关于软件工程中 AI 现状的一篇牢骚文
A grumpy screed about AI in software engineering

原始链接: https://sam.sutch.net/posts/a-grumpy-ai-screed

这位拥有 20 年经验的资深软件工程师感叹,整个行业已被“人工智能垃圾内容”(AI slop)彻底淹没。从代码合并请求(PR)和技术文档,到领导层的路线图及市场营销,现代工程工作流已完全被 AI 生成的内容所主导。 作者认为,这种转变起初或许还有些新奇,但如今已剥夺了该职业的技艺感与乐趣。尽管失去了成就感,但拒绝使用 AI 在职业生涯中是不可行的;采用 AI 现已成为一种文化和表演性的期望,招聘流程实际上已将那些善用生成式工具的工程师作为筛选标准。 虽然作者希望能找到行业中“无垃圾内容”的角落——例如围绕 Zig 等项目的社区,但他承认这些地方很可能仍将保持小众。归根结底,这篇文章反映了一种深刻的幻灭感,暗示 AI 生成内容的普遍存在已从根本上且不可逆转地削弱了软件工程这一职业。

这篇 Hacker News 帖子围绕着对当前软件工程中 AI 现状的挫败感展开。讨论主要涉及三个核心观点: * **存在性担忧:** 一些用户担心“AI 垃圾内容”阶段标志着软件工程作为一门技艺的衰落,甚至可能使这一职业变得面目全非或走向消亡。 * **技术演进:** 另一些人则认为这仅仅是抽象层面的又一次转变,将其比作历史上从汇编语言向高级编程的过渡。他们认为“代码”的定义只是在演变,而非消失。 * **工匠精神的价值:** 评论者指出,纯手工制作的软件在市场上存在潜在机遇。人们普遍渴望那些能够避开 AI 生成内容所带来的“批量化、企业化”质感的数字产品。有人建议,开发者应积极构建互联网的角落,优先展示手工打造的作品,以此作为对当前自动化趋势的一种反制。 这场交流凸显了行业内深层的分歧:一方对传统技能的丧失感到警惕,而另一方则将 AI 视为软件开发进步过程中自然演进的一步。
相关文章

原文

As a software engineer, my entire professional life has become AI slop.

Every line of every pull request? Slop.
Every word of every pull request description? Slop.
Every code review comment? Slop.
Every technical design document? Slop.
Every ticket description? Slop.
Every graphical design spec? Slop.

I know I’m not alone in this experience. From what I can ascertain by discussing with peers, looking over shoulders at my co-working space, and participating in online forums is that this is nearly every software engineer’s daily life now. In many organizations 100% of their software engineering practice is slop.

Even leadership decisions are informed in large part by slop. Slop marketing, slop emails, slop announcements, slop roadmaps. The roadmaps are at least 50% slop, at this point.

I’ve been a software developer/engineer/entrepreneur for 20 years. For most of that time, it was an actual daily pleasure to do my job. Even if I was having a rough few weeks, I could always be coaxed back into it (nerdsniped?) by working on something I was proud of. Even at the truly soulless larger companies I’ve worked for, there’s always been an interesting problem to solve, somewhere.

Recently, however, the use of AI code generation has become non optional in most of the software engineering world. It’s made this job an absolute slog. For people who enjoyed the craft, there’s little enjoyment to be derived from waiting for a coding bot to push some code for you. It’s novel, sure, but the novelty wears off quickly, especially once you realize there’s no way back.

One suggestion I’ve heard from people is that “if it’s so upsetting, then just don’t use it.” Let’s be realistic. To firmly opt out is a professionally nonviable option at this point, both from a cultural and a performative standpoint. You will stand out in a negative way in most circles. Either by your “attitude” or by your meager (in comparison) pull request/line count.

If you think there is a way to opt out professionally, I encourage you to go interview for any SWE job. Every company asks about your opinion on AI now as one of the first interviews. It's basically shorthand for determining whether you'll be difficult or not. If you want to remain employable, you better answer affirmatively and positively about embracing AI in your daily work.

Another suggestion I’ve heard is to just write my own software, my own company, completely slop free. Unfortunately, some of what is being said about the new "cost of software" development is true. Good luck standing out universe filled with AI slop apps. Even Apple is eagerly releasing slopware now.

Where does this leave us? I don’t have a conclusion. I’d love to find a corner of the world where this hasn’t happened yet, and there are some projects like Zig that have encouraging stances on AI use. I could see some communities forming around those. They won’t make much commercial impact, however. Non-AI projects may have a niche, but it’s likely they will only ever remain exactly that: a niche.

Stay sane.

Comments on Hacker News

联系我们 contact @ memedata.com