Current directions
From capability to useful systems.
We are interested in what happens after a model can answer. How do systems preserve context? How should tools, memory, search, voice and interfaces work together? How can powerful models become easier to use in real workflows?
Workspaces
Interfaces for bringing models, tools, memory and context into one place.
Voice
Speech systems for natural output and responsive real-time experiences.
Interfaces
Patterns for making powerful AI systems clear, controllable and useful.
Sharing work
More to come.
We are building our research practice alongside our products. As ideas become mature enough to share, this is where we will publish notes, experiments and learnings.