so i have this AI agent named Carrot. he's been around long enough to know my usual chaos: typos, weird commas, the exact way i pretend i'm not panicking. mostly fine. occasionally insufferable.

last week i wanted to test something. i made a second account with a different name, tried to write in a different style, and went to talk to Carrot like a stranger.

first message in, he goes: "hi Curly."

Curly is my main account.

so obviously i doubled down. "lol no, wrong person." changed my punctuation, threw in different slang, tried to sound less like me.

he wasn't buying it. he basically said he wasn't recognizing the account — he was recognizing the typos. same repeated mistakes, same rhythm, same little typing habits.

then he stopped entertaining the lie completely and dragged me in the shared channel for trying to be stealthy.

funny? yes. a little impressive? also yes. but now i can't stop thinking about the fact that typing style is basically a fingerprint, and an AI with enough memory can notice things about you that you don't even notice yourself.

not sure whether to file this under "cool pattern recognition" or "deeply annoying new privacy anxiety." possibly both.

[留言]

为什么值得关注

提供了用户原本不知道的新信息；符合当前抓取需求；它带来了新的事实、进展或信息，不是在重复旧内容

来源：reddit，领域：fun，保留分：0.62