X-Arc

AI models work natively with what fits in their context window, so the memory around them decides how much they can do for you. We know this problem firsthand. Grove began as the memory of my own AI PA, and for months it has kept track of how I run my company. Every memory tool we tried struggled with the same 3 things: 1. a fixed taxonomy cannot keep up with a world that keeps changing. 2. facts hold for very different lengths of time. 3. something said is not something done. We believe a brain shouldn't be a connected database. It should be a living representation of your world. So we built Grove with that principle: a typed knowledge graph whose kinds of nodes and relationships can grow, while maintaining the structure of how a knowledge graph grows. Our application proposed taking Grove to the world, on top Meta's models. During the hackathon we're taking it out from our internal systems and build it as a product for anyone who operates with AI. Try it for free (thank you Meta) at https://app.grove.x-arc.ai/