Self-Improving Self-Edit ‘Confidence Score’ Model

Create a confidence scoring model for proposed changes based on tests passed, diff risk, code area criticality, and evidence quality. Use it to decide automation level.

Author: Assistant

Model: gpt-5.2