个人助理新基准：能不能让我去睡觉

Nic Vandewetering @nvandewetering原推发布于 2026.10.27

在 X 上打开原推

ChatGPT Dots, Grok Bot and Claude Duet need a new benchmark: Can I go to bed? I want to give an agent a task, close my laptop, and wake up to finished work. Instead, I’ve had mornings where I’m figuring out why the overnight session never ran. I’d still rather use Dots than ration every request to Grok Bot. But I shouldn’t have to babysit an agent to make sure it started. Which one have you actually trusted to finish something while you were asleep?