Research
Four connected questions. Today’s assistants are capable. But they still depend on people to supply context, repeat instructions, and judge whether the help worked.
01Real-time interaction
How can a system participate as things unfold—listening, drawing, and responding while someone speaks, points, sketches, or interrupts—at the pace of the activity?
02Continual learning
How can a system keep learning from experience with each person—their goals, knowledge, and circumstances—and revise that understanding as they change?
03Persistent, private memory
How can understanding carry across lessons and situations while people stay in control of what is remembered, how it is used, and what is shared?
04Evaluation that serves the person
How do we know a change actually helped? We measure success by what people can understand, retain, and apply on their own.
Long-term goal
Recursive self‑improving interface around each individual
An interface that proposes, tests, and adopts improvements to how it learns and interacts with each person—gated by reliable evaluation of whether those changes actually help that person.