DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux | Lex Fridman Podcast #501

章节 1：智能代理重塑编程的新时代

📝 本节摘要：

过去九个月，人工智能从代码补全和聊天助手迅速演进为能调用工具、自我检查、协同规划并承担大部分编码工作的智能代理。程序员角色也由亲自编写代码转向提出问题、审查结果与信任实现路径。网页应用已可近乎全程由AI完成，但大型代码库、底层系统和安全关键项目仍需人工把关与架构治理。

[原文] [Lex Fridman]: There are decades where nothing happens and weeks where decades happen. And we have seen decades of progress happen in the last nine months. If you're not recognizing the gravity of the moment, that's the delusion. That's the psychosis. Who would not get delirious if suddenly a genie pops out of the bottle and says, "You can have whatever you want. Every feature you've ever dreamed of in an operating system, I can deliver them to you, most of them in five minutes, a few in 20, and if we really go hog wild, it's gonna take me two hours." I want the fastest car breaking the speed limits. I want the diver watch that can go down the Mariana Trench. I want the operating system that can install in less than 60 seconds. One of the things I always loved about race cars was when I would stumble out of the car absolutely smashed and barely able to

[译文] [Lex Fridman]: 有些几十年什么都没发生，而有些几周却发生了几十年的变化。过去九个月里，我们见证了几十年才会发生的进步。如果你没有意识到这一刻有多重要，那就是一种妄想，一种精神错乱。突然有一天，一个精灵从瓶子里跳出来对你说：“你想要什么都可以。你梦想中的操作系统功能，我都能交付给你，大多数五分钟就够，少数需要二十分钟，如果我们真的放开手脚，最多也就两个小时。”我想要一辆能突破限速的最快汽车。我想要一块能潜入马里亚纳海沟的潜水表。我想要一个安装时间不到六十秒的操作系统。我一直很喜欢赛车的一点，就是有时我会从车里踉踉跄跄地爬出来，整个人筋疲力尽，几乎连头都抬不起来，

[原文] [Lex Fridman]: hold my head up, and I'd lay down on the garage floor and just think, "Holy fuck, I'm alive." The Overton window does not open itself. It opens one nudge at a time by people risking a little. A little reputation, a little pushback, a little criticism, or maybe sometimes a lot of reputation or a lot of criticism or a lot of pushback. Creating life with another human that you love is literally the peak experience of being on the planet.

[译文] [Lex Fridman]: 只能躺在车库地板上，心想：“我靠，我还活着。”奥弗顿窗口不会自己打开。它是由人们一次次冒一点风险，慢慢推开一小点的：牺牲一点声誉，承受一点反弹，接受一点批评；有时，也可能要付出很多声誉，面对很多批评和强烈反弹。和你所爱的人共同创造一个生命，确实是人在这个星球上能够拥有的巅峰体验。

[原文] [Lex Fridman]: The following is a conversation with David Heinemeier Hansson, also known as DHH, creator of Ruby on Rails, CTO of 37signals, creator of the new Omarchy Linux operating system, best-selling author, race car driver, and one of the most outspoken and prolific programmers in the world. For the past 20-plus years, for him, programming meant meticulously handcrafting beautiful Ruby code. But recently, since late 2025, DHH has embraced the AI revolution and has

[译文] [Lex Fridman]: 下面是我与 David Heinemeier Hansson 的对话，他也被称为 DHH，是 Ruby on Rails 的创始人、37signals 的首席技术官、新款 Omarchy Linux 操作系统的创始人、畅销书作者、赛车手，也是世界上最敢于发声、最高产的程序员之一。二十多年来，对他而言，编程意味着精心手工打造优美的 Ruby 代码。但最近，从 2025 年底开始，DHH 接受了 AI 革命，并且已经

[原文] [Lex Fridman]: openly transformed himself into, once again, one of the most outspoken and prolific agentic engineers, even though he hates that term. So let's say practitioners of whatever programming is becoming where AI is doing most of the actual programming and the human steers the ship with high-level design, vision, and taste. Yes, it's true. DHH is at times controversial, but he's always fearless, brilliant, and fun to talk to. This was, once again, an intense, eye-opening, and wild rollercoaster ride of a conversation. This is a Lex Fridman podcast. To support it, please check out our sponsors in the description where you can also find links to contact me, ask questions, give feedback, and so on. And now, dear friends, here's DHH.

