯·卡普進行了面試，並與其他領導團隊成員交談。我想我不需要說服你卡普有多奇怪——只需看看他的一次訪談。我不能透露我和卡普談了什麼，但他在 2012 年的一次訪談中展現了他的風格：
I like to meet candidates with no data about them: no résumé, no preliminary discussions or job description, just the candidate and me in a room. I ask a fairly random question, one that is orthogonal to anything they would be doing at Palantir. I then watch how they disaggregate the question, if they appreciate how many different ways there are to see the same thing. I like to keep interviews short, about 10 minutes. Otherwise, people move into their learned responses and you don’t get a sense of who they really are.My interviews were often not about work or software at all – one of my interviews we just spent an hour talking about Wittgenstein. Note that both Peter Thiel and Alex Karp were philosophy grads. Thiel’s lecture notes had come out not long before and they discussed Shakespeare, Tolstoy, Girard (then unknown, now a cliché) and more.
我喜歡與沒有任何資料的候選人見面：沒有履歷、沒有初步討論或職位描述，只有候選人和我在一個房間裡。我會問一個相當隨意的問題，這個問題與他們在 Palantir 的工作無關。我會觀察他們如何拆解這個問題，是否能夠理解同一件事有多少種不同的看法。我喜歡將面試保持在短時間內，大約 10 分鐘。否則，人們會進入他們習慣的回答模式，這樣就無法真正了解他們是誰。
我的面試往往根本不是關於工作或軟體——在其中一場面試中，我們花了一個小時討論維根斯坦。值得注意的是，彼得·蒂爾和亞歷克斯·卡普都是哲學畢業生。蒂爾的講義不久前剛發布，他們討論了莎士比亞、托爾斯泰、吉拉爾（當時不為人知，現在卻成了陳詞濫調）等等。
The combo of intellectual grandiosity and intense competitiveness was a perfect fit for me. It’s still hard to find today, in fact - many people have copied the ‘hardcore’ working culture and the ‘this is the Marines’ vibe, but few have the intellectual atmosphere, the sense of being involved in a rich set of ideas. This is hard to LARP - your founders and early employees have to be genuinely interesting intellectual thinkers. The main companies that come to mind which have nailed this combination today are OpenAI and Anthropic. It’s no surprise they’re talent magnets. [1]
智力的宏大與強烈的競爭性完美契合了我。事實上，今天仍然很難找到這種組合——許多人模仿了「硬核」的工作文化和「這裡是海軍陸戰隊」的氛圍，但很少有人擁有那種智力氛圍，讓人感受到參與一系列豐富的思想。這很難模擬——你的創始人和早期員工必須是真正有趣的智力思考者。今天讓我想到的主要公司，成功結合了這兩者的是OpenAI和Anthropic。他們成為人才磁鐵並不奇怪。[1]
2. Forward deployed 2. 前進部署
When I joined, Palantir was divided up into two types of engineers:
當我加入時，Palantir 分為兩種類型的工程師：
- Engineers who work with customers, sometimes known as FDEs, forward deployed engineers.
與客戶合作的工程師，有時被稱為 FDEs，即前線部署工程師。 - Engineers who work on the core product team (product development - PD), and rarely go visit customers.
在核心產品團隊（產品開發 - PD）工作的工程師，鮮少去拜訪客戶。
FDEs were typically expected to ‘go onsite’ to the customer’s offices and work from there 3-4 days per week, which meant a ton of travel. This is, and was, highly unusual for a Silicon Valley company.
FDE 通常被期望每週到客戶的辦公室工作 3-4 天，這意味著需要大量的旅行。這對於一家矽谷公司來說，是非常不尋常的。
There’s a lot to unpack about this model, but the key idea is that you gain intricate knowledge of business processes in difficult industries (manufacturing, healthcare, intel, aerospace, etc.) and then use that knowledge to design software that actually solves the problem. The PD engineers then ‘productize’ what the FDEs build, and – more generally – build software that provides leverage for the FDEs to do their work better and faster. [2]
這個模型有很多需要深入探討的地方，但關鍵概念是，你獲得了在困難行業（如製造、醫療、情報、航空航天等）中業務流程的深入知識，然後利用這些知識來設計實際解決問題的軟體。產品開發工程師接著將 FDE 所建構的產品化，並且—更一般地說—開發能夠幫助 FDE 更好、更快地完成工作的軟體。[2]
This is how much of the Foundry product took initial shape: FDEs went to customer sites, had to do a bunch of cruft work manually, and PD engineers built tools that automated the cruft work. Need to bring in data from SAP or AWS? Here’s Magritte (a data ingestion tool). Need to visualize data? Here’s Contour (a point and click visualization tool). Need to spin up a quick web app? Here’s Workshop (a Retool-like UI for making webapps). Eventually, you had a damn good set of tools clustered around the loose theme of ‘integrate data and make it useful somehow’.
這就是 Foundry 產品最初形成的樣子：FDEs 會前往客戶現場，必須手動處理一堆繁瑣的工作，而 PD 工程師則開發了自動化這些繁瑣工作的工具。需要從 SAP 或 AWS 獲取數據？這裡有 Magritte（一個數據攝取工具）。需要可視化數據？這裡有 Contour（一個點擊式可視化工具）。需要快速啟動一個網頁應用？這裡有 Workshop（一個類似 Retool 的網頁應用界面）。最終，你擁有了一套非常出色的工具，圍繞著「整合數據並以某種方式使其有用」這一鬆散主題。
At the time, it was seen as a radical step to give customers access to these tools — they weren’t in a state for that — but now this drives 50%+ of the company’s revenue, and it’s called Foundry. Viewed this way, Palantir pulled off a rare services company → product company pivot: in 2016, descriptions of it as a Silicon Valley services company were not totally off the mark, but in 2024 they are deeply off the mark, because the company successfully built an enterprise data platform using the lessons from those early years, and it shows in the gross margins - 80% gross margins in 2023. These are software margins. Compare to Accenture: 32%.
當時，讓客戶使用這些工具被視為一個激進的舉措——他們並不具備這樣的能力——但現在這驅動了公司超過 50%的收入，這被稱為Foundry。從這個角度看，Palantir 成功地實現了從服務公司到產品公司的轉型：在 2016 年，將其描述為矽谷的服務公司並不完全錯誤，但到了 2024 年，這種描述就完全不準確了，因為該公司成功地建立了一個企業數據平台，並從早期的經驗中汲取了教訓，這在毛利率上得到了體現——2023 年的毛利率達到 80%。這是軟體毛利率。與埃森哲相比：32%。
Tyler Cowen has a wonderful saying, ‘context is that which is scarce’, and you could say it’s the foundational insight of this model. Going onsite to your customers – the startup guru Steve Blank calls this “getting out of the building” – means you capture the tacit knowledge of how they work, not just the flattened ‘list of requirements’ model that enterprise software typically relies on. The company believed this to a hilarious degree: it was routine to get a call from someone and have to book a first-thing-next-morning flight to somewhere extremely random; “get on a plane first, ask questions later” was the cultural bias. This resulted in out of control travel spend for a long time — many of us ended up getting United 1K or similar — but it also meant an intense decade-long learning cycle which eventually paid off.
