No Sloptober

No Sloptober

No Sloptober 在这个十月,我们向你发起挑战:完全停止使用任何基于大语言模型(LLM)的工具。把这当作一次给大脑的“禁食”!这并非对他人的评判,而是对你个人的挑战。细微差别(Nuance)在互联网上很难(甚至不可能)把握,平衡也是如此。请培养你对大语言模型细微差别的感知力——了解它们擅长什么,又不擅长什么。

Suggested “hard core” 🤘 approach [1] [2]: 建议的“硬核” 🤘 方案 [1] [2]: No AI/LLM tools whatsoever at home or work, do it the hard (core) way. No AI Search summary (turn off here), No Chat Bot chats/search, No Code Review, No Claude/OpenCode/Pi/Agents/Models etc. 无论在家还是工作中,完全不使用任何 AI/LLM 工具,用最“硬核”的方式完成任务。关闭 AI 搜索摘要,不使用聊天机器人进行对话或搜索,不使用 AI 进行代码审查,不使用 Claude/OpenCode/Pi/智能体/模型等。

Onarheim’s Law: “Agents can only maintain or INCREASE entropy in a system. Humans are uniquely capable of decreasing it” 奥纳海姆定律(Onarheim’s Law):“智能体只能维持或增加系统中的熵。唯有人类有能力降低熵。”

Why? 为什么? Even if you find LLMs useful, it is useful to know how to do things yourself and train your skills. Stretch your brain, don’t be a “Meat Proxy” [3]. Do you like reading LLMs PR descriptions? Mental health benefits, and avoiding ai psychosis. Gain perspective by seeing what was good or bad when using LLMs. Learn something new! LLMs are well documented to make learning difficult for humans. Security - agents are known for wild security choices, and installing malware by accident (looking at you Shai Hulud). 即使你觉得大语言模型很有用,亲自动手并锻炼自己的技能依然至关重要。拓展你的大脑,不要成为一个“肉身代理”(Meat Proxy)[3]。你喜欢读大语言模型生成的 PR(拉取请求)描述吗?这样做有益于心理健康,并能避免“AI 精神错乱”。通过回顾使用大语言模型时的优缺点来获得新的视角。去学习新东西吧!已有大量文献证明,大语言模型会阻碍人类的学习。在安全方面,智能体以做出鲁莽的安全决策而闻名,甚至可能意外安装恶意软件(点名批评 Shai Hulud)。

Write a blog post about your experiences in No Sloptober good or bad, I want to know what you learned about yourself! Bragging rights? Honor and glory!? Post #no-sloptober on socials. Screenshot your lack of usage in claude or whatever. In the words of my toddler “I do by MYSELF” 写一篇博客分享你在“No Sloptober”期间的经历,无论好坏,我想知道你对自己有了什么新的认识!炫耀的资本?荣誉与荣耀!?在社交媒体上发布 #no-sloptober 标签。截屏展示你在 Claude 或其他工具上的零使用记录。用我蹒跚学步的孩子的话说:“我自己做!”

What? 要做什么? Review and Clean up existing LLM output in your work/home codebases. They really do like to write A LOT of cruft in their comments. They also love to over-engineer/over-complicate software. Learn a new language! Pick a new language like Zig, Rust, Go, Lua, TypeScript, etc. Learn a new skill, join my Neovim cult! Pick up a new web framework, like Astro (or Mastro), or Effect.ts. Practice code challenges, remember “code katas”? Start a project! Make a game (I recommend Excalibur, or Godot). 审查并清理你工作或个人代码库中现有的 LLM 输出。它们确实喜欢在注释中写大量垃圾信息,也喜欢过度设计或过度复杂化软件。去学习一门新语言!选择像 Zig、Rust、Go、Lua、TypeScript 等语言。学习一项新技能,加入我的 Neovim 教派!尝试一个新的 Web 框架,比如 Astro(或 Mastro)或 Effect.ts。练习代码挑战,还记得“代码卡塔”(code katas)吗?开启一个项目!制作一个游戏(我推荐 Excalibur 或 Godot)。

Convince your Manager: 说服你的经理: “Hey , let’s see if we can do some cost risk analysis around our LLM usage. Let’s experiment with reduced/no LLM usage and measure our teams/orgs velocity, incident rate, and cost and see if there are any potential savings or risk mitigations we could enact in the future.” Or perhaps using the “iron triangle” proposition of software development. You can have things good, cheap, or fast… pick two. “嘿,让我们针对 LLM 的使用情况做一些成本风险分析吧。让我们尝试减少或停止使用 LLM,并衡量团队/组织的开发速度、事故率和成本,看看未来是否能实现潜在的成本节约或风险规避。” 或者,也可以利用软件开发的“铁三角”理论:你可以选择高质量、低成本或高速度……但只能三选二。

Convince yourself: 说服你自己: Find joy again in your craft. Find the flow state again. Understand where your own gaps are in your own knowledge and capability, and are you okay delegating those? Are there truly tasks that have no value? Can they be automated deterministically cheaply and fast? Learning takes mental effort and friction. 重新找回你对手艺的热爱。重新找回心流状态。了解自己在知识和能力上的差距,并思考你是否真的愿意将这些任务外包出去?是否真的存在毫无价值的任务?它们能否通过确定性的方式快速且廉价地自动化?学习需要脑力投入和克服阻力。

Unfortunately, corporations sometimes have an “AI under duress” to keep your job. Do what’s right for your life and family and adjust these suggestions as needed to keep the overlords happy. 不幸的是,企业有时会强迫你使用 AI 以保住工作。请为了你的生活和家庭做出正确的选择,并根据需要调整这些建议,以让你的“老板们”满意。


[1] 语言翻译工具非常有用,能帮助全球用户使用非母语参与交流。我认为这是一个独特的人类例外,也是一种极其支持人类的用途,但我将决定权留给你。然而,我不想贬低专业人工翻译和本地化在产品和服务中的重要性,这些领域需要机器无法替代的准确性。 [2] “肉身代理”(Meat Proxy)是一个术语,指那些通过聊天机器人处理工作对话,却不经过任何思考或最少编辑审查就直接将输出内容转发给你的人。我个人觉得这极其不尊重人,我宁愿对方不回复。