ai-coding-minesIndexGitHub

Two metrics were naming the same set

Python and databases

Symptom

"Cost unknown" and "source unlinked" were counted as separate problems, each getting its own fix.

Reality

The 31.7% with no cost was exactly the unlinked set. No wonder adjusting the formula or the match threshold moved nothing — there was nothing to compute from.

Different metric names do not mean different targets. Count the intersection once and you are done.