But the general use case is not the most efficient for a greenfield to-be fully managed by an agentic system code-base. It is built to be good around the scaffo

But the general use case is not the most efficient for a greenfield to-be fully managed by an agentic system code-base. It is built to be good around the scaffold(programming like humans) and not the actual problem space. Anthropic's target should be a codebase designed for agentic comprehension from the first commit. Here the codebase adapts to the agent. You can enforce conventions, structured metadata, semantic indexing, explicit dependency graphs. Whatever makes the agent's job trivial rather than heroic.