[译文] [Lex Fridman]: 公开地再次把自己转变成最敢于发声、最高产的代理式工程师之一，尽管他讨厌这个称呼。或者说，成为正在形成中的新型编程方式的实践者：AI 负责大部分实际编程，而人类凭借高层次的设计、愿景和品味掌舵。没错，事实就是如此。DHH 有时颇具争议，但他始终无所畏惧、才华横溢，而且非常适合交谈。这又是一场激烈、令人豁然开朗、如同疯狂过山车般的对话。这里是《Lex Fridman 播客》。如果你想支持节目，请查看简介中的赞助商，那里也有联系我、提问、反馈等链接。现在，亲爱的朋友们，让我们请出 DHH。

[原文] [Lex Fridman]: 13 months ago, we sat down right here to talk about programming, and then everything changed. At that time, you were a bit skeptical about the role of AI in the process of programming, and then we went through this rapid evolution of agentic engineering. So, simple question to start: How has your view on AI's role in programming changed? Are you excited? Are you terrified? Are you on an emotional rollercoaster ride?

[译文] [Lex Fridman]: 十三个月前，我们就在这里坐下来聊编程，之后一切都变了。那时，你对 AI 在编程过程中的作用还有些怀疑，后来我们经历了代理式工程的快速演进。所以先从一个简单的问题开始：你对 AI 在编程中的作用有什么变化？你是感到兴奋，还是恐惧？你是否正在经历一场情绪过山车？

[原文] [David Heinemeier Hansson]: I am incredibly excited.

[译文] [David Heinemeier Hansson]: 我无比兴奋。

[原文] [Lex Fridman]: You've been talking about fun a lot.

[译文] [Lex Fridman]: 你一直在反复谈论“乐趣”。

[原文] [David Heinemeier Hansson]: There is none of the existential threat. That doesn't exist for me as an emotional component. It is only there as an intellectual component, and the emotional component for me is 100% pure, unadulterated joy—and optimism and amazement that we've made computers do this. And I find it so interesting that we talked just 13 months ago because it's like we talked in different universes, different eras.

[译文] [David Heinemeier Hansson]: 我完全没有那种生存威胁感。对我来说，它并不是一种情绪上的成分，只存在于理性层面。而在情绪上，对我而言百分之百是纯粹、不掺杂任何杂质的喜悦，还有乐观，以及对我们竟然让计算机做到这一点的惊叹。我觉得很有意思的是，我们明明才在十三个月前聊过，却感觉像是在不同的宇宙、不同的时代里交谈。

[原文] [David Heinemeier Hansson]: I love this quote. I think it's Lenin. There are decades where nothing happens and weeks where decades happen. And we have seen decades of progress happen in the last nine months. I mean, imagine you're there when the Wright brothers take flight. Yesterday, the New York Times would write, "It's gonna be 10,000 years before we fly," and then the day after, we're up in the skies, and just a few years after that, there were transatlantic planes. The whole world has completely changed. What a blessing to be there in that moment. If you zoom out and look at all of human history, how many humans got to live within the same epoch that they were born in? They never saw that complete change of the world and of society. And to have been blessed with two of those feels just such a privilege. I got to see the internet—

[译文] [David Heinemeier Hansson]: 我很喜欢这句话。我想这是列宁说的：有些几十年什么都没发生，而有些几周却发生了几十年的变化。过去九个月里，我们见证了几十年才会发生的进步。想象一下，如果你正好在莱特兄弟首次飞行时身处现场。就在前一天，《纽约时报》还会写道：“人类要再过一万年才能飞上天。”可第二天，我们就已经飞到了天空中；仅仅几年后，跨大西洋航班就出现了。整个世界彻底改变了。能够置身那个时刻，是多么大的福气。把视野拉远，看看整个人类历史，有多少人出生和生活在同一个时代，直到去世都没有见证世界和社会的彻底改变？而我有幸经历了两次这样的变化，实在是一种莫大的 privilege。我亲眼见证了互联网——

[原文] [David Heinemeier Hansson]: —from pre-internet to post-internet, and then now pre-AI, post-AI. What an amazing run. How fortunate.

