问HN：LLM懒惰开发者
现在我发现，与公司其他开发人员合作变得越来越困难，因为我们似乎都生活在一个充满人工智能的混乱之中。
我在工作中使用大型语言模型（LLMs），认为它们可以很有用，但我仍然会遇到一些连强大的Claude都无法解决的问题，这时我会想：“这个人对这个领域有很多知识，或者可能以前见过类似的错误，也许他们会知道！”
但结果往往不是这样！即使他们曾经知道，花费脑力去记住这些信息所需的精力也太宝贵了，他们宁愿把我的问题直接扔回他们选择的LLM中，然后用那个来回复我。另一个可以接受的回答是问我是否尝试过询问LLM该怎么做，因为也许那样就能解决问题。为什么还要费心呢？不如不断按下虚拟老虎机的生成按钮，直到得到一个有效的答案。
我曾有一个来自其他团队的人主动提出提交代码来修复我们遇到的问题，审查后我惊讶地发现，这几乎是我们自己向LLM询问后尝试并失败的内容的逐字复制。这是在回答“我们是否尝试过使用LLM来解决这个问题”时回答“是”的情况下发生的，但你永远不知道，这个人这次可能问对了方式……
我最喜欢的另一个例子是被问到在我们开始构建之前，是否已经用AI检查过我们的“设计”。这可能是最基础的REST API设计，但如果我们遗漏了什么呢？我们怎么知道？如果我们请求AI的帮助，它还会帮我们吗？
合作在哪里？学习和与他人分享的愿望在哪里？我们对自己所做的事情的信心在哪里？我们到底在这里做什么？
查看原文
More often than not now I find it harder to work with other devs at my company because of the AI slop trough we all seem to live in now.<p>I use LLMs at work and think they can be useful, but I will still hit problems that even the almighty Claude cannot solve and think to myself: “this person has a lot of domain area knowledge here or has probably seen an error like this before, maybe they will know!”<p>Nope! Or even if they did know at one point the amount of calories required to use their brain long enough to remember are too precious to be spent when they can just blast my question right back into an LLM of their choosing and reply with that. The other acceptable answer is asking me if I’ve tried to ask an LLM what to do yet or not because maybe that will solve it. Why even bother anymore? Just keep hitting the generate button on the virtual slot machine until you get something that works.<p>I had someone from another team offer to commit code to fix an issue we were having and upon reviewing it imagine my surprise when it was almost verbatim what we had just tried and failed with from asking the LLM ourselves. This was after answering “yes” to the question of if we had tried using an LLM or not yet to solve the issue as well, but you never know this person might have asked it the right way this time…<p>One of my other favorite ones is being asked if we had checked our “design” with AI before we committed to building it. This was for probably the most textbook and basic REST API designs you could have, but what if we missed something?? How would we know? Would the AI still help us later if we asked it to?<p>Where is the collaboration? Where is the desire to learn and share with others? Where is the confidence in what we do? What the hell are we doing here?