Kevin Liao argues that retrieving similar fragments from past conversations alone cannot reliably preserve a project’s current constraints and reasons for decisions, and proposes document-based memory. Released as Operator Memory, the workflow reads guidance, specifications and research notes before a task and updates them after implementation. The repository describes Markdown knowledge organized into personal-project, shared-project and user spaces, with Git sharing and no separate vector database or background agent required. This is a developer’s experience and tool design, however, not a comparative experiment demonstrating superiority over every retrieval-based approach.