Our Investment in Supermemory

LLMs are getting smarter—but their memory is still stuck in the past.

LLMs memory is often shallow, hard to customize, and fragmented across tools. What they really need is structured, interoperable, high-fidelity memory that works across apps, agents, and workflows.

This is the gap Supermemory is solving.

That’s why we’re thrilled to lead the $2.6M pre-seed round in Supermemory, an open-source project turned infrastructure company building the most interoperable, scalable memory layer for LLMs and AI agents.

The story starts with a “second brain” consumer app– a tool to capture and connect knowledge across apps and devices. Built by Dhravya Shah, Supermemory quickly gained a passionate user base—50K+ users, millions of saved items, and 10K GitHub stars in under a year. But under the hood, something more powerful was emerging: a new kind of memory engine that could serve not just people, but machines.

Dhravya, at 19, already has two exits under his belt. He filed a patent in AI infra at Cloudflare and dropped out of college to build Supermemory full-time. His insight: memory isn’t just search—it’s understanding, context, and recall. And LLMs don’t have that today.

Supermemory is fixing that.

They’ve built an open-source, full-stack memory system—vector database, content parser, extractors—designed to sync, store, and serve context across agents and platforms. It’s performant, developer-friendly, and modeled on how human memory works. Early users are already processing billions of tokens a day on the platform.

The long-term ambition is bold: to become the default memory layer for AI, enabling agents that remember you, learn from you, and get better over time.

You can try it yourself at supermemory.ai. We’re proud to be early partners to Dhravya and the team on this journey.