Idea: "hide complexity inside" reads differently when the reader is an agent

[ortonom]

Hi Matt — not a bug, just an idea/consequence we ran into using codebase-design in an agent-heavy setup, in case it's useful.

The deep-modules framing is great. One prompt gave us pause once the primary reader and editor of the code became an agent rather than a human:

Can I hide more complexity inside?

For a human, hiding complexity behind a small interface buys working-memory relief. For an agent it cuts the other way — hiding complexity hides it from the agent too. The agent holds the whole context and has to read and edit behind the interface, so concealment removes visibility it actually needs, and accidental complexity just gets a nicer wrapper.

Suggested rewrite — one judgment line, no extra checklist:

Can I eliminate this complexity, not just move it behind the interface? Hide only what's essential — the reader (increasingly an agent) still has to see through the seam.

Thanks for publishing these.