Ref/ect: Self-Improving RL layer on top of Observability
Reflect. RL layer built on top of observability. It’s not a prank; we actually made observability and traces useful. Today, we’re releasing Reflect. Similarity is not enough for retrieval. We’re taking agents from searching what’s most similar to searching what actually gets the right trajectory and, thus, the right outcome. Here’s how it works. Built as a reinforcement learning layer on top of an observability platform, Reflect doesn’t just retrieve; it reasons about what to remember and plans […]