Amazon Web Services – Four Years and Out

Amazon Web Services – Four Years and Out

亚马逊云科技 (AWS) —— 四年离职感言

Today marks four years since I joined AWS. My last day will be Friday. I have to say being fired from AWS is actually a relief. There have been a lot of changes to the company since I joined in 2022, and the company I wanted to work for is no longer the same company. 今天是我加入 AWS 四周年的日子，而本周五将是我在这里的最后一天。不得不说，被 AWS 解雇反而让我感到一种解脱。自 2022 年我入职以来，公司发生了巨大的变化，我当初想要效力的那家公司已不复存在。

This past year, while I was doing my best to make AWS play nice in open source communities, there were two main drivers making me unhappy with my job: organizational change and the acceleration of the focus on Generative AI. The organizational change came in the form of the man who hired me, David Nalley. I was skeptical about joining AWS, especially since I work in open source, but David convinced me that his team, OSSM (Open Source Strategy and Marketing), was dedicated to making AWS a better citizen in open source communities. 过去一年里，我一直在尽力让 AWS 在开源社区中表现得更友好，但有两个主要因素让我对这份工作感到不满：组织架构的变动以及对生成式 AI 的过度聚焦。组织变动始于聘用我的主管 David Nalley。起初我对加入 AWS 持怀疑态度，尤其是考虑到我深耕开源领域，但 David 说服了我，他表示他的团队——开源战略与营销部 (OSSM)——致力于让 AWS 成为开源社区中更好的参与者。

Amazon has a really odd viewpoint when it comes to the people who work there. They view almost all employees as “fungible”. Now the first time I had ever heard the term “fungible” was in reference to non-fungible tokens (NFTs), but it basically means “replaceable”. Amazon built a huge retail business on processes that could take someone who was relatively healthy and relatively intelligent, and turn them in to a productive fulfillment center employee in a couple of weeks. 亚马逊在对待员工方面有着非常奇特的观点。他们几乎将所有员工视为“可替代的”(fungible)。我第一次听到“可替代”这个词是在非同质化代币 (NFT) 的语境下，但它在这里的基本含义就是“可替换”。亚马逊建立庞大的零售业务所依赖的流程，能够让一个相对健康且聪明的人在几周内就变成一名高效的物流中心员工。

While that may work for a shipping business, it doesn’t translate all that well to information technology, since so much of being successful in that business relies on institutional knowledge that must be earned over time. It also assumes that there is a limitless supply of people with the required skills, and a willingness to work for Amazon. In any case, during the interview process David called me “non-fungible” (which still sounds dirty in my mind but did make me proud) and I got the job. 虽然这套逻辑在物流行业行得通，但在信息技术领域却并不适用，因为该行业的成功很大程度上依赖于需要长期积累的机构知识。这种模式还假设拥有所需技能且愿意为亚马逊工作的人才储备是无限的。无论如何，在面试过程中，David 称我为“不可替代的”（虽然这个词在我听来依然有些怪异，但确实让我感到自豪），随后我得到了这份工作。

While my official role was to act as a liaison between AWS and customers who were commercial open source companies, I simplified that to mean bring a human face to a huge, faceless corporation. David was a very good manager. In fact, he is in the running to be the best manager I’ve ever had, although that title still belongs to a man named Jay Clapsadle (who is long since retired). He has an innate understanding of how AWS works, and he would always nudge me into those situations where my unique but limited talents would be put to good use. 虽然我的正式职责是担任 AWS 与商业开源公司客户之间的联络人，但我将其简化为：为这家庞大且冷冰冰的企业带来一丝人性化的温度。David 是一位非常出色的经理。事实上，他是我职业生涯中最好的经理人选之一，尽管这个头衔目前仍属于早已退休的 Jay Clapsadle。他天生了解 AWS 的运作方式，总是能引导我将自己独特但有限的才能发挥在最合适的地方。

Well, last year David, being very good at his job, got promoted to run the entire AWS Developer Experience organization. OSSM is a part of it, but I no longer interacted with him in a meaningful way. My “David Time” went almost to zero. Also, last year the focus at AWS turned fully and almost desperately toward GenAI. This post is already too long so I won’t pull out all of the examples I was going to bring up at this point in the narrative, but we started being driven to use as much AI as possible. 然而去年，由于工作表现出色，David 被提拔去管理整个 AWS 开发者体验部门。OSSM 虽然隶属于该部门，但我与他之间不再有实质性的互动。我与他共事的时间几乎降为零。此外，去年 AWS 的重心完全且近乎绝望地转向了生成式 AI。这篇文章已经太长了，所以我不会列举我原本打算提到的所有例子，但我们开始被要求尽可能多地使用 AI。

