Research and principles.
We build real-time voice interaction and a trustworthy action layer for personal intelligence.
Principles
Presence
Beyond voice presence to life presence. Understanding your whole context, not just the current conversation.
Continuity
One assistant across all devices and time. Memory persists. Context carries over. Nothing is ever lost.
Action
Not just conversation—execution. The assistant can do things on your behalf, with permission.
Control
Permission-first behavior. User-controlled memory. Transparent actions. No hidden autonomy.
Voice presence vs. Life presence
Voice Presence
- Natural-sounding speech
- Emotional expressivity
- Handles interruptions
- Session-based memory
- Single device
- Conversation only
Sounds human
Life Presence
- Natural voice + deep context
- Understands your whole life
- Persistent memory over time
- Cross-device continuity
- Takes action on your behalf
- Genuinely useful daily
Understands and helps
Research posts
Building real-time voice interaction
Our approach to low-latency, natural conversation that feels human.
Conversational speech model architecture
How we achieve real-time speech-to-speech with unified understanding and generation.
Beyond voice presence: The case for life presence
Why natural-sounding voice is just the beginning of truly useful AI.
Trust through transparency
How we design for user control and predictable behavior in AI systems.
Cross-device continuity
Technical foundations for seamless context transfer across surfaces.
The action layer
Building an AI that can act on your behalf—safely, transparently, and with permission.