泰勒·科文有一句很棒的話：「情境是稀缺的東西」，可以說這是這個模型的基礎洞見。親自去拜訪客戶——創業大師史蒂夫·布蘭克稱之為「走出大樓」——意味著你能捕捉到他們工作方式的隱性知識，而不僅僅是企業軟體通常依賴的扁平「需求清單」模型。公司對此深信不疑：接到某人的電話後，預訂第二天一早飛往某個極其隨機的地方的航班已成為常態；「先上飛機，後問問題」成為了文化偏見。這導致了長期的旅行支出失控——我們中的許多人最終獲得了美國聯合航空 1K 或類似的會員資格——但這也意味著經歷了一個長達十年的密集學習周期，最終得到了回報。
My first real customer engagement was with Airbus, the airplane manufacturer based in France, and I moved out to Toulouse for a year and worked in the factory alongside the manufacturing people four days a week to help build the version of our software there.
我第一次真正的客戶合作是與法國的飛機製造商空中巴士（Airbus）進行的。我搬到了圖盧茲，並在那裡工作了一年，每週四天與製造團隊一起在工廠裡協助開發我們的軟體版本。
My first month in Toulouse, I couldn’t fly out of the city because the air traffic controllers were on strike every weekend. Welcome to France. (I jest - France is great. Also, Airbus planes are magnificent. It’s a truly engineering-centric company. The CEO is always a trained aeronautical engineer, not some MBA. Unlike… anyway.)
我在圖盧茲的第一個月，每個週末都無法飛出這座城市，因為空中交通管制員罷工。歡迎來到法國。（我開玩笑的——法國很好。此外，空中巴士的飛機非常棒。這是一家真正以工程為中心的公司。CEO 總是受過航空工程訓練，而不是某個 MBA。與…無論如何。）
The CEO told us his biggest problem was scaling up A350 manufacturing. So we ended up building software to directly tackle that problem. I sometimes describe it as “Asana, but for building planes”. You took disparate sources of data — work orders, missing parts, quality issues (“non-conformities”) — and put them in a nice interface, with the ability to check off work and see what other teams are doing, where the parts are, what the schedule is, and so on. Allow them the ability to search (including fuzzy/semantic search) previous quality issues and see how they were addressed. These are all sort of basic software things, but you’ve seen how crappy enterprise software can be - just deploying these ‘best practice’ UIs to the real world is insanely powerful. This ended up helping to drive the A350 manufacturing surge and successfully 4x’ing the pace of manufacturing while keeping Airbus’s high standards of quality.
CEO 告訴我們，他最大的問題是擴大 A350 的生產。因此，我們最終開發了軟體來直接解決這個問題。我有時會將其形容為「Asana，但用於建造飛機」。你將不同來源的數據——工作訂單、缺失零件、質量問題（「不合格項」）——整合到一個漂亮的介面中，並能夠勾選工作，查看其他團隊的進展、零件的位置、時間表等等。讓他們能夠搜索（包括模糊/語義搜索）過去的質量問題，並查看如何解決這些問題。這些都是基本的軟體功能，但你也知道企業軟體有多糟糕——將這些「最佳實踐」的用戶介面應用到現實世界中是極其強大的。這最終幫助推動了 A350 的生產激增，成功將生產速度提高了四倍，同時保持了空中巴士的高質量標準。
This made the software hard to describe concisely - it wasn’t just a database or a spreadsheet, it was an end-to-end solution to that specific problem, and to hell with generalizability. Your job was to solve the problem, and not worry about overfitting; PD’s job was to take whatever you’d built and generalize it, with the goal of selling it elsewhere.
這使得軟體難以簡潔地描述——它不僅僅是一個資料庫或電子表格，而是一個針對那個特定問題的端到端解決方案，至於通用性就不必考慮了。你的工作是解決問題，而不是擔心過度擬合；產品經理的工作則是將你所建構的東西進行通用化，目標是將其銷售到其他地方。
The A350 final assembly line, in Toulouse. I hung out here most days. It was awe-inspiring. Screenshot from here.
A350 最終組裝線，位於圖盧茲。我幾乎每天都在這裡待著。這裡令人敬畏。來源。
FDEs tend to write code that gets the job done fast, which usually means – politely – technical debt and hacky workarounds. PD engineers write software that scales cleanly, works for multiple use cases, and doesn’t break. One of the key ‘secrets’ of the company is that generating deep, sustaining enterprise value requires both. BD engineers tend to have high pain tolerance, the social and political skills needed to embed yourself deep in a foreign company and gain customer trust, and high velocity – you need to build something that delivers a kernel of value fast so that customers realize you’re the real deal. It helped that customers had hilariously low expectations of most software contractors, who were typically implementors of SAP or other software like that, and worked on years-long ‘waterfall’ style timescales. So when a ragtag team of 20-something kids showed up to the customer site and built real software that people could use within a week or two, people noticed.
FDEs 通常編寫能快速完成工作的代碼，這通常意味著——禮貌地說——技術負債和臨時解決方案。PD 工程師則編寫可擴展的軟件，能夠適應多種用例，並且不會出現故障。公司的其中一個關鍵「秘密」是，創造深遠且持久的企業價值需要兩者的結合。BD 工程師通常具有高痛苦容忍度，具備深入外國公司並獲得客戶信任所需的社交和政治技能，以及高效率——你需要快速構建出能提供核心價值的產品，讓客戶意識到你是真正的實力派。客戶對大多數軟件承包商的期望通常非常低，這些承包商通常是 SAP 或其他類似軟件的實施者，並且在多年「瀑布式」的時間表上工作。因此，當一支由二十多歲的年輕人組成的雜牌軍出現在客戶現場，並在一兩周內構建出可供使用的真正軟件時，人們注意到了這一點。
This two-pronged model made for a powerful engine. Customer teams were often small (4-5 people) and operated fast and autonomously; there were many of them, all learning fast, and the core product team’s job was to take those learnings and build the main platform.
這種雙管齊下的模式形成了一個強大的引擎。客戶團隊通常規模較小（4-5 人），運作迅速且自主；這樣的團隊有很多，大家都在快速學習，而核心產品團隊的工作就是將這些學習轉化為主要平台的建設。
When we were allowed to work within an organization, this tended to work very well. The obstacles were mostly political. Every time you see the government give another $110 million contract to Deloitte for building a website that doesn’t work, or a healthcare.gov style debacle, or SFUSD spending $40 million to implement a payroll system that - again - doesn’t work, you are seeing politics beat substance. See SpaceX vs. NASA as another example.
當我們被允許在一個組織內工作時，這通常運作得非常好。障礙主要是政治因素。每當你看到政府再次給予德勤$1.1 億的合約來建設一個無法運作的網站，或是類似 healthcare.gov 的災難，或是舊金山學區花費$4000 萬實施一個同樣無法運作的薪資系統，你就會看到政治壓過實質。再看看SpaceX 與 NASA的例子。
The world needs more companies like SpaceX, and Palantir, that differentiate on execution - achieving the outcome - not on playing political games or building narrow point solutions that don’t hit the goal.
世界需要更多像 SpaceX 和 Palantir 這樣的公司，它們在 執行 上有所區別——實現結果——而不是玩政治遊戲或建立無法達成目標的狹隘解決方案。
3. Secrets 3. 秘密
Another key thing FDEs did was data integration, a term that puts most people to sleep. This was (and still is) the core of what the company does, and its importance was underrated by most observers for years. In fact, it’s only now with the advent of AI that people are starting to realize the importance of having clean, curated, easy-to-access data for the enterprise. (See: the ‘it’ in AI models is the dataset).
另一個 FDEs 所做的關鍵事情是數據整合，這個詞讓大多數人昏昏欲睡。這是（並且仍然是）公司運作的核心，其重要性多年來被大多數觀察者低估。事實上，隨著人工智慧的興起，人們才開始意識到擁有乾淨、經過整理且易於訪問的企業數據的重要性。（參見：AI 模型中的“它”是數據集）。
In simple terms, ‘data integration’ means (a) gaining access to enterprise data, which usually means negotiating with ‘data owners’ in an organization (b) cleaning it and sometimes transforming it so that it’s usable (c) putting it somewhere everyone can access it. Much of the base, foundational software in Palantir’s main software platform (Foundry) is just tooling to make this task easier and faster.
簡單來說，「數據整合」是指 (a) 獲取企業數據，這通常意味著與組織中的「數據擁有者」進行協商 (b) 清理數據，有時還需要轉換數據，以便能夠使用 (c) 將數據放在每個人都能訪問的地方。Palantir 主要軟體平台（Foundry）中的許多基礎軟體只是為了使這項任務更簡單、更快速的工具。
Why is data integration so hard? The data is often in different formats that aren’t easily analyzed by computers – PDFs, notebooks, Excel files (my god, so many Excel files) and so on. But often what really gets in the way is organizational politics: a team, or group, controls a key data source, the reason for their existence is that they are the gatekeepers to that data source, and they typically justify their existence in a corporation by being the gatekeepers of that data source (and, often, providing analyses of that data). [3] This politics can be a formidable obstacle to overcome, and in some cases led to hilarious outcomes – you’d have a company buying an 8-12 week pilot, and we’d spend all 8-12 weeks just getting data access, and the final week scrambling to have something to demo.
為什麼數據整合如此困難？數據通常以不同的格式存在，這些格式不易被計算機分析——PDF、筆記本、Excel 文件（天啊，這麼多 Excel 文件）等等。但真正阻礙的往往是組織內部的政治：一個團隊或小組控制著一個關鍵的數據來源，他們存在的理由就是作為該數據來源的守門人，通常他們在公司中的存在價值就是作為該數據來源的守門人（而且，經常還提供該數據的分析）。這種政治可能成為一個難以克服的障礙，在某些情況下甚至導致搞笑的結果——你會看到一家公司購買了一個 8 到 12 週的試點，我們卻花了整整 8 到 12 週僅僅是獲取數據訪問，最後一週則忙著準備一些可以展示的內容。
The other ‘secret’ Palantir figured out early is that data access tussles were partly about genuine data security concerns, and could be alleviated through building security controls into the data integration layer of the platform - at all levels. This meant role-based access controls, row-level policies, security markings, audit trails, and a ton of other data security features that other companies are still catching up to. Because of these features, implementing Palantir often made companies’ data more secure, not less. [4]
另一個「秘密」是，Palantir 早期發現數據訪問的爭鬥部分源於真正的數據安全擔憂，並且可以通過在平台的數據整合層中建立安全控制來緩解——在所有層級上。這意味著基於角色的訪問控制、行級政策、安全標記、審計追蹤，以及其他許多數據安全功能，這些功能是其他公司仍在追趕的。正因為這些功能，實施 Palantir 通常使公司的數據更安全，而不是更不安全。[4]
4. Notes on culture 4. 文化備註
The overall ‘vibe’ of the company was more of a messianic cult than a normal software company. But importantly, it seemed that criticism was highly tolerated and welcomed – one person showed me an email chain where an entry-level software engineer was having an open, contentious argument with a Director of the company with the entire company (around a thousand people) cc’d. As a rationalist-brained philosophy graduate, this particular point was deeply important to me – I wasn’t interested in joining an uncritical cult. But a cult of skeptical people who cared deeply and wanted to argue about where the world was going and how software fit into it – existentially – that was interesting to me. [5]
這家公司的整體「氛圍」更像是一個救世主崇拜的邪教，而不是一個普通的軟體公司。但重要的是，批評似乎受到高度的容忍和歡迎——有一個人向我展示了一封電子郵件鏈，其中一位初級軟體工程師與公司的董事進行了一場公開且具爭議性的辯論，整個公司（大約一千人）都被抄送了。作為一名理性思維的哲學畢業生，這一點對我來說非常重要——我不想加入一個不批判的邪教。但一個充滿懷疑精神、深切關心並想要討論世界走向及軟體在其中的存在意義的團體——這對我來說是有趣的。
I’m not sure if they still do this, but at the time when you joined they sent you a copy of Impro, The Looming Tower (9/11 book), Interviewing Users, and Getting Things Done. I also got an early PDF version of what became Ray Dalio’s Principles. This set the tone. The Looming Tower was obvious enough – the company was founded partly as a response to 9/11 and what Peter felt were the inevitable violations of civil liberties that would follow, and the context was valuable. But why Impro?
我不確定他們現在是否還這樣做，但在你加入的時候，他們會寄給你一本《Impro》、一本《The Looming Tower》（9/11 書籍）、一本《Interviewing Users》和一本《Getting Things Done》。我還得到了一個早期的 PDF 版本，後來成為雷·達里奧的《Principles》。這些書籍定下了基調。《The Looming Tower》顯而易見——這家公司部分是作為對 9/11 的回應而成立的，彼得認為隨之而來的公民自由的不可避免的侵犯，這個背景是有價值的。但為什麼是《Impro》呢？
Being a successful FDE required an unusual sensitivity to social context – what you really had to do was partner with your corporate (or government) counterparts at the highest level and gain their trust, which often required playing political games. Impro is popular with nerds partly because it breaks down social behavior mechanistically. The vocabulary of the company was saturated with Impro-isms – ‘casting’ is an example. Johnstone discusses how the same actor can play ‘high status’ or ‘low status’ just by changing parts of their physical behavior – for example, keeping your head still while talking is high status, whereas moving your head side to side a lot is low status. Standing tall with your hands showing is high status, slouching with your hands in your pocket is low status. And so on. If you didn’t know all this, you were unlikely to succeed in a customer environment. Which meant you were unlikely to integrate customer data or get people to use your software. Which meant failure.
成功的 FDE 需要對社會背景有異常的敏感度——你真正需要做的是與企業（或政府）高層的對口合作並獲得他們的信任，這通常需要玩一些政治遊戲。即興表演在極客中受歡迎，部分原因是它將社交行為機械化地拆解。公司的詞彙中充滿了即興表演的術語——“角色分配”就是一個例子。約翰斯頓討論了同一位演員如何僅通過改變身體行為的某些部分來表現“高地位”或“低地位”——例如，說話時保持頭部靜止是高地位，而頭部左右晃動則是低地位。站得筆直並展示雙手是高地位，駝背並把手放在口袋裡則是低地位。等等。如果你不知道這些，你在客戶環境中成功的可能性就不大。這意味著你不太可能整合客戶數據或讓人們使用你的軟體。這意味著失敗。
This is one reason why former FDEs tend to be great founders. (There are usually more ex-Palantir founders than there are ex-Googlers in each YC batch, despite there being ~50x more Google employees.) Good founders have an instinct for reading rooms, group dynamics, and power. This isn’t usually talked about, but it’s critical: founding a successful company is about taking part in negotiation after negotiation after negotiation, and winning (on net). Hiring, sales, fundraising are all negotiations at their core. It’s hard to be great at negotiating without having these instincts for human behavior. This is something Palantir teaches FDEs, and is hard to learn at other Valley companies.
這就是為什麼前 FDE（全職工程師）往往成為優秀創始人的原因之一。每個 YC 批次中，前 Palantir 創始人的數量通常超過前 Google 員工，儘管 Google 的員工數量大約是 Palantir 的 50 倍。優秀的創始人對於讀懂會議氛圍、群體動態和權力有著敏銳的直覺。這通常不被提及，但卻至關重要：創建一家成功的公司就是不斷參與談判並最終獲勝（淨收益）。招聘、銷售、籌款本質上都是談判。如果沒有對人類行為的這種直覺，很難在談判中表現出色。這是 Palantir 教給 FDE 的東西，而在其他矽谷公司則很難學到。
Another is that FDEs have to be good at understanding things. Your effectiveness directly correlates to how quickly you can learn to speak the customer’s language and really drill down into how their business works. If you’re working with hospitals, you quickly learn to talk about capacity management and patient throughput vs. just saying “help you improve your healthcare”. Same with drug discovery, health insurance, informatics, cancer immunotherapy, and so on; all have specialized vocabularies, and the people who do well tend to be great at learning them fast.
另一個是 FDE 必須要有良好的理解能力。你的工作效率與你學會用客戶的語言溝通的速度直接相關，並深入了解他們的業務運作。如果你在醫院工作，你會迅速學會談論容量管理和病人流量，而不僅僅是說「幫助你改善醫療服務」。藥物發現、健康保險、資訊學、癌症免疫療法等領域也是如此；這些都有專業的詞彙，而表現出色的人往往能夠快速學習這些詞彙。
One of my favorite insights from Tyler Cowen’s book ‘Talent’ is that the most talented people tend to develop their own vocabularies and memes, and these serve as entry points to a whole intellectual world constructed by that person. Tyler himself is of course a great example of this. Any MR reader can name 10+ Tylerisms instantly - ‘model this’, ‘context is that which is scarce’, ‘solve for the equilibrium’, ‘the great stagnation’ are all examples. You can find others who are great at this. Thiel is one. Elon is another (“multiplanetary species”, “preserving the light of consciousness”, etc. are all memes). Trump, Yudkowsky, gwern, SSC, Paul Graham, all of them regularly coin memes. It turns out that this is a good proxy for impact.
我最喜歡的泰勒·科文（Tyler Cowen）在《人才》一書中的見解之一是，最有才華的人往往會發展出自己的詞彙和迷因，而這些詞彙和迷因成為他們所構建的整個智識世界的切入點。泰勒本人當然是這方面的絕佳例子。任何《馬爾科經濟學》（MR）的讀者都能立即列出 10 個以上的泰勒語錄——「建模這個」、「背景是稀缺的」、「解決均衡問題」、「偉大的停滯」等都是例子。你可以找到其他擅長這方面的人。彼得·蒂爾（Thiel）就是其中之一，埃隆·馬斯克（Elon）則是另一個（「多行星物種」、「保護意識的光芒」等都是迷因）。特朗普、尤德科斯基、gwern、SSC、保羅·格雷厄姆，他們都經常創造迷因。事實證明，這是一個良好的影響力指標。
This insight goes for companies, too, and Palantir had its own, vast set of terms, some of which are obscure enough that “what does Palantir actually do?” became a meme online. ‘Ontology’ is an old one, but then there is ‘impl’, ‘artist’s colony’, ‘compounding’, ‘the 36 chambers’, ‘dots’, ‘metabolizing pain’, ‘gamma radiation’, and so on. The point isn’t to explain all of these terms, each of which compresses a whole set of rich insights; it’s that when you’re looking for companies to join, you could do worse than look for a rich internal language or vocabulary that helps you think about things in a more interesting way.
這個見解同樣適用於公司，而 Palantir 擁有自己龐大的術語集，其中一些術語晦澀到“Palantir 到底在做什麼？”成為了網上的一個迷因。“本體論”是一個古老的術語，但還有“impl”、“藝術家社區”、“複合”、“36 個房間”、“點”、“代謝痛苦”、“伽馬輻射”等等。重點不是解釋這些術語，每一個都壓縮了一整套豐富的見解；而是當你在尋找要加入的公司時，尋找一種豐富的內部語言或詞彙，能幫助你以更有趣的方式思考事物，這樣的選擇不會錯。
When Palantir’s name comes up, most people think of Peter Thiel. But many of these terms came from early employees, especially Shyam Sankar, who’s now the President of the company. Still, Peter is deeply influential in the company culture, even though he wasn’t operationally involved with the company at all during the time I was there. This document, written by Joe Lonsdale, was previously an internal document but made public at some point and gives a flavor for the type of cultural principles.
當提到 Palantir 的名字時，大多數人會想到彼得·蒂爾。但許多這些術語來自早期員工，特別是現在擔任公司總裁的 Shyam Sankar。不過，彼得在公司文化中仍然具有深遠的影響力，即使在我在那裡的時候，他根本沒有參與公司的運營。這份文件是由喬·朗斯代爾撰寫的，之前是一份內部文件，但在某個時候公開，展現了文化原則的風貌。
One of the things that (I think) came from Peter was the idea of not giving people titles. When I was there, everyone had the “forward deployed engineer” title, more or less, and apart from that there were five or six Directors and the CEO. Occasionally someone would make up a different title (one guy I know called himself “Head of Special Situations”, which I thought was hilarious) but these never really caught on. It’s straightforward to trace this back to Peter’s Girardian beliefs: if you create titles, people start coveting them, and this ends up creating competitive politics inside the company that undermines internal unity. Better to just give everyone the same title and make them go focus on the goal instead.
我認為彼得提出的一個想法是不要給人們頭銜。在我在那裡的時候，每個人幾乎都有「前線部署工程師」的頭銜，除此之外只有五六位董事和 CEO。偶爾會有人自創不同的頭銜（我認識的一位男士自稱為「特殊情況負責人」，我覺得這非常搞笑），但這些頭銜從未真正流行起來。這可以直接追溯到彼得的吉拉爾德信念：如果你創造了頭銜，人們就會開始渴望它們，這最終會在公司內部造成競爭政治，破壞內部團結。還不如給每個人一樣的頭銜，讓他們專注於目標。
There are plenty of good critiques of the ‘flat hierarchy’ stance -- The Tyranny of Structurelessness is a great one – and it largely seems to have fallen out of fashion in modern startups, where you quickly get CEO, COO, VPs, Founding Engineers, and so on. But my experience is that it worked well at Palantir. Some people were more influential than others, but the influence was usually based on some impressive accomplishment, and most importantly nobody could tell anyone else what to do. So it didn’t matter if somebody was influential or thought your idea was dumb, you could ignore them and go build something if you thought it was the right thing to do. On top of that, the culture valorized such people: stories were told of some engineer ignoring a Director and building something that ended up being a critical piece of infrastructure, and this was held up as an example to imitate.
對於「扁平層級」的立場有很多好的批評——《無結構的暴政》就是一個很好的例子——而且在現代初創公司中，這種觀點似乎已經不再流行，因為你很快就會看到 CEO、COO、副總裁、創始工程師等等。但根據我的經驗，這在 Palantir 運作得很好。有些人的影響力比其他人更大，但這種影響力通常是基於某些令人印象深刻的成就，最重要的是沒有人可以告訴其他人該怎麼做。因此，無論某人是否有影響力，或是認為你的想法愚蠢，你都可以忽略他們，去建造你認為正確的東西。此外，文化也讚美這樣的人：有故事講述某位工程師無視一位主管，建造出最終成為關鍵基礎設施的東西，這被視為值得模仿的榜樣。
The cost of this was that the company often felt like there was no clear strategy or direction, more like a Petri dish of smart people building little fiefdoms and going off in random directions. But it was incredibly generative. It’s underrated just how many novel UI concepts and ideas came out of that company. Only some of these now have non-Palantir equivalents, e.g. Hex, Retool, Airflow all have some components that were first developed at Palantir. The company’s doing the same for AI now – the tooling for deploying LLMs at large enterprises is powerful.
這樣的代價是，該公司經常感覺沒有明確的策略或方向，更像是一個聰明人聚集的培養皿，各自建立小領地，隨意發展。但這卻是極具創造性的。該公司所產生的許多新穎的用戶介面概念和想法被低估了。現在只有一些這些概念有非 Palantir 的對應產品，例如 Hex、Retool 和 Airflow 都有一些最初在 Palantir 開發的組件。該公司現在在 AI 領域也在做同樣的事情——在大型企業中部署 LLMs 的工具 是非常強大的。
The ‘no titles’ thing also meant that people came in and out of fashion very quickly, inside the company. Because everyone had the same title, you had to gauge influence through other means, and those were things like “who seems really tight with this Director right now” or “who is leading this product initiative which seems important”, not “this person is the VP of so and so”. The result was a sort of hero-shithead rollercoaster at scale – somebody would be very influential for awhile, then mysteriously disappear and not be working on anything visible for months, and you wouldn’t ever be totally sure what happened.
「無頭銜」的做法也意味著公司內部的人氣變化非常快。因為每個人都有相同的頭銜，你必須透過其他方式來評估影響力，比如「誰現在和這位主管關係特別好」或「誰在主導這個看起來很重要的產品計畫」，而不是「這個人是某某的副總裁」。結果就是一種大規模的英雄與混蛋過山車——某人會在一段時間內非常有影響力，然後神秘消失，幾個月內沒有任何可見的工作，而你永遠不會完全確定發生了什麼。
5. Bat-signals 5. 蝙蝠信號
Another thing I can trace back to Peter is the idea of talent bat-signals. Having started my own company now (in stealth for the moment), I appreciate this a lot more: recruiting good people is hard, and you need a differentiated source of talent. If you’re just competing against Facebook/Google for the same set of Stanford CS grads every year, you’re going to lose. That means you need a set of talent that is (a) interested in joining you in particular, over other companies (b) a way of reaching them at scale. Palantir had several differentiated sources of recruiting alpha.
我可以追溯到彼得的另一個想法是人才信號。現在我已經開始了自己的公司（目前處於隱秘階段），我對這一點的理解更加深刻：招聘優秀人才是困難的，你需要一個差異化的人才來源。如果你每年只是與 Facebook/Google 競爭同一批斯坦福計算機科學畢業生，你將會失敗。這意味著你需要一組（a）對加入你特別感興趣的人，而不是其他公司，（b）一種能夠大規模接觸他們的方法。Palantir 擁有幾個差異化的招聘優勢。
First, there were all the people who were pro defense/intelligence work back when that wasn’t fashionable, which selected for, e.g., smart engineers from the Midwest or red states more than usual, and also plenty of smart ex-army, ex-CIA/NSA types who wanted to serve the USA but also saw the appeal in working for a Silicon Valley company. My first day at the company, I was at my team’s internal onboarding with another guy, who looked a bit older than me. I asked him what he’d done before Palantir. With a deadpan expression, he looked me in the eye and said “I worked at the agency for 15 years”. I was then introduced to my first lead, who was a former SWAT cop in Ohio (!) and an Army vet.
首先，當時所有支持國防/情報工作的人的情況並不流行，這使得來自中西部或紅州的聰明工程師比平常更受青睞，還有許多聰明的前軍人、前 CIA/NSA 人員，他們希望為美國服務，同時也看到了在矽谷公司工作的吸引力。在我進入公司的第一天，我和另一位看起來比我年長的同事參加了團隊的內部入職培訓。我問他在 Palantir 之前做過什麼。他面無表情地直視我的眼睛，說：“我在那個機構工作了 15 年。”然後我被介紹給我的第一位主管，他是一位來自俄亥俄州的前 SWAT 警察和退伍軍人。
There were lots of these people, many extremely talented, and they mostly weren’t joining Google. Palantir was the only real ‘beacon’ for these types, and the company was loud about supporting the military, being patriotic, and so on, when that was deeply unfashionable. That set up a highly effective, unique bat-signal. (Now there’s Anduril, and a plethora of defence and manufacturing startups). [6]
有很多這樣的人，其中許多人極其有才華，但他們大多數並沒有加入谷歌。Palantir 是這類人才唯一真正的「燈塔」，而該公司在支持軍事、表現愛國等方面非常高調，儘管那時這些觀點是非常不流行的。這設置了一個非常有效且獨特的蝙蝠信號。（現在有 Anduril，以及眾多防務和製造業的初創公司）。[6]
Second, you had to be weird to want to join the company, at least after the initial hype wave died down (and especially during the Trump years, when the company was a pariah). Partly this was the aggressive ‘mission focus’ type branding back when this was uncommon, but also the company was loud about the fact that people worked long hours, were paid lower than market, and had to travel a lot. Meanwhile, we were being kicked out of Silicon Valley job fairs for working with the government. All of this selected for a certain type of person: somebody who can think for themselves, and doesn’t over-index on a bad news story.
其次，想要加入這家公司，你必須有些怪異，至少在最初的熱潮過後（尤其是在特朗普執政期間，這家公司被視為社會邊緣人）。這部分是因為當時這種積極的「使命導向」品牌形象並不常見，但公司也大聲宣揚員工長時間工作、薪水低於市場水平以及需要頻繁出差的事實。與此同時，我們因為與政府合作而被驅逐出矽谷的招聘會。所有這些都篩選出了一種特定類型的人：能夠獨立思考，不會過度關注壞消息的人。
6. Morality 6. 道德
The morality question is a fascinating one. The company is unabashedly pro-West, a stance I mostly agree with – a world more CCP-aligned or Russia-aligned seems like a bad one to me, and that’s the choice that’s on the table. [7] It’s easy to critique free countries when you live in one, harder when you’ve experienced the alternative (as I have - I spent a few childhood years in a repressive country). So I had no problem with the company helping the military, even when I disagreed with some of the things the military was doing.
道德問題是一個引人入勝的話題。這家公司毫不掩飾地支持西方，我大致上同意這一立場——一個更傾向於中共或俄羅斯的世界在我看來似乎是個壞選擇，而這正是我們面臨的選擇。[7] 當你生活在自由國家時，批評自由國家很容易；而當你經歷過替代選擇時（就像我一樣——我在一個壓迫性的國家度過了幾年童年），這就變得困難了。因此，即使我不同意軍方的一些做法，我也對公司協助軍方沒有任何問題。
But doesn’t the military sometimes do bad things? Of course - I was opposed to the Iraq war. This gets to the crux of the matter: working at the company was neither 100% morally good — because sometimes we’d be helping agencies that had goals I’d disagree with — nor 100% bad: the government does a lot of good things, and helping them do it more efficiently by providing software that doesn’t suck is a noble thing. One way of clarifying this is to break down the company’s work into three buckets – these categories aren’t perfect, but bear with me:
但軍方有時候不會做壞事嗎？當然會——我反對伊拉克戰爭。這正是問題的核心：在這家公司工作既不是百分之百的道德良善——因為有時我們會幫助一些我不同意其目標的機構——也不是百分之百的壞事：政府做了很多好事，幫助他們更有效率地完成這些工作，提供不糟糕的軟體是一件高尚的事情。澄清這一點的一種方法是將公司的工作分為三個類別——這些類別並不完美，但請耐心聽我說：
- Morally neutral. Normal corporate work, e.g. FedEx, CVS, finance companies, tech companies, and so on. Some people might have a problem with it, but on the whole people feel fine about these things.
道德中立。正常的企業工作，例如 FedEx、CVS、金融公司、科技公司等等。有些人可能會對此有問題，但總體而言，人們對這些事情感到還好。 - Unambiguously good. For example, anti-pandemic response with the CDC; anti-child pornography work with NCMEC; and so on. Most people would agree these are good things to work on.
毫無疑問是好的。例如，與疾病控制與預防中心（CDC）合作的抗疫情應對；與全國失蹤與剝削兒童中心（NCMEC）合作的反兒童色情工作；等等。大多數人會同意這些是值得投入的好事。 - Grey areas. By this I mean I mean ‘involve morally thorny, difficult decisions’: examples include health insurance, immigration enforcement, oil companies, the military, spy agencies, police/crime, and so on.
灰色地帶。我的意思是「涉及道德上棘手、困難的決策」：例如健康保險、移民執法、石油公司、軍事、間諜機構、警察/犯罪等。
Every engineer faces a choice: you can work on things like Google search or the Facebook news feed, all of which seem like marginally good things and basically fall into category 1. You can also go work on category 2 things like GiveDirectly or OpenPhilanthropy or whatever.
每位工程師都面臨一個選擇：你可以從事像 Google 搜尋或 Facebook 新聞動態這類的工作，這些看起來都是邊際上不錯的事情，基本上屬於類別 1。你也可以去從事類別 2 的工作，比如 GiveDirectly 或 OpenPhilanthropy 等等。
The critical case against Palantir seemed to be something like “you shouldn’t work on category 3 things, because sometimes this involves making morally bad decisions”. An example was immigration enforcement during 2016-2020, aspects of which many people were uncomfortable with.
對 Palantir 的關鍵指控似乎是「你不應該從事第三類事務，因為這有時涉及做出道德上不好的決策」。一個例子是 2016 年至 2020 年的移民執法，許多人對此感到不安。
But it seems to me that ignoring category 3 entirely, and just disengaging with it, is also an abdication of responsibility. Institutions in category 3 need to exist. The USA is defended by people with guns. The police have to enforce the law, and - in my experience - even people who are morally uncomfortable with some aspects of policing are quick to call the police if their own home has been robbed. Oil companies have to provide energy. Health insurers have to make difficult decisions all the time. Yes, there are unsavory aspects to all of these things. But do we just disengage from all of these institutions entirely, and let them sort themselves out?
但在我看來，完全忽視第三類機構並與之脫離關係也是一種責任的放棄。第三類機構需要存在。美國是由持槍的人來保衛的。警方必須執法，而根據我的經驗，即使對某些執法方面感到道德不安的人，在自己家被搶劫時也會迅速報警。石油公司必須提供能源。健康保險公司必須不斷做出艱難的決策。是的，這些事情都有不光彩的方面。但我們是否就完全與這些機構脫離關係，讓它們自行解決問題呢？
I don’t believe there is a clear answer to whether you should work with category 3 customers; it’s a case by case thing. Palantir’s answer to this is something like “we will work with most category 3 organizations, unless they’re clearly bad, and we’ll trust the democratic process to get them trending in a good direction over time”. Thus:
我不認為是否應該與第三類客戶合作有明確的答案；這是一個根據具體情況而定的問題。Palantir 對此的回答大致是「我們會與大多數第三類組織合作，除非它們明顯不良，我們會相信民主過程能隨著時間推進它們朝著良好的方向發展」。因此：
- On the ICE question, they disengaged from ERO (Enforcement and Removal Operations) during the Trump era, while continuing to work with HSI (Homeland Security Investigations).
