Clean Code for AI Agents

Premise

In 2008 Clean Code optimized code for the next human reader. The primary reader is now the AI agent that greps, reads in truncated ranges, edits, and tests the code. Same principles, re-ranked; one inverts (comments), one is demoted with a nuance (DRY), and the persistence layer is now evidence-gated: every line an agent re-reads each session must earn its tokens.

Evidence base

This version is calibrated against measured results, not taste: