[anze3db]

I personally wouldn't trust it with anything that involves submitting forms. What if it gets any data wrong? Even watching the demo video made me feel anxious!

I guess I'm finally on the other side of this Douglas Adams quote:

> Anything that's invented between when you're fifteen and thirty-five is new and exciting and revolutionary and you can probably get a career in it. Anything invented after you're thirty-five is against the natural order of things.

I do think it's a cool project though!

[quickthrower2]

Yeah RE: Douglas Adams, it is like the early days of internet, where people might not want to enter their credit card details.

I think services like this might mean the web needs a way to say "this form is reversible, and so OK for AI to drive it". Example might be unsubscribing from an email, or even ordering something if an order can be cancelled within 24 hours. Maybe the API says "this action can be cancelled within X hours" and then you as a user can decide what sort of actions the AI can do.

[_xkxo]

Totally fair point! We're planning to add a feature where you can record yourself doing some task and then [name redacted] can repeat it for you. Would that you give you more confidence that the AI will do the right thing? We're thinking this feature will useful for automating repetitive actions reliably.

[quickthrower2]

How about something similar to how websites say "Allow this site to use your camera" you make it so "Allow this form the be submitted by Genji", once they say yes, next time it doesn't need to ask.

[_xkxo]

Love that idea! And not just forms, there's other actions where users might want to have the opportunity to confirm like making payments.