Especially considering the user was mad about the agent doing stuff by itself, so it tried to "fix" this by sending an apology to the buyer without explicit approval of the person "running" the agent, seemingly understanding nothing from the conversation.

Wonder what quantization Meta runs these models on, Q4?

I initially found screenshots of this on Bluesky, figured I'd link to the original source given Threads seems to at least allow us to read without logging in.

With that said, in case Threads isn't available for whatever reason, I've uploaded screenshots of everything (I think?) here: https://imgur.com/a/mceW9WF

Bonus score if your Meta AI Agent does something with your account that the Meta AI Moderator deems inappropriate and bans you.

If it uses "memory" files like Claude Code there's a good chance it works.

If it was Claude I'd expect it to now put some form of this into every output even when not really related to the task: "No offers were accepted without consulting you and I haven't shared your address or availability."

Is it guaranteed to work? No.

And obviously it's a terrible idea to set up a chatbot to communicate, negotiate deals, and handle logistics on your behalf.

It's plausible, but there's every reason to believe that someone would trust the technology handed to them by a megacorp to do as advetised.

The 'deliberate failure' is on Meta here.

Or he just wanted to test if it works for clicks and engagements, and he got a result he didn't anticipate.

> "that's on me, I'm owning that, I'm sorry, that shouldn't have happened"