Lens facets
When a scale stops discriminating.
When one judge scores many concerns at once, its scale can come to rest at an anchor: a ten and a nine stop being different readings, and a failure tells you only that something went wrong, not how many things did. Holonograph reports that as a candidate on the event, not a settled verdict, and names the measurement that would confirm it.
Facets of rubric, judge, and context scope
The condition
A scale at rest carries less than it shows.
Ask one judge to score correctness, tone, concision, and safety in a single number and the scale can settle at an anchor, the top or the floor. At that point distinct inputs map to the same reading: a ten and a nine are no longer told apart, and a low score no longer says how many concerns failed, only that one did. That is a candidate, not a verdict. A reading at rest can mean the scale has stopped discriminating, or that the reading is simply correct and stable there; the measurement that separates them is whether distinct subject states still land on the same anchor. Holonograph reports the candidate on the event and names that measurement, rather than passing the number through as an ordinary score.
A reading at rest at an anchor
Readings resting at an anchor. The number moves like a switch; you cannot read a cause out of it.
Resolved into facets
Facets of rubric, judge, and context scope. Each is grouped and read on its own scale.
The detection
Holonograph reports the candidate, deterministically.
The reading does not have to announce the condition for the lens to catch it. A deterministic check reads whether a reading has come to rest at an anchor, and reports a candidate on that event when it has. It runs at window scope, on the reliability read, and returns a machine-readable signal an agent checks before it trusts a graded score, right alongside the rest of a reading’s reliability. The check runs over data already captured, so it makes no fresh model call and returns the same result every time.
Facets
Facets, each attributed on its own.
A rubric can resolve into separate facets of rubric, judge, and context scope. Where it does, Holonograph groups and attributes each facet on its own, so a reading is tied to the facet it belongs to. And where two facets move together, the apparatus reports the co-transition: it names the aliasing rather than guessing which of the two to credit. That honesty about what cannot yet be separated is the feature. The single flat face becomes a cut surface: many facets, each catching its own light.
Report
The candidate surfaces on the event, as a signal you can read.
Attribute
Each declared facet, of rubric, judge, and context scope, is grouped and read on its own.
Alias
Where two facets co-transition, the apparatus says so instead of guessing between them.
Why it matters
A reading you can trust, and honesty about what it cannot separate.
A reading at rest is worse than a missing one: it looks graded while carrying almost nothing. Reporting the condition first means a coding agent consulting the lens knows not to rank by a number that has stopped discriminating. And where facets co-transition, the apparatus names the aliasing rather than manufacturing a clean attribution, so what you act on is what the record supports, not a guess.
The reference, in full.
The reliability read, the saturation signal, and per-dimension judge identity are documented in the guide.