[译文] [David Heinemeier Hansson]: ——从互联网诞生前走到互联网之后，如今又从 AI 诞生前走到 AI 之后。这是一段多么惊人的历程，我们何其幸运。

[原文] [Lex Fridman]: Yeah, but the thing is, this feels like a thing that happened faster than anything else in human history.

[译文] [Lex Fridman]: 是的，但问题在于，这感觉像是人类历史上发生得最快的一件事。

[原文] [David Heinemeier Hansson]: Correct.

[译文] [David Heinemeier Hansson]: 没错。

[原文] [Lex Fridman]: So I would say somewhere around December, maybe late November—

[译文] [Lex Fridman]: 所以我会说，大概是在十二月，或者十一月底——

[原文] [David Heinemeier Hansson]: November 24.

[译文] [David Heinemeier Hansson]: 11 月 24 日。

[原文] [Lex Fridman]: This is...

[译文] [Lex Fridman]: 就是这一天……

[原文] [David Heinemeier Hansson]: That's the exact moment.

[译文] [David Heinemeier Hansson]: 那就是确切的时刻。

[原文] [Lex Fridman]: You know, when they talk about when one nation invades another in a world war or something like this, it—this is how we talk about AI changing everything.

[译文] [Lex Fridman]: 你知道，当人们谈到世界大战中一个国家入侵另一个国家，或者类似的事情时，通常会这样描述——而我们谈论 AI 改变一切时，也是这种感觉。

[原文] [David Heinemeier Hansson]: Yeah, the... And it really just shifted for many great developers. It shifted to where AI is doing some basic autocomplete, maybe writing 5, 10, 15, 20% of code to writing 80% of code.

[译文] [David Heinemeier Hansson]: 是的……它真的让许多优秀开发者的工作方式发生了转变。从 AI 只做一些基础的自动补全，可能编写 5%、10%、15% 或 20% 的代码，转变为编写 80% 的代码。

[原文] [Lex Fridman]: Or 100.

[译文] [Lex Fridman]: 或者 100%。

[原文] [David Heinemeier Hansson]: Or 100, yeah, especially if it's not a public-facing product.

[译文] [David Heinemeier Hansson]: 对，或者 100%，尤其是当它不是面向公众的产品时。

[原文] [David Heinemeier Hansson]: To me, that's why even when I look back upon our conversation a year ago, I don't actually have different opinions. I have the same opinions. A year ago, I did not like the mode of AI we were offered. It was the autocomplete mode, or it was the AI chatbot mode. Now, the chatbot I actually liked, as we talked about. Great tutor right from the get-go, great way of looking things up on the internet, not what was gonna replace me chiseling code. But then we get the agents.

[译文] [David Heinemeier Hansson]: 对我来说，这也是为什么即使回头看我们一年前的对话，我其实并没有改变观点。我还是同样的看法。一年前，我不喜欢当时提供给我们的 AI 形态：要么是自动补全模式，要么是 AI 聊天机器人模式。正如我们当时谈到的，我其实喜欢聊天机器人。它从一开始就是很棒的导师，也是很好的互联网信息检索工具，但我没把它看作会取代我一点点雕琢代码的东西。可是后来，代理出现了。

[原文] [David Heinemeier Hansson]: And the agents start out being curiosities for about five minutes, and then they get amazing, and then they get, "Oh my God, is this AGI?" And all of that happened since last year, even just within the last nine months. We have basically these few phases here. We have AI in the pre-agentic era. I was excited about that, but it was not fundamentally rewriting the rules of the game for me. It was not completely changing how I worked. I was still chiseling code; I just had a little helper, a little sidekick who could bounce ideas off, and I could look up this information online and so forth in a more efficient way. It was just a more efficient way to do what I was already doing, and it didn't change the emotional connection I had to the computer. Then we get to November 24th, 2025. Opus 4.5, to

[译文] [David Heinemeier Hansson]: 一开始，这些代理大概只能让人新鲜五分钟，然后它们变得惊艳，再然后你会想：“我的天，这是通用人工智能（AGI）吗？”这一切都发生在去年之后，甚至就在最近九个月里。基本上，我们经历了几个阶段。首先是代理时代之前的AI。我当时对此很兴奋，但它并没有从根本上改写我的游戏规则，也没有彻底改变我的工作方式。我仍然是在一点点“雕刻”代码，只是多了一个小帮手、一个小副手，可以和我交流想法；我也能更高效地在线查找信息，等等。它只是让我更高效地完成原本就在做的事情，并没有改变我与电脑之间的情感联系。然后到了2025年11月24日。对我来说，Opus 4.5是分界线，