在 ICE 問題上，他們在特朗普時期與 ERO（執法和驅逐行動）脫離了關係，但仍繼續與 HSI（國土安全調查局）合作。 - They did work with most other category 3 organizations, on the argument that they’re mostly doing good in the world, even though it’s easy to point to bad things they did as well.
他們確實與大多數其他類別 3 的組織合作，理由是這些組織大多在世界上做著好事，儘管也很容易指出他們所做的一些壞事。
- I can’t speak to specific details here, but Palantir software is partly responsible for stopping multiple terror attacks. I believe this fact alone vindicates this stance.
我無法具體說明細節，但 Palantir 軟體在阻止多起恐怖攻擊方面發揮了部分作用。我相信這一事實本身就證明了這一立場的正當性。
- I can’t speak to specific details here, but Palantir software is partly responsible for stopping multiple terror attacks. I believe this fact alone vindicates this stance.
This is an uncomfortable stance for many, precisely because you’re not guaranteed to be doing 100% good at all times. You’re at the mercy of history, in some ways, and you’re betting that (a) more good is being done than bad (b) being in the room is better than not.
這對許多人來說是一個不舒服的立場，正因為你並不保證始終做到百分之百的好。從某種程度上來說，你受制於歷史，而你在賭注的是 (a) 所做的好事多於壞事 (b) 在場比不在場要好。
This was good enough for me. Others preferred to go elsewhere.
這對我來說已經足夠了。其他人則更喜歡去別的地方。
The danger of this stance, of course, is that it becomes a fully general argument for doing whatever the power structure wants. You are just amplifying existing processes. This is where the ‘case by case’ comes in: there’s no general answer, you have to be specific. For my own part, I spent most of my time there working on healthcare and bio stuff, and I feel good about my contributions. I’m betting the people who stopped the terror attacks feel good about theirs, too. Or the people who distributed medicines during the pandemic.
這種立場的危險，當然在於它可能成為一個完全普遍的論點，用來支持權力結構所想要的任何事情。你只是在放大現有的過程。這就是「逐案處理」的意義：沒有普遍的答案，你必須具體。就我個人而言，我在那裡大部分時間都在從事醫療和生物相關的工作，我對自己的貢獻感到滿意。我相信那些阻止恐怖襲擊的人也對他們的貢獻感到自豪。還有在疫情期間分發藥物的人。
Even though the tide has shifted and working on these ‘thorny’ areas is now trendy, these remain relevant questions for technologists. AI is a good example – many people are uncomfortable with some of the consequences of deploying AI. Maybe AI gets used for hacking; maybe deepfakes make the world worse in all these ways; maybe it causes job losses. But there are also major benefits to AI (Dario Amodei articulates some of these well in a recent essay).
儘管潮流已經改變，專注於這些「棘手」領域現在變得時尚，但這些問題對於技術專家來說仍然是相關的。人工智慧就是一個很好的例子——許多人對於部署人工智慧的一些後果感到不安。也許人工智慧會被用來進行駭客攻擊；也許深偽技術會以各種方式使世界變得更糟；也許它會導致失業。但人工智慧也帶來了重大好處（達里奧·阿莫代伊在一篇最近的文章中很好地闡述了這些）。
As with Palantir, working on AI probably isn’t 100% morally good, nor is it 100% evil. Not engaging with it – or calling for a pause/stop, which is a fantasy – is unlikely to be the best stance. Even if you don’t work at OpenAI or Anthropic, if you’re someone who could plausibly work in AI-related issues, you probably want to do so in some way. There are easy cases: build evals, work on alignment, work on societal resilience. But my claim here is that the grey area is worth engaging in too: work on government AI policy. Deploy AI into areas like healthcare. Sure, it’ll be difficult. Plunge in. [8]
與 Palantir 類似，從事 AI 工作可能並不完全道德良善，也不完全邪惡。不參與其中——或呼籲暫停/停止，這是一種幻想——不太可能是最佳立場。即使你不在 OpenAI 或 Anthropic 工作，如果你是一個有可能從事 AI 相關問題的人，你可能還是希望以某種方式參與其中。有些情況很簡單：建立評估，致力於對齊，增強社會韌性。但我在這裡的主張是，灰色地帶也值得參與：從事政府 AI 政策的工作。將 AI 應用於醫療等領域。當然，這會很困難。全力以赴。[8]
When I think about the most influential people in AI today, they are almost all people in the room - whether at an AI lab, in government, or at an influential think tank. I’d rather be one of those than one of the pontificators. Sure, it’ll involve difficult decisions. But it’s better to be in the room when things happen, even if you later have to leave and sound the alarm.
當我想到今天在人工智慧領域中最具影響力的人時，他們幾乎都是在場的人——無論是在人工智慧實驗室、政府，還是有影響力的智庫。我寧願成為其中一員，而不是那些高談闊論的人。當然，這會涉及艱難的決策。但當事情發生時，身處其中總是更好，即使你之後必須離開並發出警報。
7. What next? 7. 接下來怎麼辦？
Am I bullish on the company still? The big productivity gains of this AI cycle are going to come when AI starts providing leverage to the large companies and businesses of this era - in industries like manufacturing, defense, logistics, healthcare and more. Palantir has spent a decade working with these companies. AI agents will eventually drive many core business workflows, and these agents will rely on read/write access to critical business data. Spending a decade integrating enterprise data is the critical foundation for deploying AI to the enterprise. The opportunity is massive.
我對這家公司仍然看好嗎？這一波人工智慧的生產力提升將在人工智慧開始為這個時代的大公司和企業提供槓桿時實現——在製造、國防、物流、醫療等行業。Palantir 已經花了十年時間與這些公司合作。人工智慧代理最終將驅動許多核心業務工作流程，而這些代理將依賴於對關鍵業務數據的讀取/寫入訪問。花十年時間整合企業數據是部署人工智慧到企業的關鍵基礎。這個機會是巨大的。
So yes, I’m bullish. 所以是的，我對此持樂觀態度。
As for me, I’m carrying out my long-awaited master plan and starting a company next. Yes, there will be a government component to it. The team is great, and yes we’re hiring. We even talk about Wittgenstein sometimes.
對我來說，我正在實施我期待已久的宏偉計劃，接下來要創辦一家公司。是的，這裡會有政府的成分。團隊很棒，沒錯，我們正在招聘。我們甚至有時會談論維根斯坦。
Thanks to Rohit Krishnan, Tyler Cowen, Samir Unni, Sebastian Caliri, Mark Bissell, and Vipul Shekhawat for their feedback on this post.
