The heuristic memory-risk score can move in unintuitive directions and sits next to the benchmarked scanner, which is confusing.
Decision needed
Either:
- Make it a meaningful, defensible measure, or
- Shrink it to an unmistakably cosmetic indicator,
and surface the real benchmark number (benchmarks/results.json) distinctly from the heuristic.
Partially mitigated: the static map is now explicitly labeled heuristic/preliminary and separated from the benchmarked scanner — but the score's semantics still need work.
Source: docs/issues/inspect-followups.md (Risk score)
The heuristic memory-risk score can move in unintuitive directions and sits next to the benchmarked scanner, which is confusing.
Decision needed
Either:
and surface the real benchmark number (
benchmarks/results.json) distinctly from the heuristic.Partially mitigated: the static map is now explicitly labeled heuristic/preliminary and separated from the benchmarked scanner — but the score's semantics still need work.
Source:
docs/issues/inspect-followups.md(Risk score)