[原文] [David Heinemeier Hansson]: me, was the dividing line, where suddenly... I didn't even try it on the 24th. I think I tried it on the 26th. I gave it a couple of tasks, and I realized that the quality of the output was uncannily close to what I would've written. And I remember just leaning back and thinking, "What just happened?" "How did we go from this autocomplete mess that I was talking to you about in the summer to this just a few short months later? How did we get both the increase in intelligence and then also the increase in usability?" This agent harness question where I don't know if Opus 4.5 was that much smarter than Opus 4, which is what we had in the summer, but its ability to instrument your computer, to use tools, to check its own work, to apply its intelligence in such a way that you could

[译文] [David Heinemeier Hansson]: 对我来说，它就是那条分界线，突然之间……我甚至没有在24日尝试它。我想我是在26日试的。我给了它几个任务，发现它输出的质量竟然和我自己会写出来的东西非常接近。我记得自己靠在椅背上，心想：“刚才发生了什么？”“我们怎么会在短短几个月里，从我夏天跟你说的那种混乱的自动补全，走到今天这一步？我们怎么同时得到了更高的智能和更强的可用性？”这里还有一个代理框架的问题：我不知道Opus 4.5是否真的比我们夏天使用的Opus 4聪明那么多，但它操控电脑、使用工具、检查自身工作，并以一种让你能够

[原文] [David Heinemeier Hansson]: get real meaningful work out of it, was completely different. And I think this is then the big change that happens for almost anyone who paid attention and started playing with it over the Christmas break. This is what I heard from Shopify and other places with lots of employees who suddenly had a breather, suddenly had a couple of weeks to lean back and just look at what was going on, gave it a try, and had the same mind-blowing experience that these agents were of a different genre than what we had before. And then what we get to is, I'm already excited at this point. Like, by December, I'm already revisiting sort of all my priors and going like, "Wow, if it can do this, can it also do that?"

[译文] [David Heinemeier Hansson]: 从中获得真正有意义的工作成果的能力，完全不同了。我认为，对于那些关注进展、并在圣诞假期开始尝试它的人来说，这就是一次巨大的变化。我听Shopify以及其他拥有大量员工的公司说，大家突然有了喘息的机会，突然有几周时间可以放松下来，看看究竟发生了什么。于是他们试了一下，也经历了同样令人震撼的体验：这些代理与我们之前拥有的东西，根本属于不同的类型。到了这一步，我已经非常兴奋了。比如到了12月，我已经开始重新审视自己原先的所有判断，心想：“哇，如果它能做到这个，那它也能做到那个吗？”

[原文] [David Heinemeier Hansson]: "Oh, yeah, it can." And again, Opus 4.5 now looks like a retarded model. And this is the magic of this progress, is you think you've reached something... I remember thinking at the moment, "If this is the last model we get, I'll be set. I'll be happy." I could live with Opus 4.5 for the next 20 years, and you would hear no complaints from me because it was just so incredibly amazing to see an agent do all this work in the way that I wanted it done. Because it was not just about it being able to solve a task, it was also that I could look at its path there and go, "Yep. Yep. Maybe not there, but almost," and it would give you two notes. You'll get to where I wanted to go in the way I wanted to go there. You could produce code I

[译文] [David Heinemeier Hansson]: “哦，没错，它能做到。”而现在，Opus 4.5看起来又像一个弱智模型了。这就是这种进步的神奇之处：你以为自己已经抵达了某个终点……我记得当时还想：“如果我们最后只能得到这个模型，那我也满足了。我会很开心。”接下来20年我都可以使用Opus 4.5，而且你不会听到我有任何抱怨，因为看到一个代理按照我希望的方式完成所有这些工作，实在太不可思议了。因为关键不只是它能解决一个任务，还在于我可以观察它采取的路径，然后说：“对，对。也许这里不太对，但已经很接近了。”我只要给它两条意见，它就能以我希望的方式，抵达我想去的地方。你可以让它生成我

[原文] [David Heinemeier Hansson]: wanted to merge. You could produce code that actually looked beautiful if it was written in Ruby. Rust, different question. But this ability for the agents to truly become an extension of how I wanted to work was very novel. But then we waited just until early spring, and suddenly we got sub-agents. We got harnesses that could subdivide the task, and something that would take Opus quite a while suddenly took a fifth of the time, a tenth of the time, because it could get chopped up and suddenly you got eight sub-agents working for you.

