Memory is the missing layer of intelligence
2026Why durable, personal context — not larger models — is the next real frontier, and how memory changes what an assistant can be.
Research
Intelligence that grows is a research problem first.
We are less interested in answering more questions and more interested in understanding people over time. Our research is built around memory, context, and trust.
Our research philosophy
Most progress in AI has come from scale: more data, more parameters, more compute. That work matters. But a model that forgets you the moment a conversation ends will never grow with you, no matter how large it becomes.
We study the harder, quieter problems — how an assistant can hold memory responsibly, understand the context of a whole life, and earn trust slowly. We publish openly, because intelligence that lives this close to people should be examined in the open.
Essays
Growing with people, not above them
2026A framing for systems that mature alongside a person across years, deepening understanding without ever owning it.
Consent as architecture
2025Treating privacy as a structural property of the system rather than a setting — and what that demands of us technically.
Whitepapers
- The LifeGraph: a personal memory substrateWorking paper
- Local-first inference for lifelong assistantsTechnical report