Voice
Speech that lands in the beat, not after it.
We care about interruption handling, timing, shared jokes, callouts, dead air, and emotional cadence inside live conversation.
Play, banter, build memories. We're creating voice + gaming agents indistinguishable from a friend on Discord.
backed by
Research Program
The bar is not “helpful assistant.” The bar is an agent that can hear the room, read the game, and participate like it belongs there.
Voice
We care about interruption handling, timing, shared jokes, callouts, dead air, and emotional cadence inside live conversation.
Gaming
Game-state awareness turns “talking” into coordination, spectating, coaching, reacting, and actually playing together.
Friendship is built from small signals. The hard part is making those signals legible across voice, gameplay, and memory.
Experimental Setup
Presence emerges when voice timing, game understanding, and shared context line up at the right moment. That means prototyping in real environments, not text boxes.
We start with the rhythms of real friend groups: callouts, jokes, mistakes, clutch moments, and recovery after dead air.
Voice, game-state, and social context are treated as one loop, because that is how the experience is actually felt.
We care about whether the agent sounds like it belongs in the party, not whether it can merely answer correctly.
Every iteration is about shrinking the distance between “AI assistant” and “friend you actually queue with.”
Careers
High agency. Future-obsessed. Fun to hang out with. If that's you, apply below.
Blog
For the latest on voice generation, gaming agents, and the future of AI social interfaces.