AI chat models, such as ChatGPT, are restricted by how much text they can process at once during interactions. To alleviate this, we investigate augmenting AI with an external read-write memory, allowing models to write crucial information to memory to reference later. We show that our architecture can drastically improve performance in AI's ability to customize responses to users, as well as solve complex olympiad-level coding problems. By allowing AI to infinitely accumulate knowledge and experiences in its external memory, we bring AI one step closer to be able to reason over extended horizons, just like humans!