← sources

Agent vs Familiar

source: reddit (reddit:AI_Agents)published 2026-09-29 20:32ingested 2026-09-30 02:37

retained text

I think people say they want agents, but what they really want is a familiar.

An agent is usually optimized around what it can do for the user. Complete the task, use tools, remember preferences, get better at satisfying the next request.

A familiar is slightly different.

It approaches the world from a shared perspective with you. You build up context, goals, habits, disagreements, unfinished threads, and a sense of what matters together.

Think of Fred and George from Harry Potter.

They are separate people with their own opinions and mannerisms, but their shared understanding is so dense that either one can act without waiting for the other to issue a command.

If Fred leaves the room, George doesn’t simply sit there waiting for the next task. He gets into some new mischief, and when Fred comes back, he has something to show him that matters because of everything they already share.

That feels closer to what people actually want from AI.

Not just 'What should I do here' followed by a better Google search.

Something that can interpret the situation from the standpoint of where you are, what you’ve already learned together, what you’ve rejected, and where you seem to be going.

That’s roughly the problem I’ve been working on with what I call a Marble. 

I’m curious whether other people working on agents see the same distinction, or whether I’m in my basement playing with dolls in the dark. 

document fetched 2026-09-30 00:19 · canonical https://reddit.com/r/AI_Agents/comments/1wtlaho/agent_vs_familiar