The author argues that literate programming deserves reconsideration in the age of AI coding agents. Traditional literate programming, where code and prose are intertwined, has struggled with maintenance overhead - developers must keep both the code and explanatory text synchronized. While tools like Jupyter notebooks and Emacs Org Mode have supported this approach, adoption has remained limited due to the burden of managing parallel narratives.

However, AI agents fundamentally change this equation. Large language models excel at translation and summarization, making them perfectly suited to handle the prose-code synchronization problem. When using Org Mode with agents like Claude or Kimi, developers can write runbooks that combine executable code blocks with explanatory text. The agent can be instructed to treat the Org file as source of truth, handle tangling automatically, and maintain prose synchronization without human intervention.

The practical benefits are compelling. Developers can now test features by asking agents to generate Org-based runbooks, review the generated prose and code, execute interactively, and have results captured directly in the document. This eliminates the tedious manual maintenance that historically plagued literate programming while leveraging AI's strengths in understanding and generating human-readable explanations.