提问 HN:你们都在怎么使用 LLM?

3作者: FailMore10 个月前
我:我不喜欢和大型语言模型(LLM)进行基于事实的对话(例如,我的猫可以吃带骨头的熟鸡肉吗?)。我喜欢和 LLM 进行开放式对话,在这种对话中,大致正确的方向就很有用了(例如,用 ChatGPT 的语音模式练习中文)。 我使用 Claude Code,它有时很有用(因为我对特定函数的记忆力一般),但当我要花大量时间清理应用程序中糟糕的逻辑/流程时,又会有点沮丧。 也很好奇你们是如何与 LLM 交互的,想了解其中的细节。
查看原文
Me: I don’t like to have factual based conversations with LLMs (e.g. can my cat eat cooked chicken on the bone?). I like having open ended conversations with LLMs where being vaguely in the right direction will be useful (e.g. practicing Chinese with ChatGPT voice mode).<p>I use claude code, and verge between it being useful (because my memory of specific functions is average), but then a bit bummed out when I have to spend a bunch of time cleaning up poor logic&#x2F;flows in the application.<p>Curious to know the intricacies of how you are interacting with LLMs too.