Our main goal is to build memory that can support intelligent systems.

Memory today, in the popular interpretation of intelligence, seems to live somewhere between retrieval, search, and longer context windows built around current architectures.

We think the problem is deeper.

First, today’s interpretation of intelligence is mostly agents, LLMs, and generative models, where reasoning is often represented through next-token prediction, probabilistic sampling, or iterative generative search-like processes. But human reasoning seems to be more than this. There is an implicit model of how the world works, a sense of continuity, and some deeper structure that is not captured by prediction alone.

Second, the naive understanding of memory often assumes memory is either inside the model or stored outside it in a database. But this does not fully reflect what memory does. To remember is not only to preserve information. It is to know what remains true, what changed, what matters now, and what should no longer guide the next decision.

The path to intelligence requires both directions to be explored. Reasoning is increasingly being studied, scaled, and optimized, but memory still feels like a dark map with terrains that have not yet been marked down.

We believe memory is itself an optimization problem. It deserves a real theory, an excellent architecture, and applications that make current systems more capable while supporting the future systems that come next.

This is the goal, and it is what excites us: contributing to true, safe intelligence and to the systems that may one day carry it into the world.

We are a team of next-generation researchers and builders, fascinated by the unexplored and eager to help guide the world into a new era of understanding.

We are looking for researchers, applied scientists, engineers, and sleepless people to help us solve this generational problem.

If you work on something near this, write to us.