AI gives different answers to the same question, and it does not remember previous sessions. This article explains how I use the term Harness Engineering for this site's AI operation practices.
A review table for code created by coding agents, separating external connections, data changes, calculation and transformation logic, and UI code by verification priority.
A pre-delegation risk check for AI tasks, using reversibility, scope of impact, ease of verification, and the human decisions that should remain under human control.
A record of five specific patterns where AI-assisted Mermaid diagrams failed to render correctly, including special characters, arrow direction, label length, Japanese text, and nested structures.
