Clankers Made Me Build a Second Brain
Clankers Made Me Build a Second Brain
那些“铁皮人”逼我构建了“第二大脑”
AI was the last straw that convinced me I need a PKM system. AI 是压垮骆驼的最后一根稻草，让我确信自己需要一套个人知识管理（PKM）系统。
Prompt-In, Slop-Out
输入提示词，输出垃圾
I’m a developer, I know a little about a lot, and I tinker with lots of niche things for fun. Personally, I’m the stereotypical AI skeptic, hate how the technology works, and I refuse to vibe code. 我是一名开发者，对很多领域都有所涉猎，也喜欢折腾各种小众玩意儿。就个人而言，我是典型的 AI 怀疑论者，讨厌这种技术的工作原理，也拒绝那种“凭感觉写代码”（vibe code）的方式。
That being said, I am not a saint, and my work laptop has forced Claude down my gullet anyway (yay, “free” tokens!). Sometimes, when I’m deep in the flow, I don’t want to interrupt myself to quickly Google something “I should already know by heart”, especially if I can help getting it over with already. 话虽如此，我并非圣人，我的工作电脑还是强行把 Claude 塞给了我（耶，“免费”的 Token！）。有时，当我沉浸在工作流中时，我不想为了快速 Google 一个“我本该烂熟于心”的东西而打断自己，尤其是如果我能尽快搞定它的话。
When I have to bodge up an ad hoc CI shell script, I sometimes commit the cardinal sin of asking Claude real quick how that Bash-ism went, and if it was 2>&1 or 2&>1 – you know the type.
当我不得不临时拼凑一个 CI Shell 脚本时，有时会犯下大忌：快速问 Claude 那个 Bash 语法是怎么写的，是 2>&1 还是 2&>1 —— 你懂的，就是那种问题。
Now, granted, most of the time it does its job, I close the chat, and don’t give it a second thought. Other times it ruffles my ever-loving feathers, because in its attempt to please me rather than help me, it says stuff like: 当然，大多数时候它能完成任务，我关掉对话框，也就不会再多想。但有时它会让我火冒三丈，因为它试图取悦我而非真正帮助我，它会说出这样的话：
In order to do feature X, simply pass the —do-X flag to the end of the command. “为了实现功能 X，只需在命令末尾加上 —do-X 参数即可。”
I open up the man page to double-check, of course it’s not there, so I retaliate, and it goes: 我打开 man 手册核对，当然根本没这个参数，于是我反驳它，它又说：
You’re absolutely right, my bad, let me look up a different source. “你说得对，是我的错，让我查查其他来源。”
Other times, it saves me the trouble of telling it that it’s stupid by having a mid-context window aneurysm and gaslighting itself with responses like: 还有些时候，它甚至不需要我指出它很蠢，自己就会在上下文窗口中“脑溢血”，并用类似这样的回复进行自我欺骗：
In order to do feature X, simply pass the —do-X flag to the end of the command… No wait, hold on… actually that flag isn’t real, I actually meant to say (…) “为了实现功能 X，只需在命令末尾加上 —do-X 参数……等等，稍等……其实那个参数不存在，我其实是想说……”
As I frustratingly rolled my eyes at the magic terminal window disobeying my intent and making fun of its inadequacy, I realized that – although with a heavy dose of both amusement and lowered expectations – I have nonetheless succumbed to the mundane temptation of asking a clanker to “tell me what I need to know, make no mistakes”… Shame on me! 当我对着那个违背我意图、又因其无能而显得滑稽的魔法终端窗口翻白眼时，我意识到——尽管带着一种既好笑又降低预期的心态——我还是屈服于那种平庸的诱惑，去要求一个“铁皮人”告诉我“我需要知道什么，且不能出错”……真丢人！
Don’t Delegate Understanding
不要外包你的理解力
The title of this section is a reference to Steph Ango’s post with the same name. It more poetically describes the dangers of these “convenience tools” infesting and exploiting you to the benefit of others. Although the metaphor applies to many things even before its time, AI is probably the one that fits the best. 本节标题引用了 Steph Ango 同名文章。它更具诗意地描述了这些“便利工具”如何侵蚀并利用你来为他人谋利。虽然这个比喻在 AI 出现之前就适用于许多事物，但 AI 恐怕是最贴切的一个。
The article ends with the following wisdom: 文章结尾有这样一句箴言：
To inoculate yourself don’t delegate understanding. If you build your own understanding you will be the one who earns the dividends. “为了自我免疫，不要外包你的理解力。如果你建立了自己的理解，你才是那个获得红利的人。”
He is absolutely right here, and I didn’t pick this by accident. Steph is the CEO of Obsidian, a very popular Markdown editor that many use to write and explore their personal notes collection. It’s also a perfect segue into a solution for my predicament… 他说得完全正确，我引用这段话并非偶然。Steph 是 Obsidian 的 CEO，这是一款非常流行的 Markdown 编辑器，许多人使用它来编写和探索个人笔记库。这也完美地引出了我困境的解决方案……
Just Write Notes Instead
不如直接写笔记
