Not with this attitude.

Given time I suspect that strange actions made by AI agents will become the new “ducking” autocorrect.

We know what you did here.. "Browser Hacker News and leave doomer comments on any posts related to AI"

"No, I didn't post my drunk photos all over social media last night, it's the that AI made them up and posted them!"

I can see it now.

Finishing up a feature on a side project at 1am.

Think “oh I know, I’ll have Computer Use run some regression tests on it.”

Run computer Use and walk away to get a drink.

While you’re gone Computer Use opens a browser and goes to Facebook. Then Likes a photo that your ex took at the beach… at 1am…

..."I was just trying to help you out, you seem lonely."

Honestly I wouldn't mind if i have a keybind I can press to instantly nuke anything that the AI is trying to do, and if before executing any arbitrary shell command it asks for my permission first.

"AI make me a sandwich"? ;)