周末花时间用几个 AI 工具跑了一些真实任务——总结 PDF、起草邮件，以及清理混乱的 CSV。令人惊讶的是，一些更小众的工具在速度上竟然超过了那些大牌模型。不过准确性仍然很大程度上取决于提示词的结构。很好奇大家都在用什么样的配置。
Original
Spent the weekend running a few AI tools through real tasks—summarizing PDFs, drafting emails, and cleaning messy CSVs. Surprisingly, some smaller niche tools beat the big-name models in speed. Accuracy still swings a lot depending on prompt structure though. Curious what setups others are using.