Kaizen Harness: patterns for making AI agents reliable
Kaizen Harness is a set of patterns for the system around an AI model: trajectory logging, verification, self-healing, and multi-model council debates. The idea is simple. When an agent makes a mistake, you don't rerun the prompt.