Evidence tiers
The tiers
Every claim-bearing element on this site carries a status pill. The vocabulary is inherited from our internal five-tier system, and a status only ever moves by decision — never by copyediting.
| Tier | Meaning |
|---|---|
| IN CODE | Implemented, tested, exercised in the repository |
| UNTESTED | Implemented but cannot run: missing dependency or hardware |
| TO VERIFY | Claimed but not yet backed by a reproducible artifact |
| ASPIRATIONAL | Direction and intent, not yet built |
| OPEN PROBLEM | Genuinely unresolved: we don't have the answer |
Extra pills
Two extra pills you'll see
MEASURED marks a result backed by a reproducible repository artifact on real data. SYNTHETIC marks results demonstrated on synthetic sources only — demonstrated on synthetic sources only — not real data. We treat the distinction as load-bearing: SYNTHETIC never silently becomes MEASURED.
Report a bug
If you ever find a claim on this site without a tier, or a tier that seems upgraded from what the evidence supports, tell us. That is a bug of the most serious kind we recognize.