AI faces significant challenges in contextual memory, as current models find it difficult to maintain and recall pertinent information over time. Unlike humans, who can form long-term semantic connections, AI systems depend on fixed context windows, causing the loss of crucial prior interactions.
Zep is a startup that is creating a memory layer for AI agents by utilizing temporal Knowledge Graphs, allowing agents to preserve long-term contextual information. Founded in 2023, it was part of Y Combinator’s Winter 2024 batch.
Daniel Chalef, the Founder of Zep, is featured on the show alongside Kevin Ball to discuss the complexities of contextual memory in AI, temporal knowledge graphs, ambient AI agents, and more topics.
Visit Graphiti on GitHub and read the Zep White Paper on arXiv for further information.
Full Disclosure: This episode is sponsored by Zep.
Kevin Ball, also known as KBall, is the vice president of engineering at Mento and an independent engineering coach. He co-founded and was CTO of two companies, initiated the San Diego JavaScript meetup, and manages the AI inaction discussion group via Latent Space.