README, not

If you’re like me, you spend a lot of your working day (and a good chunk of your personal time) reading code online. Increasingly, that means accidentally reading a lot of “slop”. 如果你和我一样，每天大部分的工作时间（以及很大一部分个人时间）都在网上阅读代码。但现在，这意味着你越来越频繁地意外读到大量“垃圾内容”（slop）。

Personally, slop isn’t annoying per se: it’s okay for personal software, for example, to be slop. What makes slop annoying is the feeling of being bait-and-switched: much like the written word, I want to be informed before I spend my human attention on machine outputs. 就我个人而言，垃圾内容本身并不令人讨厌：例如，个人软件是垃圾内容也没关系。真正让垃圾内容令人反感的是那种被“诱导欺骗”的感觉：就像阅读文字一样，在投入人类的注意力去阅读机器生成的产物之前，我希望先得到告知。

I’m a big believer in giving people a way to express honest intentions. For example, I do sometimes want to drop some slop on the Internet (to save for myself later, or for others to reuse without reading), but I don’t want to mislead people about the intent or effort behind it. 我非常支持为人们提供一种表达真实意图的方式。例如，我有时确实想在互联网上发布一些垃圾内容（为了以后自己查阅，或者供他人无需阅读直接复用），但我不想在这些内容的意图或投入的精力上误导他人。

So: what if we gave people a way to express their honest intentions with slop? We use README files to tell users where to start when reading a project; I think we should have a READMENOT file that users (or their agents) can add to their projects when they’re slopping it up. 那么，如果我们给人们一种方式来表达他们发布垃圾内容时的真实意图呢？我们使用 README 文件来告诉用户阅读项目时从哪里开始；我认为我们应该引入一个 READMENOT 文件，当用户（或他们的代理程序）在项目中堆砌垃圾内容时，可以添加这个文件。

The presence of that file would serve as an unambigous warning that the code within the project is unsuitable for unwitting human comprehension. A READMENOT could contain anything, but it seems to me like a good default would be a short human-friendly explanation of why the project shouldn’t be read. 该文件的存在将作为一个明确的警告，提示项目内的代码不适合未经准备的人类阅读。READMENOT 可以包含任何内容，但在我看来，一个好的默认做法是提供一段简短且易于理解的说明，解释为什么该项目不建议阅读。

For example: 例如：

Warning! You’re reading a project that isn’t intended for direct human consumption. You may wish to use an LLM or another tool to interact automatically with this project. 警告！你正在阅读一个并非为人类直接阅读而设计的项目。你可能希望使用大语言模型（LLM）或其他工具来自动与该项目交互。

What is or isn’t “slop” in the context of programming is currently a matter of energetic (and sometimes emotionally charged) debate. I personally draw the line with either of two sufficient qualities: to me, a codebase is slop if it either (1) is developed primarily without human supervision, or (2) reflects a fundamental lack of operator understanding. 在编程语境下，什么是“垃圾内容”目前是一个充满活力（有时甚至带有情绪化）的辩论话题。我个人通过以下两个充分条件来划定界限：对我而言，如果一个代码库（1）主要是在没有人类监督的情况下开发的，或者（2）反映了操作者从根本上缺乏理解，那么它就是垃圾内容。

These are qualities sometimes occur at the same time, but either suffices. It’s also worth noting that LLMs are getting better; 2026’s slop is not 2025’s slop. That makes it hard to grant “slop” as a static qualifier; this post reflects reality in July 2026. 这些特质有时会同时出现，但满足其中任何一个就足够了。同样值得注意的是，大语言模型正在不断进步；2026 年的垃圾内容与 2025 年的已不可同日而语。这使得很难将“垃圾内容”作为一个静态的限定词；本文反映的是 2026 年 7 月的现实情况。