One agent, with permission
Instinct founder Noah Shinn described the privacy controls on X. A user’s personal agent must first ask permission to communicate with the Instinct agent in a group. Users choose which groups to trust and can revoke that permission at any time.
The group agent is isolated from participants’ personal accounts and cannot access them. A personal agent also asks before sharing information or taking an action. If someone new joins the group, it pauses sending any pending responses to the chat.
Shinn’s use case is mundane by design: friends trying to settle where and when to meet can compare options, agree on a plan and carry it out in one thread.
A crowded race for the group interface
Instinct and Muse have been moving at a similar pace in recent weeks; both learned to make phone calls over the past couple of weeks. The competition is for a share of the consumer AI market, and group conversations are becoming another place to put an agent in front of users.
OpenAI introduced ChatGPT Dots last week: personal AI agents that can also be used in ChatGPT Spaces, collaborative workspaces for people and agents. Instinct’s move is narrower, focused on bringing an agent into an existing group chat, even when some participants are not customers.
I think the key test is not whether the agent can help a group make a plan, but whether people are comfortable letting it participate. Instinct has described permission gates and a pause when a new person joins; the announcement says less about what happens when a group’s members disagree about letting the agent in. The feature is coming first to early-access users and will be available to everyone “soon.” Users can ask their Instinct agent to add them to the waitlist.
Daily AI news
Every day we pick what actually matters in AI and explain it plainly — no hype, no filler. Subscribe if you want to follow where the industry is going.
Only what matters — every day
Follow on X