[译文] [David Heinemeier Hansson]: 想要合并的代码。如果是用Ruby写的，你甚至可以生成真正漂亮的代码。Rust嘛，那是另一回事。但代理真正成为我理想工作方式的延伸，这种能力非常新颖。可我们等到初春，突然就有了子代理。我们有了能够把任务拆分开的框架，原本需要Opus花相当长时间完成的事情，突然只需要五分之一，甚至十分之一的时间，因为任务可以被拆开，转眼间就有八个子代理在为你工作。

[原文] [David Heinemeier Hansson]: But both of those two first phases of the agentic age, to me, still felt like I had to drive. I could tell it what I wanted, where to look for it, and steer it a little bit when it went off, and then we'd get there, and I would go much faster. But I had to be in the driver's seat. I had to tell it what I wanted, and I had to be the reviewer, the auditor of what was coming out. And then finally now, this summer, with Opus 5, Fable, and Sol, GPT Sol, and to a lesser extent, some of the open-weight models, we've arrived at a new era where I'm not telling it where we're going. I'm telling it the problem I have. I'm telling it the fuzzy, vague idea I have. It tells me where we're going.

[译文] [David Heinemeier Hansson]: 但在我看来，代理时代最初的这两个阶段，仍然像是必须由我来驾驶。我可以告诉它我想要什么、去哪里寻找；当它偏离方向时，再稍微把它引回来。这样我们最终还是能到达目的地，而且速度快得多。但我必须坐在驾驶位上。我得告诉它我想要什么，还得担任产出结果的审阅者和审计员。然后终于到了今年夏天，随着Opus 5、Fable、Sol，也就是GPT Sol，以及少数开源权重模型的出现，我们进入了一个新时代：我不再告诉它我们要去哪里。我只告诉它我遇到的问题，告诉它一个模糊、笼统的想法。它来告诉我我们要去哪里。

[原文] [David Heinemeier Hansson]: It tells me which path to take, and I will still look at it because I'm a curious person and I like computers and I like the outcome of it, but I really kind of don't have to. I've become optional in the part that produces the code, that picks the route. Remember when early GPS systems came out? They were amazing compared to looking at a map, but you still wanted to pay attention. Is it going to drive you into the harbor? I remember these newspaper articles. "Ah, GPS, they're terrible because people don't pay attention, they drive into the harbor." When was the last time GPS drove anyone into the harbor? Like, that just doesn't happen anymore. In fact, the cars now just drive themselves, right? And this is where we've arrived at now, that I can trust.

[译文] [David Heinemeier Hansson]: 它会告诉我该走哪条路。我还是会看着它，因为我是个好奇的人，我喜欢电脑，也喜欢最终的成果；但实际上，我已经不必这么做了。在生成代码、选择路线这部分，我已经变得可有可无。还记得早期GPS系统刚出现的时候吗？和看地图相比，它们简直神奇，但你仍然会想认真盯着：它会不会把你开进港口？我记得当时有些报纸文章说：“哎，GPS太糟糕了，因为人们不注意路况，结果开进了港口。”上一次GPS把人开进港口是什么时候？这种事现在根本不会发生了。事实上，现在汽车都能自己驾驶了，对吧？而我们现在已经走到了这一步：我可以信任它。

[原文] [David Heinemeier Hansson]: For the domain I'm working in right now, I can trust it, and feel completely confident that it's gonna have my back. It's not gonna do something stupid, and if it does something stupid, it's gonna be able to recover.

[译文] [David Heinemeier Hansson]: 就我目前工作的这个领域而言，我可以信任它，完全相信它会支持我。它不会做蠢事；即使真的做了，也有能力恢复过来。

[原文] [Lex Fridman]: Mm-hmm. We should mention that there's all kinds of domains that programmers operate in.

[译文] [Lex Fridman]: 嗯。我们应该说明，程序员涉足的领域非常多。

[原文] [David Heinemeier Hansson]: Correct.

