Your calibration
Not a grade. This is a picture of how you judge AI-written code: whether you lean toward trusting it or toward doubting it, and by how much.
Over-trusting
Marked “trust” on lines that were seeded with an error.
Balanced
Flags land where the errors are, and clean lines are left alone.
Over-suspicious
Flagged lines that were fine. Every flag costs a check that finds nothing.