People were writing things like “I use AI to summarize my email!”. I mentally responded to that with “why don’t we just write better emails?”. And one that really bothered me was “I used one prompt to create my conference presentation!” In the modern economy, the most valuable commodity is attention. I really appreciate the attention my three readers give to my posts, even when I lose them halfway through. I love giving conference talks and I spend a considerable amount of time creating them, and when someone still wants to speak but doesn’t want to put in the work, it makes me angry. 人们开始写下诸如“我用 AI 来总结邮件！”之类的话。我内心对此的回应是：“为什么我们不能直接把邮件写得更好呢？”还有一件事让我非常困扰，那就是“我用一个提示词就生成了我的会议演示文稿！”在现代经济中，最宝贵的商品是注意力。我非常感谢那三位读者对我的文章给予的关注，即使他们中途可能就离开了。我热爱在会议上演讲，并花费大量时间准备内容，当有人既想演讲又不愿付出努力时，这让我感到愤怒。

Seriously, why do it? It has gotten better, but I used to see AI generated images with lots of unintelligible writing or misspelled words in slides, but the speaker left them in anyway. “Good enough” is not customer obsession. In this whole pivot to GenAI, AWS has lost its focus on the customer. Instead of working backwards from a genuine customer need, the goal seems to be to create as many things as fast as possible, throw them into the world and see which ones gain traction, whether or not they serve a real need. 说真的，为什么要这么做？情况虽然有所好转，但我过去常看到幻灯片里充斥着 AI 生成的、带有乱码或拼写错误的图片，而演讲者却视而不见。“差不多就行”绝不是“客户至上”。在全面转向生成式 AI 的过程中，AWS 已经失去了对客户的关注。目标似乎不再是从真实的客户需求出发进行反向推导，而是尽可能快地制造尽可能多的东西，把它们扔向市场，看看哪些能火，而不去管它们是否真的解决了实际需求。

There is this push to use AI to create content which will ultimately be consumed by AI, and we’ve lost the human being in the process. When AWS first introduced a viable cloud to the world, it was amazing. Back in the 1990s when you wanted to implement an enterprise software solution, you first had to take a guess at what computing power you would need. Next, you would have to order hardware from companies like Sun Microsystems or Dell and that could take weeks if not months to be delivered. 现在有一种趋势是利用 AI 去创造最终会被 AI 消费的内容，在这个过程中，我们丢失了“人”的因素。当 AWS 最初向世界推出可行的云服务时，那是令人惊叹的。回想 20 世纪 90 年代，如果你想实施一套企业软件解决方案，首先得猜测自己需要多少计算能力。接着，你必须向 Sun Microsystems 或 Dell 等公司订购硬件，这可能需要数周甚至数月才能到货。

It would then need to be racked, powered and provisioned, and then you were screwed if you happened to undersize it or criticized if you spent too much and oversized it. The cloud solved those problems, and AWS set the standard with services such as S3, EC2, RDS, etc. Go to re:Invent these days and try to find a session on those tools. Even when you can, AI will still dominate the presentation. This whole thing made me question my role. 然后还需要上架、供电和配置。如果你估算不足，项目就会搞砸；如果你投入过多导致资源过剩，又会受到指责。云服务解决了这些问题，而 AWS 通过 S3、EC2、RDS 等服务树立了行业标准。如今去参加 re:Invent 大会，试着找一场关于这些工具的会议看看吧。即便你能找到，AI 依然会占据演讲的主导地位。这一切让我开始质疑自己的角色。

My personal goal is to make AWS the default choice for running open source workloads, but what does that mean when you can simply “vibe code” the same functionality, bypassing the license? The customer focus at AWS has also changed. Instead of appealing to those people focused on the infrastructure required to build stable and feature-rich applications, it has become abstracted to focus on a level above that, since the whole promise of GenAI is to make those people no longer necessary; to make those people “fungible”. 我的个人目标是让 AWS 成为运行开源工作负载的首选，但当人们可以简单地通过“直觉编程”(vibe code) 来绕过许可协议实现相同功能时，这还有什么意义呢？AWS 对客户的关注点也变了。它不再吸引那些专注于构建稳定且功能丰富的应用程序所需基础设施的人，而是抽象到了更高的层面，因为生成式 AI 的核心承诺就是让这些人变得不再必要，让这些人变得“可替代”。

Last year the achievement I am most proud of involved getting a suspended AWS account reinstated. The financial impact to the company was negligible as this customer wasn’t a huge spender, but they are one of those people that made AWS successful in the first place. A man in northern Africa posted that his decade-old AWS environment had been shut down with little notice and no recourse. In fact, he was told that his data had been deleted. I reached out to him to see if I could help, but I wasn’t optimistic. If his data was gone, it was gone, but I really wanted to capture as much as I c 去年我最引以为傲的成就是帮助一个被封禁的 AWS 账户恢复了使用。这对公司的财务影响微乎其微，因为该客户并非大额付费用户，但他们正是当初让 AWS 走向成功的那类人。一位北非的男士发帖称，他使用了十年的 AWS 环境在几乎没有通知且无法申诉的情况下被关闭了。事实上，他被告知数据已被删除。我联系了他，想看看能否提供帮助，但我并不乐观。如果数据没了就是没了，但我真的很想尽我所能去挽回……