Opacity Ratio
NOUN · METRIC · AI PROVENANCE
The opacity ratio is the share of an AI output's claims that trace to no recorded support — the gap between what the output asserts and what its graph shows. It is measured over claims, not raw node counts, so it can't be diluted by padding — and it is the most serious quality signal in a Thread: gaps nobody can inspect, no matter how confident the output sounds.
In plain language
Every gray node is a step the agent took but didn't record. Not a wrong step, necessarily — an invisible one. Opacity is worse than error, because an error can be found and fixed; an unrecorded step can only be trusted or not. That's why opacity is the most serious signal on the scorecard: it puts a hard ceiling on how much verification is even possible.
It may be the single most important number in AI systems that nobody currently measures. Every deployed agent has an opacity ratio right now. Almost none of their operators know what it is.
See it live
Thread TT-0001 — TrustThread's own project — runs at 7% opacity: one unrecorded node out of fifteen, disclosed in gray rather than dressed up as analysis.
Inspect the gray node in Thread TT-0001 →Related terms
TERM COINED BY TRUSTTHREAD · FIRST PUBLISHED 2026-07-12 · DEFINITION V1 · THIS PAGE IS THE CANONICAL SOURCE