这个周末我没有只看演示，而是实际对几个AI工具做了压力测试。很惊讶它们处理混乱提示的方式差异这么大——一个在总结笔记方面做得很准，另一个却完全凭空捏造了来源。好奇大家在真正的工作中都在用什么，而不只是做实验。
Original
Spent the weekend actually stress-testing a few AI tools instead of just watching demos. Surprised how differently they handle messy prompts—one nailed summarizing notes, another completely hallucinated sources. Curious what others are using for real work, not just experiments.