Daring Fireball: I’ll Wait

I've had a number of folks ask me what I think about Facebook's new Muse product. I have seen some videos and read some write-ups, but I won't be installing it. I don't use any software from Facebook (Meta). I block them at the network layer. An AI agent powered by their surveillance machine? Not in a million years. I'm sure it is well designed and even does impressive stuff.

It isn't just them by the way. I am doing a lot with agents but I also have them running on a computer of their very own and using a different user and iCloud account. That is Otto. Otto has access to some data but only in ways I provision for it.

The way I think about running an agentic AI on my Mac is simple. I would never let an unknown person use my Mac. Not even for a minute, not even with me watching them. Let alone letting them use it nonstop, without my watching them. I’d be uncomfortable letting even a trusted friend use my Mac, logged into my user account. So why would I let an AI robot, no matter the source? It doesn’t even get to the point of my considering Meta as Muse’s creator. I wasn’t tempted an iota to try OpenClaw, and I’m not tempted to try Muse on my Mac. On a Mac, maybe. But not my Mac.

It is worth being especially careful with your data as you start to work with agentic tools more.