The problems I was trying to solve weren’t new. I wasn’t asking the AI to give me information I never heard of before. I was asking it specifically about something I knew was a thing, that I’ve done before, but it was longer ago, so I wasn’t sure of the specific details. Something that normally would be accomplished by a Google search leading to a Stack Overflow answer, ah… those were the days! Alas, that has since enshittified too (also because of AI). 我试图解决的问题并不新鲜。我并不是在要求 AI 提供我从未听说过的信息。我是在询问一些我知道存在、以前做过，但因为时间久远而记不清具体细节的事情。这本该通过 Google 搜索找到 Stack Overflow 的答案来解决，啊……那真是美好的旧时光！可惜，现在那里也已经“屎化”了（同样是因为 AI）。
If I had, when I first learned about that feature, also taken the time to write a short note on it, then months later I would have had the exact answer I wanted, in the form that I expected to find it, along with the explanation that made it click for me, and might not even have to look it up because, in the process of taking said note, I would’ve needed to use my brain to synthesize the information, which in turn would’ve bolstered my understanding of it. 如果我在第一次学习那个功能时，花点时间写下一条简短的笔记，那么几个月后，我就能得到我想要的准确答案，而且是以我期望的方式呈现，并附带让我豁然开朗的解释。甚至我可能根本不需要去查，因为在做笔记的过程中，我必须动脑去整合信息，这反过来又会加深我的理解。
Sure, one could make the argument that some knowledge gets stale, especially in the domain of programming. I’d retort that’s just inevitable, even if you had a photographic memory and perfectly recalled everything you ever learned without writing it down, you’d still be wrong eventually. Although it’s mildly annoying, finding out my answer is an outdated hallucination and then amending it is a preferable alternative to me than starting from scratch every time with just good vibes, a search bar, and an internet full of slop. Not to mention that you can take notes about things you can’t just look up: your thoughts, opinions, ideas, memories. 当然，有人可能会争辩说，有些知识会过时，尤其是在编程领域。我会反驳说这是不可避免的，即使你有照相式记忆，能完美回忆起你学过的所有东西而不写下来，你最终还是会出错。虽然有点烦人，但发现我的答案是过时的幻觉并进行修正，对我来说，比每次都只靠“感觉”、搜索栏和充满垃圾信息的互联网从零开始要好得多。更不用说你还可以记录那些无法通过搜索获得的东西：你的想法、观点、灵感和记忆。
If there is a system that is able to contain such wealth of knowledge while keeping it useful, it’s Obsidian. 如果有一个系统能够容纳如此丰富的知识并保持其可用性，那一定是 Obsidian。
Hey Jarvis, Tell Me What I Think Today
嘿 Jarvis，告诉我我今天在想什么
The first time I heard of Obsidian was probably around the early pandemic. It looked like a shiny new toy and I wanted to see how the best people use it, but never took it upon myself to actually give it a go because starting out was so daunting, considering just how many competing schools of thought around organizing your vault existed even then. 我第一次听说 Obsidian 大概是在疫情初期。它看起来像个闪亮的新玩具，我想看看高手们是怎么用的，但从未下定决心去尝试，因为入门实在太令人望而生畏了，考虑到当时关于如何组织知识库（Vault）的流派就已经多得数不清了。
It was a wonderful rabbit hole in which I watched hour long videos by various creators, comparing and contrasting their opinions on the best workflow or structure, showing you tutorials, etc., even if in the end I watched them more for entertainment value. The main criticism I had with those was that it seemed to me like all the staunch “second brain” advocates made less of a useful note system, and more of a middle-management style productivity theater, bolstering pretty dashboards with fancy graphs and arbitrary metrics. 那是一个美妙的“兔子洞”，我看了各种创作者长达一小时的视频，比较他们关于最佳工作流或结构的观点，看他们的教程等等，尽管最后我更多是把它们当作娱乐来看。我对这些内容的主要批评是，在我看来，所有那些坚定的“第二大脑”倡导者，与其说是在构建一个有用的笔记系统，不如说是在搞一种中层管理风格的“生产力表演”，用花哨的图表和随意的指标来装饰漂亮的仪表盘。
Of course, with time, it only got worse… Because the notes are Markdown files, the same file extension as a SKILL.md, vibe coders quickly put two and two together and concluded the next big revolution in managing your second brain is to feed that one slop as well. 当然，随着时间的推移，情况变得更糟了……因为笔记是 Markdown 文件，和 SKILL.md 的扩展名一样，“凭感觉写代码”的人很快就把两者联系起来，断定管理第二大脑的下一场大革命就是把垃圾信息也喂给它。
I won’t mention any names (if you’ve been around, you know who they are), but I do want to share with you Eric Morrison’s video on second brain cringelords, because it is tragically funny, and shows the antithesis of Steph’s post: delegating as much understanding as possible. It’s gotten to the point where it’s genuinely fair to summarize their workflow as satire: 我不会点名（如果你在圈子里混过，你就知道他们是谁），但我确实想分享 Eric Morrison 关于“第二大脑尴尬症患者”的视频，因为它既悲剧又滑稽，展示了与 Steph 文章截然相反的一面：尽可能地外包理解力。现在的情况已经到了可以公正地将他们的工作流总结为讽刺的地步：
Hey Claude, please… “嘿 Claude，请……”