Mentions you can see. Typing @username in a message now renders it as a highlighted mention chip, so it's clear at a glance who's being addressed. Your display name is what appears next to your messages; your @username is the handle people use to mention you.
Cortez knows when to stay quiet. GroupGPT's assistant replies to messages by default — but when a message directly @-mentions another human member of the room, that message is meant for that person, not the AI. Cortez now recognizes this: he still sees the message (it's there in his context, so he's aware of who's talking to whom), but he doesn't jump in with a reply. The rule is precise — only a mention that matches a real room member silences him, so a stray typo or an @-mention of Cortez himself still gets a normal response.
Why it matters
In a real group conversation, side chats happen — “@alice can you grab the link?” — and an assistant that answers every one of those quickly becomes noise. Teaching Cortez to read the room and step back when two people are talking to each other makes him feel like a thoughtful participant instead of an interrupting one, without ever losing the thread of what's going on.