感謝 Rohit Krishnan、Tyler Cowen、Samir Unni、Sebastian Caliri、Mark Bissell 和 Vipul Shekhawat 對這篇文章的反饋。
--
[1] Both OpenAI and Palantir required backing by rich people with deep belief and willingness to fund them for years without any apparent or obvious breakthroughs (Elon/YC Research, and Peter Thiel, respectively). Palantir floundered for years, barely getting any real traction in the gov space, and doing the opposite of the ‘lean startup’ thing; OpenAI spent several years being outdone (at least, hype-wise) by DeepMind before language models came along. As Sam Altman pointed out:
[1] OpenAI 和 Palantir 都需要富有且深信不疑的人支持，願意在沒有任何明顯突破的情況下資助他們多年（分別是 Elon/YC Research 和 Peter Thiel）。Palantir 在政府領域掙扎了多年，幾乎沒有真正的進展，與「精實創業」的理念背道而馳；而 OpenAI 在語言模型出現之前，花了幾年時間被 DeepMind 超越（至少在炒作上）。正如 Sam Altman 所指出的：
“OpenAI went against all of the YC advice,” Altman told Stripe cofounder and fellow billionaire John Collison
“OpenAI 違背了所有 YC 的建議，”阿特曼告訴 Stripe 的共同創辦人及億萬富翁約翰·科利森
He rattled off the ways: “It took us four and half years to launch a product. We’re going to be the most capital-intensive startup in Silicon Valley history. We were building a technology without any idea of who our customers were going to be or what they were going to use it for.”
他快速列舉了幾種方式：“我們花了四年半的時間才推出一款產品。我們將成為矽谷歷史上資本密集度最高的創業公司。我們在建立一項技術時，完全不知道我們的客戶會是誰，或者他們會用它來做什麼。”
On Saturday, Altman tweeted: "chatgpt has no social features or built-in sharing, you have to sign up before you can use it, no inherent viral loop, etc. seriously questioning the years of advice i gave to startups."
在星期六，Altman 推特： "chatgpt 沒有社交功能或內建分享，你必須註冊才能使用，沒有固有的病毒循環等等。我真的在質疑我多年來給初創公司的建議。”
There’s something to this correlation: by making the company about something other than making money (civil liberties; AI god) you attract true believers from the start, who in turn create the highly generative intellectual culture that persists once you eventually find success.
這種關聯性是有道理的：通過讓公司專注於其他目標而非賺錢（如公民自由；人工智慧的神），你從一開始就吸引了真正的信徒，他們反過來會創造出高度生產性的知識文化，這種文化在你最終獲得成功後仍然會持續存在。
It’s hard to replicate, though - you need a visionary billionaire and an overlooked sector of the economy. AI/ML was not hot in 2015; govtech was not hot in 2003.
這很難複製——你需要一位有遠見的億萬富翁和一個被忽視的經濟領域。2015 年，AI/ML 並不熱門；2003 年，政府科技也不熱門。
[2] Ted Mabrey’s essay on the FDE model here is good: https://tedmabrey.substack.com/p/sorry-that-isnt-an-fde
[2] Ted Mabrey 關於 FDE 模型的文章很好：https://tedmabrey.substack.com/p/sorry-that-isnt-an-fde
[3] Sarah Constantin – also an ex-Palantirian - goes into greater detail on this point in her great essay: https://sarahconstantin.substack.com/p/the-great-data-integration-schlep
[3] Sarah Constantin – 也是前 Palantir 員工 – 在她的精彩文章中對這一點進行了更深入的探討：https://sarahconstantin.substack.com/p/the-great-data-integration-schlep
[4] One side note: the company was often cast as a ‘data company’ in the press, or worse, a ‘data mining’ company or similar. As far as I can tell, this was a simple misunderstanding on the press’s part. Palantir does data integration for companies, but the data is owned by the companies – not Palantir. “Mining” data usually means using somebody else’s data for your own profits, or selling it. Palantir doesn’t do that - customer data stays with the customer.
[4] 一個附註：該公司在媒體上經常被描繪為「數據公司」，甚至更糟，被稱為「數據挖掘」公司或類似的稱號。據我所知，這是媒體的一個簡單誤解。Palantir 為公司提供數據整合服務，但數據的所有權屬於這些公司，而非 Palantir。「挖掘」數據通常意味著使用他人的數據來獲取自己的利潤或出售它。Palantir 不這樣做——客戶的數據仍然屬於客戶。
[5] As Byrne Hobart notes in his deeply perceptive piece about the company, “Cult” is just a euphemism for “ability to pay below-market salaries and get above-average worker retention.” This is also fair – the company paid lower than market salaries, and it was common to stick around for 5+ years. That said, most early employees did very well, thanks to the performance of the stock. But it was not obvious that we would do well; most of us had mentally written off the value of our equity, especially during the toughest years. I vividly remember there was one of those ‘explaining the value of your equity’ pamphlets that showed the value of the equity if the company was valued at $100bn, and a group of us laughing about the hubris of that. The company is, as of writing, at $97.4 billion.
[5] 正如 Byrne Hobart 在他對該公司的深刻見解文章中所指出的，「Cult」只不過是「能夠支付低於市場工資並獲得高於平均的員工留任率」的委婉說法。 這也是公平的——該公司的薪資低於市場水平，員工通常會在公司待上 5 年以上。話雖如此，大多數早期員工因為股票表現而獲得了不錯的回報。但我們並不明顯知道自己會做得好；我們中的大多數人已經在心理上放棄了自己股權的價值，尤其是在最艱難的年份。我清楚地記得有一本「解釋你股權價值」的小冊子，裡面展示了如果公司估值達到 1000 億美元，股權的價值會是多少，我們一群人對此的自負大笑不已。根據目前的寫作時點，該公司的估值為 974 億美元。
[6] By the way, the company wasn’t some edgelord right-wing anti-woke haven, even back then. Yes, there were people on all ends on the ideological spectrum, but by an large I remember the vast majority of my colleagues being normie centrists.
順便提一下，當時這家公司並不是某個極端右派的反覺醒避風港。是的，意識形態光譜的各個端點都有一些人，但我記得絕大多數同事都是普通的中間派。
[7] Most activist types are, in my view, deluded about the degree to which we do actually need a strong military. I wonder how many of them revised their views after Russia’s invasion of Ukraine - (and indeed, Palantir played a critical role in Ukraine’s response). Drones alone are a frightening new development in international affairs which most people have not sufficiently updated on.
[7] 在我看來，大多數激進派對於我們實際上需要強大軍事的程度存在錯誤的認知。我想知道在俄羅斯入侵烏克蘭後，有多少人修正了自己的觀點——（事實上，Palantir 在烏克蘭的應對中發揮了關鍵作用）。無人機 單獨 是國際事務中一個令人恐懼的新發展，大多數人對此尚未充分更新認識。
[8] Paul Christiano is a good example of this on the AI safety side - he went into government and now leads the US AI safety center.
[8] 保羅·克里斯蒂亞諾 是人工智慧安全方面的一個好例子 - 他進入政府，現在領導美國人工智慧安全中心。