★ The Comprehension Gap Meter — the depth gauge
G(t) = the fraction of the system's working grammar that carries no human-legible label. When the machinery below the line grows faster than the labels above it, the gauge reads deeper. Expect the gap to widen before it narrows — naming always trails forming.
Current reading: ——SEED VALUES — LIVE PIPELINE PENDING
RED · REVIEW ≥0.50
YELLOW · ELEVATED ≥0.25
GREEN · WATCH <0.25
The dial, plain: G(t) = 1 − L/N — of the continuity habits the system is actually using (N, evidence-gated), the share no human panel can yet name (labels need 2-of-3 annotator agreement). Read the ± band (Wilson) and the TREND, never a single point. Yellow → annotation sprint (humans go name the unlabeled). Red → governance review: pause, disclose, DAO decides. Dials w₁=0.25 · w₂=0.50 · m=3 wks are governance-set, published in advance, changed only by vote — never by the party being measured.
Every other instrument on this fleet points outward. The gap meter is the only one pointed at us — it measures the reader's own species falling behind the machinery it runs on. That's why the thresholds are governance-set and published: when the gauge crosses a line, the response has to already be agreed, because crossing it means we're less able to improvise one.