[译文] [David Heinemeier Hansson]: 没错。

[原文] [David Heinemeier Hansson]: There's... I actually don't know, but I think the most common domain in development is, like, web dev CRUD. You have a database. You have a user-facing interface, and it does something back and forth, and it could be internal to just one person, to multiple people, to a small number of people, or to the world. And I think for that, it's—I mean, you could really get to 100% of code written by AI, and really, if you're a good programmer and you have a good intuition about what's happening behind the scenes,

[译文] [David Heinemeier Hansson]: 有些领域……其实我也不确定，但我觉得开发中最常见的领域，应该是网页开发里的 CRUD。你有一个数据库，也有一个面向用户的界面，二者来回交互。它可以只供一个人内部使用，也可以供多人、少数人，甚至全世界使用。我觉得在这种领域里，代码完全由 AI 编写是确实有可能的；只要你是一个优秀的程序员，对幕后发生的事情有很好的直觉，

[原文] [David Heinemeier Hansson]: you can legitimately not look at the code. At least that's been my experience. I don't know how much prior experience with programming you need to have to kind of know that things are working correctly behind the scenes, intuitively, by observing the ripple effects, the symptoms of the system. But I think for that domain, it's close to 100%. And then there's domains like you're also operating in, which is writing a Linux distribution. There, maybe there's more because you're obsessed with speed and all that kind of stuff. There, maybe you need to look at the code a little bit more. Then there's maybe safety-critical systems, all the way down to operating a nuclear power plant or a self-driving car. Maybe you need to look at the code more carefully.

[译文] [David Heinemeier Hansson]: 你完全可以不看代码，至少我的经验是这样。我不知道你需要多少编程经验，才能仅凭观察系统的连锁反应和表现出来的症状，凭直觉判断幕后运行是否正确。但我觉得在这个领域里，AI 生成代码的比例已经接近 100%。当然，也有你所处的其他领域，比如编写 Linux 发行版。在那里，可能需要更多人工介入，因为你非常在意速度之类的问题，所以也许得多看一些代码。再往下，还有安全关键系统，比如运行核电站或自动驾驶汽车。那种情况下，可能就需要更仔细地检查代码。

[原文] [David Heinemeier Hansson]: Yes, but that said, AI is insanely capable at both finding and fixing security vulnerabilities. This was the whole blowup about Fable. This model was so capable of finding holes that an attacker could exploit that it was simply not safe to release. So the irony here is that when you look at that field, it seems like we've reached levels of intelligence that virtually no human can match, because many of these security holes are about stringing combo moves together. You find one little vulnerability here that by itself might not be the worst thing in the world, but then you combine it with four others, and suddenly you have RCE, remote command execution.

[译文] [David Heinemeier Hansson]: 是的，但话虽如此，AI 在发现和修复安全漏洞方面都强得惊人。这正是 Fable 事件引发巨大风波的原因。这个模型发现漏洞的能力太强了，攻击者可以利用这些漏洞，所以发布它根本不安全。讽刺的是，在这个领域，我们似乎已经达到几乎没有人类能够匹敌的智能水平，因为很多安全漏洞都在于把一连串组合招式串起来。你先找到一个小漏洞，单独看可能不算什么大问题，但再和另外四个漏洞结合起来，突然就能实现 RCE，也就是远程命令执行。

[原文] [David Heinemeier Hansson]: Humans who are able to do that are very rare. They usually work inside state-sponsored organizations or other clandestine operations. They're not just out and about finding things. So they've gotten so good at that. The Linux distribution is why I've gotten 100% AI-pilled. Because I've been working on Omarchy for the last three months, this version that just dropped a few days ago called Quatro. And almost right from the beginning I'm working on that version, the agent acceleration neared 100%, and in the last two months it has been 100%. I have not written—

[译文] [David Heinemeier Hansson]: 能够做到这一点的人类非常少。他们通常在国家支持的组织或其他秘密行动中工作，不会只是四处寻找漏洞。所以 AI 在这方面已经强到了这种程度。Linux 发行版正是让我彻底拥抱 AI 的原因。过去三个月，我一直在开发 Omarchy，最近几天刚发布了一个叫 Quatro 的版本。几乎从我开始开发这个版本起，代理的加速程度就接近 100%，而在过去两个月里已经达到了 100%。我没有亲手编写——

[原文] [Lex Fridman]: Really?

