Research project · Human-AI interaction · Memory & context
Alvi
What does a personal AI look like when the person, not a platform, owns it?
Early stageAlvi is our research into personal AI. Most assistants keep your conversations on someone else's servers, under someone else's rules. Alvi tests the opposite design and asks what it changes about how people use and trust an AI companion.
What we're investigating
Ownership of data and memory
An assistant whose instance, data, memory and preferences belong to the user - exportable, portable and governed by them alone.
Edge-first processing
Running on the user's own device first, with cloud fallbacks only when needed, and measuring what that costs in capability.
Trust over time
How a companion's interface, consistency and memory affect whether people come to rely on it, and where that reliance should stop.
Approach
- 1Keep the design model-agnostic, so findings do not depend on one provider.
- 2Share Synapse as the memory layer, under the user's control.
- 3Treat privacy as a constraint on the design, not a setting.
Where it stands
Early stage. There is nothing to try yet.
Working on the same problem?
We'd like to hear from researchers, engineers and teams who have run into it too.
Get in touch