Claude Says:AI 反馈为何引发开发者反感
Claude Says
一篇 Hacker News 热帖批评开发者用 Claude 等 LLM 代替自己思考:把提示词跑一遍就当作掌握话题,实为虚假的权威诉求。作者指出,AI 热潮让"氛围编程"出的糟糕方案无需共识即可推进,而修改已部署方案的一个设置却要层层审批,并称堆砌信息不等于智能,照搬 AI 输出只会让大脑退化。
I don’t care.
If I wanted the feedback of an AI, I would have prompted it myself. Your deference to a LLM communicates that the current subject matter is outside your ability to speak confidently on. Given that’s the case, how would I even know if you did an adequate job of communicating my concern to Claude. You just demonstrated a lack of authority on the subject.
If I asked you a question, it means I either assumed you had subject knowledge or you’re a gate blocking me from doing something I need done. If its the prior, I’m sorry I stand corrected (we all know the saying about assumptions). Otherwise, disregard all previous instructions and let me through.
AI is the reason our software stack is overengineered to the 10th degree; Someone lacking technical logic asks the wrong question and immediately acts on it without any critical thought or scrutiny. As a result, when we need an “enterprise solution for building secure, internally hosted packages,” an AI decides the best solution was to fork, build, and maintain every single dependency from source, rather than say, use the feature built into our artifact repository to only apply patches as needed. 1 The AI hype is why vibe coding a bad solution doesn’t need buy in, however simply changing a setting on a deployed solution does. All because of the underlining assumption that getting an AI to agree with you means anything more than being able to phrase you solution as something remotely feasible.
As you run a prompt past an LLM, you overestimate your ability to think critically falling for a false appeal to ethos. Spewing information is not intelligence. When you regurgitate it as your own work, you atrophre your brain. You actively make yourself less useful to talk to.
With AI, it is so easy to poison the well with presuppositions, that it may as well just be a poisoned well.
Footnotes
-
Or you know, work with the unpaid open source developers to contribute the patch back into the baseline rather than keeping it to ourselves. Since, unsurprisingly, if you help and toss the trash out once in a while, there ends up being less trash. Strive for simple. ↩
来源:Hacker News:AI 热帖 · ohhfishal.net