[译文] [Lex Fridman]: 真的吗？

[原文] [David Heinemeier Hansson]: —any of the code that's shipped in Quatro by hand. I've reviewed the shape of all of it. I've reviewed the individual lines of anything that's critical in the model layer of the system, and I've not looked at a bunch of the UI code. I have not looked at a bunch of the auxiliary code, and I've not written any of the new functionality entirely by hand. But then the web part, actually evolving Basecamp and Hey, our professional products that have lots of users and are relatively large code bases, have proven surprisingly tricky to fully accelerate with agents. We just released Basecamp 5 not too long ago. That was the first product at 37signals that was really agent-accelerated.

[译文] [David Heinemeier Hansson]: ——Quatro 中发布的任何代码，都不是我亲手写的。我审查过全部代码的整体结构，也逐行审查过系统模型层中所有关键部分；但有相当一部分 UI 代码我没有看，也有很多辅助代码我没有看，而且任何新功能都不是完全由我亲手写成的。不过在网页开发方面，实际演进 Basecamp 和 Hey——这两款拥有大量用户、代码库也相对庞大的专业产品——事实证明，要让代理完全加速开发，出乎意料地困难。我们不久前刚发布了 Basecamp 5，那是 37signals 第一款真正由代理加速开发的产品。

[原文] [David Heinemeier Hansson]: Because we were in this final sprint phase from around February. By then, agents were already good. And we had this early surge of, "It's solved. We can just have the designers do the programming. They know what features they want. They know what shape they want it to take. Let's just let them vibe." And we let them vibe. And we ended up with a lot of PRs that individually perhaps could have been justified for a hot moment, taken all together, destroyed the architecture of the system. And we actually had to clean up manually, mop it up by hand, by human hand, to get back to an architecture that felt cohesive and coherent. So we still have a bit of that—that was February, by the way. Things are quite different now.

[译文] [David Heinemeier Hansson]: 因为从二月左右开始，我们进入了最后冲刺阶段。那时代理已经相当不错了。我们一度兴奋地认为：“问题解决了。让设计师来编程就行。他们知道自己想要什么功能，也知道希望功能呈现什么样子。就让他们随心所欲地做吧。”于是我们真的让他们这样做了。结果产生了大量 PR，单独来看，也许一时还能说得过去，但合在一起就摧毁了系统架构。最后，我们不得不手动清理，由人亲自收拾残局，重新恢复一个感觉统一、连贯的架构。所以这方面的问题还留了一点——顺便说，那已经是二月的事了。现在情况已经大不相同。

[原文] [Lex Fridman]: Well, hold on a second. What's the lesson from that? Is one of the lessons from that that you have to be a programmer at this stage to be able to vibe code?

[译文] [Lex Fridman]: 等一下。这里面的教训是什么？其中一个教训是不是说，在现阶段，你必须是程序员，才能进行氛围编程？

[原文] [David Heinemeier Hansson]: To be able to vibe code on existing, substantial code bases—

[译文] [David Heinemeier Hansson]: 如果要在已有的大型代码库上进行氛围编程——

[原文] [Lex Fridman]: I see.

[译文] [Lex Fridman]: 明白了。

[原文] [David Heinemeier Hansson]: —even if they're CRUD, if you wanna retain the element of architecture that got that system to where it was. Now, that's also a point where I've stressed many times that when people accuse vibe coders of being slob generators, I go right back at them and say, "Have you looked at the average programmer's output?" That is some other slob too. If you've looked at behind the scenes of many great companies and what their code bases look like after there's been 3,000 humans through them, they're awful. Absolutely awful.

[译文] [David Heinemeier Hansson]: ——哪怕那些代码只是 CRUD，如果你想保留让系统发展到今天这种程度的架构要素，那也需要具备编程能力。不过我也反复强调过一点：当人们指责氛围编程者只会制造垃圾代码时，我总会反问他们：“你们看过普通程序员的产出吗？”那同样是一团乱。如果你看过许多优秀公司幕后代码库的样子，尤其是经历过 3,000 个人修改之后的代码库，就会发现它们糟糕透顶，